Adult Ciliary Epithelial Cells, Previously Identified as Retinal Stem Cells with Potential for Retinal Repair, Fail to Differentiate into New Rod Photoreceptors§


  • Author contributions: S.G.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; M.B.: collection and/or assembly of data, data analysis and interpretation; J. L.: collection and/or assembly of data, data analysis and interpretation; S. D.: collection and/or assembly of data, data analysis and interpretation, financial support; A.S.: provision of study material or patients, collection, and/or assembly of data; R.A.P.: conception and design, manuscript writing, financial, support; R.R.A.: conception and design, manuscript writing, financial support; J.C.S.: conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript, financial support.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS March 31, 2010.


The ciliary margin in lower vertebrates is a site of continual retinal neurogenesis and a stem cell niche. By contrast, the human eye ceases retinal neuron production before birth and loss of photoreceptors during life is permanent and a major cause of blindness. The discovery of a proliferative cell population in the ciliary epithelium (CE) of the adult mammalian eye, designated retinal stem cells, raised the possibility that these cells could help to restore sight by replacing lost photoreceptors. We previously demonstrated the feasibility of photoreceptor transplantation using cells from the developing retina. CE cells could provide a renewable source of photoreceptors for transplantation. Several laboratories reported that these cells generate new photoreceptors, whereas a recent report questioned the existence of retinal stem cells. We used Nrl.gfp transgenic mice that express green fluorescent protein in rod photoreceptors to assess definitively the ability of CE cells to generate new photoreceptors. We report that CE cells expanded in monolayer cultures, lose pigmentation, and express a subset of eye field and retinal progenitor cell markers. Simultaneously, they continue to express some markers characteristic of differentiated CE and typically lack a neuronal morphology. Previously reported photoreceptor differentiation conditions used for CE cells, as well as conditions used to differentiate embryonic retinal progenitor cells (RPCs) and embryonic stem cell-derived RPCs, do not effectively activate the Nrl-regulated photoreceptor differentiation program. Therefore, we conclude that CE cells lack potential for photoreceptor differentiation and would require reprogramming to be useful as a source of new photoreceptors. STEM Cells 2010;28:1048–1059


Degeneration of photoreceptor cells in the neural retina (NR) is a major cause of untreatable blindness. Multiple retinal diseases such as retinitis pigmentosa, cone dystrophy and age-related macular degeneration are all characterized by photoreceptor degeneration. Recent studies have demonstrated that photoreceptor transplantation is feasible in mice [1–3]. To establish photoreceptor transplantation therapy in humans, a source of postmitotic photoreceptor precursors is required [1]. Human fetal and embryonic material, not withstanding ethical implications, is very limited and the access is strictly regulated. Stem cells, with properties of self-renewal and with the potential to produce large numbers of neurons in vitro, offer an ideal donor source of cells to generate photoreceptors. Recently, independent groups reported the differentiation of human and mouse embryonic stem (ES) cells to a retinal progenitor-like fate, which were further committed to generate cells expressing photoreceptor markers in vitro [4–7]. Although ES-derived retinal cells may be useful for transplantation, the problem of host rejection remains.

Ideally, stem cells would be acquired from the patient to avoid the problem of graft rejection. Many investigators have recently focused their research interest on the reprogramming of somatic cells by virus gene transfer to generate induced pluripotent stem (iPS) cells [8–10]. However, the mutagenesis-risk associated with random lentivirus integration and the teratoma-forming potential of such cells requires further attention to make iPS cells practicable.

Adult tissue-specific stem cells theoretically do not bear these risks, but could be harvested from the patient and expanded for auto-transplantation. It has long been known that lower vertebrates, such as teleosts or amphibians, display the ability to generate new retinal neurons throughout life from a region called the ciliary marginal zone (CMZ), at the anterior rim of the retina [11–13]. A less potent CMZ was also discovered in the chicken [14] which displays a limited capacity for regeneration. Although the adult mammalian eye does not display these regenerative features, in 2000, Tropepe et al. reported that the ciliary epithelium (CE) of the murine eye contains a population of multipotential retinal stem cells [15]. Single pigmented ciliary epithelial cells clonally proliferate in vitro in the presence of mitogens to form sphere colonies of cells (neurospheres) [15]. When cultured under differentiating conditions (high levels of fetal bovine serum (FBS) and without mitogens), they were reported to differentiate into retinal-specific cell types, including rod photoreceptors, bipolar neurons, and Müller glia [15]. Multipotential neural progenitors in the adult mammalian CE were independently reported in the same year [16]. Follow-up research with human and pig CE cells introduced techniques to expand these cells in monolayer cultures [17, 18]. These cells may be ontogenetically closer to multipotential retinal progenitor cells (RPCs), which during development generate all of the NR cell-types, than ES or neural stem cells and thus easier to differentiate into photoreceptors. Indeed, Coles et al. reported that when these cells were transplanted into adult mice they generated new photoreceptors, albeit in small numbers. These studies indicated that if the numbers of CE-derived cells could be increased and their differentiation optimized, then they offered the prospect of providing a cell source for photoreceptor cell replacement. Most importantly, they could raise the possibility of autologous transplantation when derived from the patient's own eye. A number of laboratories have pursued these objectives and have reported the expansion of cell numbers and/or generation of photoreceptors from CE-derived cells [17–24]. However, a recent report showed evidence that the clonogenic spheres derived from the mouse and the human CE are made up of differentiated pigmented CE cells rather than a distinct retinal stem cell population [25] Cicero et al. conducted a rigorous analysis of the phenotype of cells in CE-derived spheres and concluded that differentiated pigmented CE cells proliferate in culture, and express low levels of pan-neuronal markers such as β-3-tubulin without forming retinal neurons. Both theories are based on studies relying solely on the analysis of the cellular morphology and expression profiles (reverse transcription polymerase chain reaction (RT-PCR) and immunocytochemistry) of the differentiated photoreceptor cells.

In this study, we use instead an Nrl.gfp transgenic reporter line that expresses green fluorescent protein (GFP) in developing photoreceptors to provide a definitive assessment of the ability of murine CE cells, previously designated retinal stem cells [15], to generate new photoreceptors in vitro. The Nrl gene is essential for rod differentiation and is expressed in developing and mature rods [26–28].



Mice were maintained in the animal facility at University College London. All experiments have been conducted in accordance with the Policies of the Use of Animals and Humans in Neuroscience Research, approved by the Society of Neuroscience in January 1995. The Nrl.gfp line was a kind gift of A. Swaroop and has been described previously [26]. Mice defined as “adult” were at least 6 weeks old.

CE Dissociation, Spheres, and Monolayer Cultures

Dissection and cultures of adult murine CE to generate spheres were performed as previously described [29] and monolayer cultures were established as detailed in the Supporting Information Methods. CE monolayer cultures were grown for 3 (early passage) to 5 (late passage) weeks in proliferation medium before transfer to standard differentiation medium or retinal differentiation medium. Cells were examined daily by epifluorescence and transmitted light.

Differentiation of Embryonic RPC Spheres, CE Spheres, and CE Monolayers Cultures

CE monolayer cultures or embryonic RPCs (prepared as described in Supporting Information Methods) were grown at high densities on poly-L-ornithine (100 μg/ml)/laminin (5 μg/ml)-coated 24-well dishes in either of the following media.

Standard Differentiation Medium

Dulbecco's modified Eagle's medium (DMEM)-F12 plus Glutamax (Invitrogen, U.K.) containing N2 supplement (1:100; Invitrogen, U.K.), Penicillin–Streptomycin solution (1:100; Invitrogen, U.K.), and 10% FBS. Medium was changed every second day and the cells were allowed to differentiate over the course of 14–21 days.

Retinal Differentiation Medium (as Described in Osakada et al. [6])

DMEM-F12 plus Glutamax (Invitrogen, U.K.) containing N2 supplement (1:100; Invitrogen, U.K.), Penicillin–Streptomycin solution (1:100; Invitrogen, U.K.), and 10 μM N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butylester (DAPT; Calbiochem, U.K.) for 3 days. Medium was then replaced with complete medium DMEM-F12 plus Glutamax (Invitrogen, U.K.) containing N2 supplement (1:100; Invitrogen, U.K.), Penicillin–Streptomycin solution (1:100; Invitrogen, U.K.), 10 μM DAPT (Calbiochem, U.K.), FGF2 (10 ng/ml; Peprotech EC, U.K.), heparin (5 ng/ml; Peprotech EC, U.K.), Sonic hedgehog (Shh, 3 nM; R&D Systems Inc., Minneapolis,, Taurine (100 μM; Sigma, U.K.), all-trans Retinoic acid (500 nM; Sigma, U.K.), and 1% FBS. Medium was changed every second day and the cells were allowed to differentiate over the course of 14 days.

Immunofluorescence on CE Cultures and Immunohistochemistry on Retinal Sections

As described in Supporting Information Methods.

RNA Isolation and RT-PCR

As described in Supporting Information Methods.

Lentivirus Production and CE Monolayer Infection

The coding sequences of Dsred, Gfp, mouse NeuroD, Chx10, Nrl, and human CRX genes were cloned into lentiviral vectors. Viruses were produced and harvested as previously described [30] and detailed in Supporting Information Methods. CE monolayer cells grown in proliferation medium for 2 weeks were replated in differentiation medium, at 90% confluence. After 24 hours, each lentivirus was applied to the cells at a multiplicity of infection (MOI) of 15. Medium was exchanged after 1 day and every other day for a total time of 14 days.


Nrl.gfp Labels Rod Photoreceptors in Adult Retina Whereas It Is Not Detected in the CE

In rodents, rod photoreceptor cells are predominantly born in the late embryonic and early postnatal period [31] and make up ∼ 70% of the adult neural retinal cell population [32]. The Nrl.gfp transgene is specifically expressed in rod photoreceptors in the adult retina as previously described [26] and GFP is localized throughout the cell (Fig. 1A, 1B). Nrl.gfp expression is detected in newly born photoreceptors (Supporting Information Fig. 1), but it is not detected in the mature CE, or the retinal pigmented epithelium (RPE) (Fig. 1A–1D). Notably, the adult CE expresses Pax6 and Chx10 (Fig. 1E–1L), both markers of RPCs in the embryonic NR (Supporting Information Fig. 2), and of bipolar (Fig. 1I, 1J), amacrine and ganglion cells (Fig. 1E, 1F), respectively, in the adult retina. In addition, we found that Connexin 43, a component of Gap junctions, is highly expressed in the adult CE (Fig. 1M–1P). As only a few adult CE markers have been identified, Connexin 43 may be used to better define this cell population. Cytokeratin expression was recently described as a CE marker [25]. Here, weak cytokeratin expression was detected in the CE (arrowhead; Fig. 1Q–1T), although it was difficult to distinguish from nonspecific staining of the vasculature (open arrowhead; Fig. 1T, 1X), also seen with the control secondary antibody A594 (open arrowhead; Fig. 1Y–1β). As expected, the RPE marker RPE65 is restricted to the RPE and does not extend into the CE (Fig. 1U–1X). Based on the Nrl.gfp expression pattern, the adult CE shows no evidence of photoreceptor genesis or differentiation in vivo. These findings are in contrast to a recent article which described new photoreceptor generation within the CE in adult mice with inherited retinal degeneration [33].

Figure 1.

Expression of epithelial and retinal progenitor markers in mouse adult retina and CE. (A–D): Sections of adult retina from an Nrl.gfp transgenic mouse show green fluorescent protein expression specifically in the ONL (A, B). The transgene is not expressed in the CE (C, D). (E–X): Immunostaining on sections of adult retina stained with the RPC, amacrine, ganglion cell marker, Pax6 (E–H), the RPC and bipolar cells marker Chx10 (I–L), the epithelial and RPE markers, Connexin 43 (M–P), Cytokeratin (Q–T), and RPE65 (U–X). The inset in (L) shows Chx10 single channel immunostaining. Cytokeratin specific signal is shown by a white arrowhead (T) and nonspecific staining of the vasculature is indicated by open arrowhead (T, X, B). Control using the secondary antibody anti-mouse A594 is shown (Y–β). Higher magnification pictures of the central neural retina and CE (white arrows) are shown for each panel. Sections were counter-stained with the nuclear dye Hoechst 33342 (blue). Abbreviations: RPE, retina pigmented epithelium; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer; CE, ciliary epithelium. Scale bars: 25 μm.

Figure 2.

Development of CE monolayer. (A): Diagram of the mouse adult eye showing the anatomical position of the CE and associated structures. (B): CE cells were grown as neurospheres for 7 days in vitro (D), before expansion as monolayers in proliferating medium. Scale bar 50 μm. (C–G): Electron microscopy images of primary RPE cultures (C), early (2 weeks; D,F) and late (5 weeks; E, G) passage CE monolayers showing stage IV mature melanosomes (C,D), and multivesicular organelles with scattered melanin deposition (E), white arrows. Inset in (E) shows a higher magnification image of typical melanosome at late stages in culture. (F,G) Plasma membrane at early passages (F) does not show membrane interdigitations, whereas at late passages (G) they are clearly visible (black arrows). Scale bars: 0.5 μm (C–E), 200 nm (inset in E), 1 μm. Abbreviation: CE, ciliary epithelium.

Expansion and Passaging of Ciliary Epithelium-Derived Neurosphere-Forming Cells in Monolayer Culture

Whereas brain neural stem cells can be efficiently expanded as neurospheres, the neurospheres derived from adult mouse CE tissue do not expand well in vitro [29]. To improve the expansion of CE-derived cells and test their differentiation potential using the Nrl.gfp reporter, we first developed a monolayer culture system based on previous protocols used in pig and human CE-derived cultures [17, 18]. Addition of low levels (1–4% [v/v]) of FBS allowed neurospheres to adhere to a poly-L-ornithine/fibronectin-substratum and expand into high-density monolayer cultures (Fig. 2B). The morphology of monolayer cultured cells was heterogeneous; the majority of cells possessed several processes and were of variable size, depending on cell density. Optimal survival and proliferation was achieved by plating at high density, whereas cultures with a low seeding density proliferated poorly. CE-derived monolayers were readily passaged using Accutase, whereas culture growth rapidly retarded with high amounts of cell death observed after passaging using trypsin. Typically, cell counts of a few million cells were reached after 2–3 weeks and 8–10 million after 6–8 weeks in culture (>5 independent experiments). An important concern regarding the use of stem cells maintained for long periods in culture is their potential to undergo transformation and immortalization [24]. Immortalized cells continue to divide even after reaching confluence. By contrast, CE monolayers require passaging to continue dividing and stop dividing when they are cultured in differentiation medium (removal of mitogens and addition of 10% FBS). Beyond 10–12 weeks in culture the number of dividing cells was dramatically reduced compared with early passage numbers. Immunostaining for the mitotic marker phospho-Histone3 (pH3) showed 21% ± 8% (SD) of CE cells, cultured at low density at early stages in proliferating conditions, to be in M-phase (1–2 weeks, n = 3). On the contrary, only 2.5% ± 3.5% (SD) of cells cultured either at high density, or in long-term cultures (10–12 weeks, n = 3) in proliferating conditions, or in differentiating conditions are pH3 positive (data not shown). These features are typical of primary progenitor cells, rather than stem cells, maintained in culture as monolayers [34, 35].

Although CE-derived neurospheres are always pigmented, either partially or heavily, the cells in monolayer cultures lose this dense pigmentation over time in culture (Fig. 2B–2E). To better define the pigmentation state of monolayer cells, we analyzed the melanosome morphology of monolayer cells at early and later stages (2 weeks and 8 weeks) by electron microscopy (EM). Although the function of melanosomes in the CE and RPE is not well-understood, it is thought that they have a protective function against light and oxidation [36]. The mammalian adult CE and RPE do not normally synthesize melanin. The majority of melanin synthesis occurs during embryonic development, in which immature unpigmented melanosomes (stages I, II) undergo modifications that lead to complete maturation into pigmented functional melanosomes (stage IV). The pigmented CE monolayer cells analyzed at early stages retain stage IV melanosomes, in which melanin deposition is complete and the organelle striations are obscured (Fig. 2D). In late stage CE cultures, we saw no evidence of mature stage IV melanosomes or any sign of immature elongated unpigmented stage II melanosomes. Instead, we detected round-shaped organelles with varying amounts of membrane accumulation and scattered melanin deposits (Fig. 2E). These latter organelles were never present in the early stage cultures. We therefore assume that pigmentation in CE monolayers is lost by melanosome degeneration over time. These data suggest that a process of dedifferentiation is occurring with loss of the mature differentiated pigment epithelium cell phenotype.

Differentiated epithelial cells, including CE cells in vivo, display epithelial-specific morphological features, such as basal and lateral membrane interdigitations and epithelial cell–cell junctions. EM analysis of CE-derived cells in monolayer cultures did not reveal cell–cell junctions (tight junctions, adherens junctions). After time in culture (late passage stage), cells developed membrane interdigitations typical of epithelial cells in vivo (Fig. 2F, 2G), although they remained depigmented.

Expression Profile of CE-Derived Proliferating Cells in Monolayer Culture

To characterize the extent to which the CE-derived monolayer cells display a retinal progenitor expression profile, we performed RT-PCR and immunocytochemical analyses. After expansion as CE monolayers (late and early passages), mRNA was isolated and analyzed for expression of pluripotency markers (Fig. 3A), forebrain and eye field markers (Fig. 3B), and for genes involved in retinal histogenesis (Fig. 3C) and pigmentation pathways (Fig. 3D). This CE expression profile was compared with that of ES cells and embryonic RPCs from E11.5 eyes.

Figure 3.

mRNA expression profile of CE monolayer cells at early and late passages in vitro. RT-PCR of CE monolayers cultures, maintained in proliferating conditions, shows consistent expression of proliferation markers (Klf4, c-Myc) (A). They express only a subset of forebrain and retinal progenitor cell markers (Tbx3, Tll1,Lhx2, Pax6, Chx10, Otx2, Mitf) (B, C). In addition, they lose the pigmentation markers Tyrosinase, Rpe65, and the Mitf isoformD (D). E11.5 eye, embryonic stem cells, dissected CE, RPE, and CNR are used as positive controls. Abbreviations: CE, ciliary epithelium; CNR, central neural retina; early, early passages; ES, embryonic stem cells; late, late passages; RPE, retinal pigmented epithelium.

Recent work has identified Klf4, Oct3/4, c-Myc, Nanog, and Sox2 as pluripotency genes, essential in ES cell maintenance [37, 38]. Transcripts of c-Myc and Klf4 were expressed in CE cells and immunostaining revealed that 98% of cells express Sox2 (Supporting Information Fig. 3). The presumptive eye field is defined by expression of a group of transcription factor genes, including, Rx, Pax6, Six3, Lhx2, Tll1, Tbx3, and Six6 [39, 40], all of which were detected by RT-PCR in embryonic RPCs (Fig. 3). In Xenopus embryos, simultaneous overexpression of these transcription factors and Otx2 is sufficient to promote eye development including retinal cell differentiation [40]. We found that Tbx3, Tll1, Lhx2, Six6, and Pax6 were expressed in CE monolayers (Fig. 3B), although Rx was absent (Fig. 3C). Chx10 is expressed in RPCs throughout retinal development from the earliest stage of retinal specification in the embryonic optic vesicle, whereas Mitf becomes restricted to the presumptive RPE and Pax6 and Otx2 are expressed in both RPE and NR (Supporting Information Fig. 2). The CE monolayers expressed Pax6, Otx2, and Mitf, but Chx10 mRNA was barely detectable, and a number of other markers of retinal differentiation, including Fgf15, NeuroD and the photoreceptor marker Crx were not expressed (Fig. 3C). Early and late passage cultures showed similar profiles indicating that the phenotype is maintained in monolayer cultures. These data suggest that proliferative CE monolayer cells display aspects of the gene expression profile of optic cup progenitors.

As EM analysis indicates loss of a pigmented phenotype, we next sought to establish whether any mature pigment and epithelial markers were present. Dissected uncultured adult CE, RPE, and NR were used as controls. Palmdelphin was recently described as a differentiated CE marker [25], but was found to be expressed in all control tissues (RPE, CE, and NR). Tyrosinase and RPE65 mRNA were detected in CE and RPE, whereas the D isoform of Mitf was only detected in CE. RPE65 protein was only detected in dissected RPE and not in CE (Fig. 1U–1X), suggesting post-transcriptional regulation. In CE monolayer cells, we only detected palmdelphin, whereas all pigmentation and other RPE marker were absent, consistent with the nonpigmented appearance of these cells (Fig. 3D).

To assess the significance of these mRNA profiles for protein localization at single cell resolution, early (2 weeks) and late (>5 weeks) stage cultures were analyzed by immunocytochemistry. Significant and consistent expression of markers characteristic of neural progenitors (Pax6, Sox2, Nestin, pH3) and of neurogenesis (β3-tubulin) were detected (Fig. 4 and Supporting Information Fig. 3), although levels decreased over time in culture (data not shown). Despite presence of the mRNA, Otx2 protein was never detected. Chx10 expression is a defining characteristic of RPCs throughout development (Supporting Information Fig. 2). Although Chx10 protein was found in neurospheres [15, 29], it was not detected by immunocytochemistry in monolayer cultures (Supporting Information Fig. 3). Neither RPE65 nor the other epithelial markers including ZO-1 and E-cadherin were detected (data not shown). CE monolayers consistently expressed cytokeratin and Cx43 (Fig. 4), which although not CE-specific markers, are both expressed in the adult CE in vivo. Taken together, these data show that although CE-derived monolayers proliferate and exhibit aspects of a retinal gene expression profile, a number of critical markers of RPCs, such as Rx and Chx10 are largely absent, whereas some characteristics of differentiated CE persist. These results indicate a lack of evidence that proliferating CE cells exhibit a bona fide RPC state.

Figure 4.

Protein expression profile of CE monolayer cells at early passages in vitro. Immunostaining of CE monolayers cultures, maintained in proliferating conditions, shows expression of neural progenitor (Nestin), retinal progenitor (Pax6), epithelial (Cytokeratin), and the gap junction marker (Connexin 43). By contrast, cells are negative for RPE65. RPE cultures are used as positive control for RPE65 staining. Scale bar: 25 μm (Connexin43 panel) and 50 μm (Nestin, Pax6, Cytokeratin, and RPE65). Abbreviations: CE, Ciliary epithelium; RPE, retinal pigmented epithelium.

Differentiation of Nrl.gfp Ciliary Epithelium-Derived Cells

Previously we showed that Nrl.gfp-positive cells isolated from the developing retina are able to mature into functional rod photoreceptors after transplantation into the adult retina [1]. Hence, Nrl.gfp transgene activation provides a reliable and robust assay for cells with the potential to differentiate into rod photoreceptors. As the expression profile of CE-derived cells suggests some potential for activation of a retinal differentiation program, we next utilized CE cells from the Nrl.gfp transgenic line to provide a simple assay for commitment to a photoreceptor differentiation program.

We first identified and validated conditions for culture of embryonic RPCs in vitro that were able to activate the Nrl.gfp transgene and initiate photoreceptor differentiation, before testing the same conditions on CE cells. RPCs dissected from the E11.5–12.5 retina were expanded and then transferred to two independent differentiation-promoting conditions. The first utilized conditions described in the original paper reporting the discovery of retinal stem cells (standard differentiation medium; addition of serum and removal of mitogens [15]). The second employed a combination of factors recently shown to effectively induce photoreceptor differentiation from ES cell-derived RPCs (retinal differentiation medium [6]). RPCs in both conditions activated the Nrl.gfp transgene after 4–6 days in culture indicating that these conditions support in vitro photoreceptor commitment. High-level GFP activation was observed after DAPT treatment, the first step of the retinal differentiation medium indicating that these conditions were highly effective at activating rod photoreceptor development in vitro (Fig. 5A).

Figure 5.

Retinal differentiation protocols do not activate Nrl.gfp transgene in CE-derived neurospheres and monolayer cultures. (A): Cultures of embryonic retinal progenitor cells (RPCs) activate the Nrl.gfp trangene and initiate photoreceptor differentiation. RPCs dissected from the E12.5 retina were expanded in proliferation medium followed by differentiation in standard or retinal differentiation media. The first step of the retinal differentiation medium alone (DAPT) was sufficient to activate Gfp expression. Gfp activation is significantly higher in retinal differentiation medium. (B): CE-derived neurospheres expanded for 7 days in vitro and transferred to proliferation medium or retinal differentiation medium for an additional 2 weeks. No Gfp activation is detected. (C): CE-derived monolayers in proliferation medium, retinal differentiation medium, or in standard differentiation medium, cultured for 2 weeks. No Gfp activation is detected. Scale bars: 75 μm and 50 μm in poliferation medium in B. Abbreviations: prolif. medium, proliferation medium; ret. diff. medium, retinal differentiation medium; st. diff. medium, standard differentiation medium.

CE-derived neurospheres were cultured for 7 days in vitro before exposure to the retinal differentiation medium. The spheres expressed Chx10, Rx, and other progenitor markers, as previously described ([29], Supporting Information Fig. 3 and data not shown). However, activation of the Nrl.gfp transgene was not observed in CE neurospheres either with DAPT alone or the full retinal differentiation condition (Fig. 5B), or the standard differentiation conditions, previously reported to activate photoreceptor differentiation from CE cells [15] (data not shown). Similarly, these same conditions failed to promote transgene activation when tested on CE monolayer cultures (n = 3 separate experiments with multiple independent cultures monitored daily; Fig. 5C). Other culture conditions, including the addition of Noggin, insulin-like growth factor (IGF), and Dkk1, previously shown to generate RPCs from ES cells were also tested [5]. These did not generate Nrl.gfp expressing cells, as assessed by GFP fluorescence (data not shown). qPCR was also performed for the early photoreceptor marker Crx, and other retinal markers, but no evidence for photoreceptor differentiation was detected (data not shown).

In addition, we performed immunostaining on the differentiated cultures to assess expression of neuronal and photoreceptor-specific markers. The cultures failed to show positive immunoreactivity for photoreceptor markers including rhodopsin, recoverin, and blue opsin. All antibodies effectively labeled photoreceptors in both retinal sections and dissociated retinal cells (Supporting Information Fig. 6C and data not shown). Furthermore, CE monolayers cultured for 2 weeks in either differentiation media lost expression of β-3 tubulin and displayed a dramatic reduction in the number of Pax6-positive cells (Supporting Information Fig. 4). When these cells were cultured in standard differentiation medium, they showed partially reduced nestin and dramatically increased vimentin expression (Supporting Information Fig. 5), whereas in retinal differentiation medium they develop distinct epithelial features, including formation of cortical actin and accumulation of ZO-1, a marker of the tight junctions at cell–cell contacts (Fig. 6). The apparent reversion of CE cultures toward an epithelial phenotype, when cultured in a medium that supports and promotes photoreceptor differentiation from RPCs, provides additional evidence that the CE-derived cell population does not contain bona fide RPCs.

Figure 6.

Ciliary epithelium (CE) monolayer cells cultured in retinal differentiation medium show an increase of cortical actin and ZO-1 at the cell–cell contacts. Immunostaining of CE monolayers cells cultured in proliferation medium, retinal differentiation medium or standard differentaiation medium using ZO-1 (green) antibody and phalloidin for F-actin staining (red). Cells were counter-stained with the nuclear dye Hoechst 33342 (blue). Insets show higher magnification images. Scale bar: 75 μm and 20 μm in the insets. Abbreviations: prolif. medium, proliferation medium; ret. diff. medium, retinal differentiation medium; st. diff. medium, standard differentiation medium.

Taken together, these data indicate that CE-derived cells do not undergo photoreceptor differentiation.

CE Progenitors Rarely Activate the Nrl.gfp Transgene Under Photoreceptor Differentiation Conditions Even with Exogenous Expression of Retinal Transcription Factors

Finally, we asked whether CE monolayer cells could differentiate into photoreceptors, upon exogenous expression of the retinal-specific transcription factor genes that they lacked. We used lentiviral vectors that encode for NeuroD, Chx10, CRX, and Nrl together with a Dsred viral vector, to transduce CE monolayer cells in an attempt to reprogram them along a photoreceptor differentiation pathway. We imaged live cells after 10–14 days, to evaluate the levels of photoreceptor differentiation, using the Nrl.gfp transgene activation assay. Neither CE cells transduced with a single transcription factor gene nor transduction with multiple transcription factor genes or the control Dsred marker alone led to a significant number of Nrl.gfp positive cells in culture. Less than 0.003% of the cells activated the Nrl promoter and presented GFP fluorescence (Fig. 7B) and there were no differences between controls and cells transduced with retinal transcription factor genes. Moreover, RT-PCR analysis showed no expression of the endogenous photoreceptor-specific transcription factor gene Crx, an essential player in photoreceptor development [41, 42] (Fig. 7A), whereas the virally expressed human CRX mRNA was detectable. Immunostaining analysis for rhodopsin and blue opsin did not show any positive cells (data not shown). No obvious change in cell morphology was observed after transduction. We also performed RT-PCR analysis for photoreceptor-specific gene expression (Supporting Information Fig. 6A, 6B). Rhodopsin and blue opsin mRNA were detected (18/24 samples; three independent experiments) using two different primer sets for each gene (Supporting Information Fig. 6A and data not shown). Additional photoreceptor markers Irbp and S-arrestin were not detected, whereas the CE markers Palmdelphin and Connexin43 were expressed (Supporting Information Fig. 6A). Real-time quantitative RT-PCR was used to assess the level of rhodopsin and blue opsin expression in the transduced CE cultures cultured in differentiating conditions. Rhodopsin and blue opsin mRNA were found to be present at around 30,000- and 300-fold lower levels, respectively, compared with adult NR (Supporting Information Fig. 6B). Palmdelphin expression was 20-fold higher in transduced CE cells compared with CE cells maintained in proliferating conditions. These data, together with a consistent lack of Nrl.gfp activation, strongly indicate the absence of a coherent rod photoreceptor differentiation program.

Figure 7.

Ciliary epithelium (CE) cells rarely activate the Nrl-gfp transgene under retinal differentiation conditions even with exogenous expression of retinal transcription factors. (A) Scheme of the experimental procedure for the transduction of CE monolayers cells using NeuroD, Chx10, CRX, and Nrl lentiviral vectors. (B) RT-PCR shows no differences between controls and retina transcription factor-transduced cells. The dsred lentiviral transduced (red) and the untransduced cells were used as controls. Live cells were imaged 10 days after transduction and GFP expression (green) was assessed. Less than 0.003% of the cells activate the Nrl promoter (C, white arrows). Scale bar: 100 μm. Abbreviation: GFP, green fluorescent protein.


As Tropepe et al. [15] proposed that the adult CE contains a population of retinal stem cells able to differentiate into retinal neurons, different groups have invested effort into repeating these experiments and investigating the potential of CE-derived retinal stem cells for retinal repair. Success in this area has been limited, with reports of 1% of opsin-positive cells generated from rat CE-derived cells [20], 30% of Rho4D2-positive cells generated from human CE-derived cells [17], no fully differentiated photoreceptors generated from porcine CE-derived cells [18] and threefold increase of photoreceptor markers such as rhodopsin and blue opsin in Crx-electroporated CE cells [22]. Recently, Cicero et al. [25] argued against the validity of the existing theory. The earlier studies relied only on analysis of the cellular morphology and the expression profile of the differentiated cells as assessed by immunocytochemistry and PCR [15-17, 20, 23]; such techniques can lead to false positives and mis-interpretation of the results due to nonspecific labeling or the presence of mRNA, but not the corresponding proteins. In this study, we used a genetic tool, the Nrl.gfp transgenic line that expresses GFP in developing and mature photoreceptors under the control of the Nrl promoter [26]. This novel approach based on the activation of a rod-specific promoter has allowed us to assess more definitively the potential for generating photoreceptors from CE-derived cells.

It has been previously reported that CE-derived neurospheres have a poor expansion potential [17, 18]. To use cells for autologous cell therapy, a high number of cells are required. In this study, we established the conditions for expanding, as monolayer cultures, the mouse CE-derived cells. CE-derived monolayer cultures are highly proliferative; a robust expansion was obtained when compared with floating neurospheres, reaching 10 million cells in 3 weeks. CE monolayer cells express a number of markers known to be associated with neural stem cells, such as Nestin and Sox2. They express the forebrain and eye field marker Pax6, although Rx and Chx10 were difficult to detect. Occasionally, the pan-neuronal marker β3-tubulin was detected in proliferating monolayer cultures, without specific differentiation conditions. Moreover, mRNA transcripts of additional early eye field-specific genes such as Otx2, Tbx3, Tll1, Lhx2, and Six6 were also identified. We found the presence of Otx2 transcript in CE cells, but we did not detect any Otx2-positive cells by immunostaining indicating post-transcriptional regulation of specific genes might be occurring. The expression profile of the mouse CE monolayer cells we have identified here is consistent with the previous findings from human and porcine CE-derived cells [17, 18].

Although CE-derived neurospheres are always pigmented, either partially or heavily, as also recently described [25], we found that the cells in monolayer cultures lose their dense pigmentation with time in culture suggesting that a dedifferentiation to a proliferating progenitor-like status occurs with consequent loss of the pigmented status. This process may be analogous to the transdifferentiation of RPE to NR, which in lower vertebrates can lead to the regeneration of the retina [43–45]. Notably, Pax6 and Chx10 are both detected in the adult differentiated CE in vivo and may be important for their apparent progenitor characteristics. We found that although the CE monolayer cultures resemble progenitor cells of the developing embryonic eye (by virtue of their depigmentation, proliferation characteristics and expression of eye field and retinal-specific genes), they also simultaneously display CE features including Cx43 and cytokeratin expression and formation of membrane interdigitations. Furthermore, the lack of Chx10 protein as well as other retinal markers in the proliferating cultures argues against the existence of a bona fida RPC state.

Photoreceptor differentiation experiments were performed primarily on monolayers at the stage at which they had started to dedifferentiate and showed the highest resemblance to RPCs. We tested differentiation conditions that support the commitment of embryonic RPCs and ES cell-derived RPCs to photoreceptors in vitro, on CE neurospheres, as well as CE monolayer cultures, using the expression of the Nrl.gfp transgene as a marker of rod photoreceptor commitment. The removal of mitogens and the addition of 10% FBS was previously reported to generate photoreceptors from mouse and human CE cells [15, 17]. Different reports have recently shown that by using a combination of different factors, ES cells could be directed toward an RPC state. In particular, the use of Dkk1, a Wnt/β-catenin signaling pathway antagonist, and LEFTY-A, a nodal antagonist and noggin, a potent inhibitor of the bone morphogenetic protein (BMP) pathway, were shown to effectively increase the Rx+/Pax6+ cell population in ES cells [5]. Furthermore, inhibition of Notch signaling induces photoreceptor differentiation in vivo [46, 47], and the use of the γ-sectretase inhibitor DAPT on ES-derived Rx+ cells, increases the number of Crx-positive photoreceptor precursors [6]. Here, unequivocally we did not observe Nrl.gfp transgene activation after exposing both CE-derived neurospheres and monolayers to any of these previously reported differentiation conditions. Moreover, we did not detect immunoreactivity for any photoreceptor markers, despite testing a wide array. By contrast, these protocols effectively induced Nrl.gfp transgene activation in embryonic RPCs in control experiments. These data confirm that CE cells show limited potential to generate photoreceptor cells; on the contrary, CE cells seem to revert their phenotype toward the original differentiated epithelial status, as confirmed by the cortical actin formation and ZO-1 accumulation at the cell–cell contacts.

Viral vector-mediated transduction of retinal transcription factors in different cell types has been reported to increase the number of cells expressing photoreceptor markers [19, 22, 48]. We also evaluated this approach, however, we found that a negligible proportion of cells (less that 0.003%) transduced with retinal transcription factors activate the Nrl.gfp transgene, a finding that was independent of the vector combination used. RT-PCR data indicated a low level of ectopic transcription of some photoreceptor genes, but not the photoreceptor transcription factor gene Crx, and may explain previous reports that concluded photoreceptors could be successfully generated from adult retinal stem cells.


In summary, we conclude that CE-derived cells do proliferate in vitro, do express some of the eye field and neural/retinal progenitor markers and undergo a process of dedifferentiation including the loss of pigmentation. However, they do not effectively activate the Nrl-regulated rod differentiation program and are unable to generate new rod photoreceptors. These data do not support the conclusions drawn from earlier immunostaining experiments [15, 17] which indicated high levels of new rod photoreceptor generation by retinal stem cells derived from the adult CE. The range of strategies we have used to test for activation of rod photoreceptor differentiation in CE cultures, including the Tropepe protocol [15], combined with the reliability of Nrl.gfp transgene activation in vivo, in transplanted rod photoreceptors [1], and in embryonic RPC differentiation control experiments in vitro strengthen this conclusion. Recently, Cicero et al. reported an extensive analysis designed to identify the putative retinal stem cell located within the CE and were unable to find such a cell [25]. Instead, they reported that the mature pigmented CE cells exhibit the properties of neurophere formation and clonal expansion as reported by Tropepe et al. [15]. This latter study did not pursue the differentiation potential of these cells and left open the possibility that they could undergo retinal neurogenesis. Our study has pursued this question and concludes that the photoreceptor potential of these cells is negligible. if these cells are to be useful for cell-based therapies in the treatment of retinal disease, it is likely that they will require reprogramming and/or trans-differentiation. Other more promising alternatives are the use of fetal tissue-derived RPCs, pluripotent ES, or iPS cells [6, 7]. Recently, significant progress toward retinal differentiation was achieved by using low-molecular-mass compounds on both ES cells and iPS, indicating the possibility of avoiding animal-derived factors to induce differentiation [49].


We gratefully thank Anand Swaroop for the Nrl.gfp mice, Elena Sokolskaja for assistance with lentivirus production, and Peter Munro for electron microscopy analysis. This work was supported by the Medical Research Council UK (G03000341), the Macula Vision Research Foundation, Fight for Sight, EMBO, the Ulverscroft Foundation, the Royal Society, and the NIHR Biomedical Research Centre for Ophthalmology.


The authors indicate no potential conflicts of interest.