Simultaneous generation of transplantable RGC-like and corneal progenitor cells from hiPSCs using a dual-lineage platform

Abstract

Background

Vision loss due to retinal and corneal cell degeneration is significant clinical challenge, with current cell therapies hindered by limited donor cells. To address this, we developed a streamlined platform using human induced pluripotent stem cells (hiPSCs) that integrates 2D and 3D culture techniques to simultaneously generate retinal ganglion cells (RGC) and corneal lineages.

Methods

The non-integrated hiPSCs were used to ocular cells differentiation, four time-points cells were collected for 10×Genomics sing-cell RNA sequencing to trace RGC and corneal lineage development. The confirmed differentiated window phase to pick optic vesicles for 3D ocular organoids culture, and as well as directional induced corneal epithelium in remaining 2D dish. After FACS-based cell sorting, the enriched RGC-like and corneal cells were propagated in vitro, and these cells were transplanted into optic nerve crush (ONC) and corneal damaged mice respectively to observe the regenerative repairment capacity.

Results

Through single-cell RNA sequencing, we mapped differentiation trajectories and identified surface markers-CD184 and CD171 for RGCs, and CD104 for corneal progenitors facilitating purification. In mouse models, transplanted hiPSC-derived CD184⁺CD171⁺ RGC-like cells integrated into injured retinas, enhanced host RGC survival, and restored visual function following optic nerve injury. Concurrently, hiPSC-derived CD104⁺ corneal progenitor cells exhibited self-renewal, differentiation capabilities, and accelerated corneal repair with reduced neovascularization. Additionally, this platform enables the synchronous production of retinal and corneal organoids, which are valuable for both regenerative therapy and disease modeling.

Conclusions

Our study establishes a cost-effective surface marker-based method for deriving transplantable RGC and corneal lineage cells from hiPSCs, overcoming key obstacles in ocular regenerative medicine.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13287-025-04843-z.

Background

Degenerative eye disorders/ocular trauma refer to a range of irreversible conditions characterized by the progressive deterioration and apoptosis of various ocular cells, leading to visual impairment or blindness [1, 2]. Globally, at least 2.2 billion people experience near or distant visual impairment, posing a public health challenge [3]. In conditions such as retinal degeneration and corneal damage, the loss of specific cell types, including retinal ganglion cells and corneal epithelial cells, necessitates targeted cell replacement therapies [4, 5]. The eye is particularly suitable for such therapies due to its immunological privilege, which reduces the risk of graft rejection [6]. However, the shortage of donor cells has hindered the progress of ocular regeneration medicine and limited the achievement of vision restoration.

The retina cannot self-repair or restore vision once key cells, including RGCs, retinal pigment epithelium (RPE), and photoreceptors, are lost. Retinal degenerative diseases, such as age-related macular degeneration (AMD), retinitis pigmentosa (RP), and Stargardt disease (STGD), typically involve the loss of RPE and/or photoreceptor cells due to aging or genetic factors. While treatments like vitamin supplements, anti-VEGF agents [7, 8], and gene therapies [9], can slow disease progression, they fail to reverse damage such as subretinal fibrosis, photoreceptor loss, or vision impairment.

Among cell replacement strategies, RPE transplantation using pluripotent stem cell (PSC)-derived RPE cells is the most advanced, with clinical trials showing mid-term safety and potential biological activity without adverse reactions [10–12]. However, no established clinical treatments exist for replacing RGCs or photoreceptors in retinal degeneration patients till now. Recent studies suggest transferring cellular components between donor and host photoreceptors may offer new treatment avenues for various retinopathies, but this requires substantial donor cells [13]. RGCs, the retina’s sole output neurons, present significant challenges for cell replacement, requiring integration into the ganglion cell layer and axonal extension to the optic nerve. No human clinical trials for RGC replacement have started [14], and RGCs’ s multiple subtypes and poor in vitro survival limit obtaining sufficient transplantable cells [15]. For corneal diseases like limbal stem cell deficiency (LSCD) or severe burns, corneal transplantation or stem cell-based therapies are essential, with clinical applications reported [16, 17]. However, culturing these cells remains costly.

PSCs provide a unique solution to the shortage of donor due to their limitless supply and ability to differentiate into various ocular cells and organoids using established protocols, such as SEAM (self-formed, ectodermal, autonomous, multi-zone) differentiation [18], 2D RGC [19], retinal organoids [20], corneal organoids [21] and lacrimal organoids [22] and so on. Typically, these differentiation protocols are designed to generate a single specific lineage. For instance, the SEAM differentiated system has been particularly effective for exploring corneal lineages and their clinical applications [23, 24]. In the SEAM system, initial 2D culture facilitates the self-formation of multizone ocular progenitor cells, containing neuroectoderm, surface ectoderm, neural crest, and RPE, which could subsequently develop into multi-ocular organoids upon transitioning to 3D culture [25].

Our previous established platform for differentiating hiPSCs into retinal organoids [26–28], which integrated both 2D and 3D culture methods, has demonstrated the potential to generate multiple ocular lineages from a single differentiation process. Here, we have established a simplified and cost-effective system that enable the simultaneous generation of both retinal and corneal linage seed cells/organoids for transplantation applications. By optimizing the parallel differentiation of two lineage within the same system, we aim to address the need for multiple cell types in ocular regenerative therapies. To ensure the purity and functionality of the generated cells, we employed confirmed surface markers to purify RGC-like and corneal epithelial progenitor cells. These purified cells were then propagated in vitro and transplanted in vivo to evaluate their regenerative capacity.

Materials and methods

hiPSC cells maintain and culture

The hiPSC lines (UE022, P5 and P4) were purchased from the Chinese Academy of Sciences and was generated from urine cells of healthy human adults using episomal iPSC reprogramming vectors [29], the collection of somatic cells was approved by the donors and they had had signed informed consent. The hiPSCs were feeder-free and cultured on Matrigel-coated plates (Corning) with mTeSR1 medium (Stem Cell Technologies). Cells were passaged by 0.05 mM EDTA (Thermo Fisher) solution at nearly 80–90% confluence.

hiPSCs derived ocular cells differentiation

To differentiate hiPCs into both retinal and corneal lineage cells, a previous published protocols was applied with some adjustment [26]. In brief, hiPSCs were dissociated into small clumps with 0.05mM EDTA and were reaggregated in suspension with mTeSR1 medium and 10mM blebbistatin (Sigma Aldrich) on D0. From D1 to D15, aggregates were induced to neural differentiation in NIM medium containing DMEM/F12 (1:1), 1% N2 supplement (Thermo Fisher), 1×minimum essential media nonessential amino acids (NEAA) (Thermo Fisher), 2 µg/mL heparin (Sigma Aldrich), with gradient mTeSR1 and NIM medium change from 3:1, 1:1 ration to total NIM. The suspended aggregates were reattached to Matrigel-coated dishes on D7. For initial ocular differentiation, RDM medium was used, containing DMEM/F12 (3:1) supplemented with 2% B27 (without VA, Thermo Fisher), 1×NEAA, and 1% antibiotic-antimycotic (Thermo Fisher) from D16 to D28, neural retina and transparent vesicles domains were formed, these domains were picked and transferred to suspension culture, while the remaining adhered cells were get rid of the neural cells under microscope and changed into the corneal medium for further culture.

Droplet-based scRNA-seq

Cell samples and library construction: as hiPSCs derived series timepoint ocular cells (D0, D6, D14 and D28). The single cell suspensions in 0.4% BSA-PBS were processed for single-cell sequencing using the Chromium Single Cell 3’ Reagent Version 1 Kit and Chromium Controller (10xGenomics, CA, USA). Briefly, in each experiment, 20,000 cells per reaction were loaded for Gel Bead-in-Emulsion (GEM) generation in the GemCode instrument, followed by cell lysis, barcoded reverse transcription of RNA, cDNA amplification, shearing, adaptor and sample index attachment. Libraries were sequenced using Illumina Hiseq x Ten.

Quality control and cell clustering: The single-cell analysis and data visualization were performed using a Python environment built on Numpy, Scipy, and Matplotlib libraries, and R 4.2.1 with the Seurat [30] and Monocle 3 packages installed [31]. Cell quality control and filtering, expression normalization and scaling, integration of cells from different position/species, cell clustering was visualized using t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) by performing Seurat package. Genes expressed in very few cells were excluded in subsequent analysis using the same metric for each individual scRNA-seq datasets. Genes expressed in less than 2 cells were excluded. Cells expressing very few genes were filtered out using specific cutoffs for different scRNA-seq data sets (Cutoffs were set to top 95% quantile of the distribution of genes detected (UMI >1) per cell in each data set). We also excluded cells with the highest total amount of expressed genes to account for doublets. We calculated the top 3% quantile of the distribution of genes detected per cell in each dataset, and removed the cells in this quantile. Cells with a high percentage of mitochondrial gene expression (cutoff used: 10% of total UMIs) were also excluded.

Data integrated, dimensional reduced and visualized cell clusters: Normalization of gene expression of cells was performed using the Seurat “SCTransform” function. Relative expression level of genes was calculated using the Seurat “ScaleData” function with default parameters. The hiPSCs derived series timepoint ocular cells were integrated with Seurat “IntegrateData” function, respectively. Unsupervised and unbiased clustering of all integrated cells was performed using Seurat “FindClusters” function with the following parameter: resolution = 0.03. The clusters and non-linear dimension reduction approaches (tSNE and UMAP) were used to reduce the dimensions and visualize cells in two dimensions. FindMarker function using Wilcoxon ranksum test in Seurat was used to identify the differentially expressed genes (DEGs) in one cluster compared with other clusters, and tahe DEGs were used for cell type annotation.

Trajectory analysis: The dynamic sequence alterations of hiPSCs differentiated ocular cellular state were estimated through monocle 3 based trajectory analysis. Monocle 3 assigned the re-cluster cells to specific clusters and partitions, through learning process to calculate the pseudotime basing distance between a cell and the starting cells, then construct the trajectories, and visualize cells trajectories and pseudotime by UMAP plots [32]. The function choose_graph_segments was used to construct subset cells based on the branch of chosen path in the trajectories.

Pathway enrichment analysis and gene set enrichment analysis: Pathway enrichment analysis and GO analysis were performed using Metascape.

FACS-based isolation and culture of RGC and corneal lineage cells

Sorted hiPSCs derived RGC lineage cells: The D40-D60 aged retinal organoids were used to sort the RGC-like cells. Selected the morphological integrity with clear transparent neural retinal layer, got rid of the visible attached RPE and other cells under optical microscope. The neural retina tissues were cut into 1 mm pieces and digested with Acctuase for 10 min at 37 ◦C, and pipetted single cells and filtered with 40 μm cell strainers. These cells were incubated with fluorescein coupled flow antibodies APC-CD184 and PE-CD171 for 20 min at 4 ◦C, and washed by Ames medium for three times. The cell pellet was resuspended by 1 ml Ames medium and carried out FACS sorting procedure. Collected the four phased cells for culture, the cells were plated on the Matrigel-coated 12 well plates with medium additive BDNF.

Sorted hiPSC derived corneal lineage cells: The adherent monolayer cells were kept in the DM for further 2–3 weeks, and were disgeted into single cells and were performed for FACS after ITGB4-APC incubating. The isolated hiPSCs derived ITGB4⁺ corneal epithelial cells (CECs) were cultured in corneal epithelium medium (ATCC).

Cell transplantation experiments

Approval for all the experimental animals were obtained from the animal ethics committees of Zhongshan Ophthalmic Center, Sun Yat-Sen University, and the approved project was “Mechanism and application of endogenous cell-mediated ocular tissue regeneration and repair” and the approved code was Z2021023 (Date of approval: June 20, 2021), and as while as “Mechanism Study on in Situ Regeneration of Ocular Epithelial Adult Stem Cells in the Treatment of Important Blinding Eye Diseases” within the approval No 2019-043 (Date of approval: April 30, 2019). The experiment animals were maintained in certified animal facilities and handled in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and institutional guidelines for animal investigation, and have been reported in line with the ARRIVE guidelines 2.0. The animals were kept under a 12 h light/dark cycle and were randomized before the initiation of the studies.

For cells transplantation experiments, male C57BL/6J or rats were randomly assigned to experimental groups and weas age-matched (6–8 weeks old). The mice and rats were anesthetized with pentobarbital (50 mg/kg, intraperitoneally), and then were placed on heated pad to maintain body temperature.

RGC cells transplanted: C57 mice were anesthetized with isoflurane before each procedure. From one day before transplantation to the end of the follow-up, mice or rats received immunosuppressive cyclosporine A (210 mg/ L) treatment in the drinking water. Unilateral optic nerve crush (ONC) was conducted on 6 ~ 8 weeks old male mice, the forceps pressed about 1 mm behind the optic nerve for 10s. And average 1 × 10⁴ CM-DiI labeled cells were intravitreally injected to the ONC mice immediately.

Corneal cells transplanted: C57 mice or SD rats were anesthetized with isoflurane before each procedure. From 1 day before transplantation to the end of the follow-up, mice or rats received immunosuppressive cyclosporine A (210 mg/ L) treatment in the drinking water. The entire limbus was removed, and the central corneal stromal cells (1.5–2 mm diameter) were removed using AlgerBrush II. After rinses with PBS, the wounded beds were overlaid with CM-DiI labeled hiPSC-ITGB4⁺ CECs were dissolved in 0.5% collagen I solution, adjusted the cells density was 1 × 10⁵ cells/µL, average 1µL cell solution for mouse and 2µL for rat. The donor eyelids were sewed by 10# surgical suture.

Tissue process and immunofluorescent staining

The methods were performed as previous [26]. For cultured cells, cells were fixed with 4% PFA for 10 min and washed twice with 0.01 M PBS. For organoids, the collected organoids washed with PBS for twice and fixed with 4% PFA for 30 min at room temperature. Fixed ocular organoids were successively subjected to sucrose gradient dehydration concentrations from 6.25%, 12.5% to 25%. For frozen tissue sections, the collected eyeballs or tissues were washed with PBS, fixed with 4% PFA and dehydrated with 30% sucrose overnight at 4℃. The ocular organoids and tissues were sections into 8–10 μm slices by a Leica microtome (Leica CM1950). For retinal or corneal wholemounts, the eyes were imminently enucleated and fixed in 4% PFA overnight at 4℃, tissue flat-mount prepared for staining or taking photograph.

Cells, sections or wholemounts were incubated in blocking containing 10% donkey serum (Solarbio) only or 10% donkey serum with 0.25% Triton X-100 for 1 h at room temperature. Primary antibodies were incubated at a suitable dilution in PBS overnight at 4℃. After the primary antibody incubation, the samples were rewashed with PBS three times and incubated with secondary antibodies for 1 h at room temperature. DAPI was used for nuclear counterstaining. Images were obtained using a Zeiss fluorescence microscope (Axio Observer 7). Information on the antibodies used is shown in Table 1.

Table 1.

Antibodies

Antibodies	Dilution	Distributor (Cat. No)
Primary antibodies:
PAX6	1:100	Biolegend (901301)
ITGB4	1:100	Sigma (HPA036348)
P63	1:100	Abcam (ab735)
Ki67	1:100	Abcam (ab15580)
Krt12	1:200	Abcam (ab185627)
Krt14	1:100	Novus (NBP2-45032)
Recoverin	1:200	MERCK(ab5585)
Tuj-1	1:200	Abcam (ab78078)
Synaptophysin	1:200	Abcam (ab32127)
APC-CD184	1:50	BD (555976)
PE-CD117	1:100	Life technologies (A18361)
APC-CD104	1:100	Invitrogen (MA5-23535)
Secondary antibodies:
Donkey anti-mouse (488)	1:500	Thermo Fisher (A-21202)
Donkey anti-rabbit (555)	1:500	Thermo Fisher (A-31572)
Donkey anti-mouse (555)	1:500	Thermo Fisher (A-31570)
Donkey anti-rabbit (488)	1:500	Thermo Fisher (A-21206)
Donkey anti-sheep (555)	1:500	Thermo Fisher (A-21436)

qRT-PCR

Cells were collected, and RNA extraction was performed with TRIzol Reagent (Thermo Fisher) following the manufacturer’s instructions. One microgram of RNA was reverse transcribed using random primers, and qRT-PCR was performed with a SYBR Green reaction mixture (Vazyme) using a Quant Studio 7 Flex system (Applied Biosystems). A comparative cycle threshold (Ct) method was used to calculate the levels of target gene expression, and the Ct value was normalized to GAPDH. The fold changes were calculated using the formula 2-^ΔΔCt. The primer sequences can be found in Table 2.

Table 2.

Primers for qRT-PCR

Genes	Forward primer	Reverse Primer
PAX6	AGTGAATCAGCTCGGTGGTGTCTT	TGCAGAATTCGGGAAATGTCGCAC
Brn3a	GGGCAAGAGCCATCCTTTCAA	CTGTTCATCGTGTGGTACGTG
Brn3b	CAAGCAGCGACGCATCAAG	GGGTTTGAGCGCGATCATATT
ISL1	ATGACAAAACTAATATCCAGGGG	CTGAAAAATTGACCAGTTGCTG
RBPMS-a	AGTTCATTGCCAGAGAGCCA	AGCGGGATAGGTGAAAGCAG
RBPMS-b	ACCTCCTCCTGCTTTCACCT	GCTGGCACTATCAGGAGACG
RBPMS-c	CCTCTGTACCCAGCGGAGTT	TGTGTTGGGCTTTGCCTCAG
NEFL	ATGAGTTCCTTCAGCTACGAGC	CTGGGCATCAACGATCCAGA
NAV1.6	CTGGAGAATGGAGGCACACAC	ACGCTGCTGCTTCTCCTTGTC
BCAM	ATGCTGTCGCTCAGTTCTATC	GGCTCTATTTCCGCTGTCTT
EpCAM	GCAGTTGTTGCTGGAATTGT	GCATCTCACCCATCTCCTTTAT
GJA-1	CTGGGTCCTGCAGATCATATTT	GGCAACCTTGAGTTCTTCCT

Electroretinography (ERG) and visual evoked potential (VEP)

Prior to perform recording, each group mice were dark-adapted for more than 12 h. The mice were anesthetized with isoflurane and dilated with 0.5% tropicamide for 5 min. The integrated ERG and VEP recordings were carried out using the Celeris D430 rodent ERG testing system (Diagnosys LLC, MA, USA). Mice were placed on the heated surface of the ERG system to maintain body temperature, and the cornea were kept moist with the saline solution or 1–2% hydroxypropyl. Corneal electrodes with integrated stimulators were placed on the lubricated corneas, and subdermal platinum electrodes were in the snout and back of the head at the location of the cortex. Each eye was separately exposed to 100 flashes of 1 Hz, 0.05cds/m² white light through the corneal stimulators and responses in the visual cortex were recorded for 300ms at 2000 Hz. The oscillatory potentials on the rising part of the b-wave were retrieved and the area under the curve was calculated. N1 is the value from baseline to the most negative peak value and the amplitude P1 is the value from most negative peak to the first positive peak, the latency is the value of N1-P1 [33, 34].

Corneal clinical evaluations

The mice and rats were clinically estimated the corneal opacification, re-epithelialization and neovascularization under slit-lamp microscope. For corneal opacification, the scores were on a scale of 0–4, where 0 = completely clear; 1 = slightly hazy, iris and pupils easily visible; 2 = slightly opaque, slightly opaque, iris and pupils still detectable; 3 = severely opaque, iris and pupils hardly visible; 4 = completely opaque with no view of the iris and/or pupils [35]. For corneal hazing, 1% fluorescein sodium staining regions of the corneal surface were measured using Image J. For neovascularization, the scores were as follows: 0 = none, 1 = area less than or equal to one-quarter, 2 = one-quarter less than area less than or equal to one-half, 3 = one-half less than area less than or equal to three-quarters, 4 = area greater than three quarters [36].

Statistical analysis

Data were presented as the mean ± SEM or mean ± SD. Statistical analysis was performed using GraphPad Prism software. For multiple comparisons among groups, one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparison tests was used. For comparisons between two groups, an unpaired Student’s t-test was used. P-value of < 0.05 was regarded as statistically significant.

Results

Single-cell transcriptomes analysis identified hiPSCs derived RGC and corneal lineage differentiation trajectory

Previous studies, we have optimized the hiPSCs derived retinal organoids differentiation system [26–28], which could acquire plenty of retinal organoids for isolating RGC-like cells during D40-D60 [35, 37]. To uncover the primary ocular cell types and markers expression during initial differentiation phase, we performed single-cell RNA sequencing (scRNA-seq) of the hiPSCs derived ocular cells at four time points: D0 (hiPSCs), D6 (EBs), D14 and D28, the each timepoint cells were digested into single cells and enrolled into 10× scRNA-seq pipeline (Fig. 1A). A total of 47,721 sing-cell transcriptomes passed quality control measures, and the cells were embedded using Uniform Manifold Approximation and Projection (UMAP) and clustered using Seurat graph-based clustering (Supplementary Fig. 1A-D). The identified 17 cell clusters were grouped into ten cell types based on well-known gene markers (Fig. 1B) and the cell type compositions were also analyzed (Fig. 1C). The hiPSCs showed typical gene expressing within POU5F1 and NANOG, PCs were mainly original from EBs which showed the similar gene profiles with hiPSCs whereas had higher SIX3, OTX2 and FZD5 expression. pRPCs had the similar genes profiles with RPCs, but expressing higher PC related markers, moreover, the PC-pRPCs were combined samples from both PCs and pRPCs. More important, the RGC and corneal lineage clusters were clearly recognized within ATOH7, POU4F2, DCX, SNCG, CXCR4 and L1CAM in RGC, as while as TP63, ITGB4 and KRT19 in CEs (Fig. 1D, E).

Fig. 1.

scRNA-seq analysis hiPSCs derived RGC and corneal lineage cell differentiation trajectory. A Schematic of the RNA-seq workflow. B UMAP plot of integrated scRNA-seq hiPSCs differentiated retinal cells (D0, D6, D14 and D28, cell counts = 47721). Each cell type cluster was identified based on expression of hiPSCs or retinal specific cell markers. C Stacked bar plot with mean cell-type composition of integrated hiPSCs differentiated retinal cells for each cell type. D Dotplot showing average scRNA-seq expression of marker genes of different kinds of hiPSCs and retinal cells. E Feature plots showing the typical cell type marker gene expression. F UMAP visualization of trajectory inferred by monocle3, Cells are colored by cell type. G The selective cell clusters for further pseudotime analysis according to cell identify. H UMAP visualization of pseudotime computed by monocle3, cells were colored by pesdudotime. I Featured plots showing the gene expression along differentiation trajectory. J–O Expression dynamics along the pseudotime of selected genes were plotted by Monocles 3

To investigate the differentiation history from hiPSCs to ocular cells, the cells were ordered along differentiated stages by performing trajectory analysis using Monocle3 packages. Cells in the trajectory UMAP plot are colored by cell type identified in the above Seurat integration analysis, while the monocles trajectory showed several nodes and marked with black circular dots representing principal nodes along the trajectories (Fig. 1F). Basing the original identity, we selected the representative clusters for further pseudotime analysis, the black circular “2” was chosen as the starting node for computing pseudotime trajectory (Fig. 1G, H). Consistent with the Seurat analysis, Monocle 3 reconstructed a similar trajectory, displaying the corresponding cell clusters from hiPSCs to various ocular cells, and as while as the relative gene expression profiles (Fig. 1I). In order to tracing the genetic circuits for ocular cell fate decisions, we extracted cells from the trajectory branch points which corresponded to RGC and corneal lineages, and organized the modules within the similar genes expressing over the trajectory by monocle 3 (Supplementary Fig. 2A, B). In RGC linage modules, RGC-M2, RGC-M10, RGC-M13 and RGC-M6 were clustered the hiPSCs, RPCs, neural RPCs and RGC, respectively, and the Gene Ontology (GO) analysis was performed on their difference genes, RGC-M2 enriched the most significant term for transcriptional regulation of pluripotent stem cells, RGC-M10 and RGC-M13 enriched terms were relative to cell morphogenesis and neuron differentiation, while RGC-M6 showed the term as regard to neuron projection development, cell fate commitment and linage map for neuronal differentiation (Supplementary Fig. 2C). For corneal linage, the enriched GO terms were relative to mesoderm cell development and morphogenesis of embryonic epithelium, especially in covering integrin1 pathway and ITGA6-ITGB4-Laminin complex (Supplementary Fig. 2D).

Following the pseudotime trajectory of cell differentiation, we identified some new genes expression dynamics regarding to RGC and corneal lineages. Early in cell differentiation, POU5F1 expression decreased and PAX6 expression increased obviously (Fig. 1J). Then, SOX11 and TTC3 were significantly activated and maintained in all retinal differentiating stage (Fig. 1K). DAPL1 and SPOCK1 were activated in retinal cell development, while VSX2 and RORB were initiated expressing for neural retina development, furthermore STMN2 and ONECUT2 were closed related to the RGC development (Fig. 1L-N). During the corneal lineage development, KRT19 was one important protein to take part in corneal development, and ANXA1, TFAP2A and GABRP were also identified specific participating corneal cell commitment (Fig. 1O).

Confirmed the role timepoint of harvesting the multiple ocular cells and organoids

Meanwhile, we also charactered the single-cell transcriptome about cells collecting from D28, 17 clusters were identified by Seurat analysis and data were visualized by UMAP, which clearly identify both RGC and cornea lineage clusters (Fig. 2A). The same kind cell type shared gene expression patterns, the clusters 7, 10 and 16 represented RGC lineage and showed classical RGC-relative markers and the identified CXCR4 and L1CAM were just expressing in cluster 10 or 16, as well as cluster 9 was defined as corneal lineage within specific relative markers expression, such as TP63 and ITGB4 (Fig. 2B-D). In light of CXCR4 (CD184), L1CAM (CD171) and ITGB4 (CD104) were the surface markers, we explored the application for tissue regenerative therapy for further study.

Fig. 2.

scRNA-seq analysis of D28 hiPSCs derived retinal cells. A UMAP visualization of the D28 hiPSCs derived retinal cells. Cells are colored by Seurat clusters. Recognized the clusters 7, 10 and 16 were RGC linage and clusters 9 was corneal linage. B Violin plot showing the gene expression of the discriminative markers in the 17 clusters. C Feature plot showing the gene expression of selective RGC linage markers. D Feature plot showing the gene expression of selective corneal linage markers. E Generation of various ocular cells/optic vesicles from one dish. Scale bars: 50 μm. F The un-picked retinal optic vesicles timely were ahead extending to axons. Scale bars: 50 μm

In one dish, the self-differentiated re-attached EBs around D28 days showed various ocular cells and domains, such as clear boundary retinal optic vesicles with or without hexagonal RPE cells, the integrity and transparent corneal vesicles within alone domains or along with retinal optic vesicles (Fig. 2E). In case of exceeding the appropriate collection timepoint, we found there were several retinal optic vesicles becoming black and gradually disintegration, and while some others may ahead of extending early axons from the margin areas (Fig. 2F). Overall, these factors might affect the production yields of organoids.

Transplantable hiPSCs-RGC like cells could protect host RGC degeneration and restore vision

In our previous study, we have purified CD184⁺CD171⁺ RGC like cells from retinal organoids, these cells have certain propagation capacity [37]. Here, we verified the long-time effect of these RGC like cells for regenerative repairment. FACS-based protocol was used for purified cells from D40-D60 retinal organoids, we collected the four group cells at the same time for further culture, all group cells could propagate, and CD184 positive cells showed higher proliferation ability, while the CD171 single positive cells had the lowest capacity (Supplementary Fig. 3A, B). The qRT-PCR results showed that CD184⁺CD171⁺ cells expressing the highest RGC relative markers than the rest three groups, such as Brn3b, Brn3a, ISL1, RBPMS and NEFL and so on (Supplementary Fig. 3C). So, CD184⁺CD171⁺ RGC like cells were chosen to delivered into the optic nerve injury mice for investing long-time therapy effect.

The transplanted CM-DiI positive cells could survive in the damaged retina and extend out neurofilament (Supplementary Fig. 4A). The retinal wholemount results showed that the transplanted cells played important role of protecting the optic nerve injury induced RGC somas degeneration and retained more axons than the control group (Fig. 3A-C). The RGC somas appeared apoptosis at post-injuried 3 weeks, and nearly 80–90% RGCs were dead at post-injuried 1.5 month without cell supplement therapy, whereas the transplanted cells could safeguard the host RGC somas reservation up to 60% and 30% post-injuried 3 weeks and 1.5 month respectively (Fig. 3D). According to the delayed host RGC degeneration, the optic nerves were also acquired more chance to defend injury (Fig. 3E, F). To functionally evaluate the cells therapy influence on the optic nerve electrophysiology, the combined ERG and flash visual evoked potential detection were performed and the waveforms recorded from both control and cell transplantation mice. Three weeks after the ONC injury, the F-VEP waveforms were silent and the peaks were dramatically decreased in the PBS group, while the waveforms were restored in the cell treatment group even persisting up to 1.5 month, within the average amplitudes for N1-P1 were 5.66µV ± in PBS group verse 10.5µV ± in cell treatment group (Fig. 3G, H). Beyond that, the b waves were also well restored after cell supplement therapy timely after ON crush (Fig. 3I, J). Furthermore, the tissue sections showed that cell supplement could rescue more host RGC degeneration for long-time to 6 months, and within more synaptic proteins, but the photoreceptors seemed to be least affected (Supplementary Fig. 4B). Overall, the CD184⁺CD171⁺ RGC like cells revealed protection effect on optic nerve crush induced RGC degeneration.

Fig. 3.

Transplantable CD184⁺CD171⁺ RGC like cells for restoring vision. A Representative retinal wholemount images of normal mice retinal. The optic neural fibers were labeled by CTB (green). The lower image showing a higher magnification of the corresponding white box area. Scale bars: 500 μm. B, C Representative retinal wholemount images of the ONC mice with PBS or cells treatment for 3 weeks and 1.5 month, respectively. The optic neural fibers were labeled by CTB (green). The lower image showing a higher magnification of the corresponding white box area. Scale bars: 500 μm. D Histogram plot showing the survival ratio of the host optic neural fibers after cell treatment. E, F Frozen sections showing the CTB labeled optic nerves with PBS or cells treatment for 3 weeks and 1.5 month, respectively. Scale bars: 100 μm. G The representative VEP waveforms of normal mice retina, and ONC mice with PBS or cells treatment for 3 weeks and 1.5 month, respectively. H Histogram plot showing the amplitudes (N1-P1) of PBS or cells treatment for 3 weeks and 1.5 month. I The representative ERG b waveforms of normal mice retina, and ONC mice with PBS or cells treatment for 3 weeks and 1.5 month, respectively. J Histogram plot showing the b amplitudes of PBS or cells treatment for 3 weeks and 1.5 month

hiPSCs derived CD104⁺ corneal progenitor cells participated in corneal repairing. However, the sources of human adult corneal tissues are limited by ethic dispute and insufficient HLA matched ocular tissues. Combined the sc-RNA seq transcriptome analysis of the hiPSC generated retinal cells, we are aimed to obtain the ITGB4⁺ epithelial cells from this ocular cell differentiated system. During the cell differentiation process, there spatio-temporally emerge various ocular cells containing optic vesicle, epithelial cells and mesenchymal cells (Fig. 4A). After picked out optic vesicles for generating retinal organoids, the residual adherent cells were changed for corneal epithelial cell medium and kept cultivating for another month time long, epithelial and mesenchymal cell clusters with clear morphology were coexisted (Fig. 4B). FACS-based cell sorting was used to purified ITGB4⁺cells and the positive rate was nearly 6.42% (Fig. 4C). These cells were cultured on Matrigel-coated plates for proliferation, the adherent ITGB4⁺ cells showed typical cobblestone like structure (Fig. 4D), the proliferated cells could carry out passages for further amplification, thus help us to obtain plenty of hiPSC derived ITGB4⁺ cells. The single cell clone experiment showed they were also had self-renewal ability (Fig. 4E). Immunohistochemical staining revealed the cultured cells expressing the proliferative cell markers Ki67 and PAX6, as well as the corneal markers Krt14, Krt12 and ITGB4 (Fig. 4F). While the cultured ITGB4⁺ cells could assemble spheres after Matrigel-encapsulation, within expressing PAX6, ITGB4, Krt14 and Krt12 (Fig. 4G, H), comparing the 2D adherent cells, the 3D spheres expressed higher stem/basal cells and cell junction markers, such as PAX6, BCAM and GJA-1 and so on (Fig. 4I, J).

Fig. 4.

Characterization of the cultured hiPSCs derived ITGB4⁺ corneal cells. A Differentiation of ocular cells from hiPSCs, the representative image showed the various ocular cells, within optic vesicle and the epithelial cells area (black box). Scale bar: 200 μm. B The representative image showed the unsorted epithelial and mesenchymal cells after eliminating the neural cells. C FACS sorted ITGB4⁺ corneal progenitor cells from hiPSCs derived ocular cells. D Representative images of cultured ITGB4⁺ corneal progenitor cells. Scale bars: 50 μm. E The single cells clone forming capacity of ITGB4⁺ corneal progenitor cell. Scale bar: 10 μm. F Immunofluorescent staining of ITGB4⁺ corneal progenitor cells with ITGB4 (green), Krt14 (red), Krt12 (red), P63 (red), PAX6 (green) and Ki67(green). Scale bars: 100 μm. G The representative image of ITGB4⁺ corneal progenitor cell spheres. Scale bars: 200 μm. H Immunofluorescent staining of ITGB4⁺ corneal progenitor cell spheres with ITGB4, Krt14, Krt12 and PAX6. Scale bar: 100 μm. I, J The q-RT PCR results show the genes expression by GAPDH normalized △CT value and fold change

To verify the therapeutic effects of ITGB4⁺ cells for corneal regeneration, cells were administered to the corneal scratching damaged ocular surface via eyedrop pattern, the corneal healing process was evaluated using the referenced methodology. Images from two representative eyes of each group were showed in Fig. 5A, the degree of corneal opacity and neovascularization scores and the epithelial defect were calculated at post-transplanted 3 days, 7days and 1 month. The results showed that cell treatment significantly improved corneal transparency and epithelial repairment within the closest resemblance to normal corneas when comparing to the control group (Fig. 5B, C), the similar results were also showed the parallel experiments estimating the cell treatment effect in mild epithelial cells defect mice model (Supplementary Fig. 5A). Even more we found that some eyes seemed more sever epithelial defect at D3, but recovered rapidly after cell treatment and persisted for one month, while some eyes looked like moderate epithelial defect in early stage whereas become deteriorated at late stage in control group (Supplementary Fig. 5B). The neovascularization scores were also lower after cell treatment (Fig. 5D). The representative images showed the transplanted CM-DiI labeled cells were survived in the ocular surfaces and within ITGB4 expression (Fig. 5E, F). The bigger rat eyes were using to assess the neovascularization between cell and control groups, which were also showed the similar results (Fig. 5G). Therefore, the hiPSCs-ITGB4⁺ CECs could survive in damaged cornea, and improved the scathing cornea repairment and relieved neovascularization and helped to prevent conjunctival invasion.

Fig. 5.

hiPSCs derived CD104⁺ corneal progenitor cells participated in assistance corneal repairing after injury. A Representative slit-lamp microscope and fluorescein staining images show the ocular surface of mice within removing of limbal zones combining with central stromal cells after ITGB4⁺ CECs transplantation (D3, D7 and 1 M). The upper two lines were the control group and the lower two lines were the cell group. Each point showed two parallel experiment individuals. B-D Scores of corneal epithelial defects and NEOV scores from each group were compared. E Representative wholemount images showed the integrated CM-DiI labeled ITGB4⁺ CECs. Scale bar: 1 mm. F The rat corneal wholemount showed transplantable CM-DiI labeled hiPSC-ITGB4⁺ CECs (red) with immunofluorescent staining for ITGB4 (green). The right images were enlarged area of the white box. Scale bar: 1 mm. G The representative microphotograph images of conjunctival vascular invasion after limbus + central corneal stromal injuried in rats between control and hiPSC-ITGB4⁺ CECs transplantation for post D4 and 1 M

Moreover, we also generated hiPSCs derived corneal organoids which could modify the cornea development in vivo, the corneal organoids were transparent and could see the underlying cell morphology through the organoids layer (Fig. 6A). The corneal organoids sections showed expressing ITGB4, K14 and K12 by immunohistochemical staining (Fig. 6B). ITGB4 were co-located staining with K12 or P63 in some area, which implied that the ITGB4⁺cells might represent a population of heterogenous progenitor cells (Fig. 6C).

Fig. 6.

Generation of hiPSCs derived corneal organoids. A The representative image of hiPSCs derived corneal organoids with transparent property. scale bars: 200 μm. B, C Immunofluorescent staining of corneal organoid sections with ITGB4 (green), Krt14 (red), Krt12 (red) and P63 (red). Scale bars: 100 μm and 50 μm, respectively

Discussion

The current study presents a cost-effective platform for differentiating hiPSCs into ocular tissues and cells, enabling the simultaneous generation of retinal and corneal organoids, as well as RGC-like cells and corneal progenitor cells purified using specific surface markers (CD184 and CD171 for RGC-like cells, and CD104 for corneal progenitors). Both the hiPSC-derived RGC-like cells and corneal progenitor cells exhibited proliferative capacity, allowing for the generation of sufficient cell numbers for ocular regenerative therapies and addressing the shortage of donor cell sources. The purified RGC-like cells protected host RGCs from degeneration and restored visual function after retinal optic nerve injury, while the corneal epithelial progenitor cells accelerated corneal repair and improved outcomes in corneal injury models.

Our platform employed an integrated 2D and 3D culture approach. During the differentiation of adherent embryoid bodies (EBs), multiple ocular cell lineages were gradually generated and self-organized into corneal and retinal optic vesicles. This approach differs from differentiation systems that focus on obtaining SEAM structures for corneal progenitors [18] employ whole 3D suspension culture for RO individually [38]. Given the advantages of this system, after four weeks of differentiation, the clearly visible optic vesicle domains were manually selected and transferred to suspension culture for further development. Concurrently, the remaining adherent cells were switched to CE medium to promote the proliferation of corneal epithelial cells while eliminating neural lineage cells.

A recent study based on the SEAM protocol [39] indicated tthat a large proportion of neural and glial lineages appeared at week 4, and determined the optimal time to switch to a specific medium for corneal lineage development. Consistent with these findings, we emphasize that the differentiation timepoint is crucial for distinguishing between different lineage cells for subsequent culture. Furthermore, some retinal optic vesicles began to undergo apoptosis if not transferred to suspension culture, as proper optic vesicle evagination and optic cup formation require mechanical feedback and spatial freedom [40].

Integrated scRNA-seq analysis of time-series transcriptomic profiles from hiPSCs at day 0 (D0) to differentiated early-phase ocular cells at day 28 (D28) revealed key developmental junctures and trajectories. The hiPSC marker POU5F1 exhibited transitional expression in proliferating cells of EBs, with a rapid decrease following the expression of PAX6, which initiates the formation of the presumptive neuroectoderm [41]. SIX3, a forebrain [42] and pan-ocular transcription factor, was expressed earlier than PAX6 in single cells pseudotime analysis, suggesting its role in the primary regionalization of the first committed lineages during neuroectoderm and eye field formation [43].

By analyzing gene expression modules across lineage branches, we identified SOX11 and TTC3 as genes expressed throughout neuroectoderm and surface ectoderm development, maintaining high expression in RGC lineages, which may correlate with RGC regeneration [44, 45]. DAPL1 and SPOCK1, associated with cell metabolism [46, 47], were involved in retinal progenitor cell proliferation and development, with decreased expression in the RGC lineage. VSX2 and RORB were specifically correlated with neural retinal progenitor cell fate specification; mutations in these genes can disrupt retinal cell phenotypic diversity and stratification [48, 49]. STMN2 and ONECUT2 showed high expression during RGC lineage development. STMN2 regulates microtubule cytoskeleton dynamics, serving as an immature neuron marker and preserving axon integrity [50], while ONECUT2 is expressed in postmitotic RGC and regulates subtype formation [51].

Additionally, the differentiation trajectory identified genes correlated with corneal lineage commitment, including KRT19, TFAP2A, ANXA1, and GABRP. KRT19 is a marker of epithelial stem cells [52], TFAP2A is a non-neural ectoderm specific transcriptional factor promoting surface ectoderm lineage [53], and GABRP and ANXA1 have been reported to enhance epithelial cell viability and promote corneal epithelial repair, respectively [54, 55]. These reconstructed developmental trajectories and co-expressed gene modules highlighted both known lineage-committed genes and identified less-studied candidate genes.

Morphological and scRNA-seq analyses revealed a diverse mixture of cell types after four weeks of hiPSC differentiation, including RPCs, NC cells, RPE cells, retinal optic vesicles, and CE cells. Unlike the SEAM system, which organizes cells in distinct zones [18], our platform exhibited a more random and unconstrained spatial distribution of ocular cell types in a single dish, facilitating manual selection of specific cell populations. scRNA-seq further delineated spatially distinct RGC and corneal lineage clusters at this timepoint, with RGC lineage cells originating from RPCs and corneal lineage cells closely associated with NC and POM.

scRNA-seq identified two RGC lineage clusters expressing surface markers CD184 (CXCR4) and CD171 (L1CAM), alongside transcription factors ATOH7, POU4F2, and DCX, and proliferation markers PAX6, CCND1, and SOX11, indicating these cells are RGC precursors [56]. CD184 expression preceded CD171, with the latter marking more mature RGCs. FACS isolated four cell populations, and the CD184⁺CD171⁺ double-positive RGC-like cells exhibited large soma bodies, long axons, and higher expression of RGC-specific markers compared to other groups. These cells may represent a subtype of α-RGCs, which lack CXCL12 expression and demonstrate resilience to injury and robust axon regeneration post-optic nerve crush [30, 57–59].

CD184, a receptor for SDF1 (CXCL12), acts as a chemotactic receptor, and increased SDF1 expression in ONC retinas promotes inflammation-induced optic nerve regeneration and RGC survival via CXCR4 signaling [60, 61]. CD171, a type 1 membrane glycoprotein, supports neuronal survival and neurite extension, with mutations linked to neural disorders [62–64]. Transplanted CD184⁺CD171⁺ RGC-like cells survived in the host ganglion cell layer (GCL), effectively rescuing RGC degeneration and restoring vision long-term post-ONC. SDF1-mediated chemotaxis facilitated donor RGC migration from the subretinal space to the GCL, with CXCR4 activation enhancing RGC maturation during migration [65]. These findings suggest that CD184⁺CD171⁺ RGC-like cells achieve effective regenerative repair in a modified microenvironment.

For applications in corneal lineage regenerative therapy, several studies have been conducted to identify and isolate LSCs from adult tissues or hiPSCs. The well-known limbal marker ABCB5 is required for LSC maintenance. Isolated ABCB5⁺ cells from adult tissues possess the exclusive capacity to restore the damaged cornea after grafting [66, 67], and their safety has been evaluated in preclinical programs [68]. hiPSC-derived ABCB5⁺ corneal cells exhibit limbal cell characteristics and high wound healing capacity [69]. Additionally, SEAM-derived hiPSC-ABCB5⁺ corneal cells demonstrate holoclone-forming ability and express ITGB4, similar to adult corneal epithelium [70]. Another adult BCEM + population, which partially overlaps with ABCB5⁺ cells, also exhibits holoclone formation and ITGB4 expression [71]. Consistent with these findings, our cultured adult and hiPSC-derived ITGB4⁺ corneal progenitor cells not only formed holoclones but also demonstrated excellent corneal repair capacity upon limbus removal. Furthermore, transcriptomic analysis revealed that the ITGB4⁺ cell cluster highly expresses CXCL14 and PAX6. The PAX6/CXCL14 axis has been reported to play a critical and independent role in central corneal epithelial cells for repairing corneal injuries and maintaining homeostasis, distinct from LSCs [72].

Additionally, we obtained hiPSC-derived corneal organoids from the same differentiation platform. Frozen sections revealed that most ITGB4⁺ and P63⁺ cells were located in the basal layer, while KRT12⁺ cells predominantly resided in the apical layer. The simultaneous harvesting of retinal and corneal organoids offers multiple shared applications. First, ocular organoids can be utilized in regenerative therapy. For instance, cultured corneal organoids have been used to repair the ocular surface in patients, improving their comfort and vision [73].

Similarly, transplanted retinal organoids have demonstrated long-term survival and safety, leading to increased retinal thickness in recipients with RP [74]. A recent study found that exosome vesicles (EVs) extracted from retinal organoids promoted corneal epithelial wound healing [75]. Furthermore, these EVs were shown to remodel Müller cell fate and provide a therapeutic effect for retinal degeneration [76]. Second, ocular organoids serve as valuable in vitro models. Numerous studies have utilized retinal organoids to investigate disease mechanisms and screen drugs, particularly in high-throughput drug discovery [77, 78]. Similarly, corneal organoids have been used to predict ocular irritation, evaluate therapies for dry eye, and aid in drug development [79–81]. The integration of these organoid sources facilitates the development of a highly economical and effective comprehensive platform for ocular research and therapy. Third, hiPSC-derived ocular organoids enable autologous cell-based transplants and customized precision treatment programs, thereby improving the quality of life for patients with corneal and retinal diseases.

Conclusions

This study presents a platform for differentiating hiPSCs into ocular cells, enabling the simultaneous generation of retinal and corneal lineage progenitor cells as well as retinal and corneal organoids. Through analysis of scRNA-seq data, we identified specific surface markers-CD184 and CD171 for RGC-like cells and CD104 for corneal progenitors-to purify abundant populations of these cells. The purified cells demonstrated therapeutic potential in animal models of retinal and corneal degeneration, highlighting their application in regenerative medicine.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Abbreviations

hiPSCs

Human induced pluripotent stem cells

RGC

Retinal ganglion cells

ONC

Optic nerve crush

RPE

Retinal pigment epithelium

AMD

Age-related macular degeneration

RP

Retinitis pigmentosa

LSCD

Limbal stem cell deficiency

SEAM

Self-formed, ectodermal, autonomous, multi-zone

CECs

Corneal epithelial cells

scRNA-seq

Single-cell RNA sequencing

UMAP

Uniform Manifold Approximation and Projection

RPC

Retinal progenitor cells

PCs

Proliferative cells

RO

Retinal organoids

Author contributions

Study conception and design: LG, ZY; conceived and designed the experiments and interpreted data: LG, HQ and ZQ; the sc-RNA data and statistical analysis: LG and YJ; manuscript drafting, editing and funding: LG and ZY.

Funding

This study was supported by grants from the National Key Research and Development Program of China (2022YFC2502800); CAMS Innovation Fund for Medical Sciences (2019-I2M-5-005); the State Key Laboratory of Ophthalmology, Zhongshan Ophthalmic Center, Sun Yat-sen University; Guangdong Basic Research Center of Excellence for Major Blinding Eye Diseases Prevention and Treatment (2024-YXGG-012); Natural Science Foundation of Guangdong Province grant (to Guilan Li).

Data availability

The materials inquiries can be addressed to the corresponding author.

Declarations

Consent for publication

All authors give their consent for publication.

Competing interests

The authors declare no competing interests.