Brief Report: Self-Organizing Neuroepithelium from Human Pluripotent Stem Cells Facilitates Derivation of Photoreceptors§


  • Author contributions: C.B.: conception and design, collection and/or assembly of data, data analysis and interpretation, and manuscript writing; S.M.: provision of study material and collection and/or assembly of data; E.H. and A.J.T.: provision of study material; J.C.S.: financial support, data analysis and interpretation, and manuscript writing; R.R.A.: financial support, administrative support, data analysis and interpretation, manuscript writing, and final approval of manuscript.

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

  • §

    First published online in STEM CELLSEXPRESS November 6, 2012.


Retinitis pigmentosa, other inherited retinal diseases, and age-related macular degeneration lead to untreatable blindness because of the loss of photoreceptors. We have recently shown that transplantation of mouse photoreceptors can result in improved vision. It is therefore timely to develop protocols for efficient derivation of photoreceptors from human pluripotent stem (hPS) cells. Current methods for photoreceptor derivation from hPS cells require long periods of culture and are rather inefficient. Here, we report that formation of a transient self-organized neuroepithelium from human embryonic stem cells cultured together with extracellular matrix is sufficient to induce a rapid conversion into retinal progenitors in 5 days. These retinal progenitors have the ability to differentiate very efficiently into Crx+ photoreceptor precursors after only 10 days and subsequently acquire rod photoreceptor identity within 4 weeks. Directed differentiation into photoreceptors using this protocol is also possible with human-induced pluripotent stem (hiPS) cells, facilitating the use of patient-specific hiPS cell lines for regenerative medicine and disease modeling. STEM CELLS2013;31:408–414


Most forms of untreatable blindness result from the death of retinal neurons. The loss of photoreceptor cells is the main cause of vision loss in inherited retinal degenerations, age-related macular degeneration, and diabetic retinopathy. There is a huge impetus to develop novel tools to understand and treat these disorders. Human embryonic stem cells (hESCs) provide immense promise for cell replacement in the retina and epigenetic reprogramming has paved the way to patient-specific cells for regenerative medicine and disease modeling [1]. However, to realize the full potential of these approaches, improved retinal differentiation is required. Several important publications have shown that it is feasible to differentiate hESCs and human-induced pluripotent (hiPS) cells toward retinal fates including the generation of cells expressing photoreceptor markers using various approaches [2–11]. Blocking Wnt and Nodal pathways during neural differentiation of hESC-derived embryoid bodies produces retinal progenitors after 35 days, giving rise to 20% photoreceptor precursors after 170 days. Later treatment with retinoic acid and taurine promotes maturation of rhodopsin+ and opsin+ photoreceptors that appear between 130 days and 200 days [8]. Differentiation of hESC-derived embryoid bodies into photoreceptor precursors can be accelerated by treating pluripotent cells with DKK-1 (Wnt inhibitor), Noggin (Bmp inhibitor), and IGF-1. Using this combination of factors, hESCs converted into Crx+ photoreceptor precursors in only 3 weeks, albeit at an efficiency of only 12%, and terminal differentiation into rhodopsin+ and opsin+ mature photoreceptors was rare [10]. A modified hESC and hiPSC retinal differentiation protocol using Noggin, Dkk-1, IGF-1, Lefty A, 3,3′,5-triiodo-L-thyronine (T3), fibroblast growth factor (FGF), retinoic acid, taurine, Shh, and Activin A showed increased efficiency of expression of cone opsin (60%) and rhodopsin (18%) by day 45 of culture, although photoreceptor survival was not maintained in longer term cultures [2]. Thus to date, protocols for photoreceptor-directed differentiation of hES require long and expensive cultures and are rather inefficient. However, several lines of evidence using protocols based on formation of embryoid bodies in suspension culture suggest that acquisition of a rostral phenotype during neural differentiation might be a default pathway in hESCs, giving rise to neurons from the forebrain and the retina [6, 7, 12–14]. Among the myriad of signals that could influence stem cell differentiation, interaction with the extracellular matrix (ECM) appears most promising (reviewed by Kraehenbuehl et al. [15]). Direct signaling between stem cells and the ECM can regulate proliferation and provide instructive cues for cell fate decision [16, 17]. In addition, ECM can immobilize secreted proteins to control their activity and dosage in proximity to the cell surface [18]. In the developing brain, the ECM in the neural stem cell niche is laminin rich [19], but its composition is complex as it also contains collagen IV, nidogen, heparin sulfate proteoglycans, and chondroitin sulfate proteoglycans [20, 21]. Here, we tested whether neural differentiation devoid of exogenous morphogens (default pathway), in the presence of a complex ECM, was sufficient to bias human pluripotent stem (hPS) cells differentiation toward a retinal fate. We report that using ECM, hESCs self-organize into a neuroepithelium that acquires a retinal fate by default. Under these conditions, after 10 days, up to 60% of hESCs expresses Crx, an early photoreceptor marker. These findings provide a new approach for the rapid generation of human photoreceptor precursors, without the requirement for manipulation by extrinsic factors. After 4 weeks and addition of growth factors, the retinal cells then give rise to rod photoreceptors expressing the transcription factor Nrl and the photopigment rhodopsin, thus facilitating future transplantation therapies and disease modeling for retinal degeneration.


Cells and Culture Conditions

hESCs HUES-2 (Harvard) and hiPS cells (HC83#12-2) were cultured on hES-qualified Matrigel (BD Bioscience, Oxford, UK, according to manufacturer recommendations in feeder-independent mTeSR1 medium (Stem Cell Technologies, Grenobles, France, under 5% oxygen conditions. The medium was changed daily. Cells were passaged as clumps using TrypLE (Invitrogen, Paisley, UK, After 30-second incubation, TrypLE was removed and cell colonies were detached from the plate with a cell scratcher in mTeSR1. Cell colonies were broken down by gentle pipetting and replated at a dilution of 1:5–1:10 with 10 μm Rock inhibitor (Tocris, Abingdon, UK, for 24 hours.

hiPS Cell Generation

Passage 5 human fibroblasts from discarded corneal rim were seeded at a density of 105 cells per well of six-well plates. The following day the cells were infected with supernatant from 293T (ATCC, Teddington, UK, cells producing the STEMCCA lentivirus in the presence of 4 μg/ml polybrene for 24 hours. Three days after infection, cells were harvested and transferred to a 100 mm ES-qualified Matrigel-coated dish. The next day, fibroblast medium was changed to mTeSR1. ESC-like colonies were manually picked after around 30 days and expanded on ES-qualified Matrigel.

Retinal Induction

hESCs and hiPS cells were dissociated as cell clumps as described above and plated on 1% ES qualified-Matrigel for 1 hour. Attached aggregates of cells were covered with 2% ES qualified-Matrigel diluted in neural differentiation medium consisting of Dulbecco's modified Eagle's medium/F-12 supplemented with 1% N2 and 1% B27 (all from Gibco, Paisley, UK, After overnight incubation, fresh medium without Matrigel was added and changed every other day. From day 10, Retinoic acid (500 nM, Sigma, Gillingham Dorset, UK,, Taurine (1 mM, Sigma), basic FGF (bFGF) (10 ng/ml, R&D, Abingdon, UK), acidic FGF (aFGF) (50 ng/ml, R&D), and Sonic Hedgehog (SHH) (3 nM, R&D) were supplemented. Three different batches of Matrigel were used (with similar results for each).

Gene Expression Analysis

Total RNA was extracted using RNeasy kit (Qiagen, Crawley, UK, For each sample, 1 μg of RNA was treated with DNase and reverse transcribed using RT kit (Qiagen). All results were confirmed by three independent biological samples at each data point. Primers are listed in supporting information Table S1.

Antibodies, Microscopy, and Time-Lapse Imaging

Cells were fixed using 4% paraformaldehyde for 10 minutes, washed with Phosphate Buffered Saline (PBS), permeabilized with 0.2% Triton X in PBS and blocked using 5% milk in PBS. Primary antibodies used were TRA-1-81 (Millipore, 1:100), TRA-1-60 (Millipore, Watford, UK,, 1:100), SSEA-4 (Millipore, 1:100), Pax6 (Covence, Crawley, UK,, 1:500), Rax (Abcam, Cambridge, UK,, 1:500), Sox9 (Abcam, 1:500), Crx (a gift from Kevin Gregory-Evans, 1:2,000), Mitf (Millipore, 1:200), ZO-1 (Zymed, Paisley, UK,, 1:200), Tuj-1 (Covence, 1:1,000), pH3 (Millipore, 1:200), Caspase-3 (Abcam, 1:200), Nrl (a gift from Anand Swaroop, 1:500), and Rhodopsin clone RetP1 (Sigma, 1:2,000) or clone 4D2 (1:200) overnight at 4°. F-actin was detected with Phalloidin (Sigma, 1:1,000). Cells were incubated with secondary antibody (anti-rabbit and anti-mouse, Molecular Probes, Paisley, UK,, 1:500) for 1 hour at room temperature. Imaging was performed with a Zeiss Observer Z1 and confocal imaging with a Leica EL6000. Time-lapse imaging was performed with JuLI smart fluorescent cell analyzer (one image every hour during 48 hours).

Cell Quantification

A total of 1,000–2,500 cells from different randomly chosen fields were examined for each experiment using ImageJ. All sets of experiments were performed at least five times.


hESCs in ECM Self-Organize into Polarized Neuroepithelium

We used Matrigel in order to recreate a complex ECM around hESCs. Its composition resembles the neural stem cell niche, as it contains laminin, collagen IV, nidogen, and proteoglycans. Undifferentiated hESC colonies were partially dissociated to small-cell aggregates and replated onto 1% Matrigel-coated dishes and allowed to attach. Cells were next covered with 2% Matrigel in neural differentiation medium. After 2 days, hESC aggregates formed columnar neuroepithelia (Fig. 1A and supporting information movie S1; >35 independent experiments) composed of Sox2+ progenitors (Fig. 1B). The neuroepithelium consistently formed a polarized vesicle, as indicated by dense accumulation of F-actin, ZO-1, and phosphorylated histone (pH3) at the apical side (Fig. 1B, 1C and supporting information S1), features of polarized epithelium. These structures were similar in their polarization and display of luminal mitosis to the self-organized neural tissue generated by hESCs in suspension [13]. We found that interaction between hESCs and ECM, but not the growth factors associated with ECM, appeared to be necessary for rapid neuroepithelium organization and differentiation as omission of Matrigel blocked both the columnar neuroepithelium formation and the induction of PAX6 expression observed with standard or growth factor-reduced ECM after 5 days (Fig. 1D). These results demonstrate that hESCs can interact with ECM during neural differentiation in order to rapidly form a polarized neuroepithelium within 2 days.

Figure 1.

Neural differentiation of hESCs in extracellular matrix (ECM) allows for highly efficient retinal induction in 5 days. (A): hESCs self-organize into epithelium after 2 days in ECM. (B, C): The accumulation of ZO-1 and F-actin illustrates the apico-basal polarity of the epithelium composed of neural progenitors expressing Sox2+ (n = 2). (D): After 5 days in neural differentiation medium, the expression of the early ectodermal marker PAX6 is not detectable without ECM (Matrigel–) and this induction was independent of growth factors contained in Matrigel as illustrated by PAX6 induction with GFR matrix (n = 3). PAX6 primers detect mRNA for isoform +5a and +5a. (E–H): Neuroepithelium loses its integrity and expands as a monolayer around day 5, forming neural rosettes. (I): Gene expression study reveals the expression of NOGGIN and DKK1 during neuroepithelium formation (n = 3). (J): After 5 days, transcripts corresponding to a retinal progenitor profile PAX6, RAX, and CHX10 are expressed (n = 3). (K–M): The retinal identity of the neuroepithelium was confirmed by the observation of Pax6+, Sox9+, and Rax+ colonies (n = 3). Scale bars = 50 μm (A–C) and 100 μm (H, I). Abbreviations: GAPDH, Glyceraldehyde 3-phosphate dehydrogenase; GFR, growth factor reduced; hES, human embryonic stem; NE, neuroepithelium.

Neural Differentiation of hESCs in ECM Allows for Highly Efficient Retinal Induction

After day 5, the polarized neuroepithelial vesicles lost their structural integrity, flattened, and expanded as adherent monolayer colonies (Fig. 1E) containing neural rosettes (Fig. 1F–1H). Gene expression analysis revealed increased expression of DKK1 and NOGGIN during this proliferation (Fig. 1I). As endogenous expression of DKK-1 and Noggin converts hESCs into retinal progenitors [6], we sought to determine whether the cells had followed a retinal lineage after addition of ECM. Notably, transcripts for RAX and PAX6, markers of retinal progenitors, were detectable in our culture system as early as day 2 (supporting information Fig. S2) and increased over time together with CHX10, a marker of the developing neural retina (Fig. 1J and supporting information S2). Immunostaining confirmed that Pax6+ neuroepithelium generated by means of the addition of ECM (Fig. 1K) exhibited a retinal character after 5 days, as evidenced by staining for Sox9 [22, 23] and Rax [24, 25] in 90% ± 5% and 95% ± 2% of the colonies, respectively (Fig. 1L, 1M).

Retinal Progenitors Derived from hESCs Differentiate Efficiently into Photoreceptor Precursors

During eye development, Rax+ progenitors of the optic vesicle give rise to all cell types of the retina, including photoreceptors and retinal pigmented epithelium (RPE) [24]. We next sought to examine spontaneous differentiation of hESC-derived retinal progenitors toward these two fates. Crx is one of the earliest known photoreceptor markers, regulating development of rods and cones [26]. Nuclear localization of Crx was present in a subset of the hES-derived cells by day 5 (Fig. 2A) followed by extensive differentiation into Crx+ photoreceptor precursors after 10 days (Fig. 2B). Crx+ cells represent ∼60% of the total number of cells (Fig. 2C) by day 10 and progressively express Tuj-1, a marker of differentiating neurons (Fig. 2D). Furthermore, we observed that around 5% of rosettes at day 10 of differentiation was expressing Mitf, a transcription factor specific in the eye to RPE (Fig. 2E). RPE cell identity was confirmed by the expression of RPE65 and CRALBP along with MITF (Fig. 2F). The spontaneous expression of markers specific for photoreceptors and RPE in cells derived from progenitor cells expressing Rax+/Pax6+ suggests that the hES-derived progenitor cells generated by exposure to ECM resemble the neuroepithelium of the embryonic optic vesicle and are competent to differentiate into RPE as well as photoreceptor cells of the neural retina.

Figure 2.

Retinal progenitors derived from human embryonic stem cells (hESCs) have the potential to differentiate into rod photoreceptors. (A–C): Photoreceptor precursors (Crx+) are detected as soon as day 5 of differentiation (n = 3) and represent 60% ± SD of the total number of cells after 10 days (n = 6). (D): Crx+ cells became more mature neurons expressing Tuj-1 after 10 days (n = 6). (E, F): Discrete clusters of retinal pigmented epithelium (RPE) progenitors (Mitf+) are present after 10 days of differentiation (n = 3). Transcripts for RPE-related genes MITF, CRALBP, and RPE65 are readily detected (n = 4). (G): Gene expression analysis demonstrates that photoreceptor precursors (expressing CRX and RCVRN) start to express cone (OPN1LW and OPN1MW) and rod markers (NRL) after 5 days of differentiation. However, after 10 days in culture, cone markers are absent and photoreceptors adopt a rod identity (NRL and RHO) (n = 6). (H): Time course analysis demonstrates that cone-specific transcripts OPN1LW and OPN1MW are downregulated progressively between day 5 and day 10 of differentiation concomitant to the induction of NR2E3 and RHO expression (n = 3). (I): Schematic illustration of the strategy used for maturation of rods. (J, K): After 4 weeks, Crx+ cells express Tuj-1 and rod photoreceptors are detected by the expression of the transcription factor Nrl (n = 5). (L): Quantification of Nrl+ rod photoreceptors after 30 days of differentiation (n = 5). (M): Z-stack imaging of Rhodopsin+ cells that are observed after 30 days of differentiation from hESCs. Scale bars = 100 μm (A, B, D, E, K) and 40 μm (J, M). Abbreviations: aFGF, acidic fibroblast growth factor; bFGF, basic fibroblast growth factor; GAPDH, Glyceraldehyde 3-phosphate dehydrogenase ; RA, retinoic acid; Shh, sonic hedgehog; Tau, taurine.

Maturation of Rod Photoreceptors Is Correlated with the Repression of Cone Differentiation

The mammalian retina contains two types of photoreceptors: cones and rods, with rods being the most prevalent population. Expression of the pan-photoreceptor gene RCVRN as well as CRX was detected by day 5 (Fig. 2A, 2G). In addition, at day 5, we consistently detected expression of the cone opsin genes, OPN1SW and OPN1LW (Fig. 2G, 2H), but were unable to detect the proteins by immunohistochemistry (data not shown), presumably because of the low levels of expression. We were also able to detect expression of the rod-specific transcription factor gene, NRL, but not rhodopsin (RHO), the rod-specific photopigment (Fig. 2G, 2H). Expression of RHO was detectable only after 10 days of differentiation, at which point cone opsin expression was no longer detected (Fig. 2G). These data indicate that after 10 days, the hESCs appeared to differentiate into photoreceptor precursors biased toward a rod identity.

During development, photoreceptor precursors differentiate into cones by default and rod fate is actively induced by blocking expression of cone-related genes and promoting rod-related genes [27–30]. In order to identify the timing of the bias toward rod fate adopted by photoreceptor precursors, we investigated the profile of cone and rod-specific transcripts at days 4, 6, 8, and 10 during differentiation. Expression of the cone markers OPN1SW and OPN1LW showed a progressive decline and was totally suppressed by differentiation day 10 (Fig. 2H), without detection of significant levels of cell death (supporting information Fig. S3). This decline correlated with the expression of NR2E3 (Fig. 3H), an orphan nuclear receptor implicated in the acquisition of rod fate and repression of cone fate [27–29, 31, 32], together with the rod markers NRL and RHO (Fig. 3G, 3H). Hence, during the period between day 5 and day 10 of culture, the NRL/NR2E3 rod differentiation program gained apparent dominance over cone development.

Figure 3.

Human-induced pluripotent stem (hiPS) cells can be differentiated into photoreceptors using extracellular matrix (ECM). (A, B): Upon addition of ECM, hiPS cells form neuroepithelium at day 2 that spreads as a monolayer. (C, D): After 5 days, hiPS cells differentiate into retinal progenitors expressing Rax and Sox9. (E, F): Ten days of differentiation are sufficient to derive Crx+ photoreceptor precursors expressing Tuj-1 from hiPS cells. (G, H): hiPS cell-derived photoreceptors express the rod markers Nrl and Rhodopsin after 30 days of differentiation. Scale bars = 100 μm (A, B, C, D, E, G) and 25 μm (F, H).

In order to promote photoreceptor survival and maturation, retinoic acid, taurine, bFGF, aFGF, and sonic hedgehog [8] were added from day 10 (Fig. 2I). Without addition of these factors, large-scale cell death was observed. After 4 weeks of differentiation, most of the Crx+ cells label with Tuj-1 (Fig. 2J). Rod photoreceptors were detected by immunostaining for the transcription factor Nrl (Fig. 2K). Quantification of Nrl+ cells indicated that rod photoreceptors represented 36% of the total number of cells after 4 weeks of differentiation (Fig. 2L). Further maturation of rods was demonstrated by the localization of rhodopsin in a subset of rods by day 30 (Fig. 2M). Altogether, these data demonstrate a robust derivation of rod photoreceptors after short differentiation period compared with other protocols [6–8, 10].

hiPS Cells Can Be Differentiated into Photoreceptors by Addition of ECM

Next we wanted to determine whether retinal induction in ECM could similarly be applied to generate hiPS cell-derived rod photoreceptors. hiPS cells were generated by lentiviral transduction of adult fibroblasts using a polycistronic construct expressing Oct4, Sox2, c-Myc, and Klf4 and demonstrated features of pluripotency (supporting information Fig. S4). When subjected to the same differentiation protocol described above, hiPS cells formed neuroepithelium by day 2 (Fig. 3A) and retinal progenitors expressing Rax and Sox9 by day 5 (Fig. 3B–3D). Upon retinal induction, hiPS cells differentiated into Crx+ photoreceptor precursors (Fig. 3E). Similar to hESCs, hiPS-derived photoreceptor precursors start to mature expressing Tuj-1 at day 10 of differentiation (Fig. 3F). Subsequently, Nrl+ and Rhodopsin+ rod photoreceptors were detected after 4 weeks (Fig. 3G, 3H).


In this report, we describe a new method of retinal differentiation of hPS cells using ECM that reduces the complexity and cost of previous protocols. This accelerated differentiation is striking and confirms that in vitro differentiation of hPS cells into retinal cells can be faster than in vivo development, as already illustrated for dopaminergic neurons, motor neurons, and nociceptors [33, 34]. DKK-1 and Noggin are two well-known inducers of rostral central nervous system identity in pluripotent stem cells [8, 10, 13, 35, 36] and their endogenous expression could explain neural differentiation in our system [6]. However, we show that rapid expression of PAX6 and the formation of a neuroepithelium without exogenous morphogens needed interaction with ECM components (provided by the addition of Matrigel). Additional studies will be necessary to understand the exact function of the ECM in epithelium formation, neural differentiation, and retinal patterning from hESCs. Optic vesicle (OV)-like spheres have been previously generated at a much lower efficiency (∼20%) in 25 days from hESCs and hiPS cells in suspension, without addition of Matrigel, but including initial formation of neural aggregates by plating onto laminin-coated substrates [7]. Meyer and colleagues showed the generation of Crx+ cells only after prolonged periods of culture; 80 days were necessary to obtain ∼55% Crx+ cells from the OV-like spheres, which represent only ∼11% of the total cell population. Our study shows that ECM treatment greatly improves the production of transient neural epithelium that can differentiate subsequently into CRX-expressing retinal cells at a high efficiency after 10 days (∼60% of the total number of cells).

In our culture system, Crx+ precursors initially expressed cone markers and then shifted toward NRL/NR2E3-mediated rod differentiation, as observed during late retinogenesis. By contrast, Meyer et al. demonstrated a maturation of Crx+ cells into cones [6]. Identification of the factors regulating this bias would allow directed production of cone or rod-rich populations. Interestingly, the influence of ECM on retinal differentiation of mouse and human ESCs has been recently reported by Yoshiki Sasai's laboratory [3, 4]. ESCs have been shown to self-organize into bilayered optic cups that go on to differentiate into a multilayered neural retina. Sasai et al. generated impressive human optic cups following a complex 36-day culture protocol starting with the reaggregation of dissociated hESCs. Further culture of the human optic cups first generated photoreceptor precursors by days 42–60 and by day 126 an organized photoreceptor layer expressing visual pigments was observed. These results highlight the essential role of ECM in stem cell biology and for the development of neural differentiation protocols that recapitulate developmental processes. Our data confirm the potential of hPS cell lines to form organized neural derivatives.


In our culture system, the integrity of the neuroepithelium is lost rapidly, blocking any subsequent retinal layer formation, yet photoreceptor precursors were observed by day 5 of culture. Transplantation of photoreceptor precursors from dissociated retina has been demonstrated as an efficient tool for photoreceptor replacement and rescue of visual function [37–39]. We propose that, using ECM to support the spontaneous organization of a three-dimensional neuroepithelium, it is feasible to generate in vitro, large numbers of photoreceptor precursors for transplantation without the need to form such a complex tissue as the retina. Moreover, directed differentiation of hiPS cells into rod photoreceptors indicates the robustness and general applicability of the differentiation in Matrigel, supporting its use as a standard strategy for driving differentiation of human pluripotent cells into retinal cells. ECM-based retinal induction may facilitate the future use of patient-specific photoreceptors for regenerative medicine and disease modeling.


This work was supported by the Medical Research Council U.K. (G03000341, G0901550 MR/J004553/1) and by a grant from The Millers Trust. J.C.S. is supported by Great Ormond Street Hospital Children's Charity. R.R.A is partially funded by the Department of Health's National Institute for Health Research Biomedical Research Centre at Moorfields Eye Hospital.