Generating retinal neurons by reprogramming retinal pigment epithelial cells

Abstract

Importance of the field

Retinal degenerations cause blindness. One potential therapy is cell replacement. Because the human retina lacks regeneration capacity, much attention has been directed towards searching for cells that can differentiate into retinal neurons.

Areas covered in this review

We discuss the possibility of using transcription factor genes to channel retinal pigment epithelial (RPE) cells’ capabilities of proliferation and plasticity towards the production of retinal neurons.

What the reader will gain

Experiments with chick embryos show that RPE cells – in the eye, in explant, or in a dissociated cell culture – can give rise to cells resembling retinal neurons when reprogrammed with regulatory genes involved in retinal neurogenesis. Depending on the regulatory gene used, reprogramming generates cells exhibiting traits of photoreceptor cells, amacrine cells and/or young ganglion neurons.

Take home message

Gene-directed reprogramming of chick RPE can efficiently generate cells that exhibit traits of retinal neurons. Remaining to be addressed is the question of whether the results from chicks apply to mammals. Since the RPE is located adjacent to the neural retina, RPE reprogramming, if successful in mammals, may offer an approach to repopulate the neural retina without involving cell transplantation.

1. Introduction

Vision begins in the neural retina, a thin tissue about 300 μm thick in the back of the human eye. In the neural retina, millions of neurons act in concert to produce visual input to the brain. Retinal neurons are categorized into groups according to their morphologies and functions. The primary, light-sensitive neurons are photoreceptors, which capture photons and generate electrophysiological signals. Those electrophysiological signals are then transmitted to and modulated by secondary, inner neurons, and finally relayed to the brain by retinal ganglion cells via their axons. These axons compose the optic nerve. As terminally differentiated cells, retinal neurons that die due to any cause cannot be replenished, leading to irreversible vision loss. The low quality of life caused by impaired vision has spurred intense interest in developing therapeutic approaches, including cell-replacement with endogenous cells generated by regeneration mechanisms (for a recent review, see [1]) or with exogenous cells delivered by cell transplantation (for a recent review, see [2]).

Intrinsic retinal regeneration would offer a straightforward approach to repair retinal damage caused by cell loss. In fish and chick, injuries to the retina induce the formation of new retinal neurons from progenitor/stem cells residing at the ciliary marginal zone [3–6]. However, this regeneration mechanism is lacking in the retina of mammalian species, including humans [7,8]. Although not canonical progenitor/stem cells, Müller glia in various species (including mammals) retain certain properties of retinal progenitor cells. For instance, in response to damage to retinal neurons, Müller cells will reenter the cell cycle, a proliferative response, and may give rise to retinal neurons, a regenerative capability [9–18]. Nonetheless, the regenerative capability of Müller cells in mammals is much lower than that in fish and birds [1], and challenges exist in employing Müller glia to efficiently repopulate a mammalian retina after retinal degeneration. The lack of effective retinal regeneration in the mammalian retina has fueled research exploring cells/tissues outside the neural retina, including the retinal pigment epithelium (RPE), as potential sources of new retinal neurons.

2. The RPE as a source of new retinal neurons

The RPE consists of darkly pigmented cells organized as a single-layered, transporting epithelium with important roles in retinal physiology. Anatomically, the RPE lies immediately adjacent to neural retina and forms the outer blood–retinal barrier. This anatomical location puts the RPE at a unique position for providing new neurons to repopulate a degenerating retina without cell transplantation. But a key biological question is whether the RPE, a non-neural tissue, is amenable to reprogramming to give rise to retinal neurons.

Developmentally, the non-neural RPE and the neural retina originate from the same structure – the optic vesicle. During development, the optic vesicle invaginates to form the double-layered optic cup, thus creating the anatomical separation of the RPE (the outer layer of the cup) and the neural retina (the inner layer). This common origin may bring about shared molecular and cellular characteristics and aid fate switches. Indeed, classic experiments have demonstrated an intriguing phenomenon: the RPE can give rise to a neural retina. In chick embryos, separating the RPE from the retina with a thread [19], or by surgically removing most of the retina [20], causes the RPE to develop into a neural retina. Later investigation found that basic fibroblast growth factor (bFGF) stimulates this remarkable transdifferentiation [21]. Rodent RPE from young embryos [22,23] and amphibian RPE [24,25] are also capable of RPE-to-neural retina trans-differentiation. Unfortunately, harnessing this biological phenomenon for retinal neuron production faces major obstacles, including a strict age-limitation: it takes place only during very early stages of embryonic development, prior to day 4.5 (E4.5) in chick [26] and prior to E13 in rodents [22,23].

Unlike retinal neurons, mature RPE cells of many species, including humans, can reenter the cell cycle to proliferate. In normal eyes, most RPE cells remain quiescent, probably in response to signals from the neural retina, and a small population located in the peripheral RPE undergoes proliferation [27]. Retinal detachment results in RPE cell proliferation [28]. In addition, certain pathological conditions and physical stimulations can cause RPE cells to proliferate. For instance, rat RPE cells reenter the cell cycle and proliferate in response to laser photocoagulation [29]. In the clinical setting, RPE’s proliferative response is an undesirable side-effect of surgery, because progeny cells may differentiate into cells with tractional force, causing retinal detachment and leading to visual impairment [30]. The proliferative responses of RPE and the plasticity of its progeny cells, nonetheless, raise an intriguing possibility of exploring the RPE as a convenient source of retinal neurons for replacement in situ.

With mounting knowledge about the genetic control of retinal neurogenesis, an unconventional approach has been formulated for producing new retinal neurons – gene-directed reprogramming to channel RPE’s capabilities of proliferation and plasticity towards retinal neurogenesis [31–35]. In these studies, RPE is taken from chick embryos at E6 – E15, long after the developmental loss of the competence of bFGF-induced RPE to neural retina transdifferentiation, and the progeny cells of the RPE cultured as dissociated cells are transduced with genes encoding transcription factors that play instrumental roles in retinal neuron production. More than 20 transcription factors involved in regulating retinal development have been assayed for their ability to reprogram RPE to differentiate towards retinal neurons [35]. These include genes encoding transcription factors in the basic helix-loop-helix (bHLH) family and homeodomain family. The homeodomain genes include Pax6, Rax, six3 and chx10, which are known to play important roles in the development of the eye and/or the neural retina. Despite their demonstrated importance, these homeodomain genes displayed insignificant activities in reprogramming RPE to produce cells expressing markers for retinal neurons [35]. The negative outcome does not undermine the importance of these genes in retinal neurogenesis; rather, it shows their ineffectiveness in the context of RPE cells. Among the bHLH genes assayed, four (ash1, ngn1, ngn2 and ngn3) induced RPE cell cultures to give rise to neuron-like cells and neuron-like clusters apparent by light microscopy (Figure 1) [32,34,35].

Figure 1. Neuron-like cells and neuron-like clusters in RPE cell cultures after reprogramming by ash1 (B), ngn1 (C), ngn2 (D), and ngn3 (E).

Neuron-like cells and neuron-like clusters are absent in the GFP control (A).

3. Generating photoreceptor-like cells by RPE reprogramming

Cell typing analysis showed that photoreceptor-like cells were produced in large numbers in chick RPE cell cultures after reprogramming by ngn1, ngn2 or ngn3 (Figure 2) [32,35]. NeuroD, a bHLH gene involved in photoreceptor differentiation, also guided cultured RPE cells to differentiate into photoreceptor-like cells [31,36,37]. Among the genes with significant activity in inducing RPE to photoreceptor reprogramming, ngn1 and ngn3 emerged as the two front-runners, in terms of inducing the highest number of cells expressing visinin [35], an early marker of chick photoreceptors.

Figure 2. Photoreceptor-like cells produced from RPE reprogramming by ngn1 (A,B,E) or ngn3 (C,D,F).

A, C: Bright field view. B, D: Epifluorescence immunostaining for visinin. E, F: Immunostaining for red opsin. Arrows point to the cell body, and arrowheads point to a structural feature reminiscent of the lipid-droplet typically present in chick photoreceptors.

Analyses at the molecular, cellular and physiological levels showed that the reprogrammed cells display advanced photo-receptor traits [35,37]. The new cells expressed transcription factors crx, nr2e3, raxL, RXRγ and neuroD, which participate in initiating the photoreceptor differentiation program [35]. They also express components of phototransduction, including red opsin (Figure 2E and F) [35,37], the α-subunit of cyclic nucleotide gated channels and cone α-transducin. Red opsin⁺ cells displayed dot-like immunostaining at the apices of the cells, reminiscent of the immunolocalization of red opsin of in photoreceptors in the retina, indicating a proper localization of red opsin [35]. Morphologically, in contrast to the hexagonal RPE cells, visinin⁺ cells resembled photoreceptors, with an elongated cell body, an axon-like process, an inner segment-like compartment, and a lipid droplet-like structural feature (Figure 2A – D, arrowheads) [35]. EM analysis showed that reprogrammed cells developed cellular compartments rich in mitochondria [35,37]. On the apex of the inner segment, reprogrammed cells displayed ciliary expansions, reminiscent of the developing outer segments of retinal photoreceptors in the E17 eye or in culture [35]. Perhaps most excitingly, reprogrammed cells displayed physiological traits typical of retinal photoreceptors. Retinal photoreceptors exhibit two physiological hallmarks – light response and visual recovery. Both properties are reflected as changes in by cytosolic, free calcium (Ca²⁺) levels. Fluorescent calcium imaging showed that reprogrammed cells responded to light by decreasing their Ca²⁺ levels and responded to 9-cis-retinal by increasing their Ca²⁺ levels [35,37].

To be a convenient source of photoreceptor cells, the RPE in the eye needs to be reprogrammable. To test this, experimental chick embryos were examined for visinin expression in the RPE layer. In control E7.5 chick eyes, visinin⁺ cells were confined within the neural retina at the prospective outer nuclear layer (ONL). In eyes infected with replication competent avian sarcoma-leukosis virus long terminal repeat with splice acceptor (RCAS)-ngn3, visinin⁺ cells were present in both the RPE and the neural retina (Figure 3) [38]. The RPE layer in RCAS-ngn1-infected embryos contained far fewer visinin⁺ cells. This was unexpected, because with dissociated RPE cells ngn1 was no less effective than ngn3. To test whether the eye with an intact retina exhibited some inhibition on ngn1-RPE to produce otherwise surplus photoreceptor cells, the RPE from the embryos was separated from the neural retina and cultured as an explant, to unleash the RPE’s potential to produce visinin⁺ cells. Indeed, RPE explants from embryos infected with RCAS-ngn1 produced layers of visinin⁺ cells, while the GFP control did not [38]. As with dissociated cell cultures, reprogrammed cells from RPE explants expressed genes associated with advanced photoreceptor differentiation [38].

Figure 3. Visinin⁺ cells in the RPE layer after reprogramming by ngn3.

Shown are bright-field (A), visinin immunofluorescence (B), and simultaneous view of both (C) of an E7.5 eye infected with RCAS-ngn3. Asterisks mark the RPE.

Another important issue is whether mature RPE is reprogrammable. Li et al. [38] reported that layers of visinin⁺ cells emerged from the RPE explants from chick at E12, when the RPE is close to maturity, and abundant visinin⁺ cells emerged in RPE explants from chick at E18, when the visual system is functional. These results suggest that mature RPE is reprogrammable.

4. Reprogramming RPE towards other retinal neurons

Variations were observed in the cell products from the reprogramming by bHLH genes. Reprogramming of RPE cells by ngn1 and neuroD generates mostly cells displaying photo-receptor traits [31,35,36,38]. On the other hand, reprogramming by ngn2 results in the genesis of cells of molecularly and morphologically diverse types, which resemble photoreceptor cells, ganglion cells, and, in a smaller number, amacrine cells [32]. Similarly, reprogramming by ngn3 produces diverse types of cells, with photoreceptor cells and, to a lesser extent, ganglion cells as the main products [35], and some cells displaying similarities to amacrine cells (Figure 4A and B).

Figure 4.

An RA4⁺ cell (A) and calretinin⁺ positive cells (B, C) produced by reprogramming cultured RPE cells by ngn3 (A, B) or by ash1 (C).

Different populations of cells were produced from reprogramming by ash1 (Figure 4C) [34]. In ash1-reprogrammed RPE culture, the new cells developed elaborate processes characteristic of neurons and expressed genes/markers that identify different types of retinal neurons [34]. The most prevalently expressed neural marker was calretinin, which in the chick retina identifies amacrine, ganglion, and horizontal cells. In an assay for the presence of functional, ionotropic glutamate receptors that lead to a rise in the cytosolic free calcium (Ca²⁺) concentration, the reprogrammed cells responded to glutamate and N-methyl-D-aspartate by increasing their Ca²⁺ concentrations, and after reaching a peak level, returned to the basal level. The study [34] suggests that chick RPE progeny cells can be reprogrammed by ash1 to develop molecular, morphological, and physiological properties that are characteristic of retinal neurons, probably amacrine cells.

5. Reprogramming RPE towards retinal ganglion-like neurons

Sox2 and bFGF have also been assayed for activities to reprogram RPE cells to differentiate towards retinal neurons. Sox2 belongs to the SoxB1 subfamily of transcription factors, which are characterized by a high mobility group (HMG) DNA-binding domain. It is one of four genes used to convert adult fibroblasts into induced pluripotent stem cells in mice and in humans [39–42]. Sox2 is expressed in the neural plate and in neural stem cells and progenitors during both embryonic and adult neurogenesis, in which sox2 plays an important role in maintaining neural progenitor/stem cell properties [43–46]. In the developing retina, sox2 is mostly expressed in proliferating progenitor cells [47,48], and the RPE appears to lack sox2 expression [48]. In the mouse retina, decreasing the expression of sox2 results in progenitor cells that cannot proliferate or differentiate [49]. In humans, sox2 mutations cause anophthalmia [50]. These findings show that sox2 plays an important role in the wellbeing of retinal progenitor cells. This particular role of sox2 prompted the idea that sox2 might reprogram RPE to differentiate towards retinal neurons. This idea has been experimentally tested in chick RPE cells [48].

Ectopic expression of sox2 in the chick RPE makes the experimental eyes ‘spotty’ in general appearance [48]. Closer examination has shown that the RPE cells in the ‘spotty’ regions partially or completely lack their usual dark pigmentation (Figure 5A and B) [48]. In the affected regions, BrdU incorporation was detected, indicating that those cells had reentered the cell cycle [48]. In addition, over 50% of the cells in the depigmented regions of the RPE are positive for RA4 immunoreactivity (Figure 5C and D) [48], a marker of newborn ganglion cells [51]. Some of the RA4⁺ cells exhibited neural morphologies. Cells in the depigmented regions were also immunopositive for 3A10 and 4H6, two monoclonal antibodies against a neurofilament-associated protein and a neurofilament, respectively [48]. Thus, sox2 was able to reprogram cells in the RPE layer of the eye to lose their normal appearance, to take on neural-like morphology, and to express markers atypical of RPE but typical of retinal ganglion neurons.

Figure 5. Sox2 induced the loss of pigmentation and the presence of RA4⁺ cells in the RPE layer of chick eyes.

A: Cross-section of a control E15 retina infected with RCAS-GFP. B: Cross-section of an experimental E15 retina infected with RCAS-Sox2. C: RA4 immunostaining of an experimental retina. D: Higher magnification of C. Asterisks mark the RPE. Arrows point to RA4⁺ axons of retinal ganglion cells. Arrowheads point to RA4⁺ cells in the RPE layer.

AX: Axons of retinal ganglion cells; RGC: Retinal ganglion cells; RPE: Retinal pigment epithelium.

As in the eye, sox2 guided RPE cells in culture away from their usual path and towards becoming retinal neurons. Dissociated RPE cell cultures subjected to sox2 reprogramming failed to become repigmented, which occurs in the control (Figure 6A and B) [48]. In the sox2-reprogrammed RPE cell cultures, RA4⁺ (Figure 6C and D) and 3A10⁺ cells were present in large numbers, accounting for more than 50% and more than 30%, respectively, of the total cells present. Morphologically, some of the RA4⁺ (Figure 6E and F) or 3A10⁺ cells exhibited long processes, markedly deviating from the hexagonal or fibroblast-like morphologies of RPE cells in culture. Reprogramming by sox2 decreased the expression of Mitf, Otx2 and Mmp115, genes with important roles in maintaining RPE properties, and increased the expression of neural genes Pax6 and Npy-1 [48].

Figure 6. Sox2 induced hypopigmentation and generation of RA4⁺ cells in dissociated RPE cell cultures.

A, B: Two 35-mm dishes with RPE cell cultures infected with control RCAS-GFP (A) or RCAS-Sox2 (B). C, D: RA4 immunostaining of a control culture (C) and an experimental culture (D). E, F: Individual RA4⁺ cells viewed at a high magnification.

Notably, Sox2-initiated RPE reprogramming produced cells with a somewhat limited extent of neural differentiation, both in the eye and in RPE cell cultures. This is not surprising, considering that Sox2 is well known to promote progenitor/stem cell status and inhibit neural differentiation. This limited extent of differentiation suggests that Sox2 redirects the differentiation of RPE cells towards retinal neurons, but continued neural differentiation requires additional factors. Future studies aiming for advanced neural differentiation may include regulating Sox2 expression and incorporating factors/genes that promote neural differentiation. Previous studies have shown that ath5 and NSCL1, which are bHLH genes transiently expressed in differentiated retinal ganglion cells, promote bFGF-induced RPE reprogramming towards retinal ganglion cells [33,52]. Because of the expression of differentiation markers, the reprogrammed cells were unlikely to be progenitor cells.

The lack of re-pigmentation and the presence of RA4⁺ cells in dissociated RPE cell cultures have previously been observed with the administration of bFGF to the culture [53]. This prompted an investigation of a possible link between sox2 and bFGF. The spatial patterns of expression of sox2 and bFGF overlapped in the developing retina, and their co-expression was detected in some retinal cells, particularly in Müller glia during late stages of development. Retinal expression of both sox2 and bFGF, assayed by in situ mRNA hybridization, was reduced by Müller glia atrophies and was enhanced by NMDA damage. Furthermore, RT-PCR analyses showed that sox2 was induced in RPE cell cultures treated with bFGF, while bFGF expression was enhanced by sox2 in the retina and in Müller glia culture. These results suggest a positive relationship between sox2 and bFGF and imply their involvement in Müller glial responses to retinal damage.

Significant variations exist in terms of the efficiency at which cells expressing ganglion markers emerged from reprogrammed RPE cell cultures. Sox2 is the highest, rivaled only by bFGF. However, in bFGF-initiated RPE reprogramming, the neural differentiation, if any, is very primitive [53]. All the RA4⁺ cells in bFGF-treated RPE cell cultures lack any commensurate neural morphologies, but rather retain the flat, fibroblast-like morphologies characteristic of cultured RPE cells before reaching confluency. The bFGF-reprogrammed differentiation towards retinal ganglion cells can be enhanced by bHLH genes ath5 and NSCL1, each alone [52] or more pronouncedly in combination [33]. In ngn2- or ngn3-induced RPE reprogramming, the molecular and cellular differentiation towards retinal ganglion cells is advanced, but less than 1% of the cells in ngn2-reprogrammed cultures [32] and 5 – 10% of the cells in ngn3-reprogrammed cultures [35] exhibit such differentiation. In addition, these cells are far outnumbered by those with photoreceptor traits, creating a problem of a mixed population of cell types. In comparison, sox2-reprogrammed cultures contain no photoreceptor-like cells [48], and more than 50% of the cells are positive for retinal ganglion cell marker RA4. Thus, sox2-induced RPE reprogramming is comparatively better suited for generating non-photoreceptor neurons, particularly ganglion cells.

6. Specificity of gene-directed RPE reprogramming

The variations in the products of gene-directed RPE reprogramming indicate not only that the RPE cells are responsive to the gene-directed reprogramming, but also that the products of reprogramming display gene-specificities. A comparative analysis indicates that the product population from the RPE reprogramming more or less agrees with the known or implicated roles of each corresponding gene in retinal development of the chick (Figure 7). Studies of the expression of bHLH genes in the developing chick retina showed spatial overlaps of ngn3, ngn1 and neuroD expression [54,55]. Functional studies suggest that ngn1 may play a major role in steering a progenitor cell towards the photoreceptor path during retinal neurogenesis in the chick [54]. On the other hand, ngn2 showed a temporal and spatial pattern of expression consistent with its role in progenitor cells that may later differentiate into all major types of retinal cells, while ash1 is expressed in progenitor cells likely to adopt an amacrine cell fate [56,57]. Investigation into their genetic relationships showed induction of ngn1 by ngn3, but not by ngn2 or neuroD, while ngn3 or ngn1 inhibited the expression of ngn2 and ash1 [54,55]. Overall, there appears an inductive pathway of ngn3 to ngn1 to neuroD and an inhibition of ngn2 (and ash1) by ngn1 and by ngn3 (Figure 7A). Thus, retinal neurogenesis seems to employ a complex network of bHLH regulatory factors that are expressed during different time frames and participate in the genesis of diverse cell types, for balanced production of all retinal cell populations. In the chick retina, sox2 is expressed in progenitor cells and later in Muller glia and a small number of cells in the amacrine cell layer and ganglion cell layer [47,48], consistent with its know importance for maintaining retinal progenitor cell properties [49]. In addition, the limited neural differentiation of in sox2-induced RPE reprogramming is consistent with the well known role of sox2 in inhibiting neural differentiation.

Figure 7.

A: Sketch showing the proposed genetic relationship among bHLH genes in chick retinal neurogenesis. B: Diagram of gene-directed reprogramming of RPE showing genes that have been assayed and the cell types that the major products mostly closely resemble. Parentheses indicate that the cell type constitutes a minor component of the resultant reprogramming products. All, all other types of retinal cells.

Am: Amacrine cells; Gc: Retinal ganglion cells; Pr: Photoreceptors.

7. Other potential sources of new retinal neurons

The recently-demonstrated multipotency of adult stem cells has spurred interest in testing ocular cells outside the neural retina for their capacity to generate retinal neurons. Ocular tissues/cells being explored include the iris pigment epithelium [58–62], the ciliary body [61,63–65], the limbal epithelium [66], and the RPE [31,32,35,37,38,52,67]. With the exception of the RPE, all these ocular cells/tissues give rise to disappointingly low numbers of photoreceptor cells. Notably, early reports of the presence of retinal stem cells in the ciliary epithelium of the mammalian eye [63,64] have recently been contested [68].

Mammalian stem cells of various origins have been explored for their ability to produce retinal neurons. Initial studies tested stem cells isolated from adult brain and bone marrow and met with very limited success [69–73]. Recently, significant progress has been reported in guiding human embryonic stem (ES) cells and induced pluripotent stem (iPS) cells to differentiate into retinal neurons [74–80]. ES cells can give rise to retinal progenitor cells that exhibit similar gene expression profiles such as those derived from human fetal retina [74]. These ES cell-derived retinal progenitors can further differentiate into inner retinal neurons with functional glutamate receptors [74]. Other study shows that after transplantation into genetically modified mice, ES-derived photo-receptors can restore some visual function [74,76]. When subjected to a targeted and stepwise differentiation protocol, ES and iPS cells progressively narrow their potential fates to the eye field, then to the optic vesicle, the optic cup, retinal progenitors and finally to differentiated cells of the retina, particularly RPE cells and cones [75,77–79]. These new developments are very exciting and support the prospect of generating specific types of retinal cells from ES cells and especially from iPS cells for autologous cell transplantation as a therapy for retinal degeneration [81]. Notably, cell replacement is just one aspect of cell therapies, which may include the use of cells derived from embryonic stem cells and iPS cells to prevent or to slow down the degeneration of a particular type of retinal neuron without producing such a specific type of retinal neuron from the stem cells [82–84].

8. Conclusion

Studies with the chick system show that genes encoding transcription factors participating in retinal neurogenesis can guide RPE cells to differentiate towards retinal neurons. To produce ganglion-like cells, sox2 can be used, but additional genes/factors that promote neural differentiation are needed. To produce photoreceptor-like cells, the bHLH genes ngn1 or ngn3 are thus far the best choice to reprogram RPE cells. The new cells produced from ngn1-induced RPE reprogramming express a spectrum of photoreceptor genes, including transcriptional factors that set in motion the photoreceptor differentiation program and components of the phototransduction pathway. These cells exhibit structural features typical of developing photoreceptors. Perhaps most excitingly, they develop physiological traits that are hallmarks of photoreceptors, such as response to light and to 9-cis-retinal after light-bleaching. These results support the exciting possibility of deriving functional photoreceptors from the RPE.

9. Expert opinion

In this article we have discussed gene-directed ‘RPE to retinal neurons’ reprogramming as a possible approach to produce new retinal neurons. The theme builds upon RPE’s known abilities – cell proliferation and plasticity – and channels them towards the production of new retinal neurons, especially photoreceptor cells and ganglion cells. Photoreceptor-like cells are produced at high efficiency with RPE reprogramming by bHLH genes ngn1 and ngn3, while cells expressing early ganglion markers are efficiently produced from sox2-initiated RPE reprogramming. RPE reprogramming by bHLH genes showed advanced photoreceptor differentiation at the molecular, cellular, and physiological levels. On the other hand, reprogramming by sox2 produced cells with limited extents of neural differentiation. Importantly, ngn3 and sox2 can initiate RPE reprogramming in vivo in the eye, in addition to in vitro with cultured RPE cells, supporting the possibility of deploying the RPE as a convenient source of retinal neurons for cell replacement in situ in the eye without the involvement of cell transplantation.

While the results from studies with the chick system are interesting, it will be imperative to show that the results from chick studies are applicable to human cells. If the finding apply to humans, there is tremendous scientific interest and societal significance. The convenient location of the RPE makes the approach attractive for developing cell-replacement therapies to treat blindness due to retinal degeneration without cell transplantation. Additionally, some questions critically important to the prospect of the RPE reprogramming approach need to be addressed. i) Will there still be RPE remaining after the reprogramming? RPE plays important roles to the well-being of retinal neurons. Current information supports a belief that the RPE may regenerate itself during the process by deploying its proliferation and wound-healing traits, but studies directly addressing this concern are needed. ii) Will gene-directed reprogramming using insertional viral vectors cause mutagenesis and tumorigenesis? iii) Will advanced ganglion cell differentiation be achieved by regulating sox2 expression and incorporating factors/genes that promote neural differentiation? iv) Will the reprogrammed cells detach from the RPE, migrate to the appropriate place, and establish circuitry connections with partner cells to fulfill their functional roles? These are just some of critical questions that still need to be addressed in investigating the feasibility of RPE reprogramming as an approach to produce new retinal neurons for cell replacement studies.

Article highlights.

• The biological properties and the anatomical location of the retinal pigment epithelium (RPE) make it attractive as a convenient source of new retinal neurons to repopulate the retina afflicted with retinal degeneration without cell transplantation.

• Proneural basic helix-loop-helix (bHLH) genes that participate in the genetic hierarchy of photoreceptor genesis in the neural retina can effectively reprogram RPE to give rise to photoreceptor-like cells. This reprogramming can occur in the eye, in RPE explants, and in dissociated RPE cell cultures.

• The reprogrammed cells can develop advanced photoreceptor traits.

• Sox2-initiated RPE reprogramming gives rise to cells expressing markers of retinal ganglion cells.

• Imperative future studies will include whether human or mammalian RPE is amenable to gene-directed reprogramming to give rise to photoreceptor-like or ganglion-like neurons.

This box summarizes key points contained in the article.