Author contributions: Q.H.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript; A.M.F.: collection of data; L.V.J.: conception and design, financial support, manuscript writing, final approval of manuscript; D.O.C.: conception and design, financial support, manuscript writing, final approval of manuscript.
First published online in STEM CELLS EXPRESS September 29, 2010.
Disclosure of potential conflicts of interest is found at the end of this article.
Induced pluripotent stem (iPS) cells have been generated from a variety of somatic cell types via introduction of transcription factors that mediate pluripotency. However, it is unknown that all cell types can be reprogrammed and whether the origin of the parental cell ultimately determines the behavior of the resultant iPS cell line. We sought to determine whether human retinal-pigmented epithelial (RPE) cells could be reprogrammed, and to test the hypothesis that reprogrammed cells retain a “memory” of their origin in terms of propensity for differentiation. We reprogrammed primary fetal RPE cells via lentiviral expression of OCT4, SOX2, LIN28, and Nanog. The iPS cell lines derived from RPE exhibited morphologies similar to human embryonic stem cells and other iPS cell lines, expressed stem cell markers, and formed teratomas-containing derivatives of all three germ layers. To test whether these iPS cells retained epigenetic imprints from the parental RPE cells, we analyzed their propensity for spontaneous differentiation back into RPE after removal of FGF2. We found that some, but not all, iPS lines exhibited a marked preference for redifferentiation into RPE. Our results show that RPE cells can be reprogrammed to pluripotency, and suggest that they often retain a memory of their previous state of differentiation. STEM CELLS 2010;28:1981–1991
Differentiated somatic cells can be reprogrammed to alternative differentiated phenotypes by expression of lineage-specific transcription factors (TF) [1–6]. This approach has been used to reprogram human somatic cells to a pluripotent state, using either of two sets of TFs [7, 8]. More recently, a variety of methods have been used to generate such induced pluripotent stem (iPS) cells, some not requiring genomic insertion of expression vectors [8–13]. Numerous cell types have been reprogrammed including fibroblasts , keratinocytes , neural stem cells , hepatocytes and gastric epithelial cells , and adipose stem cells . Reprogrammed iPS cells exhibit morphologies, differentiative potential, teratoma formation, and gene expression profiles similar but not identical to ESCs. Murine iPS cells have been shown to generate viable, fertile progeny via tetraploid complementation, demonstrating a pluripotency similar to ESCs [18–20]. However, significant differences between iPS cells and ESCs have been reported. For example, the tumorigenic potential and viability of chimeras of mouse iPS cells depends on the origin of the parental cell . Other differences between iPS cells and ESCs have been noted in global gene expression patterns [8, 21], expression of imprinted genes , DNA methylation , and differentiative potential and generation of differentiated derivatives [24, 25].
Differentiated somatic cells are replete with epigenetic regulatory devices that lock characteristic gene expression patterns into place. In general, they are quite refractory to reprogramming, as indicated by the very low efficiency of iPS cells generation, and lack of complete reprogramming to the ESC state. Evidence for an iPS cell “memory” of former differentiated phenotype has come from several studies. Chin et al.  identified a group of genes and miRNAs that are differentially expressed in iPS cells compared with human embryonic stem cells (hESCs), in part accounted for by residual expression of fibroblast genes. In another study, Marchetto et al.  found that human neural stem cells that were reprogrammed without vector insertion retained expression of some neural genes. However, neither of these studies examined whether reprogrammed cells tend to redifferentiate into their former identity.
iPS cell or ESC derivatives might one day be the mainstay of regenerative medicine. Age-related macular degeneration (AMD), a candidate for cellular therapy, is a blinding eye disease caused by the degeneration of retinal-pigmented epithelial (RPE) cells. RPE cells play important roles in visual function, that include contributing to the blood/retina barrier and visual cycle, absorption of stray light, transport of nutrients, secretion of growth factors, phagocytosis of shed outer segment membrane, and support of photoreceptor growth and differentiation . RPE cells arise from the early eye field of the anterior neural plate, where surrounding cells secrete factors that induce expression of fate-determining TFs . During the progression of AMD, RPE cell death leads to secondary dysfunction and death of photoreceptors. RPE cells, derived from ESCs and iPS cells [29–35] can rescue visual function in the Royal College of Surgeons rat model of RPE dysfunction [36, 37]. However, it is difficult to isolate and maintain iPS-RPE cells in vitro, as the efficiency of spontaneous differentiation is low and iPS-RPE cells tend to senesce and transdifferentiate after several passages in vitro . Based on current understanding of RPE development, efforts have been made to improve the efficiency of differentiation of pluripotent cells into RPE [38–43], but methods for obtaining high yields are still lacking. Here, we report the reprogramming of fetal RPE cells to an iPS cell state. To test the hypothesis that these cells retain some memory of their origin, we analyzed their propensity to redifferentiate into RPE cells and found that some but not all lines show increased tendency for spontaneous re-expression of the RPE phenotype.
MATERIALS AND METHODS
hESCs and iPS cells were maintained on mytomycin-C-treated HS27 human fetal fibroblast (ATCC # CRL-1634) in Dulbecco's modified Eagle's medium (DMEM)/F12 medium supplemented with 20% KnockOut serum replacement, 0.1 mM nonessential amino acids, and 0.1 mM β-mercaptoethanol (all from Invitrogen, Carlsbad, CA, http://www.invitrogen.com). For hESCs, this stem cell medium was supplemented with 4 ng/ml human basic fibroblast growth factor (bFGF; Peprotech, Rocky hill, NJ, http://www.peprotech.com/); for iPS cells, it was supplemented with 25 ng/ml zebra fish bFGF (a kind gift from James A. Thomson). In some experiments, feeder-free culture on 0.1 mg/ml Matrigel (BD biosciences, Bedford, CA, http://www.bdbiosciences.com/) was used, with the addition of two-day-conditioned stem cell medium from human HS27 fibroblasts. ESCs and iPS cells were passaged by manual dissection without enzymatic dissociation. Human fetal RPE cells (#5949) were obtained from The Center of Study of Macular Degeneration, University of California Santa Barbara (UCSB) and cultivated in DMEM medium (Invitrogen) supplemented with 10% fetal bovine serum in tissue culture flasks coated with 0.1% gelatin (Sigma-Aldrich, St. louis, MO, http://www.sigmaaldrich.com). RPE cells derived from iPS cells were cultivated in alpha modification minimum essential medium eagle (α-MEM) (Sigma-Aldrich) containing 5% or 15% (first 3 days after enrichment) heat-inactivated fetal bovine serum (Invitrogen), 1× N1 supplement (Sigma-Aldrich), 0.1 mM nonessential amino acids (Invitrogen), 2 mM L-glutamine (Invitrogen), 0.25 mg/ml taurine (Sigma-Aldrich), 0.02 μg/ml hydrocortisone (Sigma-Aldrich), and 0.013 ng/ml triiodo thyronin (Sigma-Aldrich)  in tissue culture flasks coated with gelatin.
Lentivirus Transduction and Reprogramming Procedure
Lentiviral vectors for the expression of TFs were obtained from James A. Thomson  and virus packaging was carried out in human embryonic kidney 293T cells. Each lentivirus preparation was separately titrated by infection of human RPE cells, using puromycin selection. For reprogramming, passage 9-11 RPE cells were seeded on 10-cm-diameter tissue culture dishes (BD biosciences; 106 cells per dish), and one day postseeding were incubated with a cocktail of the lentivirus (LIN28:Nanog:OCT4:SOX2 = 5:7.5:3:1) overnight in presence of 0.6 μg/ml polybrene (Millipore, Billerica, MA, http://www.millipore.com) as previously described . At day 7 post-transduction, cells were trypsinized and seeded onto HS27 feeder cells in stem cell medium (5 × 105 cells per dish). The following day, the culture medium was changed to stem cell medium supplemented with 25 ng/ml zebra fish bFGF. Colonies with ESC morphology were manually dissected and transferred to a new HS27 feeder plate on day 25 post-transduction.
Alkaline Phosphatase Staining
Cells were washed with PBS, fixed in 4% paraformaldehyde (PFA) in PBS (4°C, 20 minutes) and then rinsed with PBS twice. Alkaline phosphatase (AP) staining was carried out using the Vector Red Alkaline Phosphatase Substrate Kit I (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com/) as per the manufacturer's instructions.
Cells were passed onto four-well Permanox polystyrene chamber slides (BD biosciences), either seeded with HS27 feeder cells or coated with matrigel, and were cultivated for 4–7 days. Cell cultures were washed with PBS, fixed in 4% PFA in PBS (4°C, 20 minutes), washed in PBS, and incubated in 3% bovine serum albumin (BSA) (Sigma-Aldrich) either in PBS (nonpermeabilized cells) or PBS containing 0.1% NP40 (permeabilized cells) for 1 hour. Samples were then incubated with primary antibody diluted in 3% BSA/PBS (25°C, 1 hour), control samples were incubated with preimmune serum or isotype-matched nonimmune serum diluted in 3% BSA/PBS (25°C, 1 hour). After three PBS washes, all samples were incubated with appropriate secondary antibody diluted in 3% BSA/PBS (25°C, 1 hour). Nuclei were stained using Hoechst 33342 (Sigma-Aldrich). Samples were mounted with Gel/Mount (Electron Microscopy Sciences, Hatfield, PA, http://www.emsdiasum.com/) and visualized with an Axioskop BX60 microscope (Zeiss, Thornwood, New York, http://www.zeiss.com/).
hESCs or iPS cells were cultivated on matrigel for 2–3 passages. Cells were washed with Dulbecco's PBS and then incubated with Nonenzymatic Cell Dissociation Buffer (Invitrogen; 37°C, 7 minutes). Cells were dislodged by repeated pipetting, harvested, and resuspended in PBS/1% BSA, following filtration through a 50-μm CellTrics Filter (ParTec, Munster, Germany, http://www.partec.com). Approximately 200 μl of cell suspension containing 2 × 105 cells were incubated with primary antibody diluted in PBS (30 minutes, on ice). As a control, a parallel cell suspension was incubated with isotype-matched nonimmune serum. Cells were collected by centrifugation, washed in PBS/1% BSA, and then incubated with secondary antibody diluted in PBS (30 minutes, on ice). Finally, cells were washed and resuspended in PBS, and analyzed on the Guava EasyCyte Plus System (Millipore) using Guava ExpressPlus protocols. A total of 5,000 events were acquired; the data and graphics were analyzed and generated using CytoSoft 5.2.5 software (Millipore).
hESCs and iPS cells were cultivated on matrigel for 2–3 passages. After washing and manual removal of differentiated colonies, undifferentiated colonies were incubated with nonenzymatic cell dissociation buffer at 37°C for 7–10 minutes. Cells were resuspended in stem cell medium preconditioned for 2 days by HS27 fibroblasts and counted using a hemacytometer; cells of equal number (10,000–20,000 cells/well) were then seeded on matrigel-coated 12-well plates. The total number of cells per well was determined at 1, 2, 3, 4 days postseeding using a hemacytometer.
Genomic DNA of iPS cells was isolated using the Wizard SV Genomic DNA Purification System (Promega, Madison, WI, http://www.promega.com) as per the manufacturer's instructions. To examine the presence of the integrated transgenes in iPS cell lines, nested polymerase chain reaction (PCR) was applied. For the first round of PCR, lentiviral vector-specific primers, EF-1aF and SP3R, were used. For the second round of PCR, one primer specific for each transgene and another primer specific for the lentivirus vector were used. PCR reactions were performed using the pfx PCR system (Invitrogen) with modified reagent concentrations: 2× amplification buffer and 3× enhancer solution.
Reverse Transcription PCR and Quantitative PCR
Total RNA was prepared using RNeasy Mini Plus Kits (QIAGEN, Valencia, CA, http://www.qiagen.com) as per the manufacturer's instructions. To remove genomic DNA contamination, 2.5–5 μg of total RNA was treated using the TURBO DNA-free kit (Ambion, Austin, TX, http://www.ambion.com/) and 100–500 ng RNA from each sample was subjected to reverse transcription (RT) reaction using Oligo(dT)12–18 and the Superscript First-Strand Synthesis System (Invitrogen). Final RT products were diluted 1:10 in water and PCR reactions were performed using GoTaq Flexi DNA Polymerase System (Promega). Annealing temperature was set to 59°C for all primers. Quantitative PCR (QPCR) reactions were performed using GoTaq Hot Start Polymerase System plus 1× SYBR green I fluorescence dye (Invitrogen) and 1% dimethyl sulfoxide (DMSO). Reactions were carried out on an iCycler Real-Time PCR Detection System (Bio-Rad, Hercules, CA, http://www.bio-rad.com). For each sample, reactions for multiple housekeeping genes, including GAPDH, 18SrRNA, and β-actin, were run as controls. QPCR results were analyzed using the MyIQ System Software 1.0 (Bio-Rad) and GeNorm 3.5 (University of Ghent, Belgium).
Analysis of DNA Methylation Using Bisulfite Sequencing
The conversion of unmethylated cytosine to uracil in genomic DNA was performed using EZ Methylation-Gold Kit (Zymo Research, Orange, CA, http://www.zymoresearch.com/). Primers were designed to amplify a 300–500 bp promoter region upstream to the first exon of the gene of interest. PCR reactions were performed using GoTaq Hot Start Polymerase System (Promega). PCR products were gel purified, followed by cloning into pGEM-T easy plasmid (Promega). The plasmids were prepared using QuickLyse Miniprep Kits (QIAGEN), followed by digestion with EcoRI. The plasmids with correct-sized inserts were subjected to direct sequencing (Eton Bioscience, San Diego, CA, http://www.etonbio.com/).
In Vitro Differentiation
hESCs and iPS cells were manually dissected to obtain small cell aggregates and were then harvested and cultured in low-attachment plates (Corning, Corning, NY, http://www.corning.com/) in stem cell medium without bFGF. After 8 days of culture, embryoid bodies (EBs) were harvested and seeded on 0.1% gelatin-coated 10-cm-diameter tissue culture dishes in stem cell medium without bFGF. RPE differentiation was evaluated by microscopic characterization on pigmented spots generated on tissue culture dishes. Each cell line was tested a minimum of three times, representative data are shown.
Cell injection, teratoma generation, and harvesting were carried out by the Animal Resource Center at UCSB. In brief, iPS A3 and H14 cells from 50% to 70% confluent six-well plates were manually dissected to aggregates of approximately 100 cells, resuspended in 100 μl DMEM/F12 medium. Fetal RPE were trypsinized and 1.5 × 106 cells were resuspended in 100 μl DMEM/F12 medium. Cells were injected into the thigh muscle of immunodeficient severe combined immune deficiency (SCID) mice (The Jackson Laboratory, Bar Harbor, ME, http://www.jax.org/). For each cell line, two animals were injected. After 4–12 weeks, teratomas were dissected and fixed in 10% neutral-buffered formalin. Samples were embedded in paraffin, sectioned at 0.5 μm, and stained with H&E. Sections were examined on an Axioskop BX60 microscope (Zeiss). All animal protocols were approved by the UCSB Institutional Animal Care and Use Committee.
Reprogramming RPE Cells to iPS Cells
To determine if fetal RPE cells can be reprogrammed to pluripotency, lentiviral vectors encoding TFs OCT4, Nanog, SOX2, and LIN28, were incubated with primary human RPE cells. One week later, granular appearing colonies were evident in both control and lentivirus-exposed cultures. After 24 days, approximately 200 colonies were presented in both control and experimental cultures. To assess the formation of iPS cell colonies, some cultures were monitored for AP activity. In the lentivirus-exposed culture, 9 of 233 (3.9%) colonies were AP positive, whereas none of the colonies were AP positive in the control culture (Supporting Information Figure S1). The AP-positive colonies were morphologically similar to spontaneously differentiating ESC colonies (Fig. 1A, Supporting Information Figure S1C and S1D). On the contrary, the morphologies of AP-negative colonies were significantly different from that typical of ESCs (Fig. 1B). These AP-negative colonies had uneven peripheries and central cell mounding. Eight embryonic stem (ES)-like colonies from experimental cultures were expanded on feeder cells; six of these were selected for further analysis
The iPS lines showed a dichotomy in their tendency to differentiate in culture. Three lines (iPS A3, A4, and A6) were similar to other iPS and ES lines with 15%–30% of colonies showing some differentiated cells after 7 days of culture, whereas three other lines (B4, B6, and C5) showed a higher incidence of differentiation (>60%), and required more frequent passaging. We focused on the first three lines for further characterization. The iPS A3, A4, and A6 cells exhibited a morphology similar to that of hESCs, characterized by tightly packed colonies (Fig. 1C) and a high nuclear to cytoplasma ratio. (Fig. 1D). We investigated their expression of human ESC markers immunocytochemically. Like human ESCs, iPS cells expressed OCT4, SSEA4, TRA-1-60, and TRA-1-81 (Fig. 1E-1H, Supporting Information Figure S2). Flow cytometry confirmed that approximately 95% of undifferentiated iPS cells express three pluripotency-associated surface markers (Fig. 1I, Supporting Information Figure S3). These iPS cells were stable for more than 70 passages and displayed normal karyotype (Fig. 2A). The calculated population doubling times of iPS cells were 19.9 ± 3.5 (iPS A3, P32-P34), 16.5 ± 1.9 (iPS A4, P30-P32), and 18.4 ± 6.0 (iPS A6, P24-P38) hours, similar to the population doubling time of H14 human ESCs cultured in parallel (21.7 ± 5.7 hours, P65-P67) and consistent with previous reports .
We examined provirus integration in iPS cells by PCR using virus-specific primers. The results revealed that OCT4, Nanog, and SOX2 were integrated in all six iPS cell lines under analysis. LIN28 was absent from iPS A3, A4, and A6 but was detected in the three lines exhibiting greater tendency to differentiate (iPS B4, B6, and C5; Fig. 2B). This result is consistent with previous studies showing that LIN28 is not necessary for reprogramming . We analyzed transgene expression levels after 14 passages using QPCR and found that the transgenes were effectively silenced (Fig. 2C), suggesting that continuous transgene expression was not required to maintain the undifferentiated state of the iPS cell lines . Expression of ESC marker mRNAs was investigated using endpoint RT-PCR. Expression of LIN28, Nanog, OCT4, SOX2, KLF4, REX1, DPPA4, and telomerase reverse transcriptase (TERT) was detected in both iPS and ES lines (Fig. 3A).
To evaluate the epigenetic profiles of iPS cells, we screened promoter DNA methylation patterns for the pluripotency-associated genes, OCT4 and Nanog. Bisulfite sequencing results revealed that the promoter regions of these two genes were hypomethylated in iPS cells, similar to ESCs. In contrast, these promoter regions in cultured primary RPE cells were hypermethylated, showing that the DNA methylation state was significantly altered in the iPS lines (Fig. 3B).
Spontaneous Differentiation of iPS Cells
To test the pluripotency of iPS cells, we examined in vitro differentiation using EB-mediated differentiation. The iPS cells formed ball-like structures after suspension culture for 8 days (Fig. 4A). When seeded onto gelatin-coated plates and cultured for 2–3 weeks, the attached EBs expanded and differentiated into cells with various morphologies (Fig. 4B–4D). Immunocytochemistry detected cells positive for alpha fetoprotein (AFP, endoderm marker, Fig. 4E), βIII-tubulin (ectoderm marker, Fig. 4F), and α-smooth muscle actin (mesoderm marker, Fig. 4G). RT-PCR analysis revealed that differentiated iPS cells expressed SRY-box containing gene 17 (SOX17, endoderm), AFP (endoderm), caudal type homeobox 2 (CDX2, endoderm), Msh homeobox 1 (MSX1, mesoderm), Brachyury (mesoderm), paired box 6 (PAX6, ectoderm), and microtubule-associated protein 2 (MAP2, ectoderm; Fig. 4H). In contrast, expression levels of undifferentiated stem cell-specific genes, including Nanog, OCT4, SOX2, REX1, ESG1, DPPA2, DPPA4, TERT, and KLF4 were decreased in differentiated iPS cells (data not shown). This result suggested that iPS cells reprogrammed from RPE cells were able to differentiate to three germ layers in vitro. Interestingly, SOX17 and MSX1 are expressed in undifferentiated and differentiated ES and iPS cells, but their expression levels are particularly high in three iPS cell lines that showed a higher incidence of differentiation (iPS B4, B6, and C5).
Teratoma Formation by iPS Cells
To test the pluripotency of iPS cells derived from RPE cells, we transplanted iPS cells into the thigh muscle of immunodeficient mice. After 4–8 weeks, tumor formation was observed and histological examination showed that the teratomas contained tissues from the three primary germ layers (Fig. 4I, Supporting Information Figure S4), including fetal intestine (endoderm), fetal respiratory tissue (endoderm), cartilage (mesoderm), muscular tissue (mesoderm), and fetal neural tubes (ectoderm), confirming that the iPS cells reprogrammed from RPE cells exhibit pluripotency in vivo.
Differentiation of iPS Cells to RPE Cells
Interestingly, during the spontaneous differentiation of EBs, we observed that pigmented cell colonies appeared after 2–3 weeks cultivation (Fig. 5A, 5B). Initially, these pigmented cells were tightly packed in small islands surrounded by other cells (Fig. 5A). Eventually, the pigmented cells spread peripherally and expanded to form a monolayer of pigmented cells (Fig. 5B). These pigmented cell sheets were manually dissected, and the enriched pigmented cells were cultivated to confluence in Miller medium . During enrichment and passaging, these cells lost pigmentation, as reported previously for RPE derived from ES or iPS cells [29, 30]. However, uniform pigmentation was reestablished, 2–3 weeks after enrichment. The pigmented cells exhibited cobblestone-like polygonal shapes with phase-bright borders (Fig. 5C, Supporting Information Figure S5B) and spontaneously formed elevated domes (Fig. 5C, red arrowheads; Supporting Information Figure S5C and S5D), a typical observation in highly differentiated and polarized RPE cell cultures. In long-term culture, these pigmented cells maintained RPE-like morphology for 61 weeks without passage, at this point the cultures were terminated.
Immunocytochemical analysis showed the pigmented cells expressed RPE marker genes, including Bestrophin (BEST, membrane protein, Fig. 5D), zona occludens 1 (ZO-1, tight junction protein, Fig. 5E), transthyretin (TTR, secreted protein, Fig. 5F), orthodenticle homeobox 2 (OTX2, TF, Fig. 5G), and PAX6 (TF, Fig. 5H). However, expression was not always uniform throughout the cultures, indicating a range of cell maturity. RT-PCR assays revealed that the pigmented cells expressed many RPE marker genes (Fig. 5I), including retina and anterior neural fold homeobox (RAX, early eye field), SIX homeobox 3 (SIX3, early eye field), PAX6, microphthalmia-associated TF (MITF, RPE-specific TF), OTX2, tyrosinase (TYR, pigment synthesis), tyrosinase-related protein 1 (TYRP1, pigment synthesis), silver homolog (SILV, pigment synthesis), BEST, extracellular matrix metalloproteinases inducer (membrane-associated), ZO-1, retinaldehyde binding protein 1 (visual cycle), retinal pigment epithelium-specific protein 65kDa (RPE65, visual cycle), pigment epithelium derived factor (PEDF, secreted protein), and TTR but not homeobox gene 10 (neural retina). Expression was similar in H14 RPE derived from H14, iPSA3, iPSA4, and fetal RPE. We also investigated phagocytosis activity in the enriched pigmented cells. The results indicated that these pigmented cells can internalize fibronectin-coated microspheres, similar to the RPE cell line ARPE19 (Supporting Information Figure S5E–S5G). Taken together, these results suggest that the phenotype and the gene expression profiles of these pigmented cells are similar to native RPE cells.
To determine whether the iPS cells derived from RPE had a higher tendency to spontaneously redifferentiate into RPE, we compared them with iPS cell lines derived from human fetal lung fibroblasts (iPS(IMR90)-3), human foreskin fibroblasts (iPS(foreskin)-2 CCD-1079SK) , and human ESC lines (H9, H14). After 40 days of spontaneous differentiation, approximately 70% of iPS A3 colonies and 60% of iPS A4 were partially pigmented (Fig. 6B, 6C). In contrast, only 16% (22 of 139) of ES H14 colonies were pigmented (Fig. 6A). These pigmented cells formed typical RPE-like cobblestone monolayers and when a subset were manually dissected and expanded, they expressed RPE markers (see Fig. 5). Moreover, pigmented spots in the iPS A3 colonies were larger in size and darker in color, than those formed by iPS(IMR90) and iPS(foreskin). After longer periods of spontaneous differentiation (60 days), approximately 63.8% ± 10.6% of iPS colonies (average of two iPS cell lines [A3, A4], three separate experiments) were pigmented (Supporting Information Figure S5A). In contrast, iPS(IMR90) and iPS(foreskin) cell lines exhibited a lower frequency of differentiation into RPE cells (5%–10% of total colonies). This is consistent with previous studies on these lines . However, we found that different RPE-derived iPS cell lines exhibited varying propensities for redifferentiation. Although most of the iPS cell lines derived from RPE cells gave rise to more pigmented cells than ES or other iPS cells, some iPS cell lines derived from RPE cells failed to differentiate any pigmented colonies (IPS A6; Fig. 6C).
We examined RPE marker gene expression levels at an early stage of spontaneous differentiation of H14 and iPS A3 (7 days) using QPCR. The results showed that some RPE precursor genes, such as RAX, MITF, and OTX2, as well as some late stage RPE-specific genes, such as DCT, were more highly expressed in iPS A3 compared with H14 (Fig. 6D). These results suggest that in reprogrammed RPE cells, epigenetic imprints might be retained, resulting in enhanced expression of some RPE marker genes early in the differentiation process and induction of the RPE phenotype.
The TFs, RAX, OTX2, and MITF, are keys in retinogenesis, and the promoter regions of these three genes contain large numbers of cytosine guanine dinucleotides, so their expression is likely to be affected by DNA methylation. Therefore, we analyzed the DNA methylation patterns in promoter regions for RAX, OTX2, MITF, and PAX6 to determine if an RPE-like methylation pattern is retained in the reprogrammed iPS lines. Bisulfite sequencing revealed that DNA methylation states of these genes in iPS cells were similar to those in ESCs, and the RPE pattern was not retained (Supporting Information Figure S6). Whereas PAX6, OTX2, and RAX promoter regions were hypermethylated in the parental RPE cells, they were hypomethylated in iPS A3, iPS A6, and H14 ESCs. The MITF promoter region was hypomethylated in all lines analyzed. Bisulfite sequencing of several other RPE-specific genes, including TYR, TYRP1, DCT, TTR, and RPE65, revealed that DNA methylation states of the promoter regions in iPS cells were also similar to those of human ESCs.
This study demonstrates that human fetal RPE cells can be reprogrammed into iPS cells using four TFs (Thomson factors), OCT4, Nanog, SOX2, and LIN28. The lines derived from RPE express markers of pluripotency can differentiate into the three germ layers in vitro and can form teratomas in vivo. This adds to the growing list of somatic cell types that can be reprogrammed to pluripotency. In mouse, iPS cell lines have been obtained from somatic derivatives of all three germ layers: fibroblasts (mesoderm) , liver and stomach epithelial cells (endoderm) , and ectodermal keratinocytes . Human iPS have been derived from keratinocytes , fetal lung epithelial cells  neural precursors , and fibroblasts . Efficacy of reprogramming varies greatly with cell type: human keratinocytes are reportedly reprogrammed 100-fold more efficiently than fibroblasts , whereas mature mouse B cells required the addition or knockdown of additional TFs to obtain iPS lines . The reprogramming efficiency we observed for RPE cell (0.0018%) was similar to that reported by Yu et al. (0.0063%) . It may be possible to increase the efficacy using additional TFs (c-MYC, KLF4)  or with improved methods and new vectors for TF expression [49, 50]. In previous studies, LIN28 integration was detected in most iPS cell lines and its inclusion increased the frequency of reprogramming . Interestingly, in our study, there was no LIN28 integration detected in three of the six reprogrammed iPS cell lines. Lines that had integrated LIN28 had a higher tendency to differentiate in culture, and required more frequent passaging. Further investigation of additional lines will be necessary to confirm the correlation between LIN28 integration and propensity for differentiation. Moreover, this result might suggest that the different TFs combinations may have different effects depending on the cell type being reprogrammed.
An important, outstanding question about iPS cells is whether they retain a memory of their origin. Reprogramming a somatic cell to an ES-like state is a tall order, requiring a reboot that wipes clean the “epigenetic memory” that defines the differentiated cell . In studies of nuclear transplantation into oocytes, epigenetic memory persists in some donor nuclei, and this has been traced in part to activity of a histone H3 variant . Although early reports focused on iPS similarities to ESCs, recent detailed examination of iPS cells has showed that iPS cells retain gene expression patterns of the cell of origin. Chin et al.  carried out an extensive, genome-wide study of human and mouse ES and iPS lines and found that iPS lines derived from fibroblasts express higher levels of fibroblast transcripts not highly expressed in ESCs. These authors found that 80% of the differences between late passage iPS and ESCs could be attributed to a combination of incomplete suppression of fibroblast transcription and inadequate induction of ES genes. In another study, an iPS line from a reprogrammed neural stem cell (immortalized with c-MYC) retained expression of neural genes that were incompletely repressed . How this retention of expression patterns affects iPS cell differentiation remained to be determined.
We used spontaneous redifferentiation as an assay for memory in iPS cells. Because RPE are pigmented, it allows for a simple assay for the detection of regenerated RPE cells in a mixed population. Transcription of RPE marker genes was also quantified to assay RPE differentiation in iPS cells cultures. We tested the hypothesis that iPS derived from RPE would retain information that would lead them to redifferentiate back into RPE-like cells. It was shown that some but not all lines showed a higher tendency to differentiate down the RPE lineage. In the A3 line, this property is particularly striking, with 5- to 10-fold more pigmented cell colonies generated compared with ES or other iPS lines [29, 30]. iPS lines derived from other tissues do not show this tendancy. iPS-IMR90-3 give rise to a low number of pigmented colonies, and iPS-foreskin-2 did not form significant numbers of pigmented colonies. Feng et al.  reported that iPS-IMR90-1 and iPS-foreskin-2 did not form RPE colonies that could be propagated. Most recently, Liao et al.  examined six iPS lines generated from endodermal and neuroectodermal lineages and found very low production of RPE colonies. Furthermore, other iPS lines driven to retinal and RPE phenotypes using factors were reported to be similar to hES . hES lines we have examined (H1, H7, and hES-3) also produce fewer RPE than H14 (data not shown).
The redifferentiated RPE cells were isolated and characterized and shown to express RPE transcripts and proteins, and they have the capacity for phagocytosis. These cells are similar to those derived previously from hESC  or iPS cells . One derivative RPE line (from iPS A3) was cultured without passage for 430 days, with stable retention of RPE morphology. Cells were capable of phagocytosis and retained pigmentation after as many as four passages. This is in contrast to the results reported by Feng et al. , who observed that iPS-derived RPE grew slowly and failed to repigment after one passage. It seems likely that RPE derived from different iPS lines will have different properties of proliferation and senescence. Future studies will determine the nature of epigenetic patterns in multiple RPE lines redifferentiated from iPS-RPE, if they show differences in proliferation and senescence. It will be interesting to ask whether the redifferentiated RPE differs from the original RPE cells in any way after the developmental exercise of dedifferentiation followed by redifferentiation.
The molecular basis underlying the differentiation preference of the RPE-derived iPS cells is not understood. We hypothesize that some epigenetic imprints might remain in the reprogrammed iPS cells, which spontaneously express genes important in the early stages of RPE differentiation such as MITF and OTX2. We studied methylation states of RPE marker genes using bisulfite sequencing but did not detect any significant difference between ESC lines and RPE-derived iPS cell lines. A comprehensive epigenetic characterization of these iPS cell lines at different stages of reprogramming and differentiation will be required to elucidate whether these iPS lines retain some lineage-specific imprints.
Previous studies have identified protocols that use culture conditions to encourage differentiation of RPE cells from ES or iPS cells. After addition of factors in a temporal sequence that mimics in vivo development, it was reported that 37.4% ± 4.6% of ESCs colonies and 26.0% ± 4.0% of iPS cell colonies (reprogrammed from fibroblasts using Yamanaka factors) showed RPE cell morphology on day 60 [38, 39]. We have been able to obtain even higher differentiation frequencies from the iPS lines derived from RPE without addition of differentiation-promoting factors (63.8% ± 10.7% on day 60). It should be possible to generate large numbers of RPE cells for study in vitro without the need for extensive expansion and passage. Future experiments will determine if adult RPE can be reprogrammed without vector integration, which might provide a source of cells for therapeutic applications.
Although this article was under review, Kim et al.  and Polo et al.  reported studies showing that the cell type of origin influences the transcriptional profile, the epigenetics and differentiation potential of mouse iPS cells. However, both groups characterized early passage iPS cells and Polo et al. found that such differences were absent in late passage cells (p10–16). In contrast, the iPS-RPE lines we describe here retain memory even after extensive passage. The reasons for the persistence of memory are not known. Furthermore, although Kim et al. found that the iPS lines tended to differentiate back into the cell of origin, levels of differentiation were not as high as that of ESCs. In our study, the iPS-RPE made significantly more RPE than hESC.
We demonstrate that human RPE cells can be reprogrammed to iPS cells, some of which manifest a memory of their prior state in their tendency to redifferentiate into RPE. These results suggest that iPS cell lines from different origins may exhibit distinct differentiation preferences. Furthermore, a general strategy for iPS generation is suggested, where, if possible, it might be fruitful to reprogram the cell type ultimately desired for later use.
We thank Dr. James Thomson, Dr. Junying Yu, and Dr. Sherry Hikita for reagents and advice. This work was supported by a training grant from the California Institute for Regenerative Medicine (CIRM; T3-00009; Q.H.) and a grant from the US Army Research Office via the Institute for Creative Biotechnologies at UCSB (UARC 2/W911NF-09-D-0001). Cell culture was carried out in the UCSB Laboratory for Stem Cell Biology and Engineering, supported by a grant from CIRM (CL1-00521-1) and gifts from private individuals.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.