The BC1 hiPSC line was a gift from Professor Linzhao Cheng (University of Science and Technology of China) (Chou et al., 2011). HiPSCs were cultured in mTeSR1 medium (Stemcell Technologies) on Matrigel (Corning) coated 6-well plates, and passaged every 5–7 days at about 80% confluence according to our previously published protocol (Guan et al., 2021).

The single guide RNA (sgRNA) was designed by the online website (http://crispor.tefor.net/). The sgRNA (ACA​GCT​GAA​CAG​TTG​GCA​TC), which targeted near the TGA stop codon of exon 3 of RCVRN and was predicted to have over 50% cleavage activity, was cloned into pSpCas9 (BB)-2A-Puro (PX459; Addgene) to generate the CRISPR/Cas9 guide-carrying plasmids PX459-RCVRN-sgRNA. For targeting plasmid construction, first, P2A-eGFP fragment was amplified from pBluescript-LA-P2A-eGFP-RA (VSX2) (a gift from Professor Mengqing Xiang, Zhongshan Ophthalmic Center, Sun Yat-sen University, China), and the left and right homologous arms were amplified from the BC1 hiPSC genome. Second, these fragments were further amplified by primers with infusion homologous sequence. Third, these three fragments were linked together with pMD19-T vector (Takara; 6013) by in-Fusion^® HD Cloning Kit (Takara; 639648). The specific structure of the donor plasmid was shown in Supplementary Figure S1, and the PCR primers for targeting plasmid construction were listed in Supplementary Table S1.

Then, the BC1 hiPSCs were digested into single cells using an accutase digestive enzyme (Gibco). Approximately 1 × 10⁶ cells in 100 μl Buffer R solution were mixed with 10 μg guide-carrying plasmids and 10 μg targeting plasmids according to the manufacturer’s instruction (Neon; MPK10025). Then the mixture was electroporated with Neon™ transfection system (MPK5000) under the manufacturer’s recommended procedure: 1,100 v; 10 ms; 3 pulses. After electroporation, cells were transferred into a Matrigel-coated 24-well plate containing mTesR1 medium supplemented with 10 mM Y-27632 (Stemcell Technologies) in an incubator at 37°C and 5% CO₂. Three days later, the cells were treated with 0.4 μg/ml puromycin for 7 days to select the resistant clones. Finally, the selected clones were further confirmed by polymerase chain reaction (PCR) and by Sanger sequencing.

The genomic DNA was extracted using an Animal Genomic DNA Quick Extraction Kit (Beyotime, China) according to the manufacturer’s protocol. PCR was performed using 2× Taq MasterMix (Takara). The following PCR primers were used: RCVRN, Forward, AGA​TGT​CAG​CCT​CAT​CGC​AG; Reverse, CAT​CGC​AAC​ACA​GCC​TTG​TC. The forward primer was designed to target the 5′ upstream outside of the left homologous arm so that the correct insertion could be identified by one round of PCR experiment. The PCR conditions were: 95°C for 3 min, followed by 35 cycles of 95°C for 15 s, 55°C for 15 s and 72°C for 1 min, and the final cycle of 72°C for 5 min. The PCR products were resolved on a 1.5% agarose gel by electrophoresis. Sanger sequencing was performed by Sangon Biotech (Shanghai, China) Co.

About 2 × 10⁶ reporter hiPSCs were collected and sent to Nanjing Cobioer Biosciences Company (Nanjing, China) for short tandem repeats (STR) analysis. The STR result was compared with that of the parental hiPSCs.

The differentiation of RCVRN-eGFP hiPSCs into ROs was performed according to our previously published protocol (Guan et al., 2021). Briefly, hiPSCs were dissociated and cultured in suspension at day (D) 0 to form embryoid bodies (EBs). On D1 to D3, EBs were gradually switched to neural induction medium (NIM) containing DMEM/F12 (1:1) (Gibco), 1% N2 supplement (Gibco), 1% minimum essential media nonessential amino acids (MEM NEAA, Gibco), and 2 mg/ml heparin (Sigma-Aldrich). During D5 to D7, EBs were seeded on Matrigel-coated dishes with NIM. The medium was changed to retinal differentiation medium (RDM) containing DMEM/F12 (3:1), 2% B27 (without vitamin A, Gibco), 1% MEM NEAA, and 1% Antibiotic-Antimycotic (Gibco) since D16. Typical optical vesicles (OVs) appeared around D25 with adjacent pigmented retinal pigment epithelium cells (RPEs). They were mechanically picked up and switched to suspension culture for the development and maturation of ROs. For long-term culture, the RDM was supplemented with 10% fetal bovine serum (FBS, Natocor), 100 μM Taurine (Sigma-Aldrich), and 2 mM GlutaMAX (Gibco) since D42, and then B27 was replaced with 1% N2 since D90. The medium was changed twice a week. Bright-field and fluorescence images were taken using an Axio observer 7 microscope (Zeiss).

The hiPSCs were fixed using 4% paraformaldehyde (PFA) for 10 min at room temperature (RT). ROs were fixed using 4% PFA for 30 min at RT and dehydrated in gradient sucrose solutions at 4°C, 6.25% for 30 min, 12.5% for 30 min, and 25% overnight before embedding in O.C.T compound. Cryosections were mounted on slides for immunofluorescence (IF) staining according to previously described protocols (Li et al., 2019). Briefly, cells or sections were blocked and permeabilized with 0.25% Triton X-100 and 10% donkey serum for 1 h at RT, then incubated with primary antibodies at 4°C overnight and incubated with the corresponding secondary antibodies for 1 h at RT in the dark. DAPI (Dojindo) was used to counterstain nuclei. All antibodies used are listed in Supplementary Table S2. IF images were taken with an LSM 880 confocal microscope (Zeiss) and an Axio Scan Z1 (Zeiss). Images were processed in Zen 2.3 (Zeiss) and Adobe Illustrator CS6 (Adobe Systems Incorporated).

Karyotype analysis was performed using G-band staining of chromosomes. The reporter hiPSCs were cultured on Matrigel-coated 6-well plates with mTeSR1 medium until reaching 50% confluences. Colchicine was added into the culture medium to a final concentration of 0.8 μg/ml. The reporter hiPSCs were cultured for another 2.5 h. Then the cells were digested with 0.25% Trypsin (Invitrogen) for 1 min at 37°C, collected, and sent to Guangzhou Daan Clinical Laboratory Center (Guangzhou, China) for further analysis. Chromosomes were classified according to the International System for Human Cytogenetic Nomenclature (Simons et al., 2013).

1.5–2.5 × 10⁶ reporter hiPSCs with 30% Matrigel were injected intramuscularly into the hind limb of 6-week old immunocompromised NOD-SCID mice. Animals were monitored each week and after 5–7 weeks, teratomas were dissected and fixed in formalin, embedded in paraffin, sectioned, and stained with Hematoxylin and Eosin. Images were taken with an Axio Scan Z1 (Zeiss). The study design and experimental protocols were approved by the Animal Ethics Committee of the Zhongshan Ophthalmic Center, Sun Yat-sen University. All experimental procedures involving animals adhered to the Association Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research.

Total RNA was extracted using TRIzol Reagent (Invitrogen), and potential DNA contamination was removed by PrimeScript RT reagent Kit with gDNA Eraser (Perfect Real Time) (Takara). The quality of RNA was evaluated using a SMA4000 Ultra Micro Spectrophotometer (Merinton). Reverse transcription was performed using the PrimeScript RT reagent Kit with gDNA Eraser (Perfect Real Time) (Takara). Quantitative PCR was performed with Hieff UNICON qPCR SYBR Green Master Mix (YeaSen) using an ABI7300 fluorescence quantitative PCR instrument (Thermo Fisher Scientific) following the manufacturer’s instructions. 2^−ΔΔCt method was used to analyze data. The BC1 hiPSCs were used as references and GAPDH was used as the internal control. All primer sequences are listed in Supplementary Table S3.

ROs derived from RCVRN-eGFP reporter line were collected at D150 and mechanically separated into small pieces by a needle with 1 ml syringe. Then the clumps were dissociated into single cells by a papain dissociation system (Worthington Biochemical) according to the manufacturer’s instructions. Cell suspensions were resuspended in sorting buffer (PBS containing 1 mM EDTA (Gibco), 2% [vol/vol] FBS). Cells were sorted at 4°C by BD FACSAria III Cytometer (BD Biosciences). Both eGFP-positive (eGFP+) and negative (eGFP-) cells were separately collected in collecting buffer [50% culture medium and 50% (vol/vol) FBS]. BD FACSDiva v8.0.1 software (BD Biosciences) was used for data file collection and data analysis.

Cells from D150-ROs derived from the RCVRN-eGFP reporter hiPSCs were used to perform RNA-seq. Unsorted, FACS-sorted eGFP+ and eGFP- cells (30 ROs per experiment, 2 independent experiments) were respectively collected in TRIzol reagent (Invitrogen) and stored in a −80°C freezer until submitted to Gene Denovo Biotechnology Co. (Guangzhou, China) for RNA extraction, library preparation, sequencing and data analyses. Total RNA was extracted using TRIzol reagent kit (Invitrogen) according to the manufacturer’s protocol. The quality of the RNA was assessed by an Agilent 2100 Bioanalyzer (Agilent Technologies) and checked using RNase free agarose gel electrophoresis. Oligo (dT) beads were used to isolate the poly mRNA from the total RNA. Then the enriched mRNA was subjected to RNA-seq library construction.

The mapped reads of each sample were assembled by StringTie v1.3.1 (Pertea et al., 2015). Gene expression levels were quantified by software RSEM (Li and Dewey, 2011). For each transcription, a fragment per kilobase of transcript per million mapped reads (FPKM) value was calculated to quantify its expression, abundance, and variations. Correlation analysis was performed by R. correlation of two parallel experiments. Differential expression analysis was performed by DESeq2 (1.18.0) between two different groups and by edgeR between two samples (Robinson et al., 2010). Unless otherwise stated, the genes/transcripts with the parameter of false discovery rate (FDR) below 0.05 and absolute fold change (FC) > 2 were considered differentially expressed genes (DEGs). The Gene Ontology (GO) enrichment analysis with a Q-value < 0.05 was defined as significantly enriched.

All data are presented as mean ± SD. Comparisons between two groups were analyzed using a two-tailed Student’s t-test. p-value < 0.05 was considered statistically significant. When more than three groups, a two-way ANOVA analysis was performed. A p-value < 0.05 was considered significant. GraphPad Prism version 9.2.0 (GraphPad Software) was used for the calculation of probability values. For statistical analysis of RNA-seq, see RNA-seq and data analysis for more details.

To generate the RCVRN-reporter hiPSC line, a viral P2A-eGFP fused sequence was designed to insert before the RCVRN exon 3 TGA stop codon of the genome in the BC1 hiPSC line (Figure 1A). After selection of puromycin resistance, one of the surviving clones was verified as the correctly targeted clone by PCR and Sanger sequencing (Figures 1B,C). This clone was expanded and confirmed with typical hiPSC morphology and normal karyotype (Figures 1D,E). STR analysis displayed the genetic matches between the reporter and the parental hiPSCs.

Subsequently, the pluripotency and multi-lineage differentiation ability of the reporter hiPSCs were characterized. The reporter hiPSCs expressed typical pluripotency marker proteins NANOG, SSEA4, OCT4, and TRA-1-60 (Figures 1F–I). qRT-PCR further showed that the reporter hiPSCs had similar mRNA expression levels in pluripotency markers to the parental ones, including OCT4, SOX2, NANOG, GDF3, and DNMT3B (Supplementary Figure S2A). Teratoma formation assays confirmed the multi-lineage differentiation ability of the reporter hiPSCs that differentiated into three germ layers, including pigment cells (ectoderm), cartilage (mesoderm), and gut-like epithelium (endoderm) (Figures 1J–L). Finally, the reporter hiPSCs were guided to induce ROs with our reported protocol (Guan et al., 2021). D28 after differentiation, 3D ROs with thick and transparent neural retina-like structures (NRs) and pigmented RPEs spontaneously formed, demonstrating the retinal differentiation ability of the reporter hiPSCs (Supplementary Figures S2B–E). Altogether, an RCVRN-eGFP reporter hiPSC line was successfully established and had the ability to differentiate into human ROs for further investigations.

To evaluate whether the RCVRN-eGFP reporter functions well, the protein expression of recoverin in the reporter hiPSC-derived ROs was dynamically monitored under the fluorescence microscope (Figure 2). RCVRN-eGFP expression initially appeared in patches of cells in the NR layers during D50 to D60 (Figure 2B), gradually increasing and becoming abundant over time (Figures 2C–F). The percentage of RCVRN-eGFP-positive (eGFP+) ROs was 63.75 ± 8.54% at D60 (5 independent experiments; 20 ROs per experiment) and increased to 100 ± 0% by D73. The eGFP+ ROs developed and matured over time. Outer segment-like protrusions appeared on the outmost side of the NR layers and remained apparent up to D217, the latest time point observed (Figure 2F). In addition, no RCVRN-eGFP expression was identified in the RPE of the ROs (Supplementary Figures S3A–C). IF staining with anti-recoverin antibody demonstrated that the eGFP expression was co-localized with recoverin in ROs at all time points tested (Figure 2G–J″). These data suggested that the RCVRN-eGFP reporter could robustly and accurately monitor the expression of recoverin in the reporter ROs.

Next, the identity of the RCVRN-eGFP+ cells was explored in ROs at different developmental stages. They started to appear in NR of ROs since D50 after differentiation and self-formed an outer nuclear layer (ONL)-like structure in NRs over time (Figure 3). All eGFP+ cells were negative for proliferation marker Ki67 during the whole differentiation progress of photoreceptors, declaring that they were in postmitosis (Figures 3A–D, Supplementary Figure S4). In early-stage ROs aged younger than D120, the eGFP+ cells expressed CRX, a cone-rod photoreceptor precursor marker (Assawachananont et al., 2018; Yamamoto et al., 2020) (Figures 3E–H, Supplementary Figure S5). Moreover, eGFP+ cells progressively expressed both cone precursor markers RXRr (Figures 3I–L, Supplementary Figure S6) and Arrestin3 (ARR3) (Figures 3M–P, Supplementary Figure S7), and rod precursor marker NRL (Figures 3Q–T, Supplementary Figure S8). All these data elucidated that RCVRN-eGFP represented postmitotic photoreceptor precursors including both cone and rod precursors in early-stage ROs.

In late-stage ROs older than D120, relatively mature photoreceptor subtypes, one rod and two cones (Xiao and Hendrickson, 2000; Hendrickson et al., 2008), started to present in ROs (Figure 4). A subpopulation of eGFP+ cells highly expressed blue cone marker S opsin (Figures 4A,B, Supplementary Figures S9A,B″), red/green cone marker L/M opsin (Figures 4C,D, Supplementary Figures S9C,D″), and rod marker Rhodopsin (Figures 4E,F, Supplementary Figures S10A,B″). The number of mature rods and L/M cones gradually increased with time increasing, up to D217 tested (Figures 4G,H, Supplementary Figures S10C,D″). In addition, functional proteins were also detected in the late-stage ROs. The α-subunit of rod transducin (Gα_t1) was shown in eGFP+ cells from D120 (Figures 4I,J), while the cone-specific phototransduction protein CNGB3 was expressed in eGFP+ cells since D150 (Figures 4K,L). These results revealed that RCVRN-eGFP robustly labeled both mature cones and rods in late-stage ROs.

However, the eGFP+ cells did not express retinal progenitor markers VSX2, SIX3, and PAX6, or ganglion cell marker BRN3 in early-stage ROs (Supplementary Figure S11), indicating that RCVRN-eGFP did not label RPCs and ganglion cells. Moreover, the eGFP+ cells were also negative for amacrine cell marker AP2α, ganglion, amacrine, and horizontal cell marker HU C/D, müller glial cell marker CRALBP, and rod bipolar cell marker PKCα (Supplementary Figure S12), implying RCVRN-eGFP did not label these cells residing in the inner nuclear layer of the retina.

Collectively, the RCVRN-eGFP reporter covered nearly all developmental stages and all subtypes of photoreceptors, from photoreceptor precursors to mature cones and rods (Figure 4M), which could facilitate the selection and enrichment of photoreceptor cells at distinct stages.

We further investigated the transcriptome profile of RCVRN-eGFP positive cells in D150-ROs by RNA-seq. The eGFP positive and negative fractions were isolated and collected respectively using FACS (Supplementary Figure S13A). The percentage of eGFP+ cells in ROs was 43.36 ± 1.77% (N = 3 experiments), and the purity of the isolated cells in each fraction was more than 97.84%, which was evaluated by cell counting with a hemacytometer. Sample correlation analysis showed a high degree of correlation and reproducibility between the two independent sets (Figure 5A) in each group. Additionally, there were significant differences between the eGFP- and eGFP+ cells with alterations in 4,546 genes (Figure 5B).

Known retinal cell-type-specific genes were used to further identify the eGFP+ cells. The expression of photoreceptor-specific genes increased in eGFP+ cells such as RCVRN, CRX, RHO, and OPN1MW, consistent with the IF results (Figure 5C). In contrast, the expression of ganglion, horizontal, amacrine, müller glial, and bipolar cell marker genes decreased in eGFP+ cells (Figure 5C). Moreover, GO analysis showed the DEGs between eGFP- and eGFP+ cells were enriched in biological processes related to photoreceptor functions such as visual perception, detection of light stimulus, phototransduction, and detection of visible light (Figure 5D). Cellular components related to photoreceptor maturation and functions were also enriched such as photoreceptor cell cilium, photoreceptor outer segments, photoreceptor disc membrane, and synapse (Supplementary Figure S13B). All these data showed that the purified RCVRN-eGFP+ cells possessed transcriptome characteristics of photoreceptors, and the upregulation of photoreceptor marker genes in eGFP+ cells further demonstrated that photoreceptors could be well selected using the reporter line.

Although the FACS-sorted RCVRN-eGFP+ photoreceptors can be used as donor cells in animal models and promote preclinical research, they are not suitable as donor cells for clinical transplantation in treating patients. With the RNA-seq data of the eGFP positive and negative cells at D150, here, we attempted to analyze the mRNA expression of all the 394 CD biomarker genes listed in the gene group of CD molecules by the Human Genome Organization (HUGO) Gene Nomenclature Committee (HGNC) (https://www.genenames.org/data/genegroup/#!/group/471) in order to extract potential CD biomarkers for photoreceptor purification in the clinical settings. Differential expression analysis identified 121 DEGs of the CD biomarkers between the eGFP- and eGFP+ groups (Figure 6A). Compared with the eGFP- group, 32 DEGs were upregulated and 87 downregulated in the eGFP+ group (Figure 6B). The top 15 DEGs, which significantly upregulated with the FPKM >1 in the eGFP+ photoreceptors, were revealed, including CD26 (DPP4), CD110 (MPL), CD246 (ALK), CD19, and CD174 (FUT3) (Figure 6C). These genes could be potential CD biomarkers for the positive selection of photoreceptors. Meanwhile, the top 15 DEGs, which were significantly downregulated in the eGFP+ cells but upregulated with FPKM >1 in eGFP- cells, were also identified, including CD106 (VCAM1), CD140a (PDGFRA), CD225 (IFITM1), CD104 (ITGB4), and CD140b (PDGFRB), which could be potential CD biomarkers for negative selection of photoreceptor (Figure 6C).

In addition, CD biomarkers previously reported for the selection of mouse or human photoreceptors were further analyzed for their specificity and expressive abundance in the D150 FACS-sorted cells (Lakowski et al., 2015; Welby et al., 2017; Lakowski et al., 2018) (Figure 6D). Surprisingly, the mRNA expression of CD73 (NT5E), a widely used CD biomarker for positive selection of photoreceptor precursors and mature rods in mice and human retina (Koso et al., 2009; Eberle et al., 2011; Lakowski et al., 2015; Reichman et al., 2017; Gagliardi et al., 2018; Lidgerwood et al., 2018), actually had no significant difference between the eGFP- and eGFP+ cells (Figure 6D), which was further confirmed in the protein level by IF staining in D150 ROs (Figures 6E–H). These data were also consistent with previous studies on hiPSC/ESC-derived ROs (Welby et al., 2017; Lakowski et al., 2018). However, our results confirmed that the mRNA expression of CD133 (PROM1), a biomarker reported purifying photoreceptors in combination with other CD biomarkers (Lakowski et al., 2015; Welby et al., 2017), was highly and significantly expressed in the eGFP+ cells, compared to the eGFP- cells. The specificity of CD133 was further confirmed by IF analysis with D150 ROs (Figure 6I-L). These results indicated that the CD133 exclusively labeled eGFP+ photoreceptors should be suitable for photoreceptor selection.