Research ArticleRetinal Disease

Clinical-grade stem cell–derived retinal pigment epithelium patch rescues retinal degeneration in rodents and pigs

See allHide authors and affiliations

Science Translational Medicine  16 Jan 2019:
Vol. 11, Issue 475, eaat5580
DOI: 10.1126/scitranslmed.aat5580

A pipeline for retinal stem cell therapy

Autologous induced pluripotent stem cell (iPSC)–derived retinal pigment epithelium (RPE) transplantation has been shown to improve visual function in animal models of age-related macular degeneration (AMD) and is currently being tested in human patients. However, oncogenic mutations might occur during the cell reprogramming process. Now, Sharma et al. used CD34+ peripheral blood cells from patients with AMD to generate oncogenic mutation-free iPSCs. These cells were used for the production of clinical-grade RPE cell patches. Transplantation of the RPE patches in rodent and pig models of retinal degeneration showed therapeutic effects. The authors suggest that the production process presented here might accelerate the development of safer iPSC-derived stem cell therapies.

Abstract

Considerable progress has been made in testing stem cell–derived retinal pigment epithelium (RPE) as a potential therapy for age-related macular degeneration (AMD). However, the recent reports of oncogenic mutations in induced pluripotent stem cells (iPSCs) underlie the need for robust manufacturing and functional validation of clinical-grade iPSC-derived RPE before transplantation. Here, we developed oncogenic mutation-free clinical-grade iPSCs from three AMD patients and differentiated them into clinical-grade iPSC-RPE patches on biodegradable scaffolds. Functional validation of clinical-grade iPSC-RPE patches revealed specific features that distinguished transplantable from nontransplantable patches. Compared to RPE cells in suspension, our biodegradable scaffold approach improved integration and functionality of RPE patches in rats and in a porcine laser-induced RPE injury model that mimics AMD-like eye conditions. Our results suggest that the in vitro and in vivo preclinical functional validation of iPSC-RPE patches developed here might ultimately be useful for evaluation and optimization of autologous iPSC-based therapies.

INTRODUCTION

Cell-based therapeutics offer the promise of “permanent” replacement of degenerative tissue. The eye is an appealing area of interest because of its ease of accessibility and the urgent need for effective therapies to help a growing elderly population experiencing vision loss (1, 2). The previous success with surgical procedures transplanting autologous retinal pigment epithelium (RPE)/choroid graft obtained from the periphery of the same patients’ eyes has provided the critical proof of principle needed to develop pluripotent stem cell–derived RPE-based cell therapies (3, 4). In recent preliminary clinical studies, embryonic stem cell (ESC)– and induced pluripotent stem cell (iPSC)–based therapies have been tested in patients with age-related macular degeneration (AMD), a leading cause of blindness among elderly (510). AMD has two advanced stages: “Dry” AMD or geographic atrophy (GA) is caused by death of the RPE, a monolayer of pigmented cells located in the back of the eye, and “wet” or choroidal neovascular AMD is caused by proliferation of choroidal vessels that penetrate through the RPE leaking fluid and blood under the retina (11, 12). Both conditions lead to photoreceptor cell death, causing serious vision loss and can lead to blindness.

To date, several studies have explored subretinal delivery of stem cell–derived RPE cells in AMD patients: An ESC-derived RPE cell suspension was tested in a phase 1 clinical trial of patients with GA (dry) stage of AMD (3, 4); an autologous iPSC-RPE (iRPE) sheet was recently transplanted in one patient with the wet form of AMD (6, 7); and ESC-derived RPE patch on a paralene scaffold was tested in four dry AMD patients, and a similar patch on polyester scaffold was tested in two wet AMD patients (9, 10). These landmark studies in the field of regenerative medicine provide evidence for the safety of stem cell–based therapies, but they also unveil potential barriers facing the current state of technology and highlight the need for further innovation to successfully bring commercially approved stem cell therapy to the clinic. For example, RPE cells in suspension do not self-organize into a confluent polarized monolayer or provide barrier function in the back of patients’ eyes, affecting the long-term survival of cells (13). Over time, patients’ immune cells can reject allogeneic ESC-derived RPE. Consistent with these observations, Schwartz et al. reported that ESC-RPE cell suspension graft was compromised in some patients (6). The use of autologous iPSCs has also raised concerns, for example, Mandai et al. (7) reported concerns about potential oncogenic changes in patient’s iPSCs that might limit their use. Furthermore, previous approaches that aimed to develop clinically compatible iRPE manufacturing processes have not provided sufficient evidence to functionally validate the clinical-grade iRPE patch derived from multiple patients (14). Recently, M’Barek et al. (15) used amniotic membrane as a substrate to deliver ESC-derived RPE monolayer in a rat model of retinal degeneration. Although this approach may increase RPE monolayer integration, evaluation in larger animal models and humans is needed to understand the translational potential of this method. These observations may limit the development of an optimal path toward using stem cell–derived RPE-based cell therapy for macular degeneration.

To address the above-mentioned concerns and to accelerate development of an autologous iPSC-derived RPE-based cell therapy for macular degeneration, we have optimized good manufacturing practice (GMP) or clinical-grade manufacturing of AMD patient–specific iPSC-derived RPE (iRPE) patch using a biodegradable scaffold and tested these patches in two different animal models. The autologous iRPE patch would avoid immune rejection by patient’s cells. These iRPE cells do not show any cellular phenotypes of AMD, suggesting that the autologous iRPE patch will likely survive and integrate in patients’ eyes. We decided to transplant the patch at the borders of dry AMD lesion to rescue photoreceptors in the transition zone, where RPE has atrophied and the photoreceptors are still alive (12).

In this study, we used CD34+ cells isolated from the peripheral blood of AMD patients that allowed us to develop oncogenic mutation-free clinical-grade iPSC banks. These iPSCs banks were used to develop a clinical-grade RPE differentiation process that was more efficient and reproducible as compared to the research-grade differentiation. We further provided evidence that iRPE patches derived from three AMD patients mature and function to a similar extent and did not show any cellular phenotypes of the disease. In addition, we demonstrated that seeding iRPE cells on a biodegradable substrate substantially improved integration of the RPE monolayer as compared to transplanted cell suspension in immunocompromised rats, in rats with RPE dysfunction–associated retinal degeneration, and in a laser-induced RPE injury pig model where the clinical dose of the iRPE patch was tested. These experiments provide a complete workflow for performing preclinical studies, supporting the feasibility of using clinical-grade autologous iRPE for treating retinal degeneration.

RESULTS

Clinical-grade triphasic protocol efficiently and reproducibly generates AMD-iRPE cells

Previous work suggested that iPSC derived from AMD patient skin fibroblasts under clinical-grade conditions attained chromosomal copy number changes, likely during the reprogramming process (7). Others have reported that skin fibroblasts are not considered an ideal starting source when deriving iPSCs from older patients (16). Therefore, we sought out to develop a manufacturing workflow for autologous iPSC-derived RPE patch from patient CD34+ peripheral blood cells. Here, we developed a pipeline for generation, functional validation, and in vivo testing of the clinical-grade AMD patient–specific iRPE patch (Fig. 1, A and B). We hypothesized that because of the progenitor and proliferative nature, CD34+ cells isolated from patients’ peripheral blood will provide an acceptable source for clinical-grade iPSC generation. Passage 10 iPSC banks were generated from CD34+ cells of three advanced dry AMD patients (85, 89, and 87 years of age) using a clinical-grade episomal reprogramming protocol (17). These iPSC banks were tested for the following: sterility; greater than 85% expression of stage-specific embryonic antigen 4 (SSEA4), tumor-related antigens (TRA1-60 and TRA1-81), and octamer-binding transcription factor 4 (OCT4); normal G-band karyotyping; loss of reprogramming plasmid; matched short-tandem repeat identity; and matched oncogene sequence to the patient (table S1). To determine whether CD34+ cell–derived iPSC acquired any sequence alterations during clinical-grade reprogramming and expansion, we sequenced coding regions of 223 oncogenes across the nine iPSC clones. Oncogene exome sequence of eight iPSC clones matched their respective donor peripheral blood mononuclear cells (PBMCs) with the exception of clone B from donor 2 (2B) that showed several sequence changes, as confirmed by kinship analysis of respective exome sequences (Fig. 1C and tables S1 and S2; dbGAP ID: SUB4785176). However, none of the sequence changes in iPSC clone D2B are associated with any known cancerous phenotype (table S2; Q2 Solutions), and no mutations were seen in the coding region of the p53 tumor suppressor gene (table S2; dbGAP ID: SUB4785176) (7, 18). Our work suggests that CD34+ cells are likely to produce iPSCs with minimal mutations during reprogramming and can be used for autologous iPSC-based therapies. Using the above criteria, three validated iPSC clones per donor [donor 2: clone A (2A), clone B (2B), and clone C (2C); donor 3: clone A (3A), clone C (3C), and clone D (3D); and donor 4: clone A (4A), clone B (4B), and clone C (4C)] were selected for iRPE patch manufacturing.

Fig. 1 Generation of clinical-grade iRPE cells.

(A) Workflow illustrating a pipeline to manufacture and test autologous clinical-grade iRPE patches with the goal of filing a phase 1 clinical trial investigational new drug (IND) application to the U.S. Food and Drug Administration. (B) Time line of clinical-grade iRPE differentiation. Clinical-grade iRPE differentiation takes 77 days, is initiated with monolayer iPSCs, and is performed using xeno-free reagents. NEIM, neuroectoderm induction medium; RPEIM, RPE induction medium; RPECM, RPE commitment medium; RPEGM, RPE growth medium; RPEMM, RPE maturation medium. (C) Coding region sequencing of 223 oncogenes at 2000× depth for all nine clinical-grade AMD iPSC clones. (D and E) Flow cytometry analysis of clinical-grade iRPE derived from three AMD patients, performed at the RPE progenitor stage (D17), RPE commitment stage (D27), and immature RPE stage (D42) (n = 6). Analysis of variance (ANOVA) was performed to determine changes in percent positive cells; ***P = 0.0001 for PAX6/MITF and ***P = 8.9 × 10−14 for MITF; Dunn’s test was performed for pairwise comparisons; P values: 2B/2C-3C/3D = 0.909, 2B/2C-4B/4C = 0.400, and 3C/3D-4B/4C = 0.319. ns, not significant. (F) RPE-specific gene expression from D5 to D42 of clinical-grade iRPE differentiation (n = 6). Dunn’s test was performed for pairwise comparisons; P values: 2B/2C-3C/3D = 0.721, 2B/2C-4B/4C = 0.719, and 3C/3D-4B/4C = 0.999.

Earlier work has shown that RPE differentiation can be induced in stem cell–derived neuroectoderm cells by the activation of transforming growth factor (TGF) or canonical wingless/integrated (WNT) pathways (1922). To further improve the efficiency and reproducibility of differentiation and to make iRPE manufacturing clinically compatible, we optimized a triphasic differentiation protocol (fig. S1A and table S3): (i) Based on previous observations that dual-SMAD (suppressor of mothers against decapentaplegic) inhibition promotes neuronal fate and fibroblast growth factor (FGF) pathway activation inhibits RPE phenotype (2326), we combined low level of dual-SMAD and FGF inhibition to promote iPSCs into RPE-primed neuroectoderm cells and increased differentiation efficiency to 81 ± 7% as compared to cultures containing FGF2 and no SMAD inhibitor that had a differentiation efficiency of 24 ± 20% (fig. S1, B to E); (ii) activation of TGF or WNT pathways induced committed RPE fate in these RPE-primed neuroectoderm at 81 ± 7% or 83 ± 8.5%, respectively (fig. S1F) (19, 23, 27); and (iii) committed RPE cells were matured by inducing primary cilium with prostaglandin E2 (PGE2) treatment in cells to actively suppress canonical WNT pathway, resulting in pigmentation in derived RPE cells (fig. S1G) (28). Overall, this triphasic protocol led to efficient iRPE differentiation, with 96% cells expressing the RPE progenitor gene PAX6 (paired box protein 6) and 71% cells expressing the RPE commitment gene MITF (microphthalmia-associated transcription factor) (fig. S1, H and I).

To test the reproducibility of the clinical-grade differentiation protocol across multiple iPSC clones and multiple patients, we tested the protocol on iPSCs derived from all three AMD patients (two iPSC clones per patient). We used single-cell [two-dimensional (2D)] differentiation of iPSCs that induced epithelial morphology in cells by day 12 (D12), prompting a shift to RPECM (fig. S1J). By D17, more than 80% of cells in four of six clones coexpressed PAX6/MITF and a small percentage expressed only MITF, confirming the RPE-primed stage of neuroectoderm cells (Fig. 1D). By D27, the number of PAX6/MITF double-positive RPE progenitors dropped to 30 to 40% and the number of MITF only–positive cells increased (20 to 60% across all six clones), suggesting a shift to the committed RPE population. By D42, more than 80% of cells across all six iPSC clones expressed MITF with a concomitant loss of PAX6/MITF double-positive population, confirming the shift to immature RPE phenotype (Fig. 1D) (19). Results with PAX6- and MITF-expressing cells were further corroborated by the analysis of downstream known targets of these genes (19, 23). A considerable percentage of cells across all six iPSC clones expressed RPE progenitor markers premelanosome protein 17 (PMEL17) (40 to 85%) and tyrosine-related protein 1 (TYRP1) (20 to 60%) at D17 and a negligible number of cells expressed RPE maturity markers cellular retinaldehyde-binding protein (CRALBP) and bestrophin-1 (BEST1) (Fig. 1E). Because the cells continue to mature, the expression of PMEL17 and TYRP1 remained stable at D27 and increased to more than 99% in all six iPSC clones by D42 after cell enrichment, confirming cell purity (Fig. 1E). In comparison, CRALBP- and BEST1-positive cells continued to increase over time, reaching more than 95% for CRALBP and more than 70% for BEST1 across all six iPSC clones (Fig. 1E). The resultant RPE cells had no detectable iPSC (no OCT4 or TRA1-81+ cells; fig. S1, K and L). Expression analysis for genes involved in RPE pigmentation [glycoprotein Nmb (GPNMB) and tyrosinase (TYR)], visual cycle [aldehyde dehydrogenase 1 family member A3 (ALDH1A3), transient receptor potential cation channel subfamily M member 1 (TRPM1), RPE-specific 65-kDa protein (RPE65)], and RPE maturation (RPE65 and BEST1) (29) confirmed that all six AMD-iPSC clones differentiated with similar efficiency and progressively attained maturity, underscoring reproducibility of the clinical-grade differentiation process (Fig. 1F). Furthermore, manufacturing process reproducibility was confirmed across four different users (fig. S1, M to O). Overall, the clinical-grade protocol leads to an efficient generation of RPE cells that express several functional and maturation genes.

Biodegradable scaffold helps clinical-grade AMD-iRPE cells to functionally mature into a monolayer tissue

We hypothesized that a biodegradable scaffold would provide suitable material for RPE cells to secrete extracellular matrix (ECM) to form a polarized monolayer. Because the scaffold degrades, ECM and cells would constitute a native-like RPE tissue that would enhance the possibility of long-term integration of the iRPE patch in patients’ eyes. Scaffolds used in the clinical-grade process were manufactured using poly(lactic-co-glycolic acid) (PLGA) (50:50 lactic acid/glycolic acid, four midpoint 1.0 dl/g), with 350 nm mean fiber diameter previously shown to be optimal for RPE growth (30, 31). A single-layer heat-fused nanofiber scaffold was selected for iRPE patch manufacturing because of its high Young’s modulus that correlated with the ease of transplantation (Fig. 2, A and B). The scaffold completely degraded in 80 to 90 days [scanning electron microscopy confirmed scaffold thickness at D49 (10 μm), D56 (5 μm), D63 (2 to 4 μm), and D80 to 90 (complete degradation); fig. S2, A to H]. iRPE derived from all three patients matured and polarized on the scaffold, as confirmed by high RPE65 and GPNMB expression and basal distribution of ECM proteins COLLAGEN IV and COLLAGEN VIII [representative donor 3 clone C (D3C) shown in Fig. 2C]. This suggested that iRPE cells synthesized a de novo Bruch’s membrane equivalent, supporting our hypothesis that the 3D architecture of PLGA scaffold would trigger the secretion of ECM proteins by iRPE cells.

Fig. 2 Generation of functionally mature AMD iRPE patch.

(A) Young’s modulus of different PLGA scaffolds. **P < 0.01, two-tailed t test. (B) SEM showing surface topology of single-layer fused PLGA scaffold. (C) Top panel: Representative immunostaining for mature RPE marker RPE65 (red) and human-specific antigen STEM121 (green). Middle panel: RPE pigmentation protein GPNMB (red) and Bruch’s membrane protein COLLAGEN IV (green). Bottom panel: COLLAGEN VIII and Bruch’s membrane marker (red) (n = 3). DAPI, 4′,6-diamidino-2-phenylindole. (D) Representative transmission electron microscopy of iRPE patch on transwell membrane or the PLGA scaffold. Basal infoldings can be seen in the case of PLGA scaffold (inset) (n = 3). (E) ∆CT values of RPE-specific genes are displayed for all eight iRPE patches from three AMD donors (n = 8). Dunn’s test was performed to determine pairwise comparisons; P values: 2B/2C-3C/3D = 0.999, 2B/2C-4B/4C = 0.150, and 3C/3D-4B/4C = 0.094. (F) Live TER measurement during the last 3 weeks (D54 to 77) of iRPE patch maturation. Representative data from three clones (3A, 3D, and 4A) are displayed (n = 8). Dunn’s test was performed to determine changes in TER overtime; P values: 3A-3D = 0.630, 3A-4A = 0.845, and 3D-4A = 0.968. (G) Graphs show phagocytosis ratio for AMD iRPE patches (n = 8). Dunn’s test was performed to compare iRPE from different donors; P values: 2B/2C-3A/3C/3D = 0.005, 2B/2C-4A/4B/4C = 0.395, and 3A/3C/3D -4A/4B/4C = 0.63. (H and I) PCA combining data from the following assays (morphometric, gene expression, TER, and phagocytosis) showing variation between clones across PC1. PCA was performed based on k-nearest neighbors, and bootstrap hierarchical clustering was performed to determine differences between iRPE samples; *P < 0.05. (H) PCA plotted without D2B (I) (n = 7 to 8).

Previously, primary RPE and iRPE monolayer maturity and functionality were validated on semipermeable transwell (polyester) membranes (28, 32). To determine whether the AMD patient–derived clinical-grade iRPE patch on a PLGA scaffold and transwell behaved similarly, we made a structural, molecular, and functional comparison between the D3C iRPE patch on two surfaces. Transmission and scanning electron microscopy confirmed the presence of dense apical processes with apically located melanosomes and tight junctions between neighboring cells; basal infoldings were only detected on PLGA scaffold grown cells (Fig. 2D, inset, and fig. S3, A and B). Consistent with structural similarity between the two patch types, iRPE patch on PLGA scaffolds and polyester membranes demonstrated similar electrical properties (fig. S3, C and D) and similar expression of key RPE-specific genes [oculocutaneous albinism (OCA2), GPNMB, TYRP1, TRPM1, ALDH1A3, RPE65, and BEST1] (fig. S3E). Together, these results demonstrate that iRPE cells mature similarly on a polyester membrane or PLGA scaffold.

Our previous work and the work of other investigators suggested that donor genetics is the biggest source of variation in cell types derived from iPSCs (33, 34). To determine whether such variation also exists in clinical-grade iRPE patches derived from different patients and to develop criterion to functionally validate iRPE patches before transplantation in patients, we derived iRPE patches from all eight iPSC clones from the three donors. We defined a performance metric for “hexagonality” to measure how close the iRPE patches were to an ideal convex regular hexagonal pattern (see Materials and Methods). Quantitative morphological assessment of iRPE patches performed on images obtained with tight junction stain (ZO-1; fig. S3, F to H) revealed similar hexagonality across all eight iRPE patches (hexagonality score, 8.1 ± 0.1; out of 10; fig. S3I), suggesting similar epithelial phenotype of iRPE patches from all three patients. Gene expression analysis of RPE patches from all eight iPSC clones demonstrated similar expression of RPE markers and suggested similar maturity of RPE monolayer across different patient iRPE cells (Fig. 2E). This conclusion was further corroborated by transepithelial resistance (TER) measurement during the last 3 weeks of iRPE patch maturation from clones D3A, D3D, and D4A. All three iRPE patches demonstrated progressively increasing TER (250 to 1000 ohms⋅cm2), suggesting gradual maturity of the iRPE patch (Fig. 2F). Consistently, the iRPE patch from six of eight iPSC clones phagocytosed POS (photoreceptor outer segments) and secreted VEGF (vascular endothelial growth factor) in a polarized fashion (Fig. 2G and fig. S3J), highlighting variation across different samples. To determine the main source of variation, we performed a principal components analysis (PCA) for data obtained from morphometric, gene expression, TER, and phagocytosis assays (Fig. 2H). The iRPE patch derived from D2B iPSC clone that differs in sequence from its donor also showed significant variation in functional output (Fig. 2H and fig. S3K), leading us to speculate that sequence changes likely account for variation in functional output from this sample. PCA performed by excluding D2B data revealed that the functional output of iRPE patches clustered by donors (Fig. 2I), suggesting that patient genetics is likely a major source of variation in iRPE patch function. Overall, this validation exercise suggests that a combination of sequencing, molecular, and functional readouts can help validate transplantable iRPE patches.

Purity of differentiated cells in the final product is one of the primary goals in developing a stem cell therapy. We hypothesized that iPSCs cannot survive culture conditions used for iRPE patch maturation. To check this possibility, we performed an in vitro spiking study. RPE cells mixed with 100%, 10%, 1%, or 0% iPSCs were seeded and cultured on PLGA scaffolds for 35 days. Flow cytometry confirmed that more than 90% iPSCs had died within 2 days of culture and no iPSCs could be detected after D14 on PLGA scaffolds (fig. S4A). Gene expression analysis also confirmed the absence of iPSCs, non-RPE cells, and the absence of non-RPE lineage markers in RPE cells in all cultures (except, as expected, for 100% iPSC; fig. S4B). In conclusion, the in vitro spiking experiment confirmed our hypothesis that iPSCs do not survive RPE maturation conditions and the iPSCs or non-RPE cells could not be detected in the iRPE patch.

Clinical-grade AMD iRPE patch safely integrates in the eye and shows improved efficacy over cell suspension in a rodent preclinical study

To test the long-term integration and the safety profile of the AMD iRPE patch, we transplanted a 0.5-mm-diameter (~2500 cells) clinical-grade patch in the subretinal space of immunocompromised (Crl:NIH-Foxn1rnu) rat eyes (fig. S5, A and B). Fundus infrared imaging and optical coherence tomography (OCT) 10 weeks after surgery confirmed successful integration of the patch under the host retina (Fig. 3, A and B, horizontal white line and red arrowhead, and Table 1). Histological analysis confirmed the OCT data that the AMD iRPE patch completely integrated on rat’s Bruch’s membrane [STEM121 (red), human cells; Fig. 3C, arrowheads]. In contrast, injected iRPE cell suspension rarely integrated into the rat RPE (fig. S5C), consistent with previous observations that suspension cells do not form contiguous monolayer in the back of the eye [arrowheads; Fig. 3D, STEM121 (purple); Fig. 3E, STEM121 (red); fig. S5D, PMEL17 (red); and Table 1] (13). AMD-iRPE cells were negative for the Ki67 proliferation marker and no cases of tumor/teratoma were noted, whereas teratomas were observed in 3 of 10 eyes when pure iPSCs were injected in the subretinal space (Fig. 3F; fig. S5, E and F; and Table 1). No signs of systemic toxicity of the transplant were noted in rats, as measured by food consumption and body weight gain in all animals throughout the 10-week period (Table 1 and table S4), suggesting the safety of human iRPE cells as a transplantable patch. The data showing successful integration of iRPE patches suggest that using the scaffold might provide effective support for iRPE cells.

Fig. 3 Safety and efficacy assessment of clinical-grade AMD iRPE patch in rodent models.

(A to C) Representative en face infrared image (A), OCT (B), and immunohistochemistry for human antigen (C; STEM121, red) showing subretinal location and integration of the 0.5-mm-diameter clinical-grade AMD iRPE patch (red arrowheads). Black arrowheads mark rat RPE cells (see inset for higher magnification) in the subretinal space of immunocompromised rat at 10 weeks after surgery (n = 20). (D) Representative immunohistochemistry for STEM121 (purple) (red arrowheads, see inset for higher magnifications) confirms the presence of clinical-grade AMD iRPE cells injected in the rat eye. Note that purple color in POS is due to hematoxylin staining. Rat RPE cells are not positive for STEM121 (black arrowheads, see inset for higher magnification). (E) Representative STEM121 (red) immunostaining (red arrowheads) showing integration of a small number of human cells in the rat RPE (black arrowheads; see inset for higher magnification) (n = 10). (F) Representative Ki67 immunostaining showing lack of positivity. Human cells are indicated by red arrowheads (see inset for higher magnification; rat RPE is marked with black arrowheads). (G and I) Representative photomontage of RCS rat retina showing ONL (arrowheads) with transplanted iRPE patch (~10,000 cells on a 1-mm-diameter patch) (G) or iRPE cell suspension (100,000 cells) (I), compared with the degenerated ONL in nontransplanted areas (arrow, n = 10). (H and J) Representative immunofluorescence staining of iRPE patch (H) or iRPE cell suspension (J) implanted retina with ONL rescue (arrowheads) [red, human nuclear antigen (HuNu); green, human-specific anti-PMEL17]. Note that red arrowheads in (H) point to iRPE cells that likely dislodged from the scaffold during transplantation. (K) OKN tracking thresholds at p90. n = 10. *P < 0.05 and **P < 0.001, ANOVA.

Table 1 Summary of preclinical rat and pig studies performed to demonstrate safety and efficacy of clinical-grade AMD iPSC-RPE patch.

PR, photoreceptor; WT, wild-type; na, not available.

View this table:

To compare their efficacy, a 1-mm-diameter iRPE patch (~10,000 cells) or 100,000 iRPE cells in suspension were transplanted in a previously established Royal College of Surgeon (RCS) rat model (3538) between postnatal day 21 (p21) and p28. Balanced salt solution (BSS+) or empty scaffold was used as control. Both the iRPE patch and cell suspension rescued overlying photoreceptors, as suggested by increased thickness and higher number of cells in the photoreceptor outer nuclear layer (ONL) in the transplanted area as compared to the nontransplanted area [arrowheads mark human cells, human nuclear antigen (HuNu) (red), and human-specific PMEL17 (green); Fig. 3, G to J, and fig. S5G]. Optokinetic (OKN) measurements (39) confirmed that at p90, the iRPE patch and iRPE cell suspension transplanted animals showed similar recovery as compared to vehicle control animals (Fig. 3K and Table 1). Overall, these rat experiments indicate that the clinical-grade AMD iRPE patch maintains its monolayer architecture after transplantation in the subretinal space of rodent eye as opposed to cell suspension that shows limited ability to form a monolayer in the same environment. The dose of cells on the patch is 1/10th of the cells in suspension (10,000 cells on a 1-mm diameter-patch as compared to 100,000 iRPE cell suspension). Despite the 10-fold difference in relative dose, the iRPE patch and cell suspension showed similar recovery as revealed by OKN analysis and ONL thickness, suggesting that the AMD iRPE patch is more efficacious compared to the cell suspension.

The human clinical dose of the AMD iRPE patch integrates in the eye of a laser-induced RPE injury pig model and rescues degenerating retina

In RCS rat, the RPE monolayer is dysfunctional, but is still present, unlike what is seen in AMD patients (11). To test the AMD iRPE patch in an animal model with atrophied RPE, using the entire human clinical dose of 4 mm × 2 mm iRPE patch, we optimized laser-induced RPE ablation in pigs. We exploited a property of melanin to efficiently absorb 532-nm wavelength and used a micropulse laser to selectively injure the pig RPE (Fig. 4A) (40). RPE injury was targeted at the pig visual streak, which contains the highest density of cone photoreceptors (fig. S6A). OCT analysis of RPE/retina injury caused by 1% or 3% laser duty cycles (DC) at 330-ms exposure times revealed RPE detachment in 1% DC at 24 hours after laser and RPE thinning at 48 hours (Fig. 4, B and C), whereas 3% DC additionally caused subretinal fluid accumulation at 24 hours after laser and notable RPE/POS interface damage by 48 hours (fig. S6, B and C). Pre-laser multifocal electroretinography (mfERG) confirmed similar electrical response across the visual streak in the retina, whereas after laser both 1% and 3% DC laser–treated areas showed comparable reduction in mfERG signals (Fig. 4, D, dotted line, and E). OCT and mfERG results were confirmed at cellular levels by terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining combined with RPE65 and peanut agglutinin (PNA) labeling, which revealed apoptotic RPE and photoreceptor cells in both 1% and 3% DC laser–treated eyes, with higher apoptosis in photoreceptors at 3% as compared to 1% DC laser (Fig. 4, F and G, arrowheads, and fig. S6, D to F). Hematoxylin and eosin (H&E) staining further confirmed that both laser powers thermally damaged the RPE, but 1% DC caused less damage to POS at both 24 and 48 hours (Fig. 4, H and I, and fig. S6, G and H, arrowheads). Together, OCT, mfERG, and histology data suggest that 1% DC micropulse laser is preferred over 3% for inducing specific damage to pig RPE while maintaining retinal electrical responses.

Fig. 4 Development of a porcine iRPE patch efficacy model.

(A) Schematics of micropulse laser injuring the pig RPE. Inset: Fluorescein angiogram depicting laser-induced outer blood-retinal barrier breakdown. (B and C) Representative OCT images at 24 and 48 hours after laser (arrowheads indicate RPE thinning). n = 3. (D) Heatmap of the P1 values of the visual streak region after 1% or 3% DC laser (laser areas outlined with dashed lines). White-red indicates the highest P1 values, and blue indicates the lowest P1 values. (E) Average mfERG waveform from healthy (black), 1% (light green), and 3% (dark green) DC laser areas. (F to I) Representative immunostaining for TUNEL (green), RPE65 (yellow), and PNA (magenta) (F and G) and H&E staining (H and I) at 48 hours (arrowheads indicate apoptotic RPE). n = 3.

To deliver a 4 mm × 2 mm patch, a specific transplantation tool was designed with an S-shaped cannula that fits human (or pig) eye curvature and allows an easy delivery of the iRPE patch while maintaining its orientation (fig. S7, A and B, arrowhead). Surgery involved a four-port vitrectomy, posterior vitreous and retinal detachment, a 2.5-mm retinotomy, sclerotomy enlargement, and subretinal delivery of the iRPE patch loaded in the tool (fig. S7, C to F, arrowheads). Intraoperative OCT (iOCT) confirmed the correct subretinal delivery of the patch (fig. S7, G to I, arrowhead). We first tested bare PLGA scaffold without cells in the subretinal space of non-immunosuppressed, non–laser-injured pigs to determine whether PLGA degradation products (lactic acid and glycolic acid) caused any inflammation in the eye. OCT of pig eye 2 weeks after surgery confirmed that the empty scaffold can be delivered with minimal retinal damage (fig. S7J, arrowheads). Ten weeks after surgery, there were no signs of ocular inflammation (cloudiness, vitritis, or retinitis) as seen by OCT, but the POS layer appeared thinner and showed retinal tubulations, suggesting damage of photoreceptors—likely because the empty scaffold interfered with nutrient supply from the host RPE, the RPE-photoreceptor visual cycle, and POS phagocytosis (fig. S7K). Coinciding with scaffold degradation (5 weeks after surgery, fig. S7L), up to 80% of the N1P1 mfERG signal over the area of the implant had recovered, suggesting minimal damage caused by the empty scaffold (fig. S7M). These in vivo results are consistent with the lactic acid release profile of degrading PLGA scaffold. Eighty-three percent of total lactic acid from the scaffold was released during in vitro culture. This included about 70% of lactic acid that was released during the bulk degradation phase of our PLGA scaffold that started at week 3 and ended at the end of week 5 while scaffold was still in culture (table S5). Furthermore, the highest amount of lactic acid released by the scaffold during the bulk degradation phase (0.0074 ± 0.0014 mM per scaffold per day) was 310× less than the systemic concentration of lactic acid (2.3 mM) in blood (41) and 513× less than the lactic acid concentration in the eye at the RPE apical surface (42). Overall, these data confirmed that the PLGA scaffold is not inflammatory in the subretinal space of a pig eye and our transplantation tool safely delivered the patch.

Fundus imaging of pig eyes transplanted with a green fluorescent protein (GFP)–expressing iRPE patch confirmed that at least 70% of transplanted human cells survive over a 10-week period (fig. S8, A to C, arrowheads). This was achieved by suppressing the systemic and resident innate immune responses using prednisone, doxycycline, and minocycline, and the adaptive immune response using tacrolimus and sirolimus (4346). We tested the possibility that once the biodegradable PLGA scaffold degrades, the human clinical dose of the AMD patient–derived iRPE patch integrates on pig Bruch’s membrane and is functional. A 4 mm × 2 mm patch was transplanted in a pig eye over an area with laser-induced RPE ablation. OCT confirmed that over 10 weeks of follow-up, as the PLGA scaffold degraded, the clinical-grade iRPE patch integrated in the pig eye and the retina above the iRPE patch maintained both inner and outer retinal layers as compared to animals transplanted with an empty PLGA scaffold, where retinal tubulations were evident (Fig. 5, A to C; arrowhead in B). Immunostaining confirmed integration of the AMD iRPE patch in laser-injured pig eye and a mature phenotype of transplanted cells, as validated by strong RPE65 immunostaining in iRPE cells (green, STEM121; red, RPE65; Fig. 5, D to F, and fig. S8, D to F; retinal tabulation marked by an arrowhead in Fig. 5E). PNA staining confirmed improved organization of POS over iRPE patch transplanted retina as compared to empty scaffold transplanted retina (Fig. 5, D to F; PNA, white). To quantify differences in photoreceptor rescue between empty scaffold and iRPE patch transplants, we counted the number of photoreceptor nuclei in the ONL above the area of both transplants and compared results to adjacent healthy retina. Our analysis demonstrates that in the ONL over the empty scaffold, the number of nuclei is 42% of the adjacent healthy area, whereas over the iRPE patch area the number of nuclei is 73% of the healthy area (fig. S8G). To rule out the possibility that STEM121 labeling was caused by pig RPE phagocytosing human iRPE cells, we performed immunostaining for a nucleus-specific human antigen (STEM101) (fig. S8H). These data showed specific nuclear labeling only in a part of RPE, providing additional evidence of integration of the human iRPE patch in the pig eye. Because laser damage and subsequent iRPE patch transplantation were performed in the visual streak area of pig eye, we asked whether human RPE cells are able to preserve pig cone photoreceptors. Immunostaining for specific cone opsins (S, L, and M) confirmed the preservation of pig cone photoreceptors above the area of the iRPE patch and showed data comparable to the healthy retina (Fig. 5G, arrowhead and fig. S8I). To test functional integration of the human iRPE patch in the pig eye, we tested whether human RPE cells are able to phagocytose pig POS. Rhodopsin staining of healthy pig retina and retina transplanted with the clinical-grade human iRPE patch revealed phagocytosed POS inside human RPE cells, similar to what is seen for the native pig RPE (Fig. 5, H and I, arrowheads, and fig. S8J). Preservation of cone photoreceptors and functional integration of human RPE inside the pig eye prompted us to test recovery of electrical responses from laser-damaged pig retina over the area of iRPE patch. Heatmaps of mfERG responses showed improved signal over the iRPE patch transplanted laser-damaged visual streak area as compared to the empty scaffold transplanted pigs (Fig. 5, J to L). To address issues regarding regional variability in the placement and extent of laser injury on the effect of the iRPE patch, we used a linear mixed effect (LME) analysis of all the mfERG components. LME analysis of all mfERG components [N1, N1 width, P1, P1 width, N1P1, area under the curve (AUC), and scalar product] revealed a significant difference between the iRPE patch and the empty patch over 10 weeks (Fig. 5M). This observation was further confirmed in the linear regression analysis of AUC that also showed a significant difference between the two groups (Fig. 5N). In summary, these results confirmed that the clinical-grade AMD iRPE patch integrated with the pig retina and rescued pig photoreceptors after laser injury of pig RPE and iRPE patches from different patients demonstrate similar efficacy responses.

Fig. 5 Efficacy assessment of clinical-grade AMD iRPE patch in a porcine retinal degeneration model.

(A to C) Comparison of OCT from retina over a healthy region, retina transplanted with an empty PLGA scaffold, or retina transplanted with clinical-grade AMD iRPE patch (horizontal lines) (n = 3). (D to F) Immunostaining for STEM121 (green, arrowhead; F) and RPE65 (red) in the pig eye. PNA staining is shown in white; white arrowhead in (E) marks retinal tubulations (n = 3). (G) Immunostaining for red, blue, and green cone opsins (white; red arrowhead) and STEM121(green) in the pig eye after iRPE patch transplantation. (H and I) Rhodopsin (green) immunostaining shows phagocytosed (white arrowheads) POS by healthy pig RPE immunostained with RPE65 (red) and by human iRPE cells immunostained with STEM121 (red). Z sections show POS localization inside pig and human RPE cells (n = 3). (J to L) Heatmaps of N1P1 mfERG responses. (M and N) Average mfERG waveform (M) and mfERG data over 10 weeks of follow-up after surgery (N) (n = 3). LME was performed for data analysis, and ANOVA was used to determine statistical significance of the data. *P < 0.05.

To perform a comparative analysis between PLGA iRPE patch, transwell iRPE patch (nondegradable transwell membrane), iRPE cell suspension, and empty scaffolds, we tested all four transplants in pigs with laser-ablated RPE (fig. S9, A to I). Two weeks after surgery, OCT confirmed correct delivery of all four transplants with minimal signs of inflammation (Fig. 6, A to D). At 5 weeks after surgery, empty scaffold and iRPE suspension showed disruptions in the ONL and external limiting membrane of the retina and structures that were reminiscent of outer retinal tubulations often seen in degenerating retina (arrowheads in Fig. 6, E and H) (47). In contrast, OCT showed that retina overlying iRPE patches (both PLGA and transwell) did not exhibit any such retinal tubulations and both the ONL and external limiting membrane appeared intact (Fig. 6, F and G). OCT analysis was consistent across all three pigs evaluated for each treatment. Immunohistochemistry of the PLGA iRPE patch confirmed integration and posttransplantation maintenance of mature phenotype of human RPE cells in the back of the pig eye (compare Fig. 6, I to K; white, STEM121, human specific; red, RPE65), and maintenance of photoreceptors above the human iRPE patch but not above the empty scaffold or cell suspension transplants (green, PNA, photoreceptors; Fig. 6, I to L; fig. S9, J to O; and Table 1). In contrast, ONL above the empty scaffold shows tubulations and degeneration as seen in OCT (arrowheads in Fig. 6E). Furthermore, unlike the iRPE patch, iRPE cells in suspension lose the expression of the RPE maturity marker RPE65 (arrowhead in Fig. 6L). Consistent with the OCT and immunostaining data, mfERG confirmed higher recovery of mfERG individual waveforms and integrated data over 5 weeks in the lasered area in both PLGA iRPE patch and transwell iRPE patch, as compared to empty PLGA scaffold or iRPE cell suspension (Fig. 6, M and N). These results demonstrate that the monolayer iRPE patch is superior than cell suspension in rescuing retinal degeneration in laser-injured pig eyes.

Fig. 6 Integration of iRPE patch in a laser-induced retinal degeneration porcine model.

(A to H) Representative OCT images of pig eyes 2 and 5 weeks after transplantation with empty scaffold, PLGA iRPE patch, transwell iRPE patch, and iRPE cell suspension. iRPE transplants are indicated by red horizontal line. Green arrowheads point to retinal tubulations. (I to L) Representative immunostaining for human antigen STEM121 (green) and RPE65 (red) in the pig eye [iRPE patch is indicated by horizontal white line in (J); white arrowheads in (J) and (L) mark iRPE transplants; green arrowheads in (I) marks retinal tubulations; and red arrowheads in (J) and (K) mark rat photoreceptors]. n = 3. (M and N) Individual and average mfERG responses from different transplant conditions (n = 3; *P < 0.05, ANOVA).

DISCUSSION

A successful autologous cell therapy requires an efficient and reproducible manufacturing process that generates a safe and efficacious product. The clinical-grade manufacturing process we describe here meets all those criteria—it is efficient and reproducible, and generates safe and efficacious clinical product. The manufacturing process was developed using CD34+ cells of three advanced-stage AMD (GA) patients (17, 48, 49). We generated passage 10 working banks of clinical-grade iPSCs and validated up to three banks per patient for iPSC critical quality attributes. In addition to their pluripotency and correct G-band karyotyping, we focused on two safety attributes: loss of reprogramming plasmid and oncogene sequencing. All nine clones from three patients lost their reprogramming plasmid by passage 10 likely because we are using a low–copy number Epstein-Barr nuclear antigen (EBNA) origin of replication-based plasmid (27). Consistent with some of the published work, none of the iPSC clones acquired any oncogenic mutations during reprogramming or expansion, and eight of the nine iPSC clones had no observed sequence changes (50). Differences between our results and the two recent reports on potentially oncogenic changes in iPSCs probably stemmed from differences in starting cell type, reprogramming plasmid, and the reprogramming process (7, 18). One iPSC clone, D2B, which showed several sequence and copy number alterations, also showed variation in iRPE patch functional output. We speculate that these sequence changes are the reason for variation in functional output at the iRPE patch level; however, further investigation is needed. The present experiments lead us to propose that a combination of iPSC oncogene sequencing and functional analysis should be a necessary requirement for determining transplantable derivatives of iPSCs.

We also successfully demonstrated clinical-grade differentiation of iPSCs into a mature and polarized RPE patch on a biodegradable PLGA-based scaffold. iPSC differentiation into a transplantable RPE patch takes about 10 weeks, comparable to a recently published time line (14). The process described here is user independent and scalable to multiple iPSC clones. The exact source of variability we noted in some assays is unclear, but PCA suggested that patient genetics is likely the largest single contributing factor, consistent with previous work (33, 34). This work demonstrates that iRPE patches derived from different patients show different functional responses and that functional validation should be performed to determine the suitability of the iRPE patches for transplant.

As compared to ESC-RPE suspensions (5, 6), iRPE patches contain a fully polarized monolayer of cells that integrate into the Bruch’s membrane of immunocompromised rats and in pigs after laser-induced RPE injury. The integration probably reflects a coordination of continuous PLGA scaffold degradation and ECM production by iRPE, which facilitates integration with the host Bruch’s membrane. Our results show that iRPE cells on PLGA scaffolds synthesize Bruch’s membrane proteins COLLAGEN IV and COLLAGEN VIII. Furthermore, unlike cell suspension that may not be able to perform most RPE functions, the polarized RPE monolayer on PLGA iRPE patch and transwell iRPE patch performs a multitude of RPE functions. Combined together, these properties of the iRPE patch suggest a mode of action for improved efficacy seen with the patch approach. RPE ablation in laser-injured pig model is similar to the loss of RPE seen at the borders of GA lesion in advanced AMD eyes (12). Thus, we suggest that in AMD patients, integration of a transplanted iRPE patch will likely be made possible by “young” iRPE cells secreting metalloproteinases that can modify “aged” Bruch’s membrane (51). In comparison to our PLGA iRPE patch and transwell iRPE patch, RPE cell suspension only demonstrated occasional integration as suggested previously (52). These results provide a path forward with iRPE patches with improved integration and functionality at the borders of GA lesion in AMD patients eye, rescuing overlying photoreceptors from atrophy.

Although we recognize that both the RCS rat model and the pig laser-induced RPE injury model do not fully recapitulate AMD pathophysiology, these models have provided critical insight in the survival, integration, and potential efficacy of AMD patient–derived iRPE cells. Retinal function recovery with RPE suspension and iRPE patches was different in rodents and pigs. Although we did not perform a complete dose response study for cell suspension or the patch, we observed a similar degree of visual function rescue from both cell suspension (100,000 cells) and the iRPE patch (10,000 cells). This outcome was obtained likely because cells in suspension do not form an intact polarized monolayer in the back of the rat eye but rather behave as a chemical bioreactor that isotopically secretes neurotrophic factors. Unlike cell suspension, the entire RPE patch is a polarized cell monolayer that integrates into the rat eye and simultaneously serves the needs of the overlying photoreceptors while maintaining the integrity of its interface with the choroid. In contrast to the rat experiment, in pigs where an equal number of cells were transplanted in a 4 mm × 2 mm patch versus iRPE cell suspension (100,000 cells), higher protection of photoreceptors was seen with the patch. Furthermore, the clinical-grade iRPE patch from different AMD patients showed similar results, suggesting a reproducible manufacturing process.

One limitation of an autologous cell therapy is a fairly long manufacturing process that requires an elaborate set of reagents and quality control measures to ensure process consistency and reproducibly. Such long manufacturing process may increase clinical product cost when manufactured at a commercial scale. Furthermore, it is not clear how patient-specific iRPE cells will behave under diseased conditions. Patient-specific cells may more likely develop AMD cellular endophenotypes once transplanted in patients.

The present data help advance the field of stem cell–based therapies for macular degeneration by providing more robust clinical-grade manufacturing process and a RPE patch on a biodegradable scaffold for better integration and functionality. This works provides a complete framework for IND-enabling studies to initiate a phase 1 clinical trial using an autologous iRPE patch. It will help leverage and improve future iPSC-based trials by providing much needed evidence to assess the safety and subsequently the efficacy of autologous iRPE against AMD.

MATERIALS AND METHODS

Study design

The main objective of this study was to develop an autologous iPSC-based cell therapy for AMD. As part of this main objective, we developed a clinical-grade iRPE patch on a biodegradable scaffold from three AMD patients, functionally validated the patch, and tested safety and efficacy of the patch in preclinical animal models. For all in vitro experiments for validation of iRPE patch, six to eight different clones derived from three different AMD donors were used. All rat and pig experiments were repeated at least three times; many of the rat experiments were repeated up to 10 times to ensure adequate sample sizes (individual n is reported). Preterm morbidity rats were excluded from analysis. Blinding was performed for RCS rat OKN analysis. All outliers were reported and included in statistical analyses. Images for the replicates described in Figs. 2 to 6 are shown in file S1. Raw data for all the figures are reported in file S2 (main figures) and file S3 (supplementary figures).

Statistical analysis

Statistical analysis for in vitro functional validation of the iRPE patch was performed using R software and, where applicable, the dunn.test package. Data were first assessed for normality by determining data skewness, kurtosis, and q-q plots. All gene expression, TER, shape metrics, and phagocytosis data were found to have skewness or kurtosis values outside of a 1 to 1 range and showed significant deviance in q-q plots and thus were treated as non-normal. Dunn’s test was therefore used for reporting multiple pairwise comparisons after a Kruskal-Wallis test for stochastic dominance among k groups was performed. A Bonferroni-Dunn correction was used for all pairwise comparisons, and adjusted P values of <0.05 (*), <0.01 (**), and <0.001 (***) were considered significant. For flow cytometry data in Fig. 1D, quartile regression models and an ANOVA were used to determine differences in slope from 0. PCA data from phagocytosis, TER, and gene expression profiles across all days and clones were scaled from 0 to 1 using the total pooled data for each metric. PCA was performed, and clustering was shown based on k-nearest neighbors. Bootstrap hierarchical clustering of PCA was performed to show similarity between different iRPE samples from three donors (*P < 0.05).

LME and ANOVA were performed to determine statistical differences between different transplant groups in pigs. The equation function used in the MATLAB fitlme function was Data ~ Week + Group*Component + (1|Pig_Name). A linear regression of the AUC mfERG waveform component was also performed in GraphPad, and the elevation and y intercepts were determined. A P value of 0.05 or less was determined to be statistically significant for both the LME and the linear regression. ANOVA was used to perform statistics on OKN data. *P < 0.05 and **P < 0.01 were considered significant. To count the number of DAPI positive nuclei in the implant region in pig eyes, the data from empty scaffold and the iRPE patch areas was normalized to a healthy region data on the same section. The same distance across the retina (~400 μm) was used for the implant area and the corresponding healthy area to count. In RCS rats, a ~500-μm region was used for counting the number of nuclei in the ONL. An unpaired two-tailed t test was performed for TUNEL-positive nuclei, and the number of nuclei in ONL of RCS and pigs resulted in a significant difference between the groups (**P < 0.01 and ***P < 0.001).

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/11/475/eaat5580/DC1

Materials and Methods

Fig. S1. Optimization of iRPE differentiation.

Fig. S2. Degradation kinetics of PLGA scaffold.

Fig. S3. Evaluation of functionally mature iRPE patch.

Fig. S4. Assessment of iPSC survival in iRPE cultures.

Fig. S5. Safety and efficacy assessment of iRPE patch in rodents.

Fig. S6. Optimization of laser-induced RPE injury in pig eyes.

Fig. S7. Optimization of subretinal transplantation procedure in pigs.

Fig. S8. Analysis of iRPE patch in pig model of laser-induced retinal degeneration.

Fig. S9. Comparative efficacy analysis of iRPE patch and iRPE suspension in the pig model.

Table S1. Validation of clinical (GMP)–grade iPSC Working Bank derived from CD34+ cells.

Table S2. Oncogene exome analysis of iPSC versus donor PBMCs.

Table S3. Detailed list of clinical-grade reagents used in iPSC generation and RPE differentiation.

Table S4. Summary of body weight change in rats transplanted with iPSC-derived RPE.

Table S5. Lactic acid release analysis of in vitro degraded PLGA scaffolds.

File S1. Images for all replicates described in Figs. 2 to 6 (provided as separate Word file).

File S2. Raw data for the main figures (provided as separate Excel file).

File S3. Raw data for the supplementary figures (provided as separate Excel file).

References (53, 54)

REFERENCES AND NOTES