Long-Term Safety and Function of RPE from Human Embryonic Stem Cells in Preclinical Models of Macular Degeneration


  • Author contributions: B.L.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; C.M.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; S.W.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing and final approval of manuscript; S.G.: collection and/or assembly of data, data analysis and interpretation; P.F.: conception; L.L.: conception and design, collection and/or assembly of data, data analysis and interpretation; R. Lanza: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing and final approval of manuscript; R. Lund: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing and final approval of manuscript. B.L., C.M., and S.W. contributed equally to this article.

  • First published online in STEM CELLS EXPRESS June 11, 2009; available online without subscription through the open access option.


Assessments of safety and efficacy are crucial before human ESC (hESC) therapies can move into the clinic. Two important early potential hESC applications are the use of retinal pigment epithelium (RPE) for the treatment of age-related macular degeneration and Stargardt disease, an untreatable form of macular dystrophy that leads to early-onset blindness. Here we show long-term functional rescue using hESC-derived RPE in both the RCS rat and Elov14 mouse, which are animal models of retinal degeneration and Stargardt, respectively. Good Manufacturing Practice-compliant hESC-RPE survived subretinal transplantation in RCS rats for prolonged periods (>220 days). The cells sustained visual function and photoreceptor integrity in a dose-dependent fashion without teratoma formation or untoward pathological reactions. Near-normal functional measurements were recorded at >60 days survival in RCS rats. To further address safety concerns, a Good Laboratory Practice-compliant study was carried out in the NIH III immune-deficient mouse model. Long-term data (spanning the life of the animals) showed no gross or microscopic evidence of teratoma/tumor formation after subretinal hESC-RPE transplantation. These results suggest that hESCs could serve as a potentially safe and inexhaustible source of RPE for the efficacious treatment of a range of retinal degenerative diseases. STEM CELLS 2009;27:2126–2135


Since their discovery over a decade ago [1], human ESCs (hESCs) have been considered a promising source of replacement cells for clinical studies. However, numerous problems continue to plague clinical translation, including the risk of teratoma formation and the need for powerful drugs to overcome the problem of immune rejection. Until somatic cell nuclear transfer or induced pluripotent stem cell technology is further developed, diseases affecting the eye—a relative immunoprivileged site [2]—are likely to be among the first hESC-derived therapies. In the retina, compromised retinal pigment epithelial cell (RPE) function can lead to deteriorated vision and photoreceptor loss in both age-related macular degeneration and some forms of retinitis pigmentosa [3–7]. Although treatment options are presently extremely limited, there is evidence in animals that implantation of RPE cells can sustain photoreceptors and control deterioration of visual function [8–13].

However, despite preliminary data generated using RPE generated from hESCs [9], the transposition of scientific protocols from basic research to clinical application is precluded by the need to adhere to guidelines set forth by the U.S. Food and Drug Administration, collectively referred to as current Good Manufacturing Practices (cGMPs) and current Good Tissue Practices (cGTPs). In the context of clinical manufacturing of a cell therapy product, such as hESC-derived RPE, cGTPs govern donor consent, traceability, and infectious disease screening, whereas the cGMPs are relevant to the facility, processes, testing, and practices to produce a consistently safe and effective product for human use.

Here we augmented previously published RPE-derivation protocols [9], implementing appropriate controls around the cell manufacturing and documentation process to achieve full compliance with the cGMPs/cGTPs. hESC-RPE cell lines were prepared for clinical application and tested for safety and efficacy in several different in vitro and animal models. Three RPE batches from two hESC lines were tested in the RCS rat, an animal in which vision deteriorates over several months because of an RPE functional defect [14, 15]. We focused on longevity of effect, dosing effect, and evidence of untoward pathology. Based on these functional data, an hESC-RPE batch was chosen for a GLP safety study using the NIH III mouse model, an immune-deficient animal model with eye pigmentation but devoid of natural killer (NK) cells and mature T and B cells. Importantly, we also explored whether vision can be rescued in the Elovl4 mouse model for Stargardt's disease (STGD) [16]. STGD is a form of early-onset macular dystrophy that leads to blindness despite an intact Bruch's membrane, which is critical for RPE cell attachment and potential clinical efficacy [17].


Animals and Experimental Designs

Pigmented dystrophic RCS rats (n = 79) and ELOVL4 mice (n = 28) were used in the main experiments. NIH III immuno-nude mice (n = 45) were used for safety study. For RCS rats, animals were divided into five groups according to the doses they received. They were 5 × 103 (5,000)/eye (n = 21), 2 × 104 (20,000)/eye (n = 21), and 5 × 104 (50,000)/eye (n = 21). Animals from all dosage groups received cells with low medium and high pigmentation. All above the dosage group animals were received cells with low, medium and high pigmentation (Table 1). For further comparison, two groups were added: one group of animals (n = 8) received 7.5 × 104 (75,000)/eye cells and another group (n = 8) received 1 × 105 (100,000)/eye cells with medium pigmentation. For ELOVL4 mice, the eyes received 5 × 104 (50,000)/eye cells with medium pigmentation. All animals in the main experiments were maintained on oral cyclosporine A administered in the drinking water (210 mg/l, resulting blood concentration of ∼300 μg/l [12]) from 1 day before transplantation until they were sacrificed. An intraperitoneal injection of dexamethasone was given for 2 weeks (1.6 mg/kg/day) after surgery in cell and control injected rats and for 2 weeks alone in untreated animals. All animals were maintained under a 12-hour light/dark cycle. The animals were housed and handled with the authorization and supervision of the Institutional Animal Care Committee at the Oregon Health and Science University.

Table 1. Number of eyes treated at each pigment level
inline image

Experiments were carried out in accordance with the guidelines laid down by the NIH regarding the care and use of animal for experimental procedures and ARVO Statement for the Use of Animals in Ophthalmic and Vision Research.

Cell Preparation

Culture of hESCs and Differentiation into Mature RPE Cells.

All cell manufacturing procedures were carried out in ISO Class 5 biosafety cabinets in a fully validated, state-of-the-art ISO Class 7 clean room facility under strict environmental control monitoring systems and a routine microbial testing regimen. Single-blastomere hESC lines MA01 and MA09 were maintained as previously reported [18]. hESCs were dissociated from the primary mouse embryonic fibroblast layer by treatment with 0.05% trypsin-EDTA (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) and were seeded in 6-well low-attachment plates to allow EB formation in a chemically defined minimal essential medium (MEM)-based medium (MDBK-GM; Sigma, St. Louis, http://www.sigmaaldrich.com) containing B-27 supplement (Invitrogen) for approximately 7 days and plated on gelatin-coated (0.1%; Stem Cell Technologies, Vancouver, BC, Canada, http://www.stemcell.com) dishes until RPE colonies were visible. RPE was purified by 3-hour exposure to 4 mg/ml type IV collagenase (Invitrogen) and manually isolated with a glass pipette. Purified RPE was seeded onto gelatin-coated tissue culture plates and expanded in EGM-2 medium (Lonza Biologics, Basel, Switzerland, http://www.lonzabiologics.com) until desired density was achieved, at which point cultures were reverted to MEM-based medium (MDBK-MM; Sigma) and cultured until the appropriate phenotype was achieved. RPE was dissociated from culture using a 1:1 mixture of 0.25% trypsin-EDTA and Hanks-based cell dissociation buffer (Invitrogen) and was cryopreserved in 90% fetal bovine serum (HyClone, Logan, UT, http://www.hyclone.com) and 10% dimethylsulfoxide (Sigma).

Quantitative, Real-Time, Reverse Transcription-Polymerase Chain Reaction.

RNA was extracted from the cells using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. Eluted RNA was quantitated by spectrophotometry, and 10 μg was subjected to DNase digestion, followed by a reverse transcription reaction using a QuantiTech reverse transcription kit (Qiagen, Hilden, Germany, http://www1.qiagen.com) with a mixture of oligo-dT and random hexamers primers. Fifty nanograms per well of cDNA was used as templates in quantitative polymerase chain reactions (qPCRs) with oligonucleotides specific for hESC and retinal genes (supporting information Table 1). All qPCR reactions were performed in triplicate, with the resultant values being combined into an average threshold cycle (CT). The efficiency of qPCR was calculated from the slope of a relative standard curve using GAPDH primers. Relative quantization was determined using a Stratagene MX3005P QPCR system measuring real-time SYBR Green (PerfeCTa SuperMix, low ROX; Quanta Biosciences, Gaithersburg, MD, http://www.quantabio.com) fluorescence and calculated by the ΔΔCT method as described recently. Fold differences are calculated using the ΔΔCT in the formula 2 – ΔΔCt. Expression profiles for the mRNA transcripts are shown as fold differences in comparison to mRNA levels in hESCs.

Microarray Gene Expression Profiling.

Global gene expression analysis was performed using the human Affymetrix HG-U133 Plus 2.0 microarray platform (Santa Clara, CA, http://www.affymetrix.com) on both of the single blastomere-derived hESC lines MA01 and MA09 and the resulting RPE cells derived from each. Additionally, fetal RPE (donated by Dr. Marco Zarbin), ARPE-19 (American Type Culture Collection, Manassas, VA, http://www.atcc.org), and retinoblastoma (American Type Culture Collection) cell lines were analyzed as controls. See supporting information data for statistical analysis methods.

Western Blot Analysis.

Immunoblot analysis was carried out using standard SDS-PAGE methods using the Bio-Rad Mini-Protean and Mini-Transblot Cell (Hercules, CA, http://www.bio-rad.com). The protein bands were visualized using Western Lightning Chemiluminescence Reagent (PerkinElmer Life and Analytical Sciences, Boston, http://www.perkinelmer.com) and a Kodak 4000MM digital imaging station (Rochester, NY, http://www.kodak.com). Antibodies used were purchased from Abcam (DPPA4 and TDGF1; Cambridge, U.K., http://www.abcam.com), Chemicon (b-actin, CHX-10, Otx2, REX1, RPE65, and all horseradish peroxidase-conjugated secondary antibodies; Temecula, CA, http://www.chemicon.com), Covance (Pax6; Princeton, NJ, http://www.covance.com), Novus Biologicals (Bestrophin, CRALBP, and Pax2; Littleton, CO, http://www.novusbio.com), and Santa Cruz Biotechnology (MitF, Nanog, Oct4, PEDF, and Tyr; Santa Cruz, CA, http://www.scbt.com).

Transplantation Protocol.

Before cell transplantation, cells were thawed and washed in balance salt solution (BSS) and suspended in BSS. Three cell lines designated low, medium, and high pigment were given in different dose groups. These are summarized in Table 1. Using techniques described previously [19], a suspension of cells was delivered into the subretinal space of one eye through a small scleral incision, suspended in 2 μl of BSS medium (Invitrogen) using a fine glass pipette (internal diameter, 75-150 μm) attached by tubing to a 25-μl Hamilton syringe (Hamilton, Reno, NV, http://www.hamiltoncompany.com). The cornea was punctured to reduce intraocular pressure and to limit the efflux of cells. A sham-surgery group was treated the same way, except the carrying medium alone was injected. Pigmented dystrophic RCS rats received unilateral subretinal injections of the cell lines (n = 79 eyes) at P21; control rats received sham alone (n = 35 eyes) or were untreated (n = 29 eyes). Elovl4 mice at P28 received cells (n = 12 eyes), sham alone (n = 8 eyes), or were untreated (n = 8 eyes). Immediately after injection, the fundus was examined for retinal damage or signs of vascular distress. Any animal showing such problems was removed from the study and excluded from the final animal counts.

Spatial Visual Acuity.

Animals were tested for spatial visual acuity using an optometry testing apparatus (CerebralMechanics, Lethbridge, Canada, http://www.cerebralmechanics.com) comprising four computer monitors arranged in a square, which projected a virtual three-dimensional space of a rotating cylinder lined with a vertical sine wave grating. Unrestrained animals were placed on a platform in the center of the square, where they tracked the grating with reflexive head movements. The spatial frequency of the grating was clamped at the viewing position by recentering the “cylinder” on the animal's head. The acuity threshold was quantified by increasing the spatial frequency of the grating using a psychophysics staircase progression until the following response was lost, thus defining the acuity. Rats were tested from P60 to P240 at monthly intervals. Elovl4 mice were also tested in this apparatus at 3, 5, 7, and 11 weeks after surgery.

Luminance Threshold.

This was studied to provide a different measure of function from the spatial acuity and was achieved by recording single and multiunit activity close to the surface of the superior colliculus (SC) using glass-coated tungsten electrodes (resistance: 0.5 MΩ; bandpass 500 Hz to 5 KHz) with previously described procedures [20]. Recordings were made only in rats, selected on the basis of good and representative optomotor results: mice were not examined with this test. The brightness of a 5° spot was varied using neutral density filters (minimum steps of 0.1 log unit) over a baseline level of 5.2 log units until a response double the background activity was obtained: this was defined as the threshold level for that point on the visual field. A total of 15-20 positions were recorded from each SC. All animals were recorded at approximately P100, and some were studied again at a second time point at approximately P190. Data are expressed as a graph of percentage of SC area with a luminance threshold below defined levels and as raw results.


At the end of functional tests, all animals were killed with an overdose of sodium pentobarbital (Sigma) and perfused with phosphate-buffered saline. The eyes were removed, immersed in 2% paraformaldehyde for 1 hour, infiltrated with sucrose, embedded in optical cutting temperature, and cut into 10-μm horizontal sections on a cryostat. Four sections (50 μm apart) were collected per slide, providing five series of every fourth section collected. One was stained with cresyl violet for assessing the injection site and integrity of retinal lamination. The remaining slides were used for antibody staining, following previous protocols [21], and were examined by regular and confocal microscopy.

Safety Study.

Cells were prepared and transplanted using the same methodology described above for the RCS rat study. A minimum of six NIH III mice per group were injected with either hESCs or hESC-derived RPE from the MA09 single blastomere cell line in three time-based cohort groups (n = 36). The animals were killed by CO2 inhalation followed by exsanguination at 1, 3, and 9+ months based on cohort. Three negative control animals were also put in the study for each cohort (n = 9). Life study assessments included routine clinical assessments and body weight analysis, plus pre-sacrifice clinical chemistry (data not shown). Post mortem, eyes were removed and immersed in cold 4% paraformaldehyde, for up to 1 week. The tissue was embedded in paraffin and sectioned. Select slides were stained with hematoxylin and eosin. Slides were examined microscopically to assess retinal lamination and tumor formation.


Differentiation and Characterization of hESC-Derived RPE

Human RPE cells were generated using a cGMP-compliant cellular manufacturing process. Three different batches of RPE were created from each blastomere-derived hESC line based on morphological assessment of pigmentation (Fig. 1), an important indicator of RPE maturation.

Figure 1.

In vitro maturation and degree of pigmentation in different batches of human ESC (hESC)-derived retinal pigment epithelium cells. hESCs were matured to yield (A) light (L1), (B) medium (L2), and (C) heavy (L3) pigmentation levels. (A): Phase contrast image; scale bar = 200 μm. (B and C): Hoffman modulation contrast image; scale bar = 100 μm.

With the current process in place, each production run generated approximately 50 × 106 hRPE cells from a single frozen ampule of 1 × 106 hESCs. This represents the lower limit to our manufacturing capabilities, yet is still sufficient to dose approximately 500 rats or 50-100 human subjects. Additionally, our process (outlined in the Materials and Methods section) is completely suitable to available scale-up technologies such as bioreactor culture or large-scale fluid handling systems.

To characterize the developmental stages during RPE differentiation, several assays were used to identify the expression levels of genes key to each stage of development. qPCR was developed to provide a quantitative and relative measurement of the abundance of cell type-specific mRNA transcripts associated with the RPE differentiation process. A panel of genes associated with hESC pluripotency (Oct-4, Nanog, Rex-1, TDGF1, Sox2, DPPA2, and DPPA4), neuroectoderm intermediates (Pax6 and Chx10), and RPE (RPE-65, Bestrophin, CRALBP, PEDF, MitF, Otx-2, Tyr, and Pax2) was established and assayed for each by qPCR. With regard to quality control of cellular manufacturing, the marked decrease in all stem-related genes and concomitant increase in all retinal-associated genes, at a level of 10- to 100-fold, was deemed acceptable release criteria (in conjunction with the additional testing described below).

Figure 2 shows the gene expression profile of the transcripts during differentiation to mature RPE, including samples from hESCs (d0), embryoid bodies (EBs, d7), plated EBs (d14), mixed population of newly formed RPE and less differentiated cells (mixed, d28), purified early RPE (eRPE, d35), and fully matured pigmented RPE (mRPE, d56). A progressive decrease in the expression level of hESC-specific genes (Fig. 2A) was accompanied by an increase in the level of neuroectoderm and RPE-specific genes. Lightly pigmented RPE (Fig. 1) expressed 1,000-fold lower quantities of Oct-4, Nanog, Sox2, and DPPA4; <10,000-fold less TDGF1; and 50-fold less Rex-1 and DPPA2 than hESC. The cells also expressed 10- to 100-fold greater quantities of RPE65, CRALBP, PEDF, Bestrophin, Pax6, and MitF and expressed >100,000,000-fold Tyr, a downstream target of MitF/Otx2 in RPE. The finding that this cell population expresses genes such as Pax6 and CHX10 is not entirely surprising, because this stage represents an “immature” population of hRPE derived from embryonic cells, and as such, may continue to express markers associated with developing cells of the neuretina and/or neurectoderm.

Figure 2.

Comparative assessment of hESC-RPE cells using real-time polymerase chain reaction (PCR) and Western blot analyses. (A): Reverse transcription-PCR analysis of genes specific to hESCs, neuroectoderm, and terminally differentiated RPE cells examined throughout the in vitro differentiation process. Time points correspond to hESCs, EBs, plated EBs representing early intermediates (EB/RPE), a mixed population of cells containing newly differentiated hRPE cells, remaining progenitors (Mixed), purified eRPE (corresponding to Fig. 1A), and fully-mature mRPE (corresponding to Fig. 1C). (B): Western blot analysis of hESC-specific and RPE-specific markers. APRE-19 cells (top lane) show an inconclusive pattern of proteomic marker expression. Actin is used as protein loading control. RPE (bottom lane) derived from hESCs (middle lane) do not express the hESC-specific proteins Oct-4, Nanog, Rex-1, TDGF1, and DPPA4. However, they express RPE65, CRALBP, PEDF, Bestrophin, PAX6, Pax2, Otx2, MitF, and Tyr; all markers of differentiated RPE. Abbreviations: hESC, human ESC; RPE, retinal pigment epithelium; eRPE, early stage RPE; mRPE, mature RPE.

The phenotypic changes that RPE undergoes during the in vitro maturation process were characterized by qPCR (Fig. 1A–1C). Figure 2A shows that RPE with a higher degree of pigmentation and polygonal cell borders (corresponding to Fig. 1C) maintains higher expression of RPE-specific genes. Notably, both pigmentation and the high level of RPE-specific gene expression are correlated with the emergence of Pax2 expression and a sharp increase in MitF, Otx2, and Tyr expression. This finding is consistent with previous reports that MitF expression, and in turn Tyr, is achieved in RPE through synergy of Pax2 and Pax6 during embryonic development [22].

Proteomic Validation of Selected Transcripts in hESC-Derived RPE

To verify that genes of interest were expressed at the protein level, all targets of the initial transcriptional profile panel were assayed by Western analysis. As an internal control, hESC-derived RPE was compared with the well-studied ARPE-19 cell line by both qPCR and Western analysis. Figure 2A shows that, although hESC-RPE expresses similar levels of RPE-specific transcripts to ARPE-19, the hESC-RPE seems to express more abundant levels of these proteins (RPE65, PEDF, Pax2, and Bestrophin). Additionally, proteins expressed by hESCs are all downregulated in the final differentiated cell product. This disappearance of stem-related proteins (by immunoblot) and concomitant emergence of retinal-associated proteins is considered necessary to determine final product identity and is essential for lot release under our current production and quality scheme.

Bioinformatic Analysis of Global Gene Expression in hESC-RPE

The biological relevance of the morphological changes observed in vitro were assessed by gene expression profiling and subsequent informatic analysis of both hESC lines, each with three different morphologies: a control and several “reference” cell lines on the human Affymetrix HG-U133 Plus 2.0 microarray platform. Figure 3 shows a principal component analysis (PCA) scatter plot, indicating the contribution to variance that the two major variables, cell type and cell line (x- and y-axis, respectively), yield on global gene expression. A linear progression was observed from the undifferentiated (hESCs) state through the three levels of RPE pigmentation. Interestingly, the depigmented RPE cells (light green markers and Fig. 1A) cluster closer to both ARPE-19 (orange marker) and fetal RPE (black marker); the latter display strikingly similar morphological characteristics to this batch of cells in vitro (data not shown). The more heavily pigmented batches of RPE cells (red and purple markers) appear to cluster farther from hESCs (green markers) and retinoblastoma cells (RB, blue marker) than any other cell type tested. Whereas the pigmented batches of hRPE from MA01 and MA09 do not overlap by PCA, they are within a similar order of magnitude to each other to that of fetal RPE and ARPE-19. Whereas the majority of genes associated with the RPE state are similar in expression between these two groups when evaluated by qPCR (data presented above and not shown), a more detailed study into which gene expression values—by microarray—are generating this result is underway, Regardless, taken together, these data suggest that the more heavily pigmented hESC-RPE cells are the most differentiated, and from a safety standpoint, the most genetically divergent from cells possessing “stemness” or expressing cancer-related genes.

Figure 3.

Principal components analysis plot. Component 1 represents 69% of the variability represents the cell type, whereas component 2 represents the cell line (i.e., genetic variability). A near-linear scatter of gene expression profiles characterizes the developmental ontogeny of hRPE derived from hESCs. Abbreviations: hESC, human ESC; RPE, retinal pigment epithelium.

Pathogen Testing and Stability of hRPE

The most important criterion to consider in release of a product for clinical application is product safety. To ensure that our hRPE product was free of contamination during the extensive culture and differentiation process, we carried out the following testing according to U.S. Food and Drug Administration and International Conference on Harmonization guidelines for applicable microbial and viral agents on our master cell bank: United States Pharmacopeia membrane filtration sterility, fluorochrome-based mycoplasma, transmission electron microscopy for viral particles, in vitro tissue culture safety testing for adventitious agents, in vivo inapparent virus detection, PCR-based reverse transcriptase detection, HIV-1, HIV-2, HBV, HCV, CMV, HTLV-1 and -2, parvovirus B19, Epstein-Barr virus, and herpesvirus 6. Additionally, the master cell bank was cytogenetically analyzed by G-banding karyotype analysis. Results confirmed that these cell lines are karyotypically stable with no presence of infectious pathogens (outsourced data not shown).

Dosing Studies in RCS Rats

The effect of different doses on efficacy was titrated using the optomotor response as an indicator. The results at P90 (70 days after transplantation) are summarized in Figure 4A. Improved rescue of spatial acuity occurred from 5,000 to 50,000, after which even doubling the dose of cells to 100,000 had no significant effect on efficacy. Best performers among the cell-injected group gave a figure of 0.536 cycles/degree (c/d) compared with 0.6 c/d in normal rats, which is approximately 90% of normal value. There was no significant difference between sham and untreated groups, which performed significantly worse than the cell-injected group (p < .01).

Figure 4.

(A): Visual acuity as measured by the optomotor response shows that animals treated with 5,000, 20,000, 50,000, 75,000, and 100,000 cells performed significantly better than those with sham injection and untreated controls (p < .01) at P90 days. The best performers gave a figure of 0.563 c/d (compared with 0.6 c/d in normal rat). It is interesting to note that the difference among the groups of 50,000, 75,000, and 100,000 is not significant (p > .05). (B): Visual acuity tested in Elovl4 mice at several time points after subretinal injection of human ESC-retinal pigment epithelium showed that cell-injected animals performed significantly better than medium-injected and untreated controls (p < .05). The best performers had a figure of 0.32 c/d at P63 (compared with 0.35 c/d in normal mice), whereas control animals had a figure of 0.28 c/d. (C–F): Luminance threshold responses recorded across the superior colliculus (SC); each curve (average ± SEM) shows the percent of retinal area (y-axis) where the visual threshold is less than the corresponding value at x-axis (log units, relative to background illumination 0.02 cd/m2). Cell-injected groups are significantly better than controls: the curves showed that 28% of the area in the SC in animals with the 20,000 dose (C) 45% with the 50,000 dose (D); 40% with the 75,000 dose (E); and only 3% in medium control had thresholds of 2.2 log units. The 100,000 in F seemed better, but the variance is rather high. Blue lines, cell-treated; red lines, medium control. Abbreviation: c/d, cycles/degree.

Luminance thresholds were also measured in a subset of rats selected by their performance on the optomotor response. The area with best sensitivity corresponded to the area of retina in which the cells were introduced, as indicated in Figure 4C–4F. For statistical comparison, however, we presented the data for this part of the study as a percentage of the area of the visual field representation from which thresholds better than designated levels were recorded without regard to position. This gives a simple indicator of overall efficacy, as well as a best response figure, dissociated from spatial considerations. It is clear that the overall sensitivity recorded at 50,000 is superior to 20,000, but as with spatial acuity, it does not change significantly between 50,000 and 75,000. For example, approximately 45% of the SC gave thresholds of 2.2 log units with 50,000 cells/eye and approximately 40% with 75,000 cells/eye. Although the mean response levels at 100,000 were better and gave more long-lasting rescue than did lower doses (Figs. 4 and 5), the high variance precluded significance.

Figure 5.

Changes in acuity and luminance threshold with time. (A): Batch and longevity of effect as measured by visual acuity: cell-injected groups at all the time points (P60–P240) had significantly higher visual acuities than controls (p < .01); however, there is no difference with difference pigment levels (p > .05). (B and C): Raw data from an animal that received cell injection: luminance threshold responses were recorded at (B) P98 and (C) P187 in the same rat from multiple points within the superior colliculus (SC). This method quantifies functional sensitivity to light across the visual field of the eye. The topographical map depicts the luminance threshold responses (measured in log units relative to background illumination of 0.02 cd/m2) at 15 and 16 points in the left and right sides, respectively, within the SC. (B): All points of luminance threshold responses in the treated side are less than 2.0 log units, whereas in the untreated side, all points are greater than 2.3 log units. (C): The same animal was recorded at P187 (>5 months after surgery); there is deterioration in sensitivity to light compared to P98; however, it is still much better than the untreated fellow eye (which has no response over half the area). Abbreviation: c/d, cycles/degree.

Batch and Longevity of Effect in RCS Rats

Although slight differences in optomotor acuity were seen between the different pigment levels (Table 1), they were not significant. In contrast, there was a significant difference at all time points studied between the cell-injected groups and medium-injected and untreated controls (Fig. 5A). Over time, there was a reduction in acuity response for all the cell groups and dose levels. There was no evidence, however, that any one group showed a preferential rescue or that a higher density of cells might lead to slower deterioration over time.

To examine how luminance responses deteriorated with time, thresholds were recorded at two time points in individual rats. An example is shown in Figure 5B, 5C. As shown, the luminance thresholds show serious deterioration on the untreated side, with more than one half the area being nonresponsive at P187 compared with P98, whereas responsiveness is still sensitive on the cell-injected side, although some reduction in thresholds has occurred (0.7 log units at P98 vs. 1.0 log units at P187).

Efficacy in Elovl4 Mice

Visual acuity in normal mice tested by the same optomotor device was lower than that in rats (0.35 vs. 0.6 c/d). In untreated Elovl4 mice, the visual acuity deteriorated as photoreceptor degeneration progressed from 0.34 c/d at P28 to 0.24 c/d at P105 (Fig. 4B). Subretinal injection of hESC-RPE improved the visual acuity over controls at all time points tested. Cell-injected eyes had a figure of 0.32 ± 0.04 c/d at P63 (5 weeks after surgery) compared with 0.26 ± 0.03 c/d in sham-injected and untreated controls (Fig. 4B). Statistical analysis indicated that the difference between cell-injected and controls was significant (t test, p < .05). Because there is significant variability in the degree of photoreceptor survival across the untreated Elovl4 retina at the terminal time point, photoreceptor rescue was hard to document. Further studies will explore morphological rescue at later time points against background levels when degeneration is more advanced.

Histological Examination of RCS Rats

General Retina Structure.

Retinal sections from cell-injected, sham, untreated, and normal control rats were stained with cresyl violet and examined under light microscopy. At P90, compared with normal control (Fig. 6A), the cell-injected retina had five to six layers of photoreceptors (Fig. 6B), whereas the untreated retina had only a single layer remaining (Fig. 6C). In accordance with the functional results, the 5,000/eye doses had slightly better photoreceptor rescue (Fig. 6D) than sham-operated (primarily with localized photoreceptor rescue around injection site, data not shown), whereas the 20,000/eye produced better photoreceptor rescue (data not shown). The 50,000/eye and greater doses gave consistent photoreceptor rescue, covering a larger area of the retina (Fig. 6E, 6F) with preserved cones. At P150, cell-injected retinas still had an outer nuclear layer two to three cells deep, and the inner retina lamination was not disrupted. In contrast, both untreated and sham-operated retinas showed a typical secondary pathology, including abnormal vascular formation, RPE cells, and inner retinal neurons migrating along abnormal vessels, leading to distortion of retina lamination [19, 20]. At P240, cell-injected retinas still had an outer nuclear layer of one to two cells deep, and the inner retina still showed an orderly lamination. In contrast, advanced degeneration was very evident in control retinas: the inner nuclear layer became very irregular in thickness, ranging from one layer to multiple layers; RPE cells had migrated into the inner retina; and abnormal blood vessels were seen (Fig. 6I).

Figure 6.

Histological examination of cell-injected and untreated RCS retinas, showing photoreceptors in (A) normal, (B) cell injected, and (C) untreated eyes at P90 (arrows in B point to rescued photoreceptors; arrows in C indicate remaining photoreceptors). (D–F): Photoreceptors rescued at (D) 5,000 and (E and F) 50,000 dose (arrows in E indicate rescued photoreceptors; cone arrestin showed rescued cone photoreceptors in F). (G): Immunofluorescence- and (H) immunohistochemical-stained human specific antibody showing donor cells (arrows) formed a layer closely contact with the host RPE layer at P240. (I): Typical untreated retina at P240 with disorganized retinal lamination (left arrow indicates RPE cells migrating into inner retina; right arrow indicates disrupted inner nuclear layer). Scale bars = 25 μm. Abbrevations: INL: inner nuclear layer; IPL: inner plexiform layer; ONL: outer nuclear layer; RPE, retinal pigment epithelium; RGC: retinal ganglion cells.

Antibody Staining.

The human specific nuclear marker, MAB1281 (Chemicon) was used to identify the donor cells. They formed a layer, one to two cells deep, and integrated into the host RPE layer (Fig. 6G, 6H), as was seen in our previous study [9]. Photoreceptor rescue continued beyond the limits of distribution of donor cells, suggesting that rescue was at least in part caused by a diffusible effect. Cone arrestin antibody showed that cone photoreceptors were preserved with disorganized segments (Fig. 6F) at P90. Donor cells were still evident up to at least P249, the longest time point studied (Fig. 6G, 6H). There was no indication of continued donor cell division (shown by the proliferating cell nuclear antigen marker; data not shown).

Safety Assessment

Studies in NIH III Mice.

The long-term risk of teratoma formation was tested in the NIH III mouse model. The NIH III mouse was chosen for its immune-deficient status; the nude mouse has three mutations rendering it devoid of T cells, NK cells, and mature T-independent B lymphocytes. However, the NIH III mouse retains eye pigmentation, which provides better visualization for subretinal transplantation surgery. The surgical technique was the same as performed in the RCS study. The study compared the hESC-RPE to undifferentiated hESCs (positive control) to determine the teratoma formation potential of the 100,000 RPE cell dose over three time points: 1, 3, and 9 months (the approximate lifespan of the animal; n = 6 per cohort). In contrast to the animals that received undifferentiated hESCs, no teratoma or tumor formation was found in any of the animals injected with the hESC-derived RPE. In addition, basic animal safety assessments were normal compared with controls.

Absence of Tumorigenic Growth in RCS Rats.

In the RCS rat transplant study, none of the 79 cell-injected retinas examined, including the longest time points, showed any evidence of uncontrolled cell proliferation. There was no evidence of teratoma and/or tumor formation.


We showed the long-term safety and efficacy of hESC-derived RPE cells produced under manufacturing conditions applicable for use in human clinical trials. In addition to the development of assays with qualified range limits (which constituted the “identity” of the final hRPE product), extensive pathogen testing was carried out to ensure that the manufacturing procedure did not introduce any infectious diseases or adventitious agents into the final product. To confirm the functionality of these GMP-compliant cells, both dose-response and long-term efficacy were evaluated in homologous models of human retinal disease. Because of the proliferative nature of hESCs, evidence of safety under Good Laboratory Practice conditions (21CFR58) is imperative for translating hESC-derived cellular products into the clinic. The extensive characterization detailed above provides assurance of cellular identity, whereas the long-term tumorigenicity study presented here provides strong evidence that the hESC-RPE cells are safe and do not form teratomas and/or tumors during the lifetime of NIH III immune-deficient mice. After introduction to the subretinal space of RCS rats, the hESC-derived cells also survived for more than 8 months without evidence of pathological consequences. Importantly, the hESC-RPE cells also rescued visual functions in a dose-dependent fashion: with increased cell concentrations from 5,000 to 50,000, there was an improvement in functional rescue measured with both visual acuity and luminance threshold response. From 50,000 to 100,000, there was no obvious change in efficacy, which suggests that there is some tolerance in numbers of cells introduced and that twice the optimal dose is still effective. This is particularly important for phase I clinical trials, because extrapolation to humans is difficult to make. Previous rodent work [21] has shown that RPE cells quickly disperse as a single or double layer and that 20,000 cells of an immortalized RPE cell line can occupy at best approximately 20% of the retinal area (12.56 mm2). For the age-related macular degeneration retina, the inner macular is 3 mm in diameter: this would mean that a dose of approximately 40,000 cells would be needed to cover the inner macular area but that a larger cell number would most likely cover a larger area and would not be detrimental.

The significantly improved visual performance in Elovl4 mice adds to the value of the hESC-RPE as the cell choice for cell-based therapy to treat macular disease (in this case, a subset of patients with Stargardt's disease caused by mutation in the Elovl4 gene) [16, 25]. Stargardt's macular dystrophy is one of the most frequent forms of juvenile macular degeneration. Although some rescue can be achieved by growth factor delivery such as direct injection [26] or factor-releasing cells (encapsulated cells) such as ARPE19 cells transduced to produce ciliary neurotrophic factor [27, 28] or Schwann cells [29], these approaches cannot replace the other functions of RPE cells. Because hESC-RPE cells have a molecular profile more closely resembling native RPE than do ARPE-19 [18], they may be able to take on a broader range of RPE functions than ARPE-19 beyond simple factor delivery. The mechanism whereby hESC-RPE cells rescue vision is not totally clear. One possibility might be that the hESC-RPE cells replace crucial functions of the host RPE, because a previous study has shown, for example, that hESC-RPE cells are able to phagocytose latex beads in vitro [18]. However, because the location of photoreceptor rescue extends beyond the area of donor cell distribution, part of the rescue effect is likely also to be mediated by a diffusible trophic factor effect.

In this study, the deterioration of function with time may be related to the inadequacy of the immunosuppressive protocol used in long-term discordant (human-to-rodent) xenografts. Previous studies have shown that that donor cells decline even under triple immunosuppression [30]. Another confounding factor is that, in RCS rats, leaky blood vessels appear during the course of degeneration [23, 24], which may abrogate the immune-privileged status of the eye and compromise cyclosporine efficacy. It is not known whether some form of immunosuppression (possibly local and/or topical administration using eyedrops) might be needed when the present cells are used in a clinical setting (under those circumstances, they would be close allogeneic rather than the xenogeneic grafts of this study). An alternative is that transplants might need to be repeated to obtain a sustained therapeutic benefit. It is very encouraging, however, that even at P240 (i.e., 220 days after transplantation), some donor cells survive, photoreceptors are rescued, and a level of visual function is preserved.


In summary, we showed the long-term safety of hESC-derived RPE cells in immune-deficient animals, as well as their long-term function in two different animal models of disease using GMP conditions suitable for clinical trials. This represents an important step toward the possibility of using hESC-derived cells to slow the progress of specific macular degenerative diseases.


We thank Benjamin Cottam and Jie Duan for technical support.


Raymond Lund and Peter Francis acted as consultants for and were employed by Advanced Cell Technology.