Author contributions: V.A.W.: manuscript writing and financial support.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS January 14, 2011; available online without subscription through the open access option.
The retina is the neural tissue located at the back of the eye that captures and processes light and transmits this information to visual processing centers in the brain, which enables us to see. Basic research in retinal development has provided important insight on the control of cell fate, proliferation, and neurogenesis in the central nervous system. This review summarizes the major cellular and molecular events that occur during retinal development and highlights how this knowledge may be harnessed for new therapeutic strategies to treat retinal disease. STEM CELLS 2011;29:412–417
The neural retina is an excellent model for investigating the regulation of neurogenesis and cell diversification in the developing central nervous system. It is accessible, structurally simple, and because it is nonessential for life, it is suitable for genetic manipulations of developmental genes that if performed in other regions of the brain would be lethal. Retinal research has also been in the spotlight recently because of the growing interest in regenerative medicine strategies for the treatment of neurodegenerative disease. Specifically, recent studies have shown that cell transplantation can restore visual function in mouse models of retinal degeneration , thus highlighting the potential feasibility of using similar approaches to treat these incurable conditions in humans. Our ability to generate photoreceptors for transplantation will require knowledge of the basic mechanisms of growth and diversification of the retina. This review highlights the key concepts of retinal development that will guide efforts to generate retinal neurons from stem cells (SCs) for therapeutic applications.
ORIGIN AND FUNCTIONAL ANATOMY OF THE RETINA
The adult retina is a laminated structure that is bordered apically by the retinal pigment epithelium (RPE), a single cuboidal layer of epithelial cells, and basally by the vitreous, the fluid-filled cavity that separates the retina from the lens. The cells in the adult retina are organized into three main layers that are separated by synaptic, or plexiform, layers (Fig. 1). Light passes through the lens and the thickness of the retina to stimulate photopigments in the outer segments of the rod and cone photoreceptors, which are located adjacent to the RPE. The photoreceptors, whose cell bodies comprise the outer nuclear layer, form synapses in the outer plexiform layer with horizontal cells and bipolar neurons, whose cell bodies are located in the inner nuclear layer (INL). Bipolar cells synapse directly or indirectly via amacrine neurons in the inner plexiform layer to ganglion cells (GCs) located in the GC layer. GCs, the projection neurons of the retina, transmit this information via their axons through the optic nerve to visual processing centers in the brain. Amacrine cell bodies are located in the inner part of the INL and also comprise roughly half of the cells in the GC layer. The retina also contains a population of radial glial cells called Müller glia whose cell bodies are located in the INL. Their apical processes form adherens junctions with photoreceptors at the outer limiting membrane and their end-feet attach to the basal lamina on the vitreal surface of the retina at the inner limiting membrane. Retinal astrocytes, which develop from a separate cell lineage at the optic disc, form a dense network on the vitreal surface of the retina and are closely associated with the vasculature.
THE DEVELOPMENTAL SEQUENCE OF RETINAL NEUROGENESIS
Initially, the retinal neuroepithelium consists entirely of proliferating progenitor cells, and the onset of neurogenesis is marked by the appearance of the first postmitotic neurons, GCs, in the center of the retina at approximately embryonic day 11 (E11) in mouse . In fact, in all vertebrate species studied, GCs are the first neurons to differentiate and the remaining retinal neurons and glia are generated in a temporal, but overlapping, sequence that is fairly well conserved across vertebrate species  (Fig. 2). GC development is followed closely by the development of cones, horizontal cells, and roughly half of the amacrine cells during embryogenesis. The remaining amacrine cells, bipolar cells, and Müller glia develop in a second postnatal wave . Rod photoreceptors are the exception, as they are generated throughout the period of retinal histogenesis . Lineage tracing in a number of vertebrate species has shown that retinal progenitor cells (RPCs) are multipotential—the progeny of single RPCs can include several, and in some cases, all the different retinal cell types (reviewed in ). Retinal neurogenesis is also spatially regulated, as differentiation is initiated in the central retina and progresses toward the periphery, such that the central retina (close to where the GC axons exit the retina through the optic nerve head) is more developmentally advanced than the periphery. The proliferation phase of retinal development ends at approximately postnatal day 7 (P7) in the mouse and is followed by a period of synapse formation and remodeling, cell death, and morphological maturation of photoreceptors that continues until P21.
Thus the sequential production of retinal cell types takes place over a relatively long time frame—approximately 2 weeks in rodents. What mechanisms are in place to balance RPC proliferation with differentiation so that the appropriate number and proportion of the different cell types are produced? The next sections summarize our current understanding of the intercellular and molecular regulation of these key processes. Because of space limitations, data from the mouse system will be emphasized with key findings from other vertebrate systems indicated where appropriate.
RPC PROLIFERATION AND LINEAGE SPECIFICATION
RPC maintenance and proliferation are critically important for ensuring the availability of the appropriate numbers of these cells throughout retinal histogenesis. Several transcription factors that regulate RPC identity, including Rax, Six6, and Pax6, also function as positive regulators of RPC proliferation . In addition, Pax6 is essential for RPC multipotency, as in its absence all RPCs adopt an amacrine cell fate . Sox2 regulates the neurogenic competence of RPCs, as Sox2 null RPCs fail to initiate neuronal differentiation . Vsx2 maintains the identity of RPCs by inhibiting the RPE differentiation program in these cells [7, 8].
RPC proliferation is also critically dependent on the expression of CyclinD1 (Ccnd1), which in partnership with cyclin-dependent kinases 4 and 6 (cdk4/6) promotes progression through the G1 or growth phase of the cell cycle . Relative to other embryonic tissues, RPCs express high levels of Ccnd1 and the retinas of Ccnd1 KO mice exhibit severe hypocellularity . The chief function of CyclinD1 in this context is to inhibit the activity of the p27Kip1(Cdkn1b), the primary cyclin-dependent kinase inhibitor (cdki) in the retina. Loss of p27Kip1 restores the normal cell count in the Ccnd1 KO retina . RPC proliferation downstream of Ccnd1 is promoted by the redundant activity of E2F transcription factors and N-Myc (Mycn) .
The specification of RPCs to the various neuronal cell types is regulated by transcription factors that belong to the bHLH, homeodomain, and forkhead families (summarized in Fig. 2; also reviewed in ). In some instances single genes are essential for lineage specification, such as Nrl for rod photoreceptors , Math5 (Atoh7) for GC [15, 16], and Vsx2 for bipolar cells . With the exception of Nrl, which is also sufficient to direct cone photoreceptors to the rod cell fate , few of these transcription factors are sufficient on their own for lineage specification, suggesting that they function with other factors to specify fate. This idea is consistent with the observation that Vsx2 on its own promotes INL identity, but in combination with Mash1(Ascl1) and Math3(NeuroD4), it promotes bipolar cell development . Thus, it is currently thought that lineage specification is a two-step process: homeodomain proteins specify layer identity and bHLH genes specify lineage identity within the layer .
RPC DIVERSIFICATION—THE IMPACT OF CELL INTRINSIC AND EXTRINSIC REGULATION
Although there has been tremendous progress toward understanding the molecular regulation of cell fate and proliferation in the retina, a key question still remains—how do RPCs make cell fate choices? A number of studies have investigated the relative contribution of cell intrinsic and extrinsic regulation of cell diversification in the retina, and there is substantial evidence that both mechanisms are at play (reviewed in ). Cell intrinsic regulation is particularly important in determining the timing of RPC competence to generate different retinal cell types. For example, RPC differentiate in age-appropriate ways that is independent of the environment in which they develop . Another important principle demonstrated in these studies is the irreversibility of the sequence of RPC competence—young RPCs cannot be induced to making older cell types precociously and old RPCs cannot return to producing young cell types even when they are exposed to younger RPCs . Finally, the cell fate choices of postnatal RPCs are remarkably similar whether they are left to develop in retinal explants or in clonal culture, demonstrating that the environment appears to have little impact on the types of cells generated by late stage RPCs .
The molecular basis for the intrinsic regulation of RPC competence is beginning to be unraveled. The transcriptional regulator Ikaros has been identified as a cell-intrinsic regulator of RPC competence that can promote early cell type production when it is ectopically expressed in late stage RPCs . MicroRNAs have also been implicated in regulating RPC competence  and in the frog retina they may do so by regulating the translation of cell fate determinants .
The plane of cell division and asymmetric inheritance of determinants is another mechanism to control RPC diversification (reviewed in ). RPCs undergo mitosis at the apical surface of the retina, adjacent to the RPE, and RPCs that divide with their spindles oriented parallel to the RPE (horizontal) generate two identical daughters, whereas progenitors dividing with their spindles oriented perpendicular to the RPE divide vertically and tend to generate two different daughters . Altering spindle orientation , or altering the expression of numb, an apically localized protein that is asymmetrically inherited in vertical divisions alters the fate of dividing cells . Signals from the RPE and retina have been shown to influence RPC division plane [27, 28]; however, the identity of these factors has not been determined. In the longer term, the challenge will be to understand how the determinants that are affected by the division plane integrate with other fate determining signals and factors, such as Notch signaling and proneural genes, to regulate RPC diversification.
Environmental cues also play a role in modulating RPC proliferation, cell fate, and the timing of differentiation. Studies in several vertebrate species have implicated GCs and the morphogen Sonic hedgehog (Shh) in the regulation of RPC proliferation (reviewed in [29, 30]). GCs express Shh and genetic ablation of GCs or Shh results in reduced RPC proliferation . Shh promotes RPC proliferation by regulating the expression of Cyclins D, A, and B [31, 32], and it also controls the timing of cell cycle exit, through the induction of p57kip2 (Cdkn1c) . However, loss of Shh signaling does not completely abrogate RPC proliferation, indicating that there are additional regulators of this process in the retina. These could include intrinsic regulators, including the transcription factors expressed in RPCs, and extrinsic signals, including EGF and FGF2 [34, 35].
Perhaps the best example for the effect of extrinsic signaling on cell fate is the evidence for feedback inhibition in the control of this process. In fish and rodents, the production of amacrine and GCs is negatively regulated by differentiated neurons of that class [36, 37]. In the case of amacrine cells, this effect is mediated, in part, by transforming growth factor beta signaling . Shh, growth and differentiation factor 11, and vascular endothelial growth factor have all been implicated as GC-derived factors that inhibit GC development from RPCs . In the example of GCs, how do extrinsic inhibitory cues integrate with the intrinsic regulation of GC competence? One possibility is that negative feedback signals control the final number of GCs at a local level until intrinsic changes take over and permanently lock RPCs out of GC production. This possibility is consistent with the evidence that after GC production has ceased in the retina, there is a short period when RPCs are competent to produce GCs when they are cultured in vitro away from the retinal environment .
At the onset of neurogenesis, RPCs are confronted with two choices—to produce a postmitotic neuron or to remain in cycle as a progenitor cell. Notch-Delta signaling is one of the best studied pathways in the context of the regulation of neurogenesis in the brain, including the retina . In the canonical Notch signaling pathway, the interaction of Notch ligands, including Delta and Jagged, with Notch receptors on neighboring cells results in the nuclear translocation of the Notch cytoplasmic domain, where it interacts with the Rbpj transcriptional coactivator to alter gene expression. The primary Notch pathway targets are Hes1 and Hes5, the two antiproneural bHLH transcription factors that block neurogenesis, in part, by inhibiting the transcriptional activity and function of proneural bHLH genes . Loss of function for Notch pathway components, including Notch1, Rbpj, and Hes1, as well as pharmacological blockade of Notch signaling result in precocious differentiation and increased neurogenesis in the retina [40–43]. Conversely, increased Notch activity delays differentiation and promotes the progenitor state . The analysis of Notch pathway mutants has also revealed nonredundant roles for these effectors on cell fate, suggesting that the function of this pathway is more complex [40, 41, 43]. Moreover, as other signaling pathways, including Shh, modulate the expression of Hes1 in RPCs , it will be important to determine how the integration of multiple extrinsic inputs influences the timing of neurogenesis and cell fate.
SCs IN THE ADULT RETINA
Once neurogenesis is complete, the adult retinal neurons are postmitotic, but Müller glia maintain expression of multiple RPC genes and on injury can be induced to re-enter the cell cycle (reviewed in ). These properties of Müller glia have raised speculation that this cell population could function as a source of endogenous SCs in the retina. Indeed, in fish, Müller glia are SCs and have the capacity to regenerate all retinal neurons . In chick, the regenerative capacity of Müller glia following injury is limited to fewer retinal cell types . In the mammalian retina, Müller glial cell proliferation following injury can be promoted with Notch, morphogens, and mitogen signaling; however, their neurogenic potential is very limited . Müller glia with SC characteristics have also been isolated from the adult human eye, and although they are not known to regenerate neurons in vivo, they can be induced to proliferate and differentiate into retinal neurons in vitro . Finally, in fish and frogs, continuous growth of the eye is also dependent on the activity of SCs located at the ciliary margin (reviewed in ). An analogous structure does not exist in higher vertebrates, and there is no evidence of a proliferating pool of neurogenic cells in the mammalian eye in vivo, but it is possible to isolate clonogenic pigmented cells from the adult mammalian ciliary margin, which can propagate in vitro and differentiate into retinal neurons , although their neurogenic potential is controversial .
APPLIED DEVELOPMENTAL BIOLOGY—THE BASIS FOR CELL TRANSPLANTATION THERAPY
Vision loss associated with the most common retinal diseases such as glaucoma, retinitis pigmentosa (RP), age-related macular degeneration (AMD), and diabetic retinopathy is caused by the death of retinal neurons and is irreversible. RP and AMD, which involve degeneration of photoreceptors and in some instances the RPE, are thought to be particularly amenable to transplantation because cell delivery to the subretinal space is relatively straightforward. Indeed, in diseases where the primary cause is RPE dysfunction, transplantation of healthy RPE cells has been shown to improve vision and this procedure is currently in clinical trials in AMD patients . Vision restoration in the case of photoreceptor degeneration would only require that transplanted healthy photoreceptors establish new connections with the existing bipolar and horizontal cells. A number of cell replacement approaches to restore vision have been attempted (reviewed in ), including subretinal implants of intact sheets of fetal retina  and transplantation of progenitor cells from the hippocampus , but these have had limited success in terms of improving vision because of a failure of either the grafted tissues to integrate with the host retinal circuitry or the grafted cells to differentiate into photoreceptors. In 2006, Maclaren etal.  reported that transplantation of dissociated photoreceptors from neonatal mice could restore some measure of visual function in mouse models of RP. Two key advances from this landmark study were the demonstration that the degenerating retina was receptive to new cellular input and postmitotic photoreceptors, rather than immature progenitor cells, were better at engrafting. Subsequently, Lamba etal.  reported that human embryonic stem (ES) cell-derived photoreceptors could integrate and improve visual function when transplanted into adult blind mice, demonstrating the feasibility of using in vitro-derived human photoreceptors for vision rescue.
There are many challenges that will have to overcome in the quest to bring cell replacement strategies to the clinic for the treatment of retinal degenerative diseases. The efficiency of photoreceptor induction and purification, whether from human ES cells, induced pluripotent cells or the adult ciliary margin, will have to be optimized to obtain clinically meaningful numbers of cells and to reduce the risk of tumor formation from contaminating immature cells. The efficiency of cell integration and synapse formation also requires optimization if we are to achieve vision rescue in patients. Some of the known barriers to cell integration are physical, such as the outer limiting membrane , as well as glial cells  and poor cell survival . Future studies will no doubt identify additional disease-specific barriers to the efficiency of this process.
Salient to this discussion are some of the many ways that developmental retina research has impacted the progress of retinal cell transplantation. The ability to identify retinal cells in these SC cultures and to manipulate photoreceptor development has been facilitated by our understanding of the lineage relationships in the retina, the extensive catalogue of cell type and stage specific markers, and the identification of photoreceptor-promoting extrinsic signals. For example, the signaling pathways that control anterior brain patterning have been exploited for the development of cell culture protocols to induce photoreceptor development from mouse and human ES cells [59, 60]. RPCs and photoreceptors can be identified on the basis of cell type-specific markers, and photoreceptor differentiation from primate ES cells is enhanced in the presence of factors such as retinoic acid and hedgehog . Moving forward, the manipulation of additional developmentally relevant signaling molecules will be an important avenue to explore in the effort to improve the differentiation process and transplantation efficiency of in vitro-derived photoreceptors. Finally, the identification of the genetic networks that specify retinal cell fates has also had an impact on the transplantation field. This information has provided the plethora of genetic tools for preclinical transplantation studies, including the Nrl-gfp mouse strain that was instrumental for the photoreceptor enrichment and tracking in the Maclaren etal.  retina transplantation study. Ectopic expression of photoreceptor inducing genes has been shown to promote photoreceptor development from SCs  raising the possibility of using genetic manipulation as a means to enhance photoreceptor development from SCs and possibly even somatic cells.
The progression from simple proliferating neuroepithelium to fully functional laminated adult retina involves a complex integration of proliferation, cell fate specification, and differentiation. In this review, we have summarized some of the key concepts and players that are involved at the earliest stages of retinal development. The challenges ahead are to tackle the issue of RPC competence at the molecular level and to address how environmental cues integrate with cell intrinsic factors to control RPC development. A detailed understanding of these processes is critical for understanding normal retinal development and for the development of efficient methods to control RPC differentiation for therapeutic purposes.
We are grateful to Drs. Nadean Brown, Hong Liu, and Alan Mears for comments on the manuscript. This work was supported by operating grants from the Stem Cell Network of Canada, the Canadian Institutes of Health Research, and the Foundation Fighting Blindness Canada to V.A.W.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The author indicates no potential conflicts of interest.