Differentiation of human pluripotent stem cells to retinal pigmented epithelium in defined conditions using purified extracellular matrix proteins


  • Teisha J. Rowland,

    1. Center for Stem Cell Biology and Engineering, University of California, Santa Barbara, CA, USA
    2. Department of Molecular, Cellular and Developmental Biology, University of California, Santa Barbara, CA, USA
    3. Neuroscience Research Institute, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
  • Alison J. Blaschke,

    1. Center for Stem Cell Biology and Engineering, University of California, Santa Barbara, CA, USA
    2. Department of Molecular, Cellular and Developmental Biology, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
  • David E. Buchholz,

    1. Center for Stem Cell Biology and Engineering, University of California, Santa Barbara, CA, USA
    2. Department of Molecular, Cellular and Developmental Biology, University of California, Santa Barbara, CA, USA
    3. Neuroscience Research Institute, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
  • Sherry T. Hikita,

    1. Center for Stem Cell Biology and Engineering, University of California, Santa Barbara, CA, USA
    2. Department of Molecular, Cellular and Developmental Biology, University of California, Santa Barbara, CA, USA
    3. Neuroscience Research Institute, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
  • Lincoln V. Johnson,

    1. Center for Stem Cell Biology and Engineering, University of California, Santa Barbara, CA, USA
    2. Neuroscience Research Institute, University of California, Santa Barbara, CA, USA
    3. Center for the Study of Macular Degeneration, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
  • Dennis O. Clegg

    Corresponding author
    1. Center for Stem Cell Biology and Engineering, University of California, Santa Barbara, CA, USA
    2. Department of Molecular, Cellular and Developmental Biology, University of California, Santa Barbara, CA, USA
    3. Center for the Study of Macular Degeneration, University of California, Santa Barbara, CA, USA
    • Neuroscience Research Institute, University of California, Santa Barbara, CA, USA
    Search for more papers by this author

Correspondence to: D. O. Clegg, Neuroscience Research Institute, University of California, Santa Barbara, CA 93106, USA.

E-mail: clegg@lifesci.ucsb.edu


A potential application of human embryonic stem cells (hESCs) and induced pluripotent stem cells (iPSCs) is the generation of retinal pigmented epithelium (RPE) to treat age-related macular degeneration (AMD), a common but incurable retinal disease. RPE cells derived from hESCs (hESC-RPEs) and iPSCs (iPSC-RPEs) express essential RPE markers and can rescue visual function in animal models. However, standard differentiation protocols yield RPE cells at low frequency, especially from iPSC lines, and the common use of Matrigel and xenogeneic feeder cells is not compatible with clinical applications. The extracellular matrix (ECM) can affect differentiation, and therefore changes in ECM composition may improve the frequency of stem cell-RPE differentiation. We selected several purified ECM proteins and substrates, based on the in vivo RPE ECM environment, and tested their ability to support iPSC-RPE differentiation and maintenance. iPSCs differentiated on nearly all tested substrates developed pigmented regions, with Matrigel and mouse laminin-111 supporting the highest pigmentation frequencies. Although iPSC-RPEs cultured on the majority of the tested substrates expressed key RPE genes, only six substrates supported development of confluent monolayers with normal RPE morphology, including Matrigel and mouse laminin-111. iPSCs differentiated on mouse laminin-111 produced iPSC-RPEs expressing RPE proteins, and hESCs differentiated on mouse laminin-111 resulted in high yields of functional hESC-RPEs. This identification of key ECM proteins may assist with future scaffold designs and provide peptide sequences for use in synthetic, xeno-free, GMP-compliant generation of RPE from human pluripotent stem cells relevant to clinical translation. Copyright © 2012 John Wiley & Sons, Ltd.

1 Introduction

Age-related macular degeneration (AMD) is an incurable retinal disease that is the leading cause of blindness in people over the age of 60, affecting 30–50 million people globally (Gehrs et al., 2006). In AMD, the dysfunction and death of the retinal pigmented epithelium (RPE) causes overlying photoreceptors in the macula of the eye to die, leading to significant, progressive vision loss (Del Priore et al., 2002; Dunaief et al., 2002). Promising strategies for treating AMD involve the use of human embryonic stem cells (hESCs) (Thomson et al., 1998) or human induced pluripotent stem cells (iPSCs) (Takahashi et al., 2007; Yu et al., 2007). Both stem cell types are virtually infinite in supply, and iPSCs in particular hold promise as a source of patient-specific pluripotent cells derived from adult tissue. RPE cells can be derived from hESCs and iPSCs, and these hESC-RPEs and iPSC-RPEs are highly similar to human fetal RPEs: they express key RPE markers, display RPE functionality, as demonstrated by phagocytosis assays and ion transport (Buchholz et al., 2009; Hirami et al., 2009; Idelson et al., 2009; Klimanskaya et al., 2004; Kokkinaki et al., 2011), and have the ability to rescue visual function through subretinal transplantation in the Royal College of Surgeons (RCS) dystrophic rat (Carr et al., 2009; Hirami et al., 2009; Idelson et al., 2009; Lu et al., 2009; Lund et al., 2006; Vugler et al., 2008). Consequently, iPSC-RPEs and hESC-RPEs are potentially valuable tools to treat AMD and other retinal diseases.

While both hESCs and iPSCs can form RPE cells, current differentiation protocols are inefficient and not optimized for clinical translation. The most established protocol involves the removal of growth factors [primarily basic fibroblast growth factor (bFGF)] from the culture medium after the cells have reached confluence (ca. 7–10 days post-seeding) to induce differentiation (for a recent review, see Rowland et al., 2012). Approximately 1–8 weeks after growth factor removal, initial pigmentation is observed. In a small fraction of the hESCs and iPSCs, small pigmented regions expand over time and after 5–13 weeks they are large enough to be mechanically dissected and used to seed cultures enriched in RPE-like cells that can be expanded by multiple passaging (Rowland et al., 2012). Most protocols for this process make use of animal-derived products, including feeder cells (primarily mouse fibroblasts), gelatin and media containing fetal bovine serum (FBS) (Buchholz et al., 2009; Rowland et al., 2012).

It has previously been shown that the extracellular matrix (ECM) may affect the differentiation of human embryonic stem cells (Ma et al., 2008), and we theorized that the ECM may play a significant role in directing the differentiation of hESCs and iPSCs into RPE. The effects of different purified ECM proteins on iPSC or hESC differentiation to RPE have not been previously reported, and the derivation and maintenance of iPSC-RPEs and hESC-RPEs on a purified protein substrate may be important for clinical translation. In this study, we investigated the effects of a variety of ECM proteins found in the ECM environment surrounding RPE in vivo (based on known RPE integrin expression profiles and the ECM composition of Bruch's membrane) on RPE differentiation and maintenance in fully-defined, feeder-free culture systems. The findings presented here should assist in the development of scaffold designs for future clinical transplantation therapies.

2 Materials and methods

2.1 Cell lines and culture

iPSC line iPS(IMR90)-3 (Yu et al., 2007) (kind gift of J. Thomson) and hESC line H9 (WiCell Research Institute) were maintained on mouse embryonic fibroblasts (MEFs) isolated from 13.5 day-old mouse embryos or human foreskin fibroblasts (hs27s) (American Type Culture), respectively, in six-well tissue culture plastic plates (Thermo Fisher Scientific). MEFs and hs27s were both treated with mitomycin C (10 µg/ml; Sigma-Aldrich). iPSCs and hESCs were maintained with hESC maintenance medium [DMEM/F12/GlutaMAX I media (Invitrogen) supplemented with knockout serum replacement (20%; nvitrogen), non-essential amino acids (NEAA) (1×; Invitrogen), and β-mercaptoethanol (0.1 mm; Invitrogen)]. This medium was supplemented with basic fibroblast growth factor (bFGF; 4 ng/ml; Peprotech) for hESCs, while iPSCs were instead cultured using zebrafish bFGF (zbFGF; 100 ng/ml; kind gift of J. Thomson) (Yu et al., 2007). For substrate experiments, plates were coated overnight at 4°C with one of the following ECM proteins or substrates, at concentrations previously shown to be supportive of hESCs, iPSCs, neuronal or epithelial cells in amounts that saturate the substratum (Aisenbrey et al., 2006; Braam et al., 2008; Rowland et al., 2010; Xu et al., 2001) or according to manufacturer's instructions: gelatin (0.1%; Sigma-Aldrich), hESC-Qualified BD Matrigel (BD Biosciences), mouse laminin-111 (50 µg/ml; Invitrogen), rat laminin-332 (5 µg/ml; Chemicon), human laminin-332 (5 µg/ml; Biodesign International), human fibronectin (40 µg/ml; BD Biosciences), human vitronectin (10 µg/ml), rat collagen I (50 µg/ml; BD Biosciences), human collagen I (10 µg/ml; BD Biosciences), mouse collagen IV (10 µg/ml; BD Biosciences), human collagen IV (10 µg/ml; BD Biosciences) and bovine elastin (10 µg/ml; Sigma-Aldrich). Coated wells were rinsed with phosphate-buffered saline (PBS) or water, as recommended. Vitronectin was purified from human plasma as previously described (Rowland et al., 2010; Yatohgo et al., 1988).

Human fetal RPEs (hfRPEs) were isolated from fetal eyes of a random donor at 21 weeks of gestation, independently procured by Advanced Bioscience Resources (Alameda, CA, USA) and cultured on plates coated with gelatin in RPE medium (Maminishkis et al., 2006): MEM-α modification (Sigma-Aldrich) supplemented with FBS (5%; 15% for the first 3 days after seeding; HyClone), N1 (1×; Sigma-Aldrich), NEAA (1×), GlutaMAX-I (2 mm; Invitrogen), taurine (250 µg/ml; Sigma-Aldrich), triiodothyronin (0.013 µg/l; Sigma-Aldrich) and hydrocortisone (20 ng/ml; Sigma-Aldrich). For experiments, isolated hfRPEs were used at passage 2 or 3. For decellularization, hfRPEs were cultured in 24-well plates for at least 4 weeks, washed with PBS, treated with 1% Triton X-100 with 0.1% ammonium hydroxide (Acros) in PBS for 30 min with occasional vigorous agitation, washed with PBS four times, and stored in PBS with protease inhibitor cocktail set V (1×; Calbiochem) at 4°C (Cho et al., 2005; El Kassaby et al., 2003).

2.2 Retinal pigment differentiation

iPSCs and hESCs were passaged and seeded onto MEFs or the ECM proteins or substrates to be tested and cultured for 7 days, as described above, at which point zbFGF or bFGF was removed from the medium to induce spontaneous differentiation (Buchholz et al., 2009). This spontaneous differentiation method has been previously suggested to be more efficient at generating RPE than protocols that generate embryoid bodies (Vugler et al., 2008; our unpublished observations). Cells were cultured until pigmented spots, as detected by visual microscopic inspection, had expanded enough to allow for pigmented cell enrichment (ca. 73–87 days post-bFGF or zbFGF removal). Pigmented areas were dissected and dissociated into a single cell suspension using trypsin/EDTA (0.05%; Invitrogen), strained through a 30 µm strainer (BD Falcon) to remove clumps, and seeded at a density of 2.5 × 105 cells/cm2 onto plates coated with 0.1% gelatin with RPE medium, or coated with substrates to be tested with serum-free RPE medium (Gamm et al., 2008): DMEM (high-glucose; Invitrogen), F12 (30%; Invitrogen), and B27 (2%; Gibco, 0080085-SA). Derived RPEs were passaged every 4 weeks using trypsin/EDTA (0.05%). To assess the substrate effects on morphology, surface area coverage, pigmentation, and RPE gene marker expression of iPSCs and iPSC-RPEs after 5 or 4 weeks of culture, respectively, 24-well tissue culture plastic plates were used and at least three separate culture replicates performed. Total area of pigment of differentiating iPSC cultures, based on image capture of every pigmented spot, and area of densely pigmented regions of iPSC-RPE cultures, based on images taken of each well in the same location, were quantified using ImageJ image analysis software (NIH) through threshold normalization. Surface area coverage for all assays was estimated based on visual inspection. In longer-term differentiation studies using hESCs or comparing pigmentation frequency of iPSCs to hESCs, six-well plates were used and 12 internal culture replicates performed. Statistical analysis was performed using two-tailed Student's t-test with unequal variance for these and all subsequent assays.

2.3 Quantitative real-time polymerase chain reaction

Total RNA was isolated from iPS(IMR90)-3 at passage 2 (three separate culture replicates) using Qiagen RNeasy Plus Mini Kit (Qiagen). 250 ng–1 µg of RNA was used to synthesize cDNA, using the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). Primer pairs were designed to create a 75–200 base pair (bp) product (Beacon Design 4.0; Premier Bio-soft International). Primer specificity was confirmed using gel electrophoresis, melting temperature analysis and direct sequencing (Iowa State DNA Facility) (Buchholz et al., 2009). Quantitative real-time polymerase chain reaction (RT–PCR) was performed using a Bio-Rad MyIQ Single Color RT–PCR Detection System, using the SYBR Green method. 20 µl reactions were run in triplicate on a 96-well plate, with 0.5 µg of the cDNA synthesis reaction used per plate. Data were normalized to the geometric mean of ‘housekeeping’ genes: glyceraldehyde phosphate dehydrogenase (GAPDH), hydroxymethylbilane synthase (HMBS), glucose phosphate isomerase (GPI) (Radeke et al., 2007). Primers against both the 3′ and 5′ ends of GPI were used to monitor RNA integrity. Primer sequences used were the same as previously reported (Buchholz et al., 2009).

2.4 Immunocytochemistry

iPSC-RPEs and hESC-RPEs (at passage 2) were seeded onto chambered slides (Thermo-Fisher Scientific) coated with Matrigel or mouse laminin-111 and cultured in serum-free RPE medium. The slides were washed with PBS, fixed with 4% paraformaldehyde in 0.1 m sodium cacodylate buffer, pH 7.4, for 15 min at 4°C. The slides were washed with PBS, blocked with bovine serum albumin (BSA; 1%; Sigma-Aldrich) with 0.1% IGEPAL (Sigma-Aldrich) and 1% goat or donkey serum (Sigma-Aldrich) in PBS for 1 h at 4°C. The slides were then incubated with primary antibodies overnight at 4°C: monoclonal mouse anti-pigment epithelium-derived factor (PEDF; 10 µg/ml; LS-C17230, Lifespan Biosciences), polyclonal rabbit anti-microphthalmia-associated transcription factor (MITF; 5 µg/ml; ab59232, Abcam), monoclonal mouse anti-orthodenticle homeobox 2 (OTX2; 20 µg/ml; clone 246826, R&D Systems), monoclonal mouse anti-melanosome (Pmel17; 0.2 µg/ml; clone HMB45, Dako), monoclonal mouse anti-zona occludens 1 (ZO-1; 15 µg/ml; clone ZO1-1A12, Invitrogen), monoclonal mouse anti-bestrophin (2 µg/ml; MAB5466, Millipore), polyclonal rabbit anti-αv integrin (4 µg/ml; clone [Q-20]-R, Santa Cruz Biotechnology [SCBT]). The slides were incubated with the appropriate AlexaFluor-conjugated secondary antibodies (10 µg/ml; Invitrogen) for 30 min at 4°C, stained with Hoechst 3342 (2 µg/ml; Invitrogen) for 5 min, mounted using Fluoro-Gel with Tris buffer (Electron Microscopy Sciences) and imaged using a BX51 fluorescence microscope (Olympus). To determine the percentage of cells positive for Pmel17 expression, three images at ×40 magnification were taken of each well, with three internal replicates, and the number of cells stained for Pmel17 relative to the total number of cells stained with Hoechst was quantified.

2.5 ROS phagocytosis assay

Rod outer segment (ROS) phagocytosis assays were performed as previously described (Lin and Clegg, 1998). ROS were purified from retinal extracts prepared from bovine eyes obtained fresh at a local slaughterhouse, and then fluorescently labelled using FluoReporter FITC Protein Labeling Kit (Invitrogen). RPE cells to be tested for phagocytotic activity (used at passage 0) were seeded in a 96-well plate at 5 × 104 cells/well and cultured for 4 weeks. hfRPE and ARPE-19 cells were used as positive controls, and hs27 fibroblasts were used as a negative control. RPE cells were then challenged with 1 × 106 FITC-labelled ROS/well, with or without monoclonal mouse anti-αvβ5 integrin (50 µg/ml; ab78873, Abcam) or isotype control monoclonal mouse IgG1 (50 µg/ml; ab9404, Abcam) for 5 h at 37°C in 5% CO2. To remove unbound ROS, the wells were washed vigorously with PBS five times. Photomicrographs of total ROS uptake (ROS bound and internalized) were obtained using an IX70 inverted fluorescence microscope (Olympus). To determine ROS internalization, 0.4% trypan blue was added to the PBS in equal volume for 10 min to quench extracellular fluorescence and then the wells were washed gently four times with PBS. Photomicrographs of internalized ROS were obtained. To quantify fluorescence intensity, pixel densiometry was performed using ImageJ software (NIH), using photomicrographs from three wells for each condition. Separate experiments were normalized to ARPE-19 phagocytosis activity, which was assayed in each experiment.

2.6 Karyotype analysis

Karyotyping of iPS(IMR90)-3 (p36 and p53) and hESC H9 (p50) were performed by Cell Line Genetics.

3 Results

3.1 hESC and iPSC differentiation capabilities

As previously described, RPE differentiation can be induced by growing iPSCs or hESCs to confluence in maintenance medium, followed by culture in bFGF-free medium to elicit overgrowth and spontaneous differentiation. We compared the propensity for RPE cell generation of an iPSC line, iPS(IMR90)-3, to that of an hESC line, H9. These lines were selected because each has been shown in multiple previous studies to differentiate into RPE and rescue visual function following transplantation into the RCS rat (Buchholz et al., 2009; Carr et al., 2009; Lund et al., 2006). Furthermore, we have observed that during RPE differentiation the iPS(IMR90)-3 line generates higher levels of spontaneous pigmentation than other WiCell iPSC lines, making it an ideal iPSC line for efficient quantification of pigmentation during RPE differentiation. We allowed H9 and iPS(IMR90)-3 cells to differentiate on Matrigel and by 44 days post-bFGF removal all culture replicates with differentiating H9 cells contained pigmented spots, whereas only 6 of the 12 with iPS(IMR90)-3 (50%) had generated any pigmented spots. Karyotype analysis confirmed that these iPS(IMR90)-3 cells had a normal karyotype (see Supporting information, Figure S1). iPS(IMR90)-3 had significantly fewer pigmented spots than H9 cells by this time (0.10 ± 0.035 spots/cm2 vs 2.3 ± 0.26 spots/cm2, respectively, or 4.1 ± 1.6%; p ≤ 3 × 10–6). Of the culture replicates that generated pigment by this time, the average date of pigment onset (when the first pigmented spots were observed) for iPS(IMR90)-3 cells was 30 ± 3.5 days post-bFGF removal, whereas for H9 cells cultured under the same conditions the average date of pigment onset was significantly earlier, 21 ± 0.8 days post-bFGF-removal (p ≤ 0.05). Additionally, the general observed pigmentation rate of iPS(IMR90)-3 cells was less consistent than the rate observed for H9 cells.

3.2 Selection of ECM proteins

After finding the pigmentation frequency of iPS(IMR90)-3 cells to be significantly lower than H9 cells, we investigated whether the pigmentation frequency of the iPS(IMR90)-3 cells could be improved through culture on ECM proteins likely to be recognized by RPE integrin receptors. Integrins are multifunctional, heterodimeric transmembrane receptors involved in mediating key interactions between the cell and the ECM, including those involved in adhesion, proliferation, and in a variety of signalling pathways, potentially including stem cell differentiation signalling (Humphries et al., 2006; Krissansen and Danen, 2007; Suzuki et al., 2005). Previous studies of integrin subunit expression in human RPE, from cultured cell lines and uncultured tissues, found that several subunits are highly expressed and/or functionally active (Table 1), although variation among different types of RPE (e.g. fetal vs adult, and native tissue or freshly isolated vs extensively cultured) has been reported (Aisenbrey et al., 2006; Gullapalli et al., 2008; Zarbin, 2003; our unpublished microarray data). Integrin subunits combine to form defined heterodimers known to bind to specific ECM ligands (Dipersio et al., 1995; Humphries et al., 2006; Krissansen and Danen, 2007; Suzuki et al., 2005; Zhang et al., 2003). Bruch's membrane, the ECM complex basal to the RPE in the eye, contains several ECM proteins that bind integrins expressed by RPE cells. These include laminins (specifically laminin-111, -332, -511 and −521, formerly laminin-1, -5, -10 and −11, respectively) (Aisenbrey et al., 2006), collagens (I and IV), fibronectin, and vitronectin (Campochiaro et al., 1986; Del Priore and Tezel, 1998; Del Priore et al., 2006) (shown in bold in Table 1). Previous studies have also shown that human RPE cells can attach to several of these substrate molecules, including laminin-111, -332, −511/521 together, collagen IV, fibronectin and vitronectin (Aisenbrey et al., 2006; Del Priore et al., 2006). We selected several commercially available ECM proteins present in Bruch's membrane, and with known integrin receptors expressed by RPE, to test their ability to improve iPSC-RPE differentiation in vitro. The selected ECM proteins included laminins-111 and −332, collagens I and IV, fibronectin and vitronectin. We additionally included commercially available substrates (gelatin and Matrigel) and feeder mouse fibroblasts (MEFs) known to support RPE differentiation and/or proliferation, a decellularized human fetal RPE surface, and elastin, which is expressed in Bruch's membrane (Del Priore et al., 2006).

Table 1. Integrin expression in RPE
α-Integrin subunits expressed in RPE*β-Integrin subunit binding partners expressed in RPE*Primary ECM integrin ligands**
  • *

    Zarbin, 2003, Aisenbrey et al., 2006; Gullapalli et al., 2008; our unpublished microarray data.

  • **

    Suzuki et al., 2005; Krissansen and Danen, 2007; Zhang et al., 2003; DiPersio et al., 1995; Humphries et al., 2006.

  • Integrin ligands shown in bold are expressed in Bruch's membrane.

α3β1Laminin-332, -511 and −521; collagen IV; fibronectin; thrombospondin
α5β1Fibronectin; fibrinogen; osteopontin
α6β1Laminin-111, -211, -221, -332, -511 and −521
β4Laminin-332 and −511
α7β1Laminin-111, -211 and −221
α11β1Collagen-I and -IV
αVβ1Vitronectin; fibronectin; fibrinogen; osteopontin
β5Vitronectin; osteopontin
β8Vitronectin; fibronectin; collagen IV; laminin-111

3.3 Substrate effects on pigmentation

iPS(IMR90)-3 cells were cultured on the selected substrates in maintenance medium for 5 weeks (bFGF was removed after the first week to allow for spontaneous differentiation). Three days post-seeding, iPS(IMR90)-3 colonies and/or dense, adherent cell clusters formed on most substrates tested (see Supporting information, Figure S2A), which expanded over the 5 week period (see Supporting information, Figure S2B). Surface area coverage over time was estimated and the substrates that allowed for the highest amount of coverage were found to be Matrigel, mouse laminin-111, MEFs, decellularized hfRPE and mouse collagen IV (see Supporting information, Figure S2C). By the end of the 5 week period, spots of pigmented cells were observed on most substrates (Figure 1A). We imaged every pigmented spot at 5 weeks post-seeding and, using these images, quantified average total area of pigmentation (Figure 1B). We found that Matrigel and mouse laminin-111 supported the most pigmentation. Total area pigmented on Matrigel was significantly higher than on gelatin, MEFs, human laminin-332, rat laminin-332, human fibronectin, human vitronectin or human collagen IV (p ≤ 0.05), but similar to mouse laminin-111 (0.28 ± 0.11 and 0.30 ± 0.20 mm2 pigment/culture well, for Matrigel and mouse laminin-111, respectively, or 7.8 ± 5.9% variance). Thus, Matrigel and mouse laminin-111 supported significantly increased production of pigment compared to feeder layers and, more importantly, to other defined substrate systems.

Figure 1.

Pigmentation of iPSCs differentiated on ECM substrates after 5 weeks. (A) Brightfield images of pigmentation spots on different ECM substrates; scale bar = 100 µm. (B) Quantification of the average total area of pigment for each substrate per 24-well ± standard error of the mean (SEM). FN, fibronection; VN, vitronectin; LM, laminin; Coll, collagen; hfRPE, human fetal RPE

3.4 Survival and morphology of iPSC-RPEs on selected substrates

We next investigated the ability of selected substrates to support iPSC-RPE survival and to maintain normal RPE morphology. iPS(IMR90)-3-RPEs (derived on MEFs in hESC maintenance medium and maintained as RPE on gelatin with RPE medium) were seeded onto selected substrates and cultured for 4 weeks in serum-free RPE medium (Gamm et al., 2008). Within 1 week post-seeding, typical RPE monolayers were observed on several substrates (see Supporting information, Figure S3A), and after 4 weeks confluent monolayers had been established on six of the tested substrates (see Supporting information, Figure S3B, C): Matrigel, mouse laminin-111, human fibronectin, human vitronectin, decellularized hfRPE and mouse collagen IV. iPS(IMR90)-3-RPE cultured on these six substrates retained normal RPE morphology (Figure 2A) and displayed similar degrees of pigmentation, regardless of which substrate was used, as measured by the total area of densely pigmented regions quantified through threshold normalization (Figure 2B).

Figure 2.

Morphology and pigmentation of iPSC-RPEs (p2) cultured on ECM substrates for 4 weeks. (A) Brightfield images of iPCS-RPE monolayers, showing typical RPE morphology; scale bar = 100 µm. (B) Quantification of the average area densely pigmented relative to iPSC-RPEs on matrigel ± SEM. FN, fibronection; VN, vitronectin; LM, laminin; Coll, collagen; hfRPE, human fetal RPE

To determine whether culture on different substrates affects RPE marker expression, we collected RNA from iPS(IMR90)-3-RPE cultured for 4 weeks on selected substrates and performed quantitative RT–PCR for key RPE markers (MITF, OTX2, PEDF and tyrosinase). RPE marker expression levels for tested substrates were similar to levels found in hfRPE samples, within two-fold increase or decrease (Figure 3), with the exception of MITF expression in iPS(IMR90)-3-RPE cultured on rat collagen I (0.442 ± 0.108-fold) and tyrosinase expression in multiple samples [human laminin-332 (2.74 ± 0.604-fold), rat laminin-332 (2.10 ± 0.356-fold), mouse collagen IV (2.22 ± 0.519-fold) and human fibronectin (2.40 ± 0.581-fold)]. Levels of tyrosinase were significantly higher in iPS(IMR90)-3-RPE cultured on gelatin (2.21 ± 0.269-fold; p = 0.03) and vitronectin (4.10 ± 0.794-fold, p = 0.05).

Figure 3.

Expression of RPE gene transcripts from iPSC-RPEs (p2) cultured on ECM substrates for 4 weeks and from hfRPE cell samples, all normalized to the geometric mean of four housekeeping genes ± SEM. RPE markers were essentially unchanged from levels in hfRPE samples, most within two-fold increase or decrease. RPE genes investigated were: (A) MITF; (B) OTX2; (C) PEDF; and (D) tyrosinase. *p ≤ 0.05 compared to hfRPE. FN, fibronection; VN, vitronectin; LM, laminin; Coll, collagen; hfRPE, human fetal RPE

3.5 Derivation of iPSC-RPEs on mouse laminin-111

Based on the ability of mouse laminin-111 to support both spontaneous pigmentation and growth of iPS(IMR90)-3-RPE (derived on feeder fibroblasts), we determined whether RPE could be derived and cultured long-term using this substrate. iPS(IMR90)-3 were differentiated on mouse laminin-111 and, 87 days post-bFGF removal, pigmented spots were manually enriched and cultured on mouse laminin-111 in serum-free RPE medium for multiple passages. Protein expression of key RPE markers by iPS(IMR90)-3-RPE derived and cultured in these conditions was investigated through immunocytochemistry, which showed that these cells expressed the nuclear localized transcription factors MITF and OTX2, the cytoplasmic proteins PEDF and Pmel17, and displayed normal RPE morphology (Figure 4). These results showed that a single, defined substrate could support the derivation and maintenance of RPE from human pluripotent stem cells.

Figure 4.

Expression of RPE markers of iPSC-RPEs (p2) differentiated and maintained on mouse laminin-111. Merged images are immunofluorescently labelled with RPE marker-specific antibodies (red) and Hoechst (blue), with inset images showing higher magnification; scale bars = 100 µm (large image), 25 µm (inset). FN, fibronection; VN, vitronectin; LM, laminin; Coll, collagen; hfRPE, human fetal RPE

3.6 Derivation of hESC-RPEs on mouse laminin-111

Given the ability of mouse laminin-111 to support the derivation of iPS(IMR90)-3-RPE, we next set to determine whether RPE could also be derived from hESC line H9 in this novel defined culture system. We conducted more extensive analysis of the derivation of hESC-RPEs on mouse laminin-111 because hESCs may currently be considered more clinically relevant than iPSCs (Gore et al., 2011; Hussein et al., 2011; Lister et al., 2011). hESC-RPEs previously has been derived through differentiation on Matrigel and feeder fibroblasts (Carr et al., 2009; Rowland et al., 2012; our unpublished data), and consequently we used these substrates as positive controls. H9s (shown to have a normal karyotype; see Supporting information, Figure S1C) seeded on mouse laminin-111 or Matrigel reached confluence an average of 5 ± 0.6 or 6 ± 0.6 days post-seeding, respectively, while significantly more time (8 ± 0.7 days) was required for H9s to reach confluence on hs27 feeder fibroblasts (see Supporting information, Table S1). Average date of pigment onset was similar for all three substrates (20–22 days post-bFGF removal). The number of pigmented spots/cm2 over time was quantified and, for all but one time point (day 28), the number of pigmented spots for H9s differentiated on Matrigel was not significantly different from the number on mouse laminin-111, indicating that the two substrates allowed for similar degrees of pigmentation. However, H9s on hs27s had significantly more pigmented spots/cm2 than on mouse laminin-111 at 21 days post-bFGF removal (p ≤ 0.03) and than on Matrigel at 28 days post-bFGF removal (p ≤ 0.003) (Figure 5A). This trend continued for the duration of pigment quantification (up to 44 days post-bFGF removal). However, by 71 days post-bFGF removal, many cell sheets differentiated on hs27s had detached from the culture wells, resulting in only 2/12 (17 ± 0.1%) culture replicates with hs27s being completely (≈ 100%) attached, while all cell sheets on mouse laminin-111 or Matrigel remained intact. To salvage the pigmented cells differentiated on hs27s, these cells were enriched an average of 13–15 days before cells on Matrigel or mouse laminin-111, respectively. Upon enrichment, RPE cell yields were not significantly different between H9 differentiated on mouse laminin-111 and on Matrigel (Figure 5B). However, significantly more RPE cells were obtained from H9 differentiated on mouse laminin-111 or on Matrigel than on hs27s: 2.58 × 104 cells/cm2 on hs27s compared to 4.58 × 104 and 6.02 × 104 cells/cm2 on mouse laminin-111 and Matrigel, respectively (178 ± 11% and 234 ± 22% of yields on hs27s, respectively, p ≤ 0.02).

Figure 5.

Differentiation of hESCs to RPE on selected ECM substrates. (A) Appearance of pigment spots over time in differentiating hESCs on hs27 feeder fibroblasts, Matrigel, or mouse LM-111. (B) RPE enrichment yields from hESCs differentiated on hs27s, Matrigel, or mouse LM-111. (C) Ability of hESC-RPEs derived on Matrigel or mouse LM-111 (p0) to internalize rod outer segments (ROS), as determined by pixel analysis of photomicrographs. hfRPE and hs27s served as positive and negative controls for phagocytosis, respectively. IgG is an isotype-matched control antibody. (D) Expression of RPE markers of hESC-RPEs (p2) differentiated and maintained on mouse LM-111 or Matrigel. Merged images were immunoflourescently labelled with RPE marker-specific antibodies (red) and Hoechst (blue), with inset images showing higher magnification; scale bars = 100 µm (large image), 25 µm (inset). **p ≤ 0.03 and *p ≤ 0.003 compared to samples on mouse LM-111; ^^p ≤ 0.02 and ^p ≤ 0.003 compared to samples on Matrigel; +p ≤ 0.003 compared to samples with αvβ5 inhibitor. Error bars represent ± SEM. FN, fibronection; VN, vitronectin; LM, laminin; Coll, collagen; hfRPE, human fetal RPE

The enriched RPE cells were cultured for multiple passages and, as an essential role of RPE function in vivo is to perform phagocytosis of rod outer segments (ROS), we investigated whether H9-RPE derived on mouse laminin-111 or Matrigel were able to phagocytose ROS (Figure 5C). We found that the H9-RPE phagocytosed ROS, and this activity was inhibited by an anti-αvβ5 integrin antibody (known to block RPE phagocytosis activity in vitro) (Lin and Clegg, 1998). Additionally, expression of key RPE markers investigated through immunocytochemistry showed these cells to express several RPE markers, including PEDF, MITF, OTX2, Pmel17, and multiple membrane-associated proteins: ZO-1, Bestrophin, and αv integrin (Figure 5D). Furthermore, a high percentage of cells were positive for RPE markers, as measured by expression of Pmel17 (98.6 ± 0.44% on Matrigel and 99.2 ± 0.23% on mouse laminin-111), indicating high-purity RPE. Thus, mouse laminin-111 supports differentiation and maintenance of both iPSC-RPEs and hESC-RPEs.

4 Discussion

In this report, we show that multiple substrates and purified ECM proteins in fully-defined, feeder-free culture systems support the growth of iPSCs, the differentiation of pigmented regions from iPSCs and the maintenance of iPSC-RPEs. The effects of different ECM proteins on RPE cell differentiation from pluripotent human stem cells have not previously been reported. Mouse laminin-111, in particular, was found to support both iPSC-RPE and hESC-RPE generation and maintenance for multiple passages, demonstrating that a single ECM protein can potentially replace feeder cell platforms and other non-defined culture conditions, such as Matrigel, in some applications.

iPSCs and hESCs show variability in their propensity to generate viable RPE cells (Buchholz et al., 2009; Feng et al., 2010; Hirami et al., 2009; Hu et al., 2010; Meyer et al., 2009). Using iPSC and hESC lines previously reported to efficiently differentiate into functional RPE cells (Buchholz et al., 2009; Carr et al., 2009; Lund et al., 2006), we found that these iPSCs pigmented at frequencies significantly lower than those observed with hESCs. Recent reports have suggested that iPSCs often contain epigenetic variations or high mutational loads, which might contribute to their varying propensities to differentiate (Gore et al., 2011; Hussein et al., 2011; Lister et al., 2011). Hu et al. (2010) showed that some iPSCs derived from RPE cells may be predisposed to redifferentiate back into RPEs (Hu et al., 2010), underscoring the potential importance of epigenetic influences on differentiation.

We found that iPSCs readily developed pigmented cells when spontaneously differentiated on most ECM-derived substrates tested. The ECM has been suspected to play a central role in the differentiation of cells during development (Hynes, 2009), but little is understood of how it functions in RPE development. Previously, iPSCs and hESCs have been differentiated to RPE cells through culture on the substrates on which they are normally maintained, primarily feeder fibroblasts or Matrigel (Rowland et al., 2012); these are complex substrates containing several different ECM proteins and soluble factors. We show here that pigmented cells can be generated through differentiation on several purified proteins. The flexibility in substrate might reflect the fact that both undifferentiated stem cells and RPEs express a variety of matrix integrin receptors. The frequency of pigmentation does not appear to simply correlate with proliferation, as substrates on which iPSCs displayed high levels of surface area coverage did not necessarily support high amounts of pigmentation. For example, iPSCs on fibronectin and vitronectin displayed coverage higher than on half of the substrates tested, but resulted in low levels of pigmentation (significantly lower than on Matrigel; p ≤ 0.05). While Matrigel supported greater pigmentation of the iPS(IMR90)-3, Matrigel was unable to improve pigmentation of low-pigmenting lines, such as iPS(foreskin)-1 (data not shown). These findings suggest many encouraging routes toward xeno-free culture systems for the generation of RPE from human pluripotent stem cells for translation to clinical therapies.

While most ECM proteins and substrates tested supported some degree of iPSC-RPE growth, only six resulted in completely confluent monolayers after 4 weeks of culture. Interestingly, even substrates only moderately supportive of iPSC-RPEs growth resulted in iPSC-RPE expressing key RPE genes at levels similar to those found in hfRPE. Although expression of tyrosinase, associated with pigmentation, was significantly elevated in two samples, gene expression levels do not necessarily fully translate into protein expression levels, and significantly higher areas of densely pigmented cells may not have been observed on vitronectin (Figure 2B) due to the RPE cells reaching a terminal level of pigmentation. We have previously observed transdifferentiation in iPSC-RPE and hESC-RPE cultures after four or five passages, at which point the cells' ability to regain normal RPE morphology and pigmentation after passaging declines and they became more fibroblastic (Buchholz et al., 2009). RPE gene expression data from our current study suggest that the degree of confluence of iPSC-RPE culture does not significantly affect RPE marker expression in established, differentiated RPE cells, indicating that actors other than cell confluence, such as possibly the act of passaging itself, may be responsible for observed transdifferentiation.

Because some pigmentation was observed on most substrates tested, it appears that pluripotent stem cells possess a strong propensity to spontaneously differentiate into RPEs, and it is likely that there are other essential factors involved. Previously, researchers have generated RPEs from iPSCs and hESCs through the formation of neurospheres or embryoid bodies, with the addition of soluble signalling factors, including WNT antagonists, NODAL antagonists, TGFβ superfamily activators or the media supplements N2 and B27 (Idelson et al., 2009; Meyer et al., 2009; Osakada et al., 2009). While some of these factors may significantly improve the frequency of RPE cell differentiation, the relative effectiveness of each of these factors on RPE cell differentiation is still unclear. It may also be important to explore the effects of multiple ECM proteins in combination on RPE differentiation. Although in our current study we found that a combination of purified ECM proteins (mouse laminin-111 and collagen IV, the primary ECM components in Matrigel) did not significantly increase pigmentation (data not shown), this area merits further investigation, as many ECM protein combinations are possible and feeder fibroblast layers contain a variety of ECM proteins that collectively may promote the high degree of hESC pigmentation observed in this study. Future studies combining key ECM proteins or substrates along with essential soluble factors may significantly improve pigmentation rates, although currently it is unclear which soluble factors are most effective.

We found that mouse laminin-111 could support the differentiation, derivation and maintenance of iPSC-RPEs and hESC-RPEs expressing key RPE markers. Mouse laminin-111 and collagen IV are the two primary ECM components of Matrigel, which also supported high degrees of pigmentation. Because relatively low levels of pigmentation were observed on mouse collagen IV, this suggests that the primary ECM component responsible for pigmentation observed on Matrigel is mouse laminin-111. Additionally, laminin, present in native human Bruch's membrane (Aisenbrey et al., 2006), is also the substrate of choice for the differentiation and culture of hfRPE cells (Hu and Bok, 2001), which further supports our suggested use of a laminin protein. We previously reported that multiple integrin subunits that form heterodimers that can bind laminin-111 (specifically α6, α7, αV, and β1) are expressed by undifferentiated hESCs and iPSCs (Rowland et al., 2010), and reported here in Table 1 that these subunits are also in RPE cells, including hESC-RPEs. Although this would suggest that these laminin-111-binding integrin subunits may be expressed throughout the RPE differentiation process, it would be of interest to see how these levels may change during differentiation, and whether levels are different on laminin-111, as the above data were generated on other substrates.

Due to their more immediate clinical relevance, hESCs were further investigated. hESC differentiation on mouse laminin-111 and Matrigel initially resulted in similar numbers of pigmented spots, but significantly fewer than on feeder fibroblasts, which is by far the most commonly used substrate for differentiation of RPE from hESCs and iPSCs (Rowland et al., 2012). However, significantly higher enrichment yields were seen with hESCs on Matrigel and mouse laminin-111 than on feeder fibroblasts, most likely due to pigmenting hESC sheets detaching from the feeder fibroblast substrate before the cells were ready for enrichment, a behaviour which we have consistently observed. This finding also highlights the importance of carrying RPE differentiation studies out to the point of enrichment, as high levels of initial pigmentation might not necessary result in high levels of enrichable RPE cells, and these are the cells needed for potential transplantation therapies. Overall, our results show that Matrigel and mouse laminin-111 may be ideal substrates for generating high yields of enriched RPE, resulting in higher yields than the current most widely used substrate, feeder fibroblasts.

These findings are important not only for an improved understanding of how the ECM affects the general process of differentiation, but also for the development of ideal culture conditions and transplantation scaffold strategies for translation of these technologies to the clinic for the treatment of retinal diseases such as AMD. Future efforts should be directed at designing xeno-free systems, not only for the differentiation of RPE from human pluripotent stem cells but also for the maintenance of the derived RPE cells. Our finding that mouse laminin-111 supports the differentiation, derivation and maintenance of RPE from iPSCs and hESCs to RPE suggests that purified ECM proteins may prove a significant part of these efforts. ECM proteins contain known peptide sequences that promote cell adhesion and are usually conserved between mouse and human, and these sequences, along with our findings, may help suggest ideal xeno-free, peptide-based scaffold designs for transplantations using human pluripotent stem cell-derived RPE.

5 Conclusions

Our results suggest that laminin-111 supports the derivation and maintenance of hESC-RPEs and iPSC-RPEs. Previous derivation systems have used complex substrates that are not defined, such as feeder fibroblast layers or Matrigel. We show here that pigmented cells may arise through differentiation on several different ECM proteins and substrates found in the RPE ECM environment in vivo. These findings may potentially serve as the basis for future RPE scaffold designs for RPE transplantations to treat retinal diseases, such as AMD.


We thank Jamie Thomson for iPSCs and zbFGF, and Andrew James Bonham for a critical reading of the manuscript. T.J.R. and D.E.B. were supported by the California Institute for Regenerative Medicine (CIRM) (Training Award Nos TG2-01151 and T3-00009, respectively) and supported by the Institute for Collaborative Biotechnologies through grant W911NF-09-0001 from the U.S. Army Research Office. The content of the information does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred.


This research was reviewed by the UCSB Embryonic Stem Cell Research Oversight (ESCRO) committee. The ESCRO committee approved this research, based on guidelines set by CIRM.