Hypoxia Increases the Yield of Photoreceptors Differentiating from Mouse Embryonic Stem Cells and Improves the Modeling of Retinogenesis In Vitro§


  • Marcela Garita-HernÁndez,

    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
  • Francisco Diaz-Corrales,

    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
  • Dunja Lukovic,

    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
  • Irene GonzÁlez-Guede,

    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
  • Andrea Diez-Lloret,

    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
  • M. Lourdes ValdÉs-SÁnchez,

    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
  • Simone Massalini,

    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
  • Slaven Erceg,

    Corresponding author
    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    • CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
    • Telephone: +34 954 468 004; Fax: +34 954 461 664

  • Shomi S. Bhattacharya

    Corresponding author
    1. CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    2. Department of Genetics, UCL Institute of Ophthalmology, London, United Kingdom
    • CABIMER (Centro Andaluz de Biología Molecular y Medicina Regenerativa), Avda. Americo Vespucio s/n, Parque Científico y Tecnológico Cartuja, Sevilla, Spain
    Search for more papers by this author
    • Telephone: +34 954 468 004; Fax: +34 954 461 664

  • Author contributions: M.G.-H.: conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript, and other (perform experiments and graphic design); S.E.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript; F.D.-C., D.L., I.G.-G., S.M., L.V., and A.D.-L.: other (perform experiments); S.S.B.: principal investigator, conception and design, 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 January 29, 2013.


Retinitis pigmentosa (RP), a genetically heterogeneous group of diseases together with age-related macular degeneration (AMD), are the leading causes of permanent blindness and are characterized by the progressive dysfunction and death of the light sensing photoreceptors of the retina. Due to the limited regeneration capacity of the mammalian retina, the scientific community has invested significantly in trying to obtain retinal progenitor cells from embryonic stem cells (ESC). These represent an unlimited source of retinal cells, but it has not yet been possible to achieve specific populations, such as photoreceptors, efficiently enough to allow them to be used safely in the future as cell therapy of RP or AMD. In this study, we generated a high yield of photoreceptors from directed differentiation of mouse ESC (mESC) by recapitulating crucial phases of retinal development. We present a new protocol of differentiation, involving hypoxia and taking into account extrinsic and intrinsic cues. These include niche-specific conditions as well as the manipulation of the signaling pathways involved in retinal development. Our results show that hypoxia promotes and improves the differentiation of mESC toward photoreceptors. Different populations of retinal cells are increased in number under the hypoxic conditions applied, such as Crx-positive cells, S-Opsin-positive cells, and double positive cells for Rhodopsin and Recoverin, as shown by immunofluorescence analysis. For the first time, this manuscript reports the high efficiency of differentiation in vivo and the expression of mature rod photoreceptor markers in a large number of differentiated cells, transplanted in the subretinal space of wild-type mice. STEM CELLS 2013;31:966–978


Stem cell therapy is a potential treatment for retinal dystrophies. Retinitis pigmentosa (RP) and age-related macular degeneration (AMD), the leading causes of permanent blindness in humans, are characterized by the progressive dysfunction and death of photoreceptor cells of the retina. For such diseases, the replenishment of functional photoreceptor precursors may be a good strategy for retinal regeneration, as gene therapy or growth factor supplement cannot regenerate dying photoreceptor cells. Retinal development is a multistep process involving cell cycle exit, migration, and changes of cell morphology [1]. These changes result from a reciprocal relationship between tissue-tissue interaction and cell intrinsic factors [2]. Additionally, accumulating evidence suggests that other components of the niche, such as oxygen tension, play an important role in cell fate determination during the development of many tissues, including the nervous system and the retina [3, 4].

Mouse embryonic stem cells (mESC) allow us to recapitulate retinal development in vitro. These cells are derived from the early embryo and are characterized by their two unique features of pluripotency and self-renewal [5]. In particular, during implantation and fetal development, stem cells live at oxygen tensions between 2% and 8% [6].

Early embryonic formation during mammalian development occurs in a precise environment, where the O2 tension plays a critical role [7]. In comparison to the atmospheric oxygen tension (20%), the uterus environment is hypoxic. Mammals including rabbits (8.7% oxygen tension) and monkeys (1.5% oxygen tension) [8] as well as humans develop embryos under low oxygen tension. Up until the second trimester in humans this ranges from 0% to 3% [9, 10]. The retina is not an exception and recent studies have shown the important role that hypoxia may play in neuroprotection and development of the human retina [11]. This relative hypoxia or tissular normoxia is relatively low compared with traditional in vitro culture conditions (20% O2) [12]. Once, the pluripotency agents such as leukemia inhibitory factor (LIF) are removed, ESC spontaneously differentiate following a reproducible temporal pattern of development, which in many ways recapitulates early embryogenesis [13]. Due to these special characteristics, ESCs are considered an unlimited source for cell replacement therapies. The formation of embryoid bodies (EBs), which are three-dimensional aggregates of ESC, is the initial step in ESC differentiation. Therefore, EB culture has been widely used as a trigger for the in vitro differentiation of ESC. Numerous groups have recently demonstrated that ESC can be converted into cells that resemble retinal progenitors [14–16], photoreceptors [17], or retinal pigment epithelium (RPE) [18, 19]. Furthermore, Meyer et al. have very elegantly mimicked the early retinal development in a stepwise fashion typical of normal retinogenesis [20] and high yield of cells differentiated toward photoreceptors was achieved by Mellough et al. [21]. Others went further attempting to obtain three-dimensional structures of early optic cup using scaffolds [22, 23]. However, evidence of fully characterized high-yield populations of photoreceptors or mature RPE cells has not yet been accomplished. The main drawback of the differentiation methods available is the very low efficiency along with the lack of reproducibility. Most of the protocols published in recent years have been geared toward the induction of expression of retina-specific transcription factors, but very few publications have included information regarding in vivo integration of differentiated cells into the mouse retina, suggesting cell survival, migration, and functionality of the grafted cells [24, 25]. In this study, we have optimized and fully characterized an original protocol of differentiation that allows us to obtain photoreceptors at a high efficiency in a reproducible way. Our approach involves the use of small molecules, growth factors, and morphogenic drugs in specific growth medium to differentiate mESC using hypoxic conditions. Our work is based on the hypothesis that hypoxia influences stem cell characteristics in vivo, showing that lower oxygen tensions in vitro could mimic the microenvironment and improve the modeling of retinogenesis in vitro. To demonstrate the efficiency of our protocol in vivo, we transplanted our cells after 20 days of differentiation in the subretinal space (SS) of wild-type mice and found that they were able to complete differentiation in vivo. More than 90% of the transplanted cells expressed mature rod-specific markers such as Recoverin (Recvn) or Rhodopsin. Furthermore, no tumoral growth was observed in any of the animals transplanted, corresponding to the absence of proliferative cells in the grafted population due to a highly efficient differentiation process.


Methods used to develop this work have been described elsewhere [26–28]. See supporting information for a detailed media description (supporting information Table S1) and supporting information methods.

Maintenance of mESC Pluripotent Cultures

All experiments conducted in this study were carried out using ES-D3 cells (ATCC CRL 1934) [29] passages 18–35. ESCs were maintained modifying already published protocols [30, 31] and were incubated at 37°C under 20% oxygen tension. Cultures were passaged every 4–7 days and grown at low confluence at a 1:1,000 split ratio. Fresh medium was exchanged every 48 hours. Appropriate Master and Working Cell Banks were generated to allow all the experiments to be accomplished using early passages.

ESC Differentiation

One week prior to starting the differentiation protocol, cells were incubated at 37°C in 5% CO2 under either Normoxic (20% oxygen tension) or Hypoxic (2% oxygen tension) conditions in a Thermo Fisher incubator (CO2/O2 WJ IR Model 3141, Thermo Electron Corporation, Fisher Scientific, Waltham Massachusetts, http://www.thermofisher.com). Both conditions were maintained during the whole time of differentiation. Oxygen tension control was monitored daily.

Spontaneous Differentiation: Model of EBs

EBs were generated following an optimized protocol of the hanging drop method described by Wobus et al. [32]. mESCs were dissociated using 0.05% Trypsin for 4 minutes at 37°C. Trypsin was washed away adding EBs medium. The cell suspension generated was spun down by centrifugation (Beckman coulter) and pelleted cells were resuspended in ES medium at the desired concentration (1,000 cells per 30 μl). Hanging drops of 30 μl were plated onto the lid of a 150 mm2 ultralow attachment plate (Soria Greiner) using a multichannel pipette (Eppendorf). ESCs were allowed to aggregate in hanging drops for 3–4 days before transfer to a suspension culture. After 3 days, identical spherical EBs were formed and each drop was collected individually with a 100 μl pipette and deposited into a 10-cm ultralow-attachment dish (Soria Greiner) containing 10 ml of EB medium to a final concentration of 100 EBs per dish. The EBs were cultured for 5 and 7 days and the medium was changed every 2–3 days.

Directed Differentiation of mESC Toward Retinal Progenitors and Retinal Mature Phenotypes

Retinal differentiation of mESC was accomplished using an optimized protocol encompassing growth factors described in previously published protocols [15, 17, 33] (all R&D Systems, unless otherwise specified) combined with the manipulation of the microenvironment. EBs generated from mESC were induced to differentiate in progenitors medium supplemented with 100 ng/ml Dickkopf-related protein 1 (Dkk1) and 500 ng/ml Lefty-A for 5 days at 37°C with 5% CO2 under Normoxic (20%) or Hypoxic (2%) conditions. Media were changed 72 hours later and fresh aliquots of the growth factors were added along with 5% fetal bovine serum (FBS) and 100 ng/ml Activin-A. On day 5, media were changed and EBs were cultured for 5 more days in Progenitors media without the addition of any growth factor. Fresh Progenitors medium was changed every 48 hours. On day 10, EBs were plated in six-well plates or coverslips coated with human recombinant 30 μg/cm2 Laminin (Sigma) and 150 μg/cm2 poly(L-ornithine) (Sigma) and cultured in retinal medium supplemented with 10 μM N-[(3,5-difluorophenyl)acetyl]-L-alanyl-2-phenylglycine-1,1-dimethylethyl ester (DAPT; Calbiochem) at 37°C with 5% CO2 for 48 hours. Ideal cell density was established in 150 EBs per 9.6 cm2. To allow better attachment of the EBs, 10% FBS was added to the medium for 48 hours. Retinal medium was exchanged every 2 days. On day 16 and until day 24, retinal medium was supplemented with 10 μM DAPT, 50 ng/ml acidic fibroblast growth factor (FGF), 10 ng/ml basic FGF (Millipore), 3 nM Sonic hedgehog homolog (Shh), 0.5 μM retinoic acid (RA; Sigma), and 100 μM Taurine (Sigma). From day 24 to day 28 retinal medium was supplemented with 10 μM DAPT, 3 nM Shh, 0.5 nM RA, and 100 μM Taurine. Samples were collected on day 0, day 5, day 10, day 16, and day 28 for molecular biology analysis and immunocytochemistry. Methods of RNA isolation and quantitative RT-PCR, immunocytochemistry, apoptosis and cytotoxicity assays, and fluorescence-activated cell sorting (FACS) are described in supporting information methods.

Statistical Analyses

Data are the mean ± SEM of at least three independent experiments, except for the immunocytochemistry, for which a representative image from three assays is depicted in the figures. Comparisons between values were analyzed using one-way analysis of variance; p ≤ .05 was considered statistically significant.

Preparation of Cells for Transplantation

Cells, after 20 days of in vitro differentiation, were trypsinized to obtain a single-cell suspension. Harvested cells were labeled using a 2 μM PKH26 solution (Sigma-Aldrich) and washed in Dulbecco's Phosphate-Buffered Saline (DPBS, Gibco 14190). Stained cells were counted with a hemocytometer and the suspension to be transplanted was diluted to an appropriate cell density of 50,000 cells per microliter.

Transplantation Procedure

Ten-week-old C57BL/6NCrl mice were used in this study. Animals were distributed in two groups of seven animals each, according to the culture conditions of the cells, Normoxia or Hypoxia. All animal procedures were accomplished following the guidelines of the local ethics committee of animal experimentation. Surgical procedures were performed under general anesthesia with 100 mg Ketamine and 5 mg diazepam per kilogram body weight. Additionally, the eye was topically anesthetized with 0.1% tetracaine and 0.4% oxybuprocaine. One drop of each 10% phenyleprine and 1% tropicamide were used to dilate the pupils. Following complete dilation, the anesthetized animal was placed in lateral recumbency under the SMZ-1 Nikon dissecting microscope and positioned with one hand holding mice. The mice fundus could be visualized with the application of a drop of 2.5% methylcellulose to the eye. The fundus observation served to evaluate the condition of the eye before injection and to compare with the postoperative condition of the retina. The needle with bevel up was advanced full thickness 1 mm posterior to the sclerocorneal limbus into the posterior chamber. At least 50% of the bevel was pushed through the choroid to produce a hole sufficiently large to insert the 33 gauge blunt needle (Hamilton Company, Reno, NV). The blunt needle tip was inserted through the choroidal puncture and advanced into the posterior chamber, avoiding trauma to ciliary body or lens. Subsequently, the needle shaft was aimed slightly nasally toward the posterior chamber and it was advanced toward the desired injection location in the posterior retina. A 10 μl syringe (Hamilton, Switzerland) with a 33-gauge needle attached to an ultramicropump (World Precision Instruments, Sarasota, FL) was used to inject 1.5 μl of cell suspension (75,000 cells) slowly, at a rate of 0.05 μl/second, into the SS. Immediately after injection, the fundus was examined and any animals with massive subretinal hemorrhage or vitreous hemorrhage were removed from the study. Injected animals developed a retinal detachment and small amount of bleeding in the same area of injection. Finally, a drop of antibiotic (0.3% ciprofloxacin) was administered on each eye and animals were kept on a 37°C pad until recovery from anesthesia.

Tissue Preparation

Animals were sacrificed by cervical dislocation after 24 hours and 1 and 4 weeks of transplantation. Eyes were enucleated and immediately fixed overnight at 4°C in freshly prepared 4% paraformaldehyde solution. Eyes were then washed in Phosphate-Buffered Saline (PBS) and transferred into 30% sucrose in PBS solution for at least 12 hours before inclusion in OCT and cryosectioning. Retinal sections (18 μm) were mounted in SuperFrost Ultra Plus slides (MENZEL-GLÄSER, Braunschweig, Germany) and stored at room temperature (RT) until further processing.

Immunohistochemical Analysis

Sections were blocked in PBS containing 10% goat serum and 0.1% Triton for 1 hour at RT and incubated with primary antibodies overnight at 4°C. Primary antibodies used are listed in supporting information (supporting information Table S3a). After incubation with primary antibodies, sections were washed with PBS containing 0.1% Triton and incubated with secondary antibodies for 1 hour at room temperature. After successive washing in PBS, nuclei were counterstained with 4′,6-diamino-2-phenylindole (dilactate; Invitrogen-Molecular Probes, Eugene, OR). Immunofluorescence was observed using a Leica DM 5500 microscope (Leica Microsystems, Wetzlar Germany) and a TCS SP5 confocal microscope (Leica Microsystems, Wetzlar Germany).


Effect of Hypoxia on Spontaneous Differentiation

We wanted to assess whether hypoxic conditions could induce retinal fate of mESC. EBs were generated from ESC using the hanging drop method followed by culture in suspension under either normoxic or hypoxic conditions (Fig. 1A). Time course (day 0, day 5, and day 7) of spontaneous differentiation of EBs was analyzed for expression of pluripotency markers along with other markers of retinal commitment. FACS analysis showed significant differences in the number of cells positive for pluripotency markers in hypoxic conditions when compared with the corresponding number under normoxia during the protocol (Fig. 1C). When the FACS analysis data in hypoxia were normalized against normoxia, data showed a significant decrease of Nanog+ (46% ± 18%), Oct4+ (50% ± 14%), and Sox2+ (42% ± 17%) cells after 7 days under hypoxic conditions (Fig. 1C, 1D).

Figure 1.

Effect of low oxygen tension during early spontaneous differentiation. (A): Schematic of the generation of embryoid bodies from mouse embryonic stem cell. (B): Immunofluorescence analysis against pluripotency markers; Oct4, Ssea-1, Sox2, and Nanog before differentiation. (C): Fluorescence-activated cell sorting analysis showing the loss of pluripotency markers is more efficient after 7 days of spontaneous differentiation in low oxygen when compared with normoxia. (D): Decrease in number of positive cells for pluripotency markers under hypoxia when data were normalized against the normoxia values. (E): qPCR analysis comparing the relative levels of expression of pluripotency marker genes after 7 days of spontaneous differentiation in normoxia (blue bar) and hypoxia (red bar). (F): qPCR analysis comparing the relative levels of expression of retinal-specific genes after 7 days of spontaneous differentiation in normoxia (blue bar) and hypoxia (red bar). *, p ≤ .05 or **, p ≤ .01 was considered statistically significant. Abbreviations: EB, embryoid body; ES, embryonic stem; LIF, leukemia inhibitory factor.

Besides, RT-PCR analysis demonstrated that retinal-specific genes, such as Pax6 and Crx, were significantly upregulated (4.9 ± 1.1 and 4.2 ± 0.6-fold, respectively) when EBs were cultured under hypoxia for 7 days compared to normoxia (Fig. 1F). The upregulation of Pax6 and Crx genes in cells cultured under hypoxia corresponded with the downregulation of pluripotency genes (Fig. 1E) and the loss of pluripotency markers at a protein level as demonstrated by FACS analysis at this time point (Fig. 1C, 1D). Other retinal genes such as Six3, a well-characterized eye field transcription factor [34], Chx10 [35], a neural retina marker, and Nrl [36], a photoreceptor precursor marker, showed a slight but not significant increase under hypoxia condition (Fig. 1F). These results have shown that hypoxia significantly decreases pluripotency and increases the expression of different retinal genes in spontaneous differentiation.

Directed Differentiation

The results observed with spontaneous differentiation led us to believe that hypoxic conditions could improve the differentiation of ESC toward a retinal fate, but obviously a more sophisticated protocol for retinal differentiation had to be applied. In order to demonstrate our hypothesis and corroborate whether we could obtain a higher yield of retinal cells, we have optimized a differentiation protocol that involves the culture of ESC under hypoxia (2% O2), and the use of small molecules previously described by different authors [17, 33]. Our three-step approach is a novelty in differentiation protocol. We included EB step where EBs were generated as a hanging drops because previously published data show [37] that this is an efficient and scalable system that allows uniform distribution of culture parameters such as oxygen tension due to the homogeneous size of the generated EBs. Briefly, ESCs characterized by the typical expression of Oct4, Nanog, Sox2, and SSEA-1 (Fig. 1B) were induced to differentiate into neural retina cells or Chx10-positive cells and more mature phenotypes of retinal commitment, such as Rhodopsin in the case of rod cells, in a sequential manner (supporting information Fig. S1A). To induce in vitro retinal differentiation, ESCs were cultured as hanging drops for 3 days in the presence of Dkk1 and Lefty A. Hanging drops containing the EBs were collected and allowed to differentiate in suspension for 7 more days in progenitors medium, after which, EBs were plated and allowed to further differentiate for 18 more days in adherent culture in retinal medium (Fig. 2B and supporting information Fig. S1B). From D3 to D5, FBS and Activin A were added in culture. Different growth factors such as DAPT, FGFs, Shh, RA, and taurine were added to the protocol of differentiation as described in Materials and Methods (Fig. 2B) to promote differentiation toward photoreceptors [17, 31, 33]. All results were compared to ESC differentiated under normoxic conditions (20% O2). During the 28 days of differentiation, the expression of various transcription factors associated to each major stage of retinogenesis was analyzed (supporting information Fig. S1C).

Figure 2.

Protocol of differentiation and loss of pluripotency. (A): Schematic diagram of the three-step differentiation protocol used to generate retinal cells. (B): Detail of the protocol of differentiation used to generate retinal cells. (C): qPCR analysis showing efficient loss of pluripotency markers. *, p ≤ .05 was considered statistically significant. Abbreviations: aFGF, acidic fibroblast growth factor; bFGF, basic fibroblast growth factor; FBS, fetal bovine serum.

Stemness, Loss of Pluripotency, and Eye Field Specification

Before the start of differentiation, undifferentiated ESC cultures were characterized. All cultures regardless of passage number showed normal morphology. ESCs grew forming colonies and expressed pluripotency markers as expected; all colonies were positive for Oct4, Ssea-1, Nanog, and Sox2 as determined by immunofluorescence (Fig. 1B) and the level of expression of these markers was determined by FACS each time to assure the purity and stemness of the culture prior to differentiation. Nanog (98.4%) expression was higher than Oct4 (94.5%) and Sox2 (83.5%) but no levels lower than 80% were found in any passage of ESC used (18–35 passages, data not shown). To direct mESC toward rostral fate, the EBs were treated with Dkk1, an antagonist of Wnt/β-catenin signaling, the nodal antagonist, LeftyA, and Activin A in presence of FBS. The loss of pluripotency, as a hallmark of ESC under differentiating pressure, was measured by RT-PCR. ESC rapidly lost expression of the pluripotency genes Oct4 and Nanog and even downregulated the neural induction marker Sox2 on day 5 (Fig. 2C). Sox2 levels of expression were higher than the levels of Oct4 and Nanog at all time points analyzed and its expression was upregulated on day 10, corresponding to differentiation efforts toward ectoderm (Fig. 2C). Indeed, we have observed a majority of cells expressing the anterior neural marker Otx2, and a greater number of Otx2-positive cells in hypoxia (Fig. 3B) which is consistent with a developmental in vivo study [38] (Fig. 3A).

Figure 3.

Hypoxia improves early retinogenesis. (A): Schematic of early neural retina (purple) on embryonic day 14, where the differencing cells are expressing the transcription factors Otx2 and Rax. (B): Comparative immunofluorescence analysis showing positive cells for anterior neural specification marker Otx2 under hypoxia and normoxia. (C): Comparative immunofluorescence analysis showing positive cells for the eye field transcription factor, Rax. (D): Quantification of Rax-positive cells derived under normoxic (blue bar) and hypoxic condition (red bar). (E–G): The time course of expression of the eye field transcription factors, Six3, Pax6, and Rax, respectively, by qPCR in normoxia (blue bar) and hypoxia (red bar). (H): Immunofluorescence analysis showing Rax-positive cells colocalized with Pax6 after 10 days of differentiation under hypoxia. *, p ≤ .05 was considered statistically significant. Abbreviation: DAPI, 4′,6-diamino-2-phenylindole.

Furthermore, ESC acquired the expression of transcription factors associated with eye field specification. RT-PCR analysis revealed that Six3, gene required for early eye field specification, as well as Pax6 and Rax expression was upregulated as early as on day 5 under hypoxia (Fig. 3E–3G, respectively). Eye field cells are characterized by coexpression of Pax6 and Rax [39, 40] (Fig. 3H). As a matter of fact, on day 10 of the differentiation protocol, high expression of Rax (Fig. 3G, 3H) coincided with the high expression of Pax6 (Fig. 3F, 3H). Not only the expression of these markers was upregulated earlier using our protocol but also the levels of expression of each individual gene at their peak of expression were higher under hypoxia when compared with normoxia. Additionally, by day 16, hypoxia significantly increased the number of Rax+ cells (91% of total cells vs. 80% in normoxia; Fig. 3C, 3D). On day 10, the fold induction of Six3 in hypoxia was six times higher than in normoxia (8.2 ± 2.9 against 2.1 ± 0.91-fold change; Fig. 3E). RT-PCR analysis on day 5 revealed a difference of 14-fold change of Pax6 gene expression under hypoxia (16.4 ± 3.8 against 2.4 ± 0.2; Fig. 3G) and the expression of Rax was induced four times (6.6 ± 1.5 against 2.2 ± 0.6; Fig. 3G) above the level observed in normoxia on day 10. Our results showed that hypoxic conditions significantly increased the expression of the eye field genes.

Retinal Commitment

To induce in vitro differentiation, ESCs were cultured in suspension as EBs for 10 days in progenitors medium (Fig. 2B) after which EBs were plated and allowed to differentiate for another 18 days in adherent culture in retinal medium. The next stage of in vivo retinal specification occurs when optic vesicles from the paired eye fields are formed. At this stage, all cells express the transcription factors Mitf and Pax6 and give rise to either neural retina or RPE [41]. Furthermore, the fact that Mitf-positive cells destined to become neural retina derive from the subset of Rax-positive cells [35, 42] was confirmed with coexpression of these two markers in culture (Fig. 4A). In fact, RT-PCR analysis revealed that Mitf expression was upregulated since day 5 coinciding with the peak of expression of Pax6 (Fig. 3F). During development, Mitf is downregulated in response to the onset of Chx10, a neural retina-specific gene [35]. The dynamic expression of Mitf and Chx10 was examined over time and it was determined that neural retina phenotype was acquired as early as day 16 (Fig. 4E, 4F). Also, our cells at day 16 coexpressed Mitf and Rax (Fig. 4A), revealing RPE commitment [43], while the clusters (Fig. 4B–4D) formed by day 16 of Chx10-positive cells maintaining expression of Pax6 revealed their commitment to retinal fate (Fig. 4C, 4D). This corresponded with the peak of expression of Chx10 as checked by RT-PCR (Fig. 4F).

Figure 4.

Hypoxia improves neural retina phenotype acquisition. (A): Immunocytochemical analysis showing that all Rax-positive cells were also positive for Mitf by day 16 under hypoxia. (B): Photomicrograph showing clusters formed by day 16 of differentiation. (C): Immunocytochemical analysis showed Chx10+/Pax6+ cells deriving radially away from the clusters by day 16 of differentiation. (D): Higher magnification showing Chx10-positive cells were also positive for Pax6. (E, F): The time course of expression of the neural retina transcription factors, Mitf and Chx10, respectively, by qPCR in normoxia (blue bar) and hypoxia (red bar). *, p ≤ .05 was considered statistically significant.

Prolonged culture of Chx10+ and Pax6+ cells allowed for further maturation of these cells toward a photoreceptor phenotype. A crucial gene for rod specification is Nrl [44], highly induced since day 16 under hypoxia (Fig. 5E) coinciding with high expression of Crx (Fig. 5C) as observed by RT-PCR analysis. These data were confirmed by immunocytochemistry analysis for these two markers at the end of the protocol where a high number of cells have shown to be positive for Crx (Fig. 5A, 5B) and Nrl (Fig. 5D). The importance of this finding is because during development, Nrl interacts with Crx to induce the expression of rod-specific genes such as Rhodopsin [45]. As it has been mentioned, the primitive cone and rod photoreceptor-specific transcription factor Crx was upregulated since day 16 as well (Fig. 5C), and the immunoreactive cells against Crx antibody accounted for 75% of the total cells (75.1% ± 1.9% in hypoxia vs. 64.4% ± 2.3% in normoxia; Fig. 5B) at the end of the protocol.

Figure 5.

Hypoxia improves late retinogenesis. (A): Immunocytochemical analysis showing Crx+ cells in hypoxic culture by day 28 of differentiation. (B): Quantification of the percentage of Crx+ cells determined by three independent experiments under both conditions of differentiation. (C): Comparative qPCR analysis of Crx gene expression over the 28 days of differentiation under normoxia (blue bars) and hypoxia (red bars). (D): Immunocytochemical analysis showing Nrl+ cells in hypoxic culture by day 28 of differentiation. (E): Comparative qPCR analysis of Nrl gene expression during differentiation in normoxic (blue bar) and hypoxic (red bar) conditions. (F): Quantification of the percentage of Opsin+ cells in normoxic (blue bar) and hypoxic (red bar) conditions. (G): Comparative immunofluorescence analysis showing positive cells for cone-specific marker Opsin-S after 28 days of differentiation under normoxia or hypoxia. *, p ≤ .05 was considered statistically significant.

Moreover, it seems that our differentiation protocol applying normoxia did not promote high yield of cone cells because from the Crx+ population, only approximately 8% (8.1% ± 5.5%) of cells was immunoreactive for the cone photoreceptor-specific protein Opsin-S (Fig. 5F, 5G). Conversely, the percentage of Opsin-S+ cells was significantly increased to 32% of total cells under hypoxic conditions (31.8% ± 11.6%; Fig. 5F, 5G).

Consistently, differentiation of ESC in hypoxic conditions significantly increased their retinal commitment toward rod photoreceptors, as determined by coexpression of Rhodopsin and Recoverin at the end of the protocol (Fig. 6A, 6B, 6D), when approximately 53% of total cells showed double positive staining for these markers (53% in hypoxia against 30% in normoxia; 52.9 ± 1.5 against 29.4 ± 3.5; Fig. 6E). The induction of Rhodopsin by hypoxia was confirmed by RT-PCR. The analysis showed significant increase in fold change of Rhodopsin since day 16 in hypoxia (28.5 ± 3.6 vs. 11.1 ± 2.9) compared to normoxia (Fig. 6C). Recoverin expression did not localize with Tuj1 indicating the presence of immature neurons and mature photoreceptors in our culture (Fig. 6G). This data reveal that the generated photoreceptors are not immature neurons but mature retinal cells. Furthermore, using hypoxic conditions, morphology typical of rods was observed in isolated cases, in sections of the culture where cell density allowed for the outgrowth of structures that suggest the formation of outer segments in vitro (Fig. 6B and data not shown).

Figure 6.

Hypoxia increase the yield of Rhodopsin-positive cells. (A): Immunocytochemical analysis showing the coexpression of rod photoreceptor markers Rhodopsin and Recoverin after 28 days in hypoxic culture conditions. (B): A representative Rhodopsin+ cell in hypoxic conditions after 28 days in culture. (C): Comparative qPCR analysis of Rhodopsin gene expression during the differentiation in normoxic (blue bars) and hypoxic conditions (red bars). (D): Comparative immunofluorescence analysis of expression of Rhodopsin during the differentiation in hypoxic and normoxic conditions. (E): Quantification of the percentage of Rhodopsin+ cells in normoxia (blue bar) and hypoxia (red bar). (F): Schematic representation of obtained retinal cells. (G): Immunofluorescence analysis of other cell types present after 28 days of differentiation Tuj1+/Recoverin cells. (H): Comparative immunofluorescence analysis showing positive cells for the proliferation marker; Ki67 after 28 days of differentiation under hypoxia and normoxia. (I): Quantification of the percentage of Ki67+ cells after 28 days of differentiation in normoxia (blue bar) and hypoxia (red bar). *, p ≤ .05 was considered statistically significant. Abbreviation: DAPI, 4′,6-diamino-2-phenylindole.

To assess the proliferation capacity of differentiated cells, we stained the cells after 28 days of differentiation against the operational marker, Ki67 [46]. We have found that after 28 days of differentiation under hypoxic conditions, there was a significant decrease in the number of proliferating cells (14.8% ± 1.8%) against the amount of proliferating cells differentiated under normoxia over the same time (22.8 ± 1.7, Fig. 6H, 6I). All cell phenotypes generated at the end of our protocol are depicted in Figure 6F. To assess the possible mechanisms underlying the increased differentiation capacity of mESC in hypoxia, we observed that, a higher proliferation rate was not related with an increased yield of Rho+ cells. Hence, we did not observe any coexpression of Ki67 and Rho after 28 days of differentiation, neither under hypoxic nor normoxic conditions (supporting information Fig. S4A). Furthermore, this phenomenon was also observed in other cellular models, RPE-1 cells showed a significant decrease in their proliferation capacity after 3 weeks of hypoxic culture, as determined by the percentage of Ki67+ cells in culture and compared to the number found in normoxic conditions (supporting information Fig. S4B). To assess the safety of low oxygen tension in our cultures, we performed cytotoxicity and apoptosis (presented by Caspase 3/7 activity) tests and found that, after 1 week of hypoxic culture, RPE-1 cells showed a significant decrease in cytotoxity and Caspase 3/7 activity compared to normoxia (supporting information Fig. S4C, S4D).

To correlate hypoxia and retinal development, we followed the expression of Vegfa and Cdkn1a. The expression of these two genes is very well-known to be induced under hypoxia and recently they have been directly related with retinal hypoxia [47, 48]. We observed that hypoxia significantly increased the fold induction of Vegfa (6.1 ± 1.0 in hypoxia vs. 4.6 ± 0.7 in normoxia; supporting information Fig. S4E) and Cdkn1a (11.7 ± 1.0 in hypoxia vs. 5.5 ± 1.1 in normoxia; supporting information Fig. S4F) at day 28 of differentiation, revealing possible mechanisms through which hypoxia promotes retinal development. The exposure of cells to hypoxia leads to the activation and stabilization of hypoxia-inducible factor Hif1α as determined by FACS analysis. The percentage of Hif1α+ cells in RPE-1 cells was higher after 3 weeks of hypoxia when compared with normoxia culture (69.6 ± 4.18 vs. 55.6 ± 10.4; supporting information Fig. S4G) and significantly higher in D3-mESC cells after 1 week of hypoxia in the presence of LIF (79.9 ± 11.5 vs. 49.1 ± 10.7; supporting information Fig. S4H).

Subretinal Injection

To evaluate whether injected cells in suspension, generated using our protocol under hypoxia, could survive and integrate in vivo, C57BL6/NCrl mice received unilateral subretinal injections of 75,000 retinal cells differentiated until day 20 or medium alone (sham). The uninjected eye served as internal control for each animal.

The PKH26 staining applied prior to transplantation was used to identify surviving retinal progenitors after 24 hours and 1 week of the injection (Fig. 7A, 7C, 7E, 7F, 7H, 7J, 7K). The location of the main graft was subretinal in seven eyes with a large cluster of transplanted cells localized between the host photoreceptors and the RPE layer. Immunohistochemical analysis demonstrated dispersed transplanted cells positive for Rhodopsin (Fig. 7A, 7B, 7E, 7K) and Recoverin (Fig. 7F, 7G, 7J) across the retina, singly or in small clusters. A large number of donor cells injected in subretinal locations migrated and integrated in host retina. Transplanted cells immunopositive for photoreceptor marker Rhodopsin and Recoverin were found in outer nuclear layer and inner nuclear layer in the mice (supporting information Fig. S7). A much smaller fraction of transplanted cells was Opsin-S positive revealing the presence of cones (supporting information Fig. S6). There was no evidence of uncontrolled growth or tumor formation at any time, suggesting that donor cell proliferation might be regulated or balanced by cell death.

Figure 7.

Transplantation of photoreceptors derived from embryonic stem cell under hypoxic conditions. (A): Anatomy of the injection site and efficiency of differentiation 24 hours after transplantation. (B–E): Higher magnification of inset in (A). Coexpression of transplanted cells (red, PKH26), Rhodopsin+ (green) and DAPI (blue). (F): The efficiency of differentiation toward photoreceptors 1 week after transplantation. (G–J): Higher magnification of inset in (F) showing transplanted cells (red, PKH26) colocalizing with Recoverin (Rcvn, green) and DAPI (blue). (K): Colocalization of PKH26 and Rhodopsin in ONL 1 week after subretinal injection. Abbreviations: DAPI, 4′,6-diamino-2-phenylindole; INL, inner nuclear layer; ONL, outer nuclear layer; SS, subretinal space.


ESC offer an excellent in vitro tool to recapitulate mechanisms activated during early development. The efficient differentiation of retinal cells from ESC is a major challenge for the development of successful cell therapy, which can be applied in different retinal dystrophies such as RP and age-related macular disease (AMD). Although the early stages of development occur in a hypoxic environment, little is known about how low O2 levels modulate the pluripotency and differentiation capacity of ESC. Our data demonstrate that mESC can be efficiently directed to retinal progenitors and other mature phenotypes, such as photoreceptors, using a combination of small molecules and lowering O2 tension can enhance this efficiency. Different protocols of generation of retinal phenotypes have been published and all of them have implied important advances in the field [17, 22]. Our main issue continues to be the accomplishment of a high yield of specific populations and the modeling of retinogenesis in vitro.

In the first phase of our protocol, we used the combination of Dkk1, Activin A, and LeftyA to direct mESC toward rostral neural progenitors, applying an approach used in a previous study [33]. This strategy is widely used to generate rostral neural progenitors [16, 49–51]. Indeed, we observed high percentage of Rax+ and Otx2+, rostral neural progenitors, and markers for early eye field. These two markers together with Pax6, Six3, Six6, and Lhx2 play an important role in the establishment of anterior neuroectodermal region which maintains high capacity for generation of future retinal progenitors [39, 52, 53]. Significant upregulation of some of these markers (Six3, Pax6, and Rax) was observed as early as by day 5 in culture. Large increase of the eye field markers coincided with rapid decrease of main pluripotency markers indicating high differentiation potential of our protocol. Our results have shown that lowering the O2 tension near the physiological level is a more effective parameter for retinal differentiation, especially for photoreceptor precursors. It seems that the effects of this parameter are expressed very early in differentiation significantly increasing the expression of Six3, Pax6, and Rax (Fig. 3E–3G).

The next phase in retinal specification in vivo occurs with the formation of the optic vesicle, determined mainly by the expression of Mitf and Pax6 giving rise to multiple cell types of the functional retina. Cells coexpressing Pax6 and Chx10 give rise to neural retina only. Experiments using different vertebrates indicated that SHH and FGF signaling play a critical role in future specification of retinal cells [54–56]. Once neural retina progenitor phenotype has been acquired, it is necessary for further maturation of these cells. As shown by others, Notch signaling pathway needs to be blocked at this point to allow for an increase in the Crx+ cells [17]. For these reasons, we supplemented the medium with the γ-secretase inhibitor DAPT from day 10. RA and taurine were added to obtain mature photoreceptors [17, 31]. Hence, further retinal specification included DAPT, FGFs, Shh, RA, and Taurine. These conditions together with low O2 tension significantly increase the population of cells coexpressing Pax6 and Chx10 as well as Crx+ cells compared to normoxia. High percentage of derived photoreceptors, reflected by coexpression of Rhodopsin and Recoverin for rods, and Opsin-S for cones, revealed that low O2 tension promoted a photoreceptor fate of mESC. With regard to efficiency, the induction of rods, namely the yield of Rhodopsin+ cells, was higher when compared with other protocols using murine ESC [17, 31] or induced pluripotent stem cells (iPS) [15, 57, 58]. The results in this study are consistent with the recently published study applying hypoxic condition [59], describing an increased yield of Pax6- and Chx10-positive cells. Here, we further define the characterization of generated cells achieving a higher yield of mature retinal phenotypes as well as in vivo study. Our results have shown not only that the population of retinal cells can be increased under hypoxia in a way that mimics normal retinal development but also we demonstrate for the first time that early rostral differentiation (Otx2), early eye field acquisition (Six3, Rax, and Pax6), and mature retinal phenotypes (Crx, Opsin-S, and Rhodopsin) are increased in yield under hypoxia. Furthermore, hypoxic condition seems to improve the timing of retinogenesis, as it has been observed by RT-PCR analysis. Eye field transcription factors are highly expressed as early as by day 5 and the important suppression of Mitf by upregulation of Chx10 occurred by day 16 instead of day 28 only when the cells differentiated under hypoxia. This allowed for a bigger population of neural retina progenitors earlier in the differentiation protocol that could mature into cells expressing photoreceptor markers, such as, Crx, Nrl, and Rhodopsin, all three populations highly increased at the end of our protocol.

Our study went further showing efficient in vivo evaluation of generated retinal precursors. Previous studies have shown that photoreceptors taken from young animals efficiently incorporate in adult retina when transplanted in the SS [25, 60]. We also successfully grafted in vitro generated retinal cells in the adult mouse retina, which resulted in cell survival. Interestingly, high percentage of transplanted cells expressed rod-specific markers, such as Rhodopsin and Recoverin, although without known possible implication of local environment on further differentiation. It seems that specific retinal niche was preferable for mature differentiation of retinal progenitors. This data demonstrate the viability of the cells and the robustness of our protocol. However, to fully validate the protocol, further in vivo functional analyses have to be performed.

Different studies on pluripotent stem cells have demonstrated improved differentiation when different hypoxic conditions were applied [59, 61]. The exact mechanism of hypoxia on retinal differentiation still remains to be elucidated. The primary transcriptional regulators of cellular hypoxic adaptation in mammals are hypoxia-induced factors (HIFs). Hypoxic preconditioning was shown to stabilize HIF-1α in the retina, further inducing the expression of target genes with neuroprotective properties like vascular endothelial growth factor (Vegfa) and erythropoietin (Epo) suggesting a link between HIF-1α-driven gene expression and neuroprotection [62, 63]. Little is known about the molecular effects of hypoxia on retinal differentiation. The published reports mainly correlate the expression of individual genes and hypoxia in different retinal functions [47]. For example, hypoxia increases vascular endothelial growth factor (Vegfa) expression in the retina [59, 62]. The identification of this gene together with p21 (Cdkn1a) is strongly suggestive of their role in general retinal neuroprotection [64, 65]. During our differentiation protocol under hypoxic conditions, both genes were significantly upregulated when compared with normoxic conditions. This data suggest that hypoxia, through activation of HIF-1α, decreasing apoptosis, and cytotoxicity, has influenced retinal differentiation [66, 67]. Although, the source of Vegfa could be RPE cells observed in retinal progenitors generated in hypoxic conditions [68] (data not shown), further investigation has to be performed to elucidate the origin of neuroprotective processes. Moreover, a significant decrease of proliferating cells at the end of our protocol in hypoxic conditions suggests that hypoxia favors postmitotic cells, increasing therefore the number of photoreceptors (Fig. 6F, 6I). It is known that p21, increased in hypoxia (supporting information Fig. 4F), not only provokes cell cycle arrest, necessary to start the differentiation process, but also can repress apoptosis [69] (supporting information Fig. S4D). Therefore, we propose that hypoxia promotes retinal differentiation through activation of p21. Higher yield of early eye field markers in hypoxic condition suggests that hypoxia is preferable for EB formation and early differentiation. This data corroborate with earlier findings where an improved differentiation of human ESCs was observed in hypoxic conditions [70, 71]. We also observed a more compact structure of EBs in hypoxia (data not shown), which could have a consequence on further differentiation. For this to be confirmed, further detailed studies have to be performed.

Conclusion and Future Perspectives

We believe the application of a new modified protocol for differentiation of mESC reported here supports the hypothesis that hypoxia is necessary to induce efficient differentiation of ESC toward a higher yield of retinal phenotypes. The timing of retinogenesis is also improved in hypoxic conditions, by decreasing the time to acquire an eye field phenotype and achieve mature population of photoreceptors in vitro. Purification of these specific retinal cells can allow us to define the conditions to expand a homogeneous population that will be further differentiated into fully mature photoreceptor cells. Further experimentation is required to elucidate the precise mechanism or mechanisms by which hypoxia exerts its effect on retinal differentiation. In summary, the novel findings of the work reported here are: (a) the most efficient protocol so far, for the differentiation of any kind of stem cells (mouse, human, or induced-pluripotent cells), toward rod photoreceptor cells (53% ± 1.5%). (b) The modeling of retinogenesis has been accomplished for the first time with mESC only under hypoxic conditions. (c) Photoreceptor precursors from mESC differentiate toward Rhodopsin/Recoverin double positive cells after transplantation in the retina, and a complete lack of tumor formation, demonstrate the importance of an efficient differentiation process and the loss of pluripotency of the transplanted cells. We believe that our findings provide the technical framework necessary for a highly efficient differentiation of mESC toward photoreceptors, which is important for advances in cell therapy and regenerative medicine.


This work was supported by funds for research from Junta de Andalucia PI-0113-2010 (S.E.) and “Miguel Servet” contract of Instituto de Salud Carlos III of Spanish Ministry of Science and Innovation (S.E.). Thanks to Dr. Anand Swaroop for the Nrl antibody.


The authors indicate no potential conflicts of interest.