Efficient Stage-Specific Differentiation of Human Pluripotent Stem Cells Toward Retinal Photoreceptor Cells§


  • Author contributions: C.B.M.: designed and performed research, analyzed data, wrote the paper, and approved final version of manuscript; E.S. and D.H.W.S.: performed research and contributed to paper writing and final approval of manuscript; I.M.G.: performed research, data writing, and final approval of manuscript; M.L.: designed and performed research and contributed to paper writing, final approval of manuscript, and fund raising.

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

  • §

    First published online in STEM CELLSEXPRESS January 20, 2012.


Recent successes in the stem cell field have identified some of the key chemical and biological cues which drive photoreceptor derivation from human embryonic stem cells (hESC) and human induced pluripotent stem cells (hiPSC); however, the efficiency of this process is variable. We have designed a three-step photoreceptor differentiation protocol combining previously published methods that direct the differentiation of hESC and hiPSC toward a retinal lineage, which we further modified with additional supplements selected on the basis of reports from the eye field and retinal development. We report that hESC and hiPSC differentiating under our regimen over a 60 day period sequentially acquire markers associated with neural, retinal field, retinal pigmented epithelium and photoreceptor cells, including mature photoreceptor markers OPN1SW and RHODOPSIN with a higher efficiency than previously reported. In addition, we report the ability of hESC and hiPSC cultures to generate neural and retinal phenotypes under minimal culture conditions, which may be linked to their ability to endogenously upregulate the expression of a range of factors important for retinal cell type specification. However, cultures that were differentiated with full supplementation under our photoreceptor-induction regimen achieve this within a significantly shorter time frame and show a substantial increase in the expression of photoreceptor-specific markers in comparison to cultures differentiated under minimal conditions. Interestingly, cultures supplemented only with B27 and/or N2 displayed comparable differentiation efficiency to those under full supplementation, indicating a key role for B27 and N2 during the differentiation process. Furthermore, our data highlight an important role for Dkk1 and Noggin in enhancing the differentiation of hESC and hiPSC toward retinal progenitor cells and photoreceptor precursors during the early stages of differentiation, while suggesting that further maturation of these cells into photoreceptors may not require additional factors and can ensue under minimal culture conditions. STEM CELLS2012; 30:673–686


Outer retinal degeneration (ORD) is the leading cause of blindness in the developed world. There is still no treatment for many forms of ORD where retinal pigmented epithelium (RPE) dysfunction and photoreceptor demise predominate. Clinical trials indicate some protective effects of dietary supplements or intraocularly injected trophic factors in slowing disease progression, but these are not a definitive cure. Gene therapy strategies for specific forms of hereditary retinal dystrophy (HRD) are under development, and the viability of this approach for partial restoration of visual function has been demonstrated [1–5]. Yet this approach remains ineffective in cases where a substantial loss of photoreceptors has already occurred and is difficult to achieve in HRD resulting from complex underlying genetic etiology. In such cases, cell replacement remains an important option.

A seminal paper by MacLaren et al. demonstrated that the only cells capable of successful integration and differentiation into functional rod photoreceptors in the mouse retina were postnatal postmitotic cells that were already committed to a photoreceptor fate [6]. By contrast, only embryonic stage Crx-positive cells were able to differentiate into cone photoreceptors following transplantation [7]. For photoreceptor replacement in humans, the equivalent donor cell type would have to be obtained from the second trimester embryo, which raises clear ethical concerns. A more viable option would be to try and differentiate human embryonic stem cells (hESC) and/or human induced pluripotent stem cells (hiPSC) into photoreceptors in vitro prior to transplantation.

Several key publications have shown that it is possible to drive the differentiation of hESC and hiPSC toward photoreceptors and RPE using various approaches ([8–15] and Supporting Information Table S1). It has been suggested that early retinal progenitor cells (RPCs) can be obtained from ventral neural populations derived from differentiating hESC and hiPSC cultures in the absence of growth factors; however, this process is heavily dependent on the cell line-specific endogenous signaling, especially those related to Dkk1 and Noggin [11, 16].

The RPE, which can be generated during the spontaneous differentiation of hESC and hiPSC upon removal of basic fibroblast growth factor (bFGF; Supporting Information Table S1), is critical to the normal function and survival of the light-sensitive photoreceptors; indeed the two cell types are interdependent on each other. It is likely that in many forms of ORD, the replacement of both cell types will be required, highlighting the need to identify new renewable cell sources that can be used for restorative purposes. The restorative potential of hESC- and more recently hiPSC-derived RPE has already been demonstrated in the dystrophic Royal College of Surgeons (RCS) rat [17–21]. RPE transplantation alone, however, although protective for remaining photoreceptors, does not act to replace photoreceptors that have already been lost. Additionally, the potential of hESC-derived photoreceptors for cell replacement has been demonstrated [22, 23], yet photoreceptor differentiation protocols remain variable and in some cases require extensive culture periods (Supporting Information Table S1). Two studies [22, 23] have shown that hESC- and hiPSC-derived retinal cell grafts can express photoreceptor markers and integrate within the outer nuclear layer in intact newborn and adult retina. However, following subretinal transplantation into Crx−/− mice (an animal model of Leber's Congenital Amaurosis), these cells failed to develop photoreceptor outer segments, indicating that their final maturation and functional integration within the compromised retina remains incompletely understood. Thus, development of protocols aimed at efficiently generating enriched population of photoreceptor precursors at the correct ontogenetic stage is key prior to any translation of cell-based restorative therapies for the treatment of photoreceptor loss.

With this aim, following on from previous reports which demonstrate the differentiation of retinal cell types from pluripotent cells, and by mimicking the chemical and cytokine microenvironmental signals known to guide retinal histogenesis during normal development, we have developed an adapted differentiation protocol to promote photoreceptor differentiation from hESC and hiPSC cultures. This method combines some aspects of previously published protocols on hESC and hiPSC differentiation toward the retinal lineage [8, 12], which we have then further modified with additional supplements that were added on the basis of reports from retinal and eye field developmental papers. We show that our serum and feeder-free differentiation method results in the generation of photoreceptor progenitors and RPE from hESC and hiPSC with higher efficiency than what has previously been achieved. Our data also suggest that at least some hESC and hiPSC lines are endowed with an endogenous potential to generate neural and retinal cells under minimal culture conditions that can be enhanced and/or accelerated by the addition of specific mitogens and metabolites (Dkk1, B27, and Noggin) during the initial phase of differentiation.


hESC and hiPSC Culture and Differentiation

hESC (line H9, WiCell, passage 34-45) and hiPSC generated from human neonatal dermal skin fibroblasts (hiPSC-NHDF, [24]) were grown under atmospheric oxygen conditions on mitotically inactivated mouse embryonic fibroblasts (MEFs) in hESC medium containing knockout-Dulbecco's modified Eagle's medium (DMEM), 1 mM L-glutamine, 100 mM nonessential amino acids (NEAA), 20% knockout serum replacement (KOSR, Gibco), 1% penicillin-streptomycin and 8 ng/ml bFGF (Invitrogen). hESC/hiPSC medium was changed daily and hESC/hiPSC colonies were passaged every 4-5 days at a ratio of 1:3-1:4 by incubation in 1 mg/ml collagenase IV. Prior to hESC/hiPSC differentiation, MEFs were removed by culture of hESCs and hiPSCs on hESC-qualified mouse matrigel (BD Biosciences, New Jersey) for two passages. Differentiation experiments were initiated using the protocol outlined in Figure 1A. hESC/hiPSC clumps were harvested following collagenase treatment and cultured in low adherence bacteriological plates (Fisher Scientific, Loughborough, UK) to form embryoid body (EB) cultures. Control cultures were differentiated in a basal ventral neural induction media (VNIM) (Fig. 1A, blue dashed line) consisting of DMEM/Ham's F-12 (DMEM/F-12, PAA, Linz, Austria) containing decreasing concentrations of KOSR (20% for the first 5 days, 15% until day 9 then 10% until day 37) supplemented with L-glutamine, NEAA, and B27 (1:50, PAA). On day 30 of differentiation, EB suspension cultures were transferred to attachment culture conditions on tissue culture were coated with poly-L-ornithine (dissolved at 10 μg/ml in sterile dH2O, Sigma, Dorset, UK) and laminin (5 μg/ml in phosphate buffered saline [PBS], Sigma) for the remaining 30 days. On day 37, culture medium was changed to a basal KOSR-free formula (Fig. 1A, blue dashed line) consisting of DMEM/F12/L-glutamine/NEAA/B27 and N2 (1:100, Gibco, Paisley, UK).

Figure 1.

Three-step induction protocol and examples of the resulting culture morphology. (A): Schematic showing the different stages of the 60-day differentiation protocol. Colored broken lines represent the different stages of media supplementation (Materials and Methods). (B–J): The morphology of human embryonic stem cells differentiating under (B–D, I, J) control and (E–G, H) photoreceptor-induction conditions at (B, E) 35 days, (C, F) 45 days, and (D, G, H) 60 days of differentiation. (H): Cells differentiating under conditions designed to induce photoreceptor production generated neural-like populations with a proportion of cells exhibiting long neurites. (I, J): The typical appearance of retinal pigmented epithelial-like cells in control cultures on day 35 of differentiation under (I) phase microscopy and (J) following confocal microscopy counterstained with Hoechst 33342. KOSR, knockout serum replacement. Scale bars: B, C, E, F = 200 μm, D, G, H = 100 μm, and I, J = 50 μm. Abbreviation: PR, photoreceptor.

To drive photoreceptor induction from hESC/hiPSC (which we have termed PR-induced cultures, shown by the red dashed line in Fig. 1A), hESC/hiPSC were differentiated as for control cultures but with VNIM that was supplemented with recombinant mouse (rm) Noggin (1 ng/ml, R&D Systems, Abingdon, UK), recombinant human (rh) Dickkopf-1 (Dkk1, 1 ng/ml, R&D Systems), Insulin-like growth factor 1 (rhIGF-1, 5ng/ml, Sigma) rhLefty A (500 ng/ml, R&D Systems), Human Sonic Hedgehog (Shh, 30 nM, Peprotech, London, UK) and 3,3′,5-triiodo-L-thyronine (T3, 40 ng/ml, Sigma) until day 37. KOSR-free media was then supplemented with rmNoggin (10 ng/ml), rhDkk1 (10 ng/ml), rhIGF-1 (10 ng/ml), rhbFGF (bFGF, 5 ng/ml), retinoic acid (500 nM, Sigma), T3 (40 ng/ml), taurine (2 μm, Sigma), and Shh (12 nM) until day 60. During days 37-41 (Fig. 1A, black dashed line), PR-induction medium was supplemented with Human Activin-A (100 ng/ml, Peprotech), to encourage the exit of photoreceptor progenitor cells from the cell cycle and their further maturation [25]. This induction method combines growth factors at defined concentrations shown to induce photoreceptor differentiation from two previous studies [8, 12] plus our own addition of Activin A, Shh (known to cause an increase in the number of retinal progenitors and photoreceptor cells) and T3 (known to promote cone photoreceptor differentiation in human fetal retinal cultures; [26]). Two different concentrations were used for Shh; 30 nM during the initial stages of differentiation (previously defined as optimal for inducing ventral forebrain specification; [27, 28]) and 12 nM in the last stage of differentiation (shown to promote differentiation of embryonic rat retinal cultures toward a photoreceptor phenotype; [29]). Minimal control cultures [con(−)] consisted of hESC/hiPSC population that were differentiated in basal media alone (Fig. 1A, green dashed line), in the absence of B27/N2/additional factors. For more information, refer to Supporting Information.


Quantitative Reverse Transcription-Polymerase Chain Reactions

Total RNA was extracted using TRIzol reagent (Invitrogen, Paisley, UK) according to manufacturer's instructions. Following DNase treatment using RQ1 DNaseI (Promega, Southampton, UK), cDNA was synthesized using SuperScript III Reverse Transcriptase (Invitrogen). Quantitative reverse transcription-polymerase chain reactions (qRT-PCR) analysis was carried out using SYBR GreenER PCR master mix (Invitrogen). All reactions were carried out and samples analyzed using an AB7900HT real-time analyzer. All primers used in this study are listed in the Supporting Information Table S2. Data were analyzed using qBase v1.3.5 and the comparative threshold cycle (Ct) method. In all samples, the results were normalized to the geometric mean of three housekeeping genes GAPDH, SDHA, and RPL13A, and referenced to hESC, 7-day differentiated hESC, human retina or human RPE in accordance with the gene of interest.

Array-Comparative Genomic Hybridization

To ensure that the hESC line (H9) used in this study had not acquired any chromosomal abnormalities during ex vivo expansion, we performed array-comparative genomic hybridization (aCGH) as outlined in Supporting Information. The analysis did not show significant abnormalities (Supporting Information Fig. S1A–S1C), with the exception of two close amplifications of 186 and 458 kb at 14q23.2 and 14q23.3, which correspond to known copy number variations (Database of Genomic Variants at http://projects.tcag.ca/variation/). We also noticed a deletion of 2MB at 6p22.2-p22.1; however, in-depth analysis suggested that the deletion observed at chromosome 6 and amplifications observed at chromosome 14 were not significant as they did not meet the criteria of significant genomic aberrations, which is the presence of three consecutive probe alterations (at chromosome 6p22.2-p22.1, three nonconsecutive deleted probes were found in a region comprising 77 probes [Supporting Information Fig. S1B], while at 14q23.2 and 14q23.3, only two consecutive probes were amplified, respectively [Supporting Information Fig. S1C]). We also investigated the copy numbers of 61 reported stem cell-associated genes [30] and this analysis showed no significant differences in DNA copy number over the entire chromosomal regions. In summary, our aCGH analysis indicated that H9 hESC used in this study did not show genomic instabilities that could impact upon their pluripotency.

Fluorescence Immunocytochemistry and Microscopy

Cells were fixed in 4% (w/v) paraformaldehyde for 15 minutes. Permeabilization (0.2% (v/v) Triton X-100 in PBS) was performed prior to staining with antibodies for internal cell markers. A blocking step was performed by incubation in antibody diluent (1% (w/v) bovine serum albumin, Sigma, with 5% (v/v) normal goat serum, Invitrogen) prior to staining for 30 minutes. Cells were incubated with anti-rhodopsin (Sigma), anti-opsin blue (OPN1LW, Millipore), anti-opsin red/green (OPN1SW, Millipore, Watford, UK), anti-RPE65 (AbCam, Cambridge, UK) antiprimary antibodies for 1 hour and secondary antibodies for 30-60 minutes. Nuclei were counterstained with 10 μg/ml Hoechst 33342 then cultures coverslipped using Vectashield (Vector Laboratories, Peterborough, UK). Bright-field and fluorescent images were obtained using a Zeiss inverted microscope and AxioVision software. Immunopositive cells were quantified manually by randomly sampling and photographing 10 fields of view at ×20 magnification. All Hoechst-labeled nuclei and immunopositive cells within these confines were counted.

Flow Cytometric Analysis

Differentiating cells were dissociated by treatment with Accutase (Gibco) and then processed in 35 μm Medicons using a Medimachine (Becton Dickinson, New Jersey). Cells were collected by centrifugation at 250g for 5 minutes and fixed first in 0.5% paraformaldehyde at 37°C for 10 minutes, then by treatment with 90% methanol for 30 minutes on ice. Cells were washed in PBS then transferred to incubation buffer (0.2% BSA and 5% goat serum in PBS) for 10 minutes at room temperature followed with unconjugated anti-Chx10, anti-CRX, anti-rhodopsin, anti-OPN1LW, anti-OPN1SW, or anti-RPE65 primary antibody for 1 hour before incubation with fluorescein (FITC) secondary antibody (1:800, Sigma) for 30 minutes at room temperature. Samples were washed in PBS and protected from light prior to analysis. Analysis was performed using the FACSCalibur system (BD Biosciences) with CellQuest Pro software (BD Biosciences). At least 10,000 events were analyzed in each experiment.

Statistical Analysis

Two-tailed paired Student's t test was used to analyze results between groups at various time points during the differentiation process and between control and PR-induced groups at the same stage of differentiation. Results were considered significant if p < 0.05.


To design a more efficient photoreceptor differentiation protocol for pluripotent cell types, we have combined some aspects of previously published protocols on hESC and hiPSC differentiation toward a retinal lineage [8, 12] and then made additional modifications on the basis of reports from retinal and eye field developmental papers. Akin to previously published methods [8, 12], our differentiation protocol (Fig. 1A) was designed to guide the transition of cells from a pluripotent state to that of the retinal lineage, more specifically, a photoreceptor fate by sequentially exposing cells to antagonists and cytokines known to direct: (a) a ventral neural fate; (b) eye field/retinal determination; and (c) photoreceptor specification and maturation. hESC/hiPSC were differentiated under this defined culture protocol (named PR-induction, red line in Fig. 1A) for a period of 60 days and analyzed by qRT-PCR, flow cytometry, and immunocytochemistry. Control cultures were set up in parallel in the absence of growth factors and/or culture supplements and named appropriately in each figure (blue and green lines in Fig. 1A).

The Loss of Pluripotency and Acquisition of a Neural Lineage

Differentiation of hESC under control and PR-induction conditions yielded no discernable differences in culture morphology at the EB stage. qRT-PCR analysis of gene expression demonstrated the rapid downregulation of pluripotency markers NANOG and OCT4 over the first 30 days, indicating the loss of pluripotency (Fig. 2). Following transfer to adherent culture conditions, EBs attached to generate proliferative cultures (Fig. 1B–1H) with many cells displaying neural-like morphology. Neural cells exhibit a variety of morphology depending on their phenotype and stage of maturation. Neural progenitor cells are often bipolar in shape with a small soma and shorter, less developed neurites than their mature neuronal and glial counterparts (e.g., Fig. 1D). Late stage neurons can differ vastly in their soma size as well as their polarity (ranging from unipolar to multipolar), dendritic arborization and axon length, some cells exhibiting axons which traverse great distances (e.g., as in Fig. 1D–1H). Pigmented foci, indicating the emergence of RPE cells, appeared within floating EB cultures over days 20-30 and following transfer to adherent culture conditions generated sheets of pigmented epithelial cells resembling native RPE (Fig. 1I and 1J). This occurred with higher frequency in control cultures, possibly due to the addition of bFGF in PR cultures, which has been shown to suppress RPE induction [10, 13].

Figure 2.

Pluripotency and germline gene expression in differentiating human embryonic stem cells (hESC) over 60 days. Assessment of pluripotency markers NANOG and OCT4 and lineage markers from all three germ layers in differentiating control (blue bars) and PR-induced population (red bars) over 60 days by quantitative reverse transcription-polymerase chain reactions. All data (Y-axis) are presented as the normalized ratio of gene expression to housekeeping genes GAPDH, SDHA, and RPL13A. NANOG and OCT4 are calibrated to hESC (hESC = 1), all other genes to hESC that had been differentiated for 7 days (H9d7 = 1). Data = mean ± SEM, n = 3. Abbreviations: FGF, fibroblast growth factor; PR, photoreceptor.

qRT-PCR analysis of PR-induced cultures revealed low levels of trophoctoderm (CDX2), endoderm (GATA4, AFP), and mesoderm (BRACHYURY) markers after day 30 of differentiation and an upregulation of the primitive ectodermal-associated gene FGF5 (Fig. 2, red bars). This occurred alongside an increase in SOX2 and NCAM1 gene expression over the differentiation period (Fig. 4A), indicating the emergence of neuroectodermal derivatives. In control cultures however, some expression of AFP, CDX2, and GATA4 was maintained over the 60 days of differentiation (Fig. 2, blue bars), with observed levels of FGF5 expression lower than in mitogen-treated population, indicating spontaneous differentiation toward other germ layers.

The Emergence of a Retinal Lineage

We went on to assess the emergence of eye field and retinal progenitor markers in cultures differentiating under our PR-induction and control regimens (Figs. 3, 4). NESTIN expression remained constant throughout the differentiation period while the early retinal marker PAX6 increased in PR-induced cultures over time (p = .048) from day 45 to reach a prominent peak at day 60 (Fig. 4A). Flow cytometric analysis detected the greatest number of cells expressing the retinal progenitor marker CHX10 on day 15 of differentiation in the PR-induced group (Fig. 4B, red line) showing an almost twofold increase over control population at this time (expressed in 25% of control population and 43% of PR-induced). After day 30, both population follow a similar expression course and decline from 13% of cells at day 45 (Fig. 3A) to only a fragment of the population expressing this antigen by day 60 (<1%, Fig. 4B). No significant changes were, however, detected in CHX10 transcript levels during the differentiation period and also between control and PR-induced groups. This might be explained by differences between quantification methods, the sensitivity of each method and the measurable outcome (protein vs. transcript) with flow cytometric analysis assessing the percentage of CHX10 immunopositive cells (and not total level of CHX10 protein expression) and qRT-PCR measuring the total level of the CHX10 transcript.

Figure 3.

The expression of retinal and photoreceptor-specific markers in differentiating human embryonic stem cells by immunofluorescence histochemistry and flow cytometry. (A): Flow cytometric analysis of CHX10, CRX, Opsin blue (OPN1SW), Opsin red/green (OPN1LW), and RHODOPSIN under control (con) and photoreceptor-induction conditions (PR) at 45 and 60 days. (B–D): Immunofluorescent staining of (B, C) RHODOPSIN (green) and ZO-1 (red) counterstained with Hoechst (blue) and (D) cone-specific OPN1SW (red) near cells immunopositive for the RPE marker RPE65 (green) at 45 days. Scale bars = 50 μm. (E): CRX, OPN1SW, and RHODOPSIN expression as detected by flow cytometry in H9 cells differentiated under control (con) and PR-induction conditions at 45 days. Values equal mean ± SEM (n = 3). Both cultures demonstrated expression of photoreceptor-specific markers at this time point. Abbreviation: RPE, retinal pigmented epithelium.

Figure 4.

The expression of differentiation-associated genes and antigens. (A): Quantitative reverse transcription-polymerase chain reaction analysis of neuroectodermal (SOX2, NCAM1, NESTIN, and NEUROD), retinal progenitor (PAX6 and CHX10), photoreceptor progenitor (CRX), phototransduction pathway (RECOVERIN, PDE6β, OPSIN, and ARR3), and retinal pigmented epithelial (RPE65 and SILV) gene expression in control (blue bars) and mitogen-treated population (PR, red bars) over the 60 days of differentiation. Data (Y-axis) is normalized to housekeeping genes and referenced to human embryonic stem cells (hESC; SOX2, NCAM1, NEUROD, NESTIN, and PAX6, hESC = 1), human retina (CHX10, CRX, RECOVERIN, PDE6β, OPSIN, and ARR3, human retina = 100), or human RPE (RPE65 and SILV, RPE = 1). Values equal mean ± SEM (n = 3). (B): Flow cytometric analysis of PR-induced (red line) and control cultures differentiated with B27/N2 (con, blue line) against control(−) cultures differentiated in the absence of B27/N2 (con(−), green line) revealed reduced immunopositivity for CHX10, RPE65, and OPN1SW in con(−) cultures and an important role for B27/N2 in the levels of retinal-specific gene expression previously observed in control population (A, blue bars) in the absence of any other mitogens or antagonists. Values equal mean ± SEM (n = 3). Abbreviations: PR, photoreceptor; RPE, retinal pigmented epithelium.

RPE65 is a cytoplasmic protein involved in retinoid metabolism [31, 32]. Flow cytometric and qRT-PCR analysis revealed an early peak in RPE65 expression at day 30 in the PR-induced group (Fig. 4A, 4B). In control cultures, we observed the emergence of RPE cells over the first 30 days which, unlike PR-induced cultures, continued to rise until day 60 (Fig. 4A, RPE65, SILV) consistent with published literature, which indicates spontaneous emergence of these cells in the absence of exogenous growth factors [10, 13, 15]. The expression of eye field and retinal progenitor markers PAX6, CHX10, and RPE-specific RPE65 alongside the appearance of pigmented foci over the first 30 days of differentiation indicates the emergence of early retinal-specific phenotypes from differentiating hESC.

The Emergence of a Photoreceptor Phenotype

Following the transfer of differentiating hESC to adherent culture conditions, we assessed the expression of photoreceptor markers by qRT-PCR, flow cytometry and immunofluorescent histochemistry (Figs. 3, 4). Flow cytometric analysis on day 45 of differentiation confirmed the persistence of some CHX10+ cells (up to 13%) and a similar number of cells were immunopositive for the early postmitotic photoreceptor precursor marker CRX (up to 16%, Fig. 3A). Over days 45-60, a marked increase was observed in CRX gene expression under control induction conditions (Fig. 4A). This led us to test whether any differentiating cells within these populations might be expressing other photoreceptor markers. This included the neuronal calcium-binding protein RECOVERIN, ARRESTIN 3 which is a marker of outer segments of red, green, and blue cone photoreceptors, and cone-specific light-sensitive G-protein-coupled receptors, or opsins. Indeed, from days 45 to 60 of differentiation, an increase in the expression of ARRESTIN 3 was detected under control conditions (Fig. 4A).

Flow cytometric analysis revealed the highest expression of photoreceptor markers RHODOPSIN, cone-specific opsin blue (OPN1SW) and opsin red/green (OPN1LW) at day 45 of differentiation (found in 18%, 52%, and 60% of cells cultured under PR-inducing conditions, respectively, Fig. 3A, 3E). By day 60, this proportion had declined, resulting in almost complete loss of RHODOPSIN (1.1%) and only a small number of OPN1SW (∼ 4%) immunopositive cells remaining, while 13% of the population remained OPN1LW+ at this time (Fig. 3A). The expression of photoreceptor-specific markers was confirmed following immunocytochemical analyses and manual cell counts (Fig. 3B–3D).

Detection of the postmitotic photoreceptor marker CRX alongside late-stage photoreceptor-specific markers indicates the emergence of hESC-derived photoreceptors within cultures differentiated using our protocol. Comparison of control and PR-induced cultures by qRT-PCR, however showed only one statistically significant increase in the expression of OPSIN in PR-induced population observed at day 60 (p = .035, Fig. 4A). In addition, flow cytometric analysis from day 30 onward showed no significant changes in the expression of photoreceptor-specific markers (Figs. 3E, 4B) and, in most cases, control and PR-induced populations followed a similar time course with regard to the onset of neuroectodermal, retinal progenitor and photoreceptor-specific markers and levels of gene expression, suggesting that perhaps an endogenous and compensatory growth factor signaling pathway is active in control cultures, corroborating findings reported by Meyer et al. [11].

Further Investigation into the Conditions Enabling Retinal Specification from Differentiating hESC Reveals an Important Role for B27 and N2

As shown above, we observed that control populations differentiated under basal media conditions were capable of exhibiting neural, eye field, retinal and photoreceptor gene expression over time. This was an unexpected result and we set out to address the role of specific media components during hESC differentiation. Human ESCs express the receptors for fibroblast growth factor (FGF), activin A, epidermal growth factor (EGF), nerve growth factor (NGF) and retinoic acid; however, the ligands for these are all absent in commercially available KOSR [33]; therefore, it is not likely to be the presence of KOSR during the first 37 days of differentiation which enabled these effects. Some constituents of the neural supplement B27 are demonstrated to promote neuralization, enhancing neural progenitor proliferation and differentiation as well as the development of the machinery necessary for phototransduction [34–36]. This suggests that B27, a component of our basal ventral neural induction medium and therefore present in control cultures during differentiation, may play a role in the specific differentiation of hESC towards retinal cells. We therefore questioned whether the removal of this supplement over the first 37 days of differentiation followed by the combined removal of B27 and N2 in the final window of differentiation (days 37-60) would affect differentiation potential.

To investigate this, we introduced an additional control into our experiments, by differentiating hESC cultures in the absence of B27 or N2 in a minimal media [named control(−) cultures, refer to Figure 1A green line]. hESC differentiating under these conditions lost their pluripotency marker expression in a similar way to control and PR-induced groups (Supporting Information Fig. S2). The removal of B27, followed by combined B27 and N2 removal from the differentiation media resulted in a reduction in early expression levels of the retinal progenitor marker CHX10 when compared with supplemented cultures; but this was, however, followed with by a steady increase in CHX10 over days 15-60 (Fig. 4B), reaching by day 60 the same expression level as observed in B27/N2 supplemented cultures at day 15 of differentiation. Similar results were also observed for the RPE marker RPE65, where the same expression levels were observed for PR-induced, control and control(−) groups on day 60 of differentiation (Fig. 4B). Despite this, hESC differentiating under control(−) conditions failed to reach the peak OPN1SW expression observed at day 45 in PR-induced and control cultures. Both sets of results shown in Figures 3E and 4B suggest no significant differences in early and mature photoreceptor marker expression between control and PR-induced populations. Given that B27 (present throughout differentiation) and N2 (added in the final window of differentiation after day 37, refer to Fig. 1A) are the only common culture components between PR-induced and control groups and are both missing from the control(−) group, this suggests a role for these two components in enhancing and accelerating the endogenous capacity of hESC cultured under minimal culture conditions to undergo neural and retinal differentiation.

Similar results were obtained during the differentiation of hiPSC-NHDF cells (Supporting Information Fig. S3). Expression of both cone-specific opsin blue (OPN1SW) and opsin red/green (OPN1LW) at day 45 of differentiation was similar across control and PR-induced cultures, but lower in control(−) cultures, while the opposite was observed for early retinal marker, PAX6 (Supporting Information Fig. S3). Together, these results indicate an endogenous, although somewhat delayed, capability of differentiating hESC cultures to produce certain neural and retinal phenotypes during prolonged culture in the absence of specific inductive cues.

The Endogenous Expression of Key Signaling Factors in hESC

The endogenous expression of the Wnt antagonist DKK1 and FGF signaling has previously been reported in hESC differentiating toward the eye field [11]. Endogenous SHH signaling in mouse ESC (mESC) can promote a ventral neural fate [37] and, although endogenous SHH is very low in hESCs, levels of SHH become enriched upon EB differentiation [38]. We therefore investigated the expression of various factors in hESCs differentiating under our protocol. qRT-PCR results indicated that endogenous levels of EGF, NGF, and SHH rose in hESC cultures over the differentiation period, NODAL was strongly expressed on day 15 but then rapidly declined, while expression of DKK1 remained relatively constant after this time point. We observed no significant difference in the endogenous expression of EGF, NGF, SHH, NODAL, or DKK1 between control and PR-induced cultures (Fig. 5A, blue and red lines).

Figure 5.

Assessment of endogenous growth factor and antagonist gene expression. (A): Quantitative reverse transcription-polymerase chain reactions (qRT-PCR) analysis of EGF, NGF, SHH, NODAL, and DKK1 expression was performed at 15-day intervals over the differentiation period in control (with B27/N2 only, blue line), minimal control (no B27/N2/mitogen supplementation, green line) and in B27/N2/mitogen-supplemented PR-induced population (PR, red line). (B): qRT-PCR analysis showing delayed onset of NEUROD expression in control(−) cultures compared with control and PR-induced population. (C): qRT-PCR analysis revealed that, of all conditions tested, the highest expression of CRX and RECOVERIN on day 60 was seen in control and PR-induced cultures. (A–C): Data (Y-axis) are normalized to housekeeping genes and calibrated to human embryonic stem cells = 1 for panel A and human retina for panels B and C (human retina = 1). Values equal mean ± SEM (n = 3). Abbreviations: EGF, epidermal growth factor; NGF, nerve growth factor; PR, photoreceptor; and SHH, human sonic hedgehog.

When these results were compared with control(−) population, some differences emerged (Fig. 5A, green lines). Akin to our observations of the delayed upregulation of early retinal markers in control(−) cultures, increased endogenous expression of EGF, NGF, NODAL, and SHH was observed from day 45 to day 60 in these cultures, reaching levels that were greatly elevated in comparison with control and PR-induced population (Fig. 5A). DKK1 levels were also higher in control(−) cultures suggesting elevated endogenous Dkk1 signaling in hESC differentiating under minimal culture conditions and corroborating previously published data [11].

We then went on to assess the effects of each of our three culture conditions; control(−), control, and PR-induced on the expression of neuroectodermal, retinal, and photoreceptor genes. Our analysis demonstrated that hESC differentiating for 60 days without B27 and N2 resulted in a delay in the onset of neuroectodermal gene expression when compared with normal controls and PR-induced population (Fig. 5B), consolidating the flow cytometry data analysis shown in Figure 4B. Nonetheless, by qRT-PCR analysis, the cultures demonstrating the best CRX and RECOVERIN expression on day 60 across all groups were the control (with B27/N2) and PR-induced populations (Fig. 5C) corresponding with the data shown in Figure 4B and once more suggesting that at least B27 and/or N2 (present in both control and PR-induced cultures) are two important media components that enhance the induction of photoreceptor-specific gene expression in differentiating hESC.

This suggests that, in the absence of external cues, a proportion of hESCs follow an intrinsic neural and retinal differentiation pathway, which is associated with the delayed onset of neural and retinal markers under minimal culture conditions. This capability, as seen in minimal controls, may be due to the upregulation of endogenous signaling factors starting as early as the first 15 days (DKK1) but from our observations in most cases between days 30 and 60 (EGF, NGF, NODAL, and SHH). Nonetheless, the induction of photoreceptor marker expression remained more efficient in control and PR-induced cultures.

Further Analysis of the Individual Contribution of Each Factor During hESC Differentiation

To test the individual contribution of each component of the PR-induction media, we supplemented differentiation media with isolated factors during floating EB culture for up to 30 days. There were some apparent differences in gross EB morphology between normal controls (with B27) and minimal controls (no B27) at this stage of differentiation (Fig. 6A; panels a and b, respectively) with more cystic EBs appearing in the minimal control group. In addition, some interesting morphological differences emerged following treatment with individual factors. The addition of T3 promoted the formation of small, highly pigmented EBs (Fig. 6A, panels c and d), while LeftyA induced the formation of polarized EBs containing a prominent axis (Fig. 6A, panel e). Interestingly, differentiation with IGF-1, shown specifically to induce eye formation in Xenopus and an enhancer of retinal progenitor gene expression [39], resulted in the appearance of disc-shaped EBs containing numerous visible internal neural rosette-like structures (Fig. 6A, panels f and g), which have also been reported under minimal culture conditions [16].

Figure 6.

Further analysis of the individual contribution of B27/N2 supplements and added factors on endogenous gene expression and embryoid body morphology. (A): Morphological differences were observed between EBs from (a) control(−) and (b) control cultures as well as during supplementation with individual factors. T3 treatment promoted the formation of small, highly pigmented EBs (c, d), LeftyA induced the formation of polarized EBs with a prominent axis (e), while in IGF-1-treated population disc-shaped EBs emerged which contained numerous embedded optic cup-like structures (f, g) only rarely seen to contain pigmented cells (g). (B):PAX6 and CHX10 expression at days 15 and 30 was enhanced in cultures treated with Dkk1, noggin, or T3 during the first 30 days of differentiation. Data (Y-axis) are normalized to housekeeping genes. Results for PAX6 and CHX10 are calibrated to human retina (human retina = 1). Values equal mean ± SEM (n = 3). Abbreviations: IGF, insulin-like growth factor; shh, human sonic hedgehog; and T3, 3,3′,5-triiodo-L-thyronine.

Our quantitative RT-PCR expression analysis detected the best CHX10 expression in cultures exposed to B27 plus Dkk1 or B27 plus noggin (Fig. 6B). In cultures supplemented with B27 and one other factor only, PAX6 expression was also found to be higher than PR cultures differentiated with all factors together (Fig. 6B). Endogenous Dkk1 signaling has been implicated in acquisition of early retinal markers (PAX6 and RAX) during differentiation of hESC [11]. To investigate this further in our differentiation protocol, we performed flow cytometric analysis in Dkk1-supplemented control cultures (but devoid of all other added factors) and observed a 76.6% increase in number of CRX expressing cells and a 7.8% decrease in number of cone-specific opsin blue (OPN1SW) expressing cells (data not shown), suggesting an important role for Dkk1 in enhancing the specification of photoreceptor precursors but not later stage photoreceptor cells.

In cultures exposed to PR-induction conditions for the first 37 days then swapped to minimal media either alone or supplemented with isolated factors, we observed that some factors elicited similar results (Fig. 7A). For example, hESC differentiated under minimal conditions (no B27/N2/factors) from days 37 to 60, or in minimal media supplemented with Dkk1, N2, or T3 resulted in the elevated endogenous expression of FGF2, SHH, DKK1, EGF, and NGF (Fig. 7A). We went on to assess the effect of each media component on the expression of retinal and photoreceptor genes. Of mitogen-treated groups, our analysis demonstrated the best expression of OPSIN, PDE6β, and ARRESTIN3 on day 60 in cultures that had been differentiated under PR-induction conditions for 37 days and then either with no factors or supplemented with N2, Dkk1, or T3 (Fig. 7B). These qRT-PCR results were confirmed by flow cytometry data, which showed an increase in the number of CRX+ (84%), OPN1SW+ (89.2%), and OPN1LW+ (34.6%) cells in cultures that had been differentiated under PR-induction conditions for 37 days and subsequently under minimal conditions compared to those subjected to PR-inducing conditions for the entire differentiation period (data not shown). Together these data suggest an important role for B27, Dkk1, and Noggin in the first 37 days of differentiation. In addition, they confirm that although the addition of N2, Dkk1, or T3 during the final stage of differentiation can facilitate the yield of the mature photoreceptors, this is not a requirement for further cell maturation to occur.

Figure 7.

Further analysis of the directed differentiation of human embryonic stem cells (hESC) with different combinations of supplements and mitogens revealed variance in the expression of mitogens, mature retinal genes, and culture morphology. Treatment of hESC with B27/N2 and all factors for 37 days and then with isolated factors for the remainder of the differentiation period induced variation in (A) endogenous gene expression and (B) photoreceptor and RPE-specific genes on day 60 of differentiation. All treatments generally yielded heterogeneous neural-like cultures (C–I) with the presence of RPE sheets displaying distinct boundaries (D, E). Cells within adherent EBs commonly generated long processes, some of which extended hundreds of microns (C). Optic-cup-like structures that were observed following IGF-1 treatment during floating culture were maintained following transfer to adherent conditions in PR-induction media (F). These were also observed in population differentiated with PR-induction media over the first 37 days and then with either T3 (G) or taurine (H), and in lower frequency in cultures grown under PR-induced conditions for the full differentiation period (I). Data are normalized to housekeeping genes and referenced to hESC (panel A, hESC = 1), human retina (panel B, human retina = 1), or human RPE (RPE = 1, RPE65, panel B). Values equal mean ± SEM (n = 3). Scale bars: C–E, G, I = 100 μm and F, H = 400 μm. Abbreviations: bFGF, basic fibroblast growth factor; EGF, epidermal growth factor; FGF, fibroblast growth factor; IGF-1, insulin-like growth factor 1; NGF, nerve growth factor; PR, photoreceptor; RPE, retinal pigmented epithelium; SHH, human sonic hedgehog; and T3, 3,3′,5-triiodo-L-thyronine.


The ability to generate expandable populations that can yield a large number of photoreceptors is a prerequisite for the successful application of cell replacement therapy for outer retinal degeneration. Our strategy for photoreceptor production from hESC and hiPSC relies upon the use of various factors and antagonists that are known to play a role in neural induction and retinal histogenesis during development and from in vitro studies [8, 11, 12, 21, 25, 40-42]. Our aim was to develop a protocol to increase the yield of photoreceptors from hESC and hiPSC in a shorter time frame than what has been achieved previously.

Our results indicate that our protocol directs the stage-specific differentiation of hESC and hiPSC toward a neural fate, followed by retinal field and then photoreceptor lineage. The appearance of these phenotypes occurred in a sequential manner consistent with retinal development in vivo [43]. We observed that CHX10+ retinal progenitors emerged during the early phase of differentiation, while the majority of CRX+photoreceptor progenitors were generated over days 30-45. The onset of mature photoreceptor markers at this time indicated that the emergence of the two distinct photoreceptor types, rods and cones, was also concordant with what one might expect developmentally; cone photoreceptors are born during the early phase of retinal histogenesis while rods emerge later [44]. Our gene expression results support this observation, showing elevated levels of cone-specific OPSIN on day 30, prior to the peak of rod-specific gene expression on day 60.

We also report an endogenous capacity of hESCs and hiPSC to generate neural and retinal phenotypes under minimal culture conditions. This capability has previously been demonstrated in hESC and hiPSC [11, 16], but is not a widely published phenomenon and it may be due to the endogenous production of various signaling molecules and growth factors from the heterogeneous mix of cells present in embryoid body cultures generated from pluripotent stem cells as demonstrated by this study and others. Trophic support for emerging retinal cell types from neighboring RPE cells is another potential explanation for the elevation in retinal-specific markers observed in populations differentiated in the absence of mitogens or supplements [45].

A recent groundbreaking paper demonstrates that structures reminiscent of the optic vesicle can arise from mESCs differentiated in low-concentration matrigel and minimal differentiation medium [46], highlighting a propensity for mESC to undergo stepwise differentiation and self-organization under very simple culture conditions. This corroborates and extends previously published data showing that mouse ESC cultured under minimal conditions and without instructive signaling can undergo neural differentiation [47]. We also observed the emergence of neural rosette-like structures from hESCs and hiPSC during the first 30 days of differentiation. These were most commonly observed during the early stages of differentiation in cultures treated with IGF-1, a known inducer of eye formation and a potent enhancer of retinal progenitor gene expression [8, 39]. During the latter stages of differentiation, we also observed numerous rosette-like structures in populations that had been treated with all supplements and mitogens for photoreceptor-induction over the first 37 days, followed by treatment with T3 or taurine only. Some were also observed in cultures differentiated under our photoreceptor-induction regime for the entire differentiation period, albeit with less frequency. These rosette-like structures are similar to optic vesicle-like structures reported by Meyer et al. [16] during hESC/hiPSC differentiation, and current work within our group is focusing on identification of the RPC types present therein and their potential to generate functionally mature photoreceptor and RPE cells.

One previous study has shown that the treatment of hESCs with B27, Dkk1, noggin, IGF-1 and subsequently with N2 and FGF can result in more than 80% of cells expressing retinal progenitor markers [8]. However, although 12% of the population expressed CRX, less than 0.01% were found to be positive for mature photoreceptor markers. Osakada et al. achieved CRX expression in 11% of the population following 120 days of differentiation, which increased to around 20% by 170 days, with 5% of cells expressing Rhodopsin by day 150 [12]. Meyer et al. achieved 55.9% of CRX expressing cells (at day 80 of differentiation), which undergo differentiation to photoreceptor-like cells displaying characteristic electrophysiological responses [16]. Each of these studies represents great achievement (Supporting Information Table S1 for more detail). Yet for this therapy to be translated to the clinic, it will be advantageous if this process was more efficient. We set out to try and improve existing protocols and from our results we report that we are able to generate 16% of CRX+ cells and 52% of cone-like photoreceptor cells within 45 days of differentiation. Although this differentiation period is longer than what has been reported in one earlier study [8], the higher efficiency of generated cells that express mature photoreceptor markers compared to previous studies after 45 days of differentiation (refer to Supporting Information Table S1) strongly suggests that our three-step differentiation protocol extends and improves current achievements in the hESC/hiPSC differentiation field. Notwithstanding the increased efficiency, our protocol has also its limitations. We observed the rapid loss of cells committed to a photoreceptor fate (CRX+/OPSIN+/RHODOPSIN+) over days 45-60. This is similar to observations in a study using cultured mouse RPCs where the rapid loss of rhodopsin-expressing cells was observed shortly after plating in RPC culture [48]. How to promote the long-term survival of postmitotic photoreceptors during prolonged culture remains an important challenge. It seems that this will not be achieved by structural means alone [46]; however, purification of various cell types during their peak production period may overcome this issue somewhat.


Our three-step differentiation protocol was designed with the aim to mimic neural and retinal development. To achieve this, we combined serum- and feeder-free culture conditions with known growth factors and metabolites that have been used with some degree of success in previous hESC differentiation studies [8, 12] alongside additional factors that are implicated to play a role in retinal development from developmental studies. To be able to understand the role of each of these factors, we performed stepwise deletions of single and/or combined factors. This detailed analysis indicated that the most important components for achieving efficient photoreceptor generation within a reasonable time frame (45 days) are B27 and/or N2. While RPC and photoreceptor precursor generation could be enhanced by the addition of Dkk1 and Noggin to B27-supplemented medium during the early stages of differentiation, the further maturation of these cells to express mature photoreceptor markers was most efficient in the absence of added supplements during the final stage of differentiation. These findings are in close resonance with the reported propensity of mESC to undergo stepwise differentiation resulting in the formation of self-organized optic cup structures under very simple culture conditions [46] and while it remains to be investigated whether human pluripotent stem cells are endowed with such intrinsic self-organizing capacity, the data presented in this manuscript on the role of various metabolites and growth factors provide a solid platform for initiating such investigations.

Endogenous growth factor signaling specific to each human pluripotent cell line has been shown to affect the differentiation outcome [16, 49]. Although we obtained very similar results from the differentiation of cells of the H9 hESC line and one hiPSC line generated in our laboratory, we are aware that the application of our differentiation protocol to other pluripotent stem cells may require optimization of these signaling pathways for efficient photoreceptor generation. Rapid developments in the field of induced pluripotency are now clearing the path toward the production of patient-derived stem cells that overcome the ethical and methodological issues surrounding the use of embryonic derivatives [50, 10, 51, 52]. These, alongside reports of iPSC production in patients with HRD [16], combined with genetic correction now set the stage for the future successful application of cell replacement therapy in patients suffering from currently incurable forms of blindness.


We are grateful to Fight for Sight UK, Prof. John Finlay, Newcastle Health Charity, Sunderland Eye Infirmary, Newcastle University, UK NIHR Biomedical Research Centre for Ageing and Age-related disease and the Newcastle upon Tyne NHS Hospitals Trust, Conselleria de Sanidad (Generalitat Valenciana), and the Instituto de Salud Carlos III (Ministry of Science and Innovation) for funding this work. We thank Dr. Owen Hughes and Ian Dimmick for their help with the flow cytometric analysis.