Molecular profiling of stem cell-derived retinal pigment epithelial cell differentiation established for clinical translation

Abstract

Human embryonic stem cell-derived retinal pigment epithelial cells (hESC-RPE) are a promising cell source to treat age-related macular degeneration (AMD). Despite several ongoing clinical studies, a detailed mapping of transient cellular states during in vitro differentiation has not been performed. Here, we conduct single-cell transcriptomic profiling of an hESC-RPE differentiation protocol that has been developed for clinical use. Differentiation progressed through a culture diversification recapitulating early embryonic development, whereby cells rapidly acquired a rostral embryo patterning signature before converging toward the RPE lineage. At intermediate steps, we identified and examined the potency of an NCAM1⁺ retinal progenitor population and showed the ability of the protocol to suppress non-RPE fates. We demonstrated that the method produces a pure RPE pool capable of maturing further after subretinal transplantation in a large-eyed animal model. Our evaluation of hESC-RPE differentiation supports the development of safe and efficient pluripotent stem cell-based therapies for AMD.

Keywords: Large-eyed model; age-related macular degeneration; cellular profiling and transcriptome; cellular therapy; clinical translation; differentiation protocol dynamics; human embryonic stem cell-derived retinal pigment epithelial cell; retinal progenitor cells; single-cell RNA sequencing; subretinal injection.

Figure 1

Global scRNA-seq characterization of hESC-RPE differentiation trajectory (A) Schematic of the hESC-RPE differentiation experimental protocol where scRNA-seq was performed at the seven time points (bolded; D, day) in three cell lines: HS980, KARO1, and E1C3. (B) Brightfield images during HS980 differentiation. Scale bars, 100 μm; inset scale bars, 20 μm. (C) Principal component (PC) representation of 26,615 single cells across three lines using 2,000 cv-mean enriched genes. (D) PC showing signature scores for pluripotency, retinal progenitors, and RPE cells. (E) Bar graphs showing average normalized gene expression of pluripotent, retinal progenitor, and RPE markers in scRNA-seq data. Error bars represent standard deviation of the mean across three lines, except for the hESC time point. (F) PC plot colored by cell line in red. (G) Plot showing cumulative explained variance curve for each time point and all lines, applied to estimate how much variance accumulates over sets of correlated genes (biological-driven variability), as opposed to uniformly across genes (white noise). (H) Line plots showing percentage of cells positive for retinal marker genes at each time point. (I) Line plots showing scRNA-seq-based cell-cycle phase assignment. Cycling: S and G₂/M; non-cycling: G₁/G₀. Intervals in (H) and (I) represent the 95% confidence intervals. See also Figure S1.

Figure 2

Evaluation of the diverse neuroepithelial cell type derivatives in early hESC-RPE differentiation (A) Uniform manifold approximation and projection (UMAP) at differentiation day 7 (D7) and D14 in three lines. Cells were grouped into retinal progenitor (RetProg), lateral neural fold-like (LatNeEp), pre-placodal-like (Pre-Plac), cranial neural crest-like (CrNeCr), mesenchyme (MesCh), pluripotent (Pluri), and endoderm-like (Endo) clusters. (B) UMAPs showing normalized gene expression of marker genes RAX (RetProg), DLX5 (LatNeEp), FOXE3 (Pre-Plac), FOXC1 (CrNeCr), HAND1 (MesCh), NANOG (Pluri), and SOX17 (Endo). (C) UMAPs in (A) colored by cell line. (D) Bar graphs showing cell type composition in each line at D7 and D14. (E) Enriched gene expression heatmap for HS980 cell types. (F) Plots showing relative expression of neural tube patterning markers in D7 (top) and D14 (bottom) cells across pseudospace in HS980. (G) Schematic of the patterned anterior neural plate at the neurulation stage. Left: putative location of the cell types corresponding to identified clusters. Right: schematic of genes patterning the rostral embryo. (H) UMAP of hESC-RPE differentiation D30 in three lines, colored by cell type. (I) Pseudotime trajectory of all D30 RPE and RetProg cells (82.5% of total at D30). (J) Scatter plots showing progenitor (SOX2, RAX, VSX2, LHX2), early (MITF, TYRP1, PMEL), mid (TYR, RLBP1), and late (RPE65, BEST1, TTR) gene expression along pseudotime. See also Figure S2.

Figure 3

Comparative analysis of RPE induction between hESC-RPE, 3D EB differentiation, and human embryonic eye (A) UMAP of 3D EB cultures at D14 (HS980 line). (B and C) UMAPs showing signature scores for brain regions (B) and neural tube cell types (C) visualized on the EB D14 UMAP. (D) Bar plots comparing cell type compositions in 2D and 3D cultures at D14 (top) and D28/30 (bottom). (E) UMAP of 3D EB cultures at D28 colored by cell type. (F) Projection of 2D D30 cells from all three cell lines onto the UMAP from (E) using pairwise correlation distances, colored by annotated cell type (see supplemental experimental procedures, cf. Figure 2H). Cells in gray are those from (E). (G) UMAP of human embryonic optic cup cells at Carnegie stages 12, 13, 14, and 15 (week 5, W5), colored by cell type (left) or stage (right). (H) Heatmap of enriched gene expression by cell type across all samples in (G). (I) Heatmap showing signature scores of in vitro cell clusters at D7, D14, and D30 illustrating the correspondence to in vivo clusters from (G). Signature scores were obtained using the top 30 genes of the respective in vivo reference population. See also Figure S3.

Figure 4

Characterization of the NCAM1-High sorted D30 hESC-RPE population (A) Bar graph of top genes from anticorrelation analysis at HS980 D30. Genes with a mean normalized expression <0.5 were excluded. (B) Brightfield and immunofluorescence stainings of D30 cells showing co-expression of VSX2, NCAM1, and Ki67 markers. Scale bars, 200 μm. (C) Representative FACS plot of NCAM1-CD140b sorting to distinguish distinct populations at D30. Negative gates were set based on fluorescence minus one (FMO) and hESC control samples. (D) Post-sort pellets of CD140b-High and NCAM1-High cells. (E) UMAP of NCAM1-High (pink), CD140b-High (blue), and unsorted (gray) D30 cells after CCA integration. (F) Dot plot illustrating the proportion of cells corresponding to each identified cell type in scRNA-seq samples from (E). (G) Dot plot of selected progenitor (FEZF2, CRB1, SOX2, FGF9, VSX2) and RPE (SFRP5, TTR, SLC35D3, TYR, RLBP1) genes enriched in the sorted samples. (H) Graphs showing qRT-PCR of retinal progenitor (SIX6, VSX2) and RPE (BEST1, RPE65) marker genes in populations from (E) at the moment of sort and at post-sort D30, D35, D40, D45, and D60. (I) UMAP of NCAM1-High (pink), CD140b-High (blue), and unsorted (gray) D60 cells. (J) Dot plot illustrating the proportion of cells corresponding to each identified cell type in scRNA-seq samples from (I). (K) Dot plots of early (MITF, TYRP1, PMEL, SERPINF1, DCT, ELN) and late (RLBP1, BEST1, RPE65, RGR, TTR, SFRP5) RPE genes in the LateRPE cell clusters from each sorted sample. (L) Brightfield and immunofluorescence stainings of unsorted, CD140b-High, and NCAM1-High populations 30 days after sorting (D60) showing co-expression of CD140b, BEST1, and ZO-1 markers. Scale bars, 100 μm. (M and N) Bar graphs showing PEDF secretion (M) and TEER measurements (N) of the unsorted, CD140b-High, and NCAM1-High populations at D60. ∗∗p < 0.0001 compared to Not Sorted and NCAM1-High. In (H), (M), and (N), error bars represent mean ± SEM from three independent experiments. See also Figure S3.

Figure 5

Neuroretinal progenitor differentiation of NCAM1-High-sorted hESC-RPE D30 cells (A) Schematic of the neuroretinal progenitor (altered) differentiation protocol (HS980 line). D30 NCAM1-High-sorted cells were sorted and replated on Matrigel containing DMEM/F12, hDKK1, Noggin, hIGF-1, and bFGF until scRNA-seq at D70. (B) Brightfield images and cobblestone junction scores of sorted and unsorted populations at D70. Scale bars, 100 μm. (C) UMAP of NCAM1-High sorted cells at D70. (D and E) CCA integration of scRNA-seq data from embryonic week 7.5 eye (D) and NCAM1-High-sorted cells subjected to the altered protocol (E). (F) Heatmap of enriched gene expression for cell types in (C). (G and H) Gene expression heatmaps of lens (G) and epithelial (H) cells identified in the reference and in vitro. Shared and differentially expressed genes are shown on the left and right plots, respectively. (I) RNA velocity of embryonic retinal ganglion cells (left) and hESC-derived neurons (right). (J–L) Heatmaps showing gene expression analysis of embryonic and hESC-derived neurons along their respective pseudotimes. RGC, retinal ganglion cell; PC, photoreceptor cell; HC, horizontal cell.

Figure 6

Late hESC-RPE differentiation profiling (A–C) UMAPs and enriched gene expression heatmaps of hESC-RPE scRNA-seq data at D38 (A), D45 (B), and D60 (C) in all three lines. (D) RNA velocity and pseudotime analysis of HS980 RPE at D60. (E) Phase portraits of upregulated RPE marker genes RPE65 and BEST1 as well as a downregulated progenitor marker PAX6. The diagonal line represents the estimated steady state of gene expression, with cells above the steady state experiencing gene upregulation and those below gene downregulation. (F) Plot showing ordinal classification of 20,682 single hESC-derived retinal progenitor and RPE cells at six differentiation time points along embryonic stages. (G) Graph representing classification distribution for seven hESC-RPE differentiation D60 biological replicates (HS980: 3,655 cells; E1C3: 61,479 cells; KARO1: 1,236 cells). See also Figure S6.

Figure 7

Phenotyping of hESC-RPE transplanted in the albino rabbit subretinal space (A) Infrared and SD-OCT images of injected hESC-RPE cells (HS980 line) in the subretinal space of albino rabbits. Green lines indicate the SD-OCT scan plane. White arrows indicate the hyper-reflective RPE layer. Scale bars, 1 mm. (B) Brightfield and immunofluorescent staining for human marker NuMA and BEST1 30 days after injection. Scale bars, 50 μm. (C) Gene expression heatmap comparing 65 single hESC-RPE cells 30 days after transplantation to embryonic week 7.5 retinal progenitors, adult photoreceptors, undifferentiated hESCs, and D60 EMT-RPE. (D) Pearson’s correlation matrix between gene expression profiles of HS980 hESC-RPEs at D30 and D60, post-transplantation (in vivo) RPE, adult RPE and melanocytes, and embryonic RPE. (E) Dot plot graph showing log₂ fold change of RPE markers between HS980 hESC-RPE D60 cells, in vivo RPE, and adult RPE. Error bars represent mean ± SEM from all cells at each time point. (F) Ordinal classification summary matrix showing the percentage of HS980 retinal cells from in vitro and in vivo time points predicted to correspond to each RPE developmental time point (embryonic weeks 5–24, adult). (G) Graph showing classification distribution for hESC-derived progenitor and RPE cells in vitro and in vivo. See also Figure S7.