Mice
rd10 (Pde6brd10, C57BL/6J genetic background) mice were purchased from Jackson Laboratory. BALB/c- Nude mice were purchased from Charles River. All mice used in this research were housed under specific pathogen-free conditions. All procedures conducted on these animals were in compliance with ethical animal license protocols and received approval from the Laboratory Animal Ethics Committee of Wenzhou Medical University (Research License wydw2023-0223).
Cell lines
H9 hESCs (WiCell) and their derivatives were cultured in accordance with established procedures68. In brief, the cells were grown on plates coated with Matrigel (Corning, 356231) in TeSR-E8 medium (Stem cell Technologies, 5990) at 37°C within a 5% CO2 incubator. The culture medium was refreshed daily, and cell morphology was observed using an inverted fluorescence microscope (ZEISS). Passaging of cells was performed routinely every 4–5 days using 0.5 µM EDTA solutions (Gibco, AM9261), and they were transferred to Matrigel-coated plates in TeSR-E8 medium supplemented with 10 µM Y-27632 (Selleck, S1049) to uphold their pluripotent state. HEK-293 cells were cultured in DMEM (Dulbecco's Modified Eagle Medium) supplemented with 10% FBS (Gibco, 10099158).
Retina samples collection
Fetal eye samples were acquired from voluntarily donated aborted fetuses between gestational weeks 16 to 22. For 10x Genomics snATAC&GEX-Seq analysis, the ciliary margin zone and neural retina regions of the eye samples were meticulously dissected separately under a microscope and stored at -80°C for subsequent analysis. For Stereo-sequencing, the eye sample was snap-frozen in liquid nitrogen pre-chilled isopentane in Tissue-Tek OCT (Sakura, 4583) and then transferred to a -80℃ freezer for storage before the experiment. For immunocytochemistry and H&E staining, the samples were fixed in 4% paraformaldehyde (Beyotime, P0099) for 1 h, followed by gradient sedimentation using 10%, 20%, and 30% sucrose solutions (Beyotime, ST1672). Subsequently, they were embedded in NEG-50 (Thermo, 6502) and stored at -80°C for further processing. For cell isolation and culture, the samples were collected in ice-cold PBS (Gibco, c14190500cp), gently fragmented into small pieces, and then centrifuged at 200g for 2 min. The supernatant was removed, and 1 mL of digestion buffer (consisting of 2 mg/mL collagenase IV (Gibco, 17104019), 10 U/µL DNase I (NEB, 11284932001), and 1 mg/mL papain (Sigma, P4762) in PBS) was added. The tissues were incubated at 37°C for approximately 15–20 min, with pipetting every 5 min to aid in tissue digestion and obtain single cells. Finally, an equivalent volume of neutralization buffer (10% FBS in PBS) was added to halt the digestion process.
Human retinal organoid differentiation
Human retinal organoid differentiation was conducted following established protocols34,68. H9 hESCs were digested into single cells using TrypLE (Gibco, 12563-011) containing 0.05 mg/mL DNase I and 20 µM Y-27632. The cell concentration was adjusted to 120,000 cells/mL using Type I differentiation medium, which consisted of 45% IMDM (Thermo, 12440053), 45% F12 (Thermo, 11765054), 10% KSR (Gibco, 10828010), 1% GlutaMAX (Thermo, 35050061), 450 µM Monothioglycero (Sigma, M6145), 1% Pen Strep (Gibco, 15140122) and 20 µM Y-27632. The cells were seeded into a low-adhesion V-bottom 96-well plate at a volume of 100 µL per well. On the 6th day of culture, half of the Type I differentiation medium was replaced with fresh medium containing 3 nM hBMP4 (R&D, 314-BP). This medium replacement was repeated on the 9th, 12th, and 15th days with fresh medium. On the 18th day, the cell aggregates were transferred to a 10 cm non-adherent culture dish, and the culture medium was switched to Type II differentiation medium. Type II medium consisted of 90% DMEM/F12-Glutamax medium (Thermo, 10565042), 10% FBS (Biological Industries, 04-001-1A), 1% N2, 0.1 mM Taurine (Sigma, T8691), 0.5 µM Retinoic acid (Sigma, R2625), and 1% Pen Strep. Long-term culture was maintained under high oxygen conditions (40% O2), and the state and morphology of the organoids were continuously monitored in real-time using an inverted fluorescence microscope.
Fetal retina sample processing and single-cell multiomic sequencing
Preparation of Single-Cell Nuclear Suspension: Freshly frozen retina samples are selected for the preparation of single-cell nuclear suspensions through a series of steps, including cell membrane lysis, filtration, washing, and density gradient centrifugation or flow cytometry sorting. Subsequently, quality control tests are performed to assess nuclear integrity and concentration. It is required that the nuclear membrane remains intact, without noticeable impurities and fragments. Cell viability should be less than 5%, the aggregation rate should be less than 5%, nuclear concentration should fall within the range of 2000–3000 nuclei/µL, and the volume should exceed 60 µL. Wash the fetal retina samples twice with PBS, use digestion solution to digest at 37°C for 10 min to prepare single cells, and perform quality control tests for nuclear integrity and concentration. Single-Cell Sequencing: Following the manufacturer's instructions provided in the CG000338 Chromium NextGEM Single Cell Multiome ATAC_GEX (10x Genomics, PN-1000283) User Guide RevE, use the Chromium Controller to generate GEMs (Gel Bead-In-Emulsions) by combining the single-cell nucleus suspension/transposase with Gel Beads and oil droplets. Within the GEM oil droplets, the gel beads are lysed, and under PCR conditions, oligos containing Poly (dT) capture RNA, while oligos containing spacers capture transposed DNA sequences, both of which are labeled with 10x barcodes corresponding to the target sequences. Subsequently, the GEMs are broken, and the target fragments captured by magnetic beads are recovered. After pre-amplification, a 160 µL mixture is obtained, including complete DNA sequences with read1N and read2N, as well as amplified transcriptomic cDNA sequences. From the pre-amplification product, 40 µL is taken for ATAC library construction, directly adding a sample index for amplification. After purification, an ATAC library is obtained, which undergoes quality inspection. At the same time, 35 µL of the pre-amplification product is taken for cDNA amplification and purification. The purified cDNA undergoes fragmentation, end repair, A-tailing, adapter ligation, and dual-index amplification to complete library construction. The constructed library undergoes quality inspection, and Illumina NovaSeq 6000 sequencer is used to perform sequencing of the transcriptome using the PE150 sequencing strategy. The ATAC library is sequenced using the PE50 sequencing strategy.
Multiomic sequencing data processing and analysis
Single-cell multi-omic sequenced files were processed for demultiplexing and analyzed using Cell Ranger-ARC (v2.0.2). We used the human reference genome GRCh38 as reference to map genes. Single- nucleus RNA sequencing (snRNA-seq) data were analyzed using Seurat (v4.1.1)59, and single- nucleus ATAC (snATAC-seq) data were analyzed using Signac (v1.9.0) 73. For snATAC-seq data, calling peaks were used the MACS2 and the genomic positions were mapped and annotated with the reference human genome EnsDb.Hsapein.v86 and hg38. To obtain high-qualified cells, we adopted a stringent criterion to filter out low quality cells with < 1000 or > 20000 expressed gene counts, < 1000 or > 20000 fragment counts, > 0.05% blacklist fraction, < 2 transcription start site (TSS) enrichment, > 2% nucleosome signal, > 5% mitochondrial rate and > 10% ribosomal rate. Due to nuclei-derived samples indeed not contain valid cytoplasmic mitochondrial content; dead cells were excluded by removing cells with very low RNA yields.
Furthermore, we performed downstream analyses of data normalization, integration, and assay build. The “SCT” transform method was leveraged to process gene expression data, and the “RunTFIDF” and “RunSVD” function were used to process ATAC data. Integrating snRNA-seq and snATAC-seq data were conducted by anchoring using “FindIntegrationAchors”. Both methods of “rpca” and “SCT” normalization were used to dimensional reduction and batch correction for integrated snRNA-seq data, respectively. The “lsi” method was applied along with “RunTFIDF” and “RunSVD” for dimensional reduction and batch correction of integrated snATAC-seq data. Based on reductions from both gene expression and ATAC data, “FindMultimodalNeighbors” was utilized to obtain a joint neighbor graph, and “FindClusters” with SLM algorithm was utilized to identify cell clusters with the cluster resolution set to 0.5. The function of “FindMarkers” based on the Wilcoxon sum-rank test method was used to determine highly expressed genes, and “LR” test method was used to determine up-regulated motif accessibility. Based on differential feature analysis results, we annotated cell types in the integrated datasets by using well-known marker genes. The Uniform Manifold Approximation and Projection (UMAP) method was applied to visualize the integrated single-cell data distribution in two-dimensional space.
To calculate RNA velocity, we used both velocyto (v0.17.17) 74 and scVelo (v0.2.4) 50 to analyze reads that passed the quality control after clustering as instructed. First, the standard velocyto pipeline was run to count spliced and unspliced reads for each sample based on the filtered CellRanger-generated bam files. Then the output loom file was used as input for scVelo based on the dynamic model to estimate velocity embedding. We also applied the partition-based graph abstraction method (PAGA, v1.2)52, a function of Scappy (v1.9.3) 75 python package, to infer the potential differential trajectory with default setting.
Stereo-sequencing and analysis
Tissue processing and in situ reverse transcription: Cryosections of the eye sample were cut at a thickness of 10 µm using a Leika CM1950 cryostat. The sections were attached to the surface of the Stereo-seq chip (BGI) and incubated at 37°C for 3 min. After that, the sections were fixed in methanol and incubated at -20°C for 40 min before preparing the Stereo-seq library. The same sections were stained with a nucleic acid dye (Thermo, Q10212). After washed with 0.1x SSC buffer (Thermo, AM9770) supplemented with 0.05 U/mL RNase inhibitor (NEB, M0314L), the sections placed on the chip were permeabilized using 0.1% pepsin (Sigma, P7000) in 0.01 M HCl buffer, incubated at 37℃ for 5 min and then washed with 0.1x SSC buffer supplemented with 0.05 U/mL RNase inhibitor. RNA released from the permeabilized tissue and captured by the DNB was reverse transcribed overnight at 42℃ using SuperScript II (Invitrogen, 18064-014). After reverse transcription, tissue sections were washed twice with 0.1x SSC buffer and digested with Tissue Removal buffer at 55℃ for 10 min. cDNA-containing chips were then subjected to Prepare cDNA Release Mix treatment for overnight at 55℃. cDNA was purified using the VAHTSTM DNA Clean Beads (0.8×).
Library construction and sequencing: The resulting cDNAs were amplified with KAPA HiFi Hotstart Ready Mix (Roche, KK2602) with 0.8 mM cDNA-PCR primer. The concentrations of the resulting PCR products were quantified by QubitTM dsDNA Assay Kit (Thermo, Q32854). A total of 20 ng of DNA were then fragmented with in-house Tn5 transposase at 55℃ for 10 min, after which the reactions were stopped by the addition of 0.02% SDS and gently mixing at 37℃ for 5 min after fragmentation. Fragmented products were amplified as described below: 25 mL of fragmentation product, 1x KAPA HiFi Hotstart Ready Mix and 0.3 mM Stereo-seq-Library-F primer, 0.3 mM Stereo-seq-Library-R primer in a total volume of 100 mL with the addition of nuclease-free H2O. The reaction was then run as: 1 cycle of 95℃ 5 min, 13 cycles of 98℃ 20 seconds, 58℃ 20 seconds and 72℃ 30 seconds, and 1 cycle of 72℃ 5 min. PCR products were purified using the AMPure XP Beads (0.63 and 0.153), used for DNB generation and finally sequenced on MGI DNBSEQ-Tx sequencer.
Stereo-seq raw data processing: Fastq files were generated using a MGI DNBSEQ-Tx sequencer. CID and MID are contained in the read 1 (CID: 1–25 bp, MID: 26–35 bp) while the read 2 consist of the cDNA sequences. CID sequences on the first reads were first mapped to the designed coordinates of the in situ captured chip achieved from the first round of sequencing, allowing 1 base mismatch to correct for sequencing and PCR errors. Reads with MID containing either N bases or more than 2 bases with quality score lower than 10 were filtered out. CID and MID associated with each read were appended to each read header. Retained reads were then aligned to the reference genome GRCh38-3.0.0 using STAR (DOI: 10.1093/bioinformatics/bts635) and mapped reads with MAPQ > 10 were counted and annotated to their corresponding genes. UMI with the same CID and the same gene locus were collapsed, allowing 1 mismatch to correct for sequencing and PCR errors. Finally, this information was used to generate a CID-containing expression profile matrix. The whole procedure was integrated into a publicly available pipeline SAW available at https://github.com/BGIResearch/SAW.
Stereo-seq clustering and analysis: Expression profile matrix was divided into non-overlapping 26521 bins covering an area of 14650 x 18400 DNB. The resulting cells were further processed by Seurat (34062119) followed by SCTransform and scaling. 2000 feature genes were selected and top 20 principal components with the highest explained variance from the PCA results were selected to enhance our analysis. To classify the cell bins accurately and identify different cell types effectively, a resolution of 4 was chosen for clustering analysis in order to achieve precise results. Subsequently, clusters located in the CMZ region were isolated based on their spatial location and marker genes specific to this region. A re-clustering analysis was performed on this isolated CMZ region using a resolution of 1 and relying on marker genes defining RPE.stem-like cells and hNRSC. ‘FindMarkers’ function was also used to calculated the highly expressed genes for different clusters and celltypes. As for cells in other regions, annotation of all clusters was conducted based on their spatial location, different expression genes as well as known cell type markers obtained from single-cell data analysis. Finally, similar cell types were merged together after careful consideration.
Stereo-seq pseudotimes analysis: To infer the cell lineage of three interested cell types, namely hNRSC, RPCs, and PCs, we employed Slingshot (https://doi.org/10.1186/s12864-018-4772-0) for pseudotimes analysis. Slingshot stands out as an exceptionally robust and versatile tool that integrates stable techniques suitable for handling noisy single-cell data while effectively identifying multiple trajectories. It also applicability in Stereo-seq data scenarios. Furthermore, given the simplicity of the targeted cell populations and the presence of a solitary trajectory path, Slingshot enables swift inference of trajectories directly based on the original dimensionality-reduced data from Seurat. The technical support related to Stereo-seq in this study was provided by Annoroad Gene Technology (Beijing) Co., Ltd.
Organoid sample preparation and single-nucleus RNA sequencing
Following the manufacturer's guidelines for 10x Genomics single-nucleus RNA Sequencing, 5–10 hRO samples were chosen, and a series of steps were executed, including homogenization, cell membrane lysis, filtration, washing, and nuclear purification, with the aim of isolating single-cell nuclei. Subsequent to these procedures, quality control tests were conducted to assess nuclear integrity and concentration. The single-cell nuclei were then processed in accordance with the 10x Genomics Chromium Next GEM Single Cell3_v3.1_Rev_D (10x Genomics, PN-1000121) protocol for library construction and quality control. In brief, single nuclei were suspended in PBS containing 0.04% BSA. The nuclei suspension was loaded onto the Chromium Next GEM Chip G (10x Genomics, PN-1000120), and the Chromium Controller was utilized to create single-cell gel beads in the emulsion (GEMs) following the manufacturer's recommendations. Captured nuclei were lysed, and the resulting RNA was barcoded through reverse transcription within individual GEMs. This process generated barcoded, full-length cDNA, and libraries were subsequently constructed according to the manufacturer's protocol. Quality assessment of the libraries was conducted using Qubit 4.0 and the Agilent 2100, and sequencing was carried out on the Illumina NovaSeq 6000, aiming for a sequencing depth of at least 50,000 reads per nucleus with 150 bp (PE150) paired-end reads.
Single nucleus analysis
Raw single-nucleus RNA sequencing (snRNA-seq) data were processed by using the standard CellRanger pipeline to demultiplex and align sequencing output to the GRCh38 human reference genome. Quality control for organoids cells was performed using Seurat (v4.1.1)59. We filtered out low-quality cells with < 500 or > 5000 detectable genes, < 1000 expressed gene counts, > 2% mitochondrial rate, and > 3% ribosomal rate. The log normalizing and scaling for unique molecular identifier (UMI) counts was performed by the SCTransform function in Seurat package. FindVariableFeatures was performed using a variance-stabilizing transformation to identify the top 3000 highly-variable genes. Principal component analysis (PCA) was conducted to calculate principal components (PCs) that could explain most of the single-cell dataset through leveraging these highly variable genes. Batch effects among samples were removed by R package harmony (v0.1.0)76 based on top 50 PCs. The shared nearest neighbor graph method was used to identify cell clusters with the resolution of 0.4. Similar with annotation of fetal eye, we assigned these clusters to organoid retinal cell types by using well-known specific marker genes. We used the UMAP method to visualize single-cell data distribution in two-dimensional space.
Developmental trajectory analysis
To infer the developmental trajectory, we applied both Monocle2 (v2.28.0)77 and Monocle3 (v1.3.1)51 to calculate the pseudotimes of cells from both fetal and hRO data snRNA-seq and snATAC-seq data, respectively. The standard protocol of Monocle2 and Monocle3 with default parameters was adopted to order cells into potentially differentiation trajectories, followed by optional statistical tests to find genes those alterations in expression over trajectories. The cell trajectories of Monocle2 were visualized for different cell types and different differentiation states, and the results of Monocle3 trajectory analysis were visualized in two-dimensional scatter plot. The pseudotime was calculated by setting hNRLCs at the root state.
Calculating the stemness scores of single cells
To evaluate the stemness of cells in lineage 1 for both fetal and hRO snRNA-seq data, we used the CytoTRACE (v0.3.3)53 to estimate transcriptional diversity of each individual cell in terms of differential or stemness status. The cells in lineage 1 were given a stemness score ranging from 0 to 1 according to their differential potential. The higher score represents higher stemness and less differentiation, and the lower score represents lower stemness and more differentiation. Furthermore, we also leveraged an orthogonal method of Stem_ID/RaceID54,78, an algorithm designed for discerning stem cells from all cell types, to validate the stemness of cells in lineage 1. By applying the “compentropy” function of the RaceID (v0.3.0) R package78, we calculated the entropy of each cell from lineage 1 in both fetal and organoid data. Since entropy inversely correlates with cell differentiation state79,80, thus, cells with high entropy indicate higher functional uncertainty and more differentiation potential (namely high stemness), while cells with low entropy indicate more differentiated states with constrained cell fates and functionalities (namely low stemness).
Regulon activity analysis
To assess the regulation intensity of transcription factors (TFs), we leveraged the workflow of single-cell regulatory network inference and clustering (pySCENIC v0.12.0) to calculate the activity of regulons based on the TFs and their target genes56. Following the standard pySCENIC protocol, we leveraged GRNboost to identify potential genes that are co-expressed with TFs. To uncover putative direct-binding targets, RcisTarget was used to perform cis-regulatory motif analysis for each co-expression module. Only significant regulons were included for further analysis, and the activity of each regulon in each cell was estimated by the AUCell method. To connect regulons with cell types, we leveraged the Wilcoxon rank-sum test to prioritize cell type-specific regulons with the AUC scores. The AUC scores for each TF in each cell type were averaged for visualization of heatmap.
Inferring gene regulatory network based on single-cell multiomic sequencing data
We applied Pando (v1.0.5)55 to fetal retinal data of multimodal single-cell measurements, where both RNA and ATAC components are measured in the same cell, to infer gene regulatory network (GRN). First, Pando selects candidate regions for GRN inference based on evolutionary conservation regions, prior cis-regulatory element regions, and data-driven accessible peaks from snATAC-seq data. Based on an extended motif database that obtains binding motifs and binding site predictions from JASPAR (2020 release)81, CIS-BP database82, and AnimalTFDB83, Pando identifies putative cis-regulatory elements and putative trans-regulators (that is, TFs) of each gene. Subsequently, Pando uses a linear regression model to carry out the inference of the regulatory interactions between TF-binding site pairs and the target gene. To fit the linear regression model, the function glm from the stats R package using Gaussian noise was used. The fitted coefficients were examined for statistical significance using analysis of variance (ANOVA) to determine significant TF-motif-target triplets for pruning the network. To visualize the inferred TF network, the pairwise Pearson correlations of log-normalized expression of all TFs in the network across all cells were calculated. Based on the correlation value and fitted coefficients, Pando computes a combined TF-gene linkage score matrix and uses this matrix to perform principal component analysis (PCA). Top 20 PCs were used to generate the UMAP embedding of the TFs via the uwot R package with default parameters.
Jaccard similarity index
Immunocytochemistry and H&E staining
To prepare the organoids, they were fixed in 4% paraformaldehyde at 4℃ for 45 min. Subsequently, they were embedded in NEG-50 and cryosectioned at 14–16 µm thickness on slides using a Leica cryostat. The cryosections were then either utilized for immunocytochemistry or stored at -80°C for later use. Retina tissue samples were processed for cryosections using standard protocols. The procedure involved dissection of the retina tissue, followed by post fixation in 4% paraformaldehyde for 1 h. Gradient sedimentation is then performed with 10%, 20%, 30% sucrose solutions. The cryoprotected samples were then embedded in NEG-50 and cryosectioned at 16 µm thickness. Monolayer cells were fixed with 4% paraformaldehyde for 15 min to preserve their morphology. For immunocytochemistry, both the cryosections and monolayer cells were initially blocked and permeabilized in a solution containing 4% BSA (Beyotime, ST2254) and 0.5% Triton X-100 (Solarbio, T8200) for 1 h at room temperature. They were then incubated with primary antibodies overnight at 4°C. After primary antibody staining, the sections and cells were rinsed three times with PBS and subsequently incubated with secondary antibodies in the dark at room temperature for 1 h. Following the removal of the secondary antibody, the nuclei were stained with a 300 nM DAPI stainning solution (Thermo, D1306) for 15 min. Finally, the stained sections and cells were visualized using confocal microscopy (Leica). H&E staining of the sections was performed using the Hematoxylin and Eosin Staining Kit (Beyotime, C0105S). The stained sections were then scanned using the ZEN 2012 (blue edition) scanning system (Zeiss). Antibodies are described in Supplementary Table 4.
RNA-Seq and analysis
For each sample, total RNA was extracted using TRIzol reagent (Thermo, 15596018CN) and subsequently purified using the RNeasy Mini Kit (Qiagen, 74104), following the manufacturer's instructions. The quality and quantity of the RNA were assessed using a NanoDrop 2000 spectrophotometer, Agilent 2100 Bioanalyzer, and Agilent RNA 6000 Nano Kit (Agilent, 5067 − 1511). Annoroad Gene Technology handled the RNA library construction and RNA sequencing. To create the sequencing libraries, the NEB Next Ultra RNA Library Prep Kit for Illumina (NEB, E7770L) was employed. Library clustering was carried out using the HiSeq PE Cluster Kit v4-cBot-HS (Illumina, PE-401-4001), following the manufacturer's recommendations. After cluster generation, the libraries were subjected to sequencing on an Illumina platform, generating 150 bp paired-end reads. The initial analysis of the data was performed utilizing BMKCloud, accessible at http://www.biocloud.net/34.
Clonal sphere assay and induced differentiation
For the RSC clonal sphere assay, select the well-preserved and fresh samples; use a sterile V-Lance blade to separate the CMZ under a microscope. Prepare a digestion solution using 0.25% TE enzyme (with the addition of 50 µg/mL DNase Ⅰ and 20 µg/mL Y-27632). Digest the separated CMZ into single cells (digestion time ≤ 15 min). Resuspend the cells in serum-free (SFM) culture medium (48% DMEM (Gibco, C11995500bt), 48% F-12, 20 ng/mL bFGF (PeproTech, AF-100-18C), 20 ng/mL EGF (PeproTech, AF-100-15), 2% B27 (Gibco, 17504044), 1% N2 (Gibco, 17502048), 5 ng/mL heparin (Sigma, H3393) and 1% Pen Strep), and seed them in a pre-prepared 12-well culture plate. Perform a half-media change on day 3 and day 5, and incubate at 37°C in a carbon dioxide incubator for 7 days. After culturing the CMZ spheres for 7 days, transfer them to a 12-well plate treated with either poly-L-lysine or laminin. Use SFM culture medium containing 5% FBS for adherent culture. After 3 days of culture, the spheres will adhere to the bottom of the dish, and cells will start migrating out of the spheres. Then, transfer the spheres to retinal differentiation medium (88% DMEM/F12, 10% FBS, 0.5 µM RA, 20 ng/mL FGF2, 20 ng/mL EGF, 2 mM L-glutamine, 1% N2 and 1% Pen Strep) for differentiation culture. Change the medium every other day and continue differentiation culture until around day 20.
Establishment of the retinal organoid regeneration model
Select visually healthy and relatively mature hROs, aged over 60 days. Under a microscope, use a sterile V-Lance knife to excise the VSX2-tdTomato-labeled retinal portion, while preserving the rest of the organoid. Maintain the cultured organoids and establish a control group (simultaneous excision of the entire retina and ciliary margin). Utilize confocal microscopy for live-cell imaging to dynamically observe the regenerative repair process mediated by the CMZ-hNRSCs by tracking the VSX2-tdTomato-positive cells. Collect the organoid samples at critical time points of retinal regeneration for subsequent analysis.
EdU cell proliferation assay
Following the manufacturer's instructions, an appropriate amount of EdU (Beyotime, C0071S) was added to the culture medium and the organoids were incubated 24 h in a cell culture incubator. This allowed EdU to be taken up and integrated into newly synthesized DNA by actively proliferating cells within the organoid. After incubation, the cells were fixed in 4% paraformaldehyde at 4°C for 45 min to preserve their morphology. Next, the fixed organoids were embedded in NEG-50 and cryosectioned at 14–16 µm thickness on slides using a Leica cryostat. The sections were then subjected to EdU staining using an EdU detection kit according to the provided instructions. Following EdU staining, the sections were further stained with DAPI for nuclear staining. Finally, the EdU-labeled cells were visualized and recorded using a fluorescence microscope.
10x Genomics visium spatial transcriptomic (ST) analysis
Staining and Imaging: The histological workflow utilized the Visium CytAssist Spatial Gene Expression kit (10x Genomics, PN-1000520) for Fresh Frozen tissues. Cryosections, measuring 10 µm in thickness, were cut and affixed onto GEX arrays. These sections were then placed on a Thermocycler Adaptor with the active surface and incubated at 37℃ for 5 min, followed by fixation in chilled methanol at -20℃ for 30 min. Subsequently, the sections were stained using H&E. Brightfield images were captured using a Leica DMI8 whole-slide scanner at 10x resolution.
Gene Expression and Transfer: Visium CytAssist Spatial Gene Expression analysis was performed using the Visium CytAssist Spatial Gene Expression for Fresh Frozen (FF) kits provided by 10x Genomics (PN-1000520). The tissue slices underwent crosslinking treatment, followed by probe hybridization and ligation on the crosslinked sections. Once successfully ligated, the probes, originally captured within the cells, were liberated through enzymatic digestion of RNA from the tissue sections using RNase, along with the tissue removal enzyme (10x Genomics, PN-3000387). Subsequently, the released probes were extended, eluted, and then transferred to new tubes for further analysis and utilization.
cDNA Library Preparation for Sequencing: The probe extension and library construction procedures followed the established Visium workflow for Fresh Frozen samples. Libraries underwent paired-end dual-indexing sequencing, and afterward, they were sequenced on an Illumina Novaseq6000 sequencer, with a targeted sequencing depth of at least 100,000 reads per spot. The sequencing strategy employed was paired-end 150 bp. CapitalBio Technology conducted this sequencing process.
Data Preprocessing and Quality Control: ST sequencing data were processed using the CellRanger V3 platform. Raw sequencing reads were aligned to the human genome (GRCh38, ENSEMBL). The Space Ranger output files were imported into Seurat (v4.1.1) for subsequent analysis. Raw data underwent quality control to filter out low-quality spots and genes (lowqspot = 0.01, mitper = 25, geneExprMin = 10). Gene expression values were then normalized using SCTransform to account for technical variability and to ensure accurate downstream analyses.
Dimensionality Reduction and Clustering: Normalized data were subjected to dimensionality reduction using Principal Component Analysis (PCA) through the RunPCA function. FindNeighbors and FindClusters (cluster_res = 0.6) functions were applied to yield clusters of spatial transcriptomic (ST) spots representing distinct cellular neighborhoods. To visually explore the spatial organization of ST spots, a Uniform Manifold Approximation and Projection (UMAP) was generated using the RunUMAP function. To uncover genes differentially expressed between spot clusters, the FindAllMarkers function was employed. DotPlot visualizations were generated to present the expression levels of these marker genes across different spot clusters. Utilizing cell type-specific marker genes derived from single-cell data, each spot cluster was annotated with the corresponding cell types. In cases where multiple cell types shared the same spot cluster, these were denoted using a '/' delimiter.
SPOTlight Deconvolution: To map single-cell data onto spatial spots, we employed the SPOTlight method, which employed a deconvolution algorithm based on the differential expression marker genes from snRNA-seq data. The spotlight_deconvolution function calculated the proportional representation of distinct cell types within each spot. This information was then used to generate the SpatialFeaturePlot, illustrating the distribution of various single-cell cell types across the spatial landscape.
Cell Type Annotation and Pseudotemporal Analysis: For insights into pseudotime of ST spots, the spatial.trajectory.pseudotime method from the stlearn package was employed. The pl.gene_plot function, using the "CumSum" approach, depicted the expression level of single-cell marker genes for hNRSCs across spatial spots.
Flow cytometry
The samples underwent two washes with DPBS solution. Subsequently, 1.5 mL of a digestion solution, comprising 0.25% Trypsin-EDTA (Thermo, 25200056), 50 µg/mL DNase Ⅰ, and 20 µg/mL Y-27632, was added to the samples. They were then incubated in a 37°C cell culture incubator for 10–15 min. Once it was confirmed that the majority of the masses had visually disappeared, 3 mL of a neutralizing solution composed of 90% DPBS, 10% FBS, 50 µg/mL DNase Ⅰ, and 20 µg/mL Y-27632 was added to halt the reaction. Subsequently, the single cells resulting from trypsinization underwent multiple washes with 3%FBS. Afterward, they were stained with antibody in PEB buffer (PBS containing 0.5% BSA and 2 mM EDTA) for 30 min while kept on ice. Following staining, the cells were filtered through a 100 µm nylon mesh and subjected to fluorescence analysis using FACSCanto II (Becton Dickinson).
Subretinal transplantation
The transplantation therapy was performed on rd10 mice, a recognized model of retinal degeneration. All experiments conducted in this study followed the guidelines and permissions provided by the Institutional Committee of Laboratory Animal Ethics. A cell suspension (1×105 cells/µL) in a volume of 1-1.5 µL was injected into the subretinal space. Prior to the procedure, 2–4-week-old mice were anesthetized with an intraperitoneal injection of sodium pentobarbital (40 mg/kg, Sigma, P3761). To facilitate the procedure, the pupils were dilated using 1% tropicamide, followed by the application of topical anesthetic proparacaine hydrochloride (0.5%). Injections were carried out using a Hamilton syringe fitted with a 33-gauge needle under an operating microscope. Following the surgery, ofloxacin ointment (Sinqi Pharmaceutical, H10940177) was applied topically for 3 days to prevent dry eyes and infection. All mice, including sham-transplanted, are fed cyclosporine (Sangon, A600352) 24 h before transplantation until 2 weeks post-transplantation to mitigate immune rejection in the mice. The efficacy of the transplantation therapy was then assessed at the specified time points to evaluate its effectiveness.
Electroretinogram Recording
Corneal scotopic flash electroretinogram recordings were conducted on the eyes of rd10 mice from postoperative weeks 2 to 20. In brief, following a 12 h period of dark adaptation, the mice were anesthetized with an intraperitoneal injection of sodium pentobarbital. The pupils of the mice were dilated using 1% tropicamide. To prevent hypothermia, the animals were maintained at a body temperature of 37°C using a heating pad. Two active gold electrodes were positioned on each cornea to serve as the recording electrodes. Reference and ground electrodes were subcutaneously placed in the mid-frontal and tail areas of the head, respectively. The ERG parameters for photopic responses were set as follows: stimulation intensities at 0.48 log candela (cd)•s/m2 (light 3.0). The ERG parameters for scotopic responses were configured as follows: stimulation intensities at − 2.02 log cd•s/m2 (dark 0.01), 0.48 log cd•s/m2 (dark 3.0), and 0.98 log cd•s/m2 (dark 10.0).
Optomotor Response-based visual function assessment
To evaluate visual performance in each experimental group, we employed the optomotor response (qOMR) as a visual-driven behavioral task. The detection system creates a virtual stimulation environment using four screens, with the animal positioned on a platform at the center of the setup. The system tracks the movement of the animal's head in real-time relative to the presented visual stimulus and objectively quantifies the OMR results. We designed and implemented a virtual stimulation protocol that included spatial frequencies ranging from 0.05 to 0.5 cycles per degree (c/deg). Each spatial frequency was randomly presented for 60 seconds, moving at a speed of 12° per second. Clockwise and counter-clockwise head tracing responses were used to assess each eye's performance. The qOMR score is calculated as the ratio of concordant-to-discordant body/head movements in response to the moving gratings displayed on the screens surrounding the animal. A qOMR score of 1.0 served as a threshold for sensory perception: animals with qOMR scores below 1.0 were considered unable to perceive the specific spatial frequency being investigated.
Plasmids Design and Construct
The sgRNA plasmids were constructed by restriction cloning of protospacers downstream of a U6 promoter using BbsI cut site on px330 (Addgene, 68807) by standard protocol. For ligation, single-stranded DNAs were annealed with 4-bp overhangs on both ends of the double-stranded DNAs, with these overhangs acting as a substrate for T4 DNA ligase. Cloning backbones were digested with either BbsI-HF (NEB, R3539S); Ligation was carried out by adding 1 µL T4 DNA ligase (5 U/µL, Thermo, EL0011) to give a total ligation reaction volume of 10 µL. The oligonucleotide sequences used for plasmid construction in this study are listed in Supplementary Table 5.
The shRNA plasmids were constructed by restriction cloning of downstream of U6 promoter using AgeI and EcoRI cut sites on pLKO.1 - TRC cloning vector (Addgene, 10878) by standard protocol. For a typical reaction, DNA oligo was diluted to 100 µM with ddH2O and annealed in NEB 2.1 buffer and heated at 95°C for 2 min before slowly cooling to room temperature. Ligation was carried out by adding 1 µL T4 DNA ligase to give a total ligation reaction volume of 10 µL. The oligonucleotide sequences used for plasmid construction in this study are listed in Supplementary Table 5.
For constructed VSX2-tdTomato plasmid, the P2A-tdTomato DNA sequence was synthesized by GENEWIZ, and inserted in-frame into VSX2 gene Exon 5 before the stop codon. The homology arm DNA sequence was obtained from genomic DNA of H9 ES cells with PCR (Vazyme, P520), gel purified with FastPure Gel DNA Extraction Mini Kit (Vazyme, DC301). Then infusion into HDR vector (Constructed by our lab), which cut by HindIII-HF (NEB, R3104V) and MfeI-HF (NEB, R3589V), together with P2A-tdTomato with the Uniclone One Step Seamless Cloning Kit (Genesand, SC612). The oligonucleotide sequences and vector sequence used for plasmid construction in this study are listed in Supplementary Table 5.
Lentiviral package and transfection protocol
For lentivirus packaging, HEK-293 cells were seeded at 2.5×106 cells per dish on 10 cm dishes in DMEM medium62. 24 h after seeding, 5 µg psPAX2 plasmid (Addgene, 12260), 2.5 µg pMD2.G plasmid (Addgene, 12259) and 5 µg shRNA plasmid were mixed and prepared with transfection reagent (Yeason, 40802ES08) following the recommended protocol from the vendor. 72 h after transfecting, lentivirus was collected following the recommended protocol, concentrated overnight using Universal Virus Concentration Kit (Beyotime, C2901S) and used within 2 days to transduce H9 hESCs without a freeze thaw cycle.
For electroporation, hESCs were dissociated into single cells using Accutase solution (Stem Cell Technologies). About 1 × 106 cells were resuspended in a nucleofector solution, prepared following the manufacturer's instructions. This solution was prepared by combining 82 µL of P3 primary cell solution with 18 µL of supplement 1 (Lonza, V4XP-3032). To this mixture, 5 µg of a plasmid mixture was added, which included 2.5 µg of sgRNA plasmid and 2.5 µg of the VSX2-tdTomato targeting vector. The cell-plasmid mixture was then transferred into a nucleofection cuvette (Lonza). The cells inside the nucleofection cuvette were nucleofected using program CA137 on the Nucleofector 4D (Lonza). Following nucleofection, the cells were gently transferred to Matrigel-coated plates containing TeSR-E8 medium supplemented with 10 µM Y-27632 and cultured in a 37°C, 5% CO2 incubator. After electroporation, the cells were exposed to 2 µg/mL puromycin (Yeason, 60210ES25) for approximately 7 days. Subsequently, puromycin-resistant clones were selected and expanded for genotyping.
Genomic DNA extraction and genotype
Genomic DNA extraction of cells is subsequently performed using FastPure Blood/Cell/Tissue/Bacteria DNA Isolation Mini Kit (Vazyme, DC112). The targeted region from collected genomic DNA was amplified using PCR (Phanta Flash Master Mix, Vazyme, P520) and sequenced using a Sanger sequencing platform (GENEWIZ). The PCR included 200 ng of genomic DNA and 0.5 µL to 10 nM of forward and reverse primers in a final reaction volume of 40 µL. In the PCR, samples were incubated for 3 min at 95°C; 10 s at 95°C, 5 s at 65°C and 15 s at 72°C for 35 cycles, and 1 min at 72°C. After the first PCR step, products were assessed on a 1.2% TAE gel; products were purified using FastPure Gel DNA Extraction Mini Kit (Vazyme, DC301).
Tissue RNA extraction, cDNA preparation, and quantitative real-time PCR
After lentivirus infection, the CMZ was collected by dissected and each was homogenized separately in a ball mill. The total RNA was extracted using FastPure Cell/Tissue Total RNA Isolation Kit V2 (Vazyme, RC112) according to the manufacturer’s instructions. The RNAs were reverse-transcribed using HiScript III 1st Strand cDNA Synthesis Kit (+ gDNA wiper) (Vazyme, R312) to synthesize cDNA. qRT-PCR was performed on a CFX Manager Real-Time PCR system (Bio-Rad) using specific primers and Taq Pro Universal SYBR qPCR Master Mix (Vazyme, Q712). The results were calculated as ΔΔCT using GAPDH as internal reference transcript. ΔΔCT was calculated as ΔCT of the Knock down–ΔCT of the control. To work out the fold of gene expression, we performed 2-ΔΔCT. The primer sequences were provided in Supplementary Table 5.
Statistical analyses
All experiments were carried out autonomously and replicated a minimum of three times. The outcomes are expressed as the mean ± SD. For comparing two groups, an unpaired two-tailed Student's t-test was employed, whereas one-way ANOVA with Tukey's test or Dunnett's multiple comparisons test was utilized for contrasting multiple groups. Functional enrichment analyses were conducted to identify differential expression genes enriched GO-term biological processes using WebGestalt84 and biological pathways using Metascape85. The statistical analysis was conducted using SPSS Statistics 19.0 software. Statistical significance was established for p values less than 0.05.