Generating Neural Retina from Human Pluripotent Stem Cells

The eye serves as the primary source of information among human sensory organs, with the retina being the principal visual sensory tissue in mammalian eyes¹. Retinopathy stands as one of the primary global causes of eye diseases, leading to blindness². Approximately 2.85 million people worldwide suffer from varying degrees of vision impairment due to retinopathy³. Consequently, investigating its pathogenesis is crucial for early diagnosis and timely treatment. Most studies on human retinopathy have primarily focused on animal models^4,5,6. However, the human retina is a complex, multi-layered tissue comprising various cell types. Traditional two-dimensional (2D) cell culture and animal model systems typically fail to faithfully recapitulate the normal spatiotemporal development and drug metabolism of the native human retina^7,8.

Recently, 3D culture techniques have evolved to generate tissue-like organs from pluripotent stem cells (PSCs)^9,10. Retinal organoids (ROs) generated from human PSCs in a 3D suspension culture system not only contain seven retinal cell types but also exhibit a distinct stratified structure similar to the human retina in vivo^11,12,13. Human PSC-derived ROs have gained popularity and widespread availability and are currently the best in vitro models for studying the development and disease of the human retina^14,15. Over the past few decades, numerous researchers have demonstrated that human PSCs, including embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs), can differentiate into ROs using various induction protocols. These advancements hold enormous potential in retinopathy for disease modeling, drug screening, and stem cell-based therapies^16,17,18.

However, the generation of neural retina (NR) from human pluripotent stem cells (PSCs) is a complex, cumbersome, and time-consuming process. Furthermore, batch-to-batch variations in tissue organoids may lead to lower reproducibility of results^19,20. Numerous intrinsic and extrinsic factors can influence the yield of retinal organoids (ROs), such as the number or species of starting cells and the use of transcription factors and small-molecule compounds^21,22,23. Since the first human RO was generated by the Sasai laboratory¹¹, multiple modifications have been proposed over the years to enhance the ease and effectiveness of the induction process^13,21,24,25. Unfortunately, to date, no gold standard protocol has been established for generating ROs in all laboratories. Indeed, there is a certain degree of discrepancy in ROs resulting from different induction methods, as well as wide variation in the expression of retinal markers and the robustness of their structure^22,26. These issues may severely complicate sample collection and the interpretation of study findings. Therefore, a more consolidated and robust differentiation protocol is needed to maximize efficiency with minimal heterogeneity of RO generation.

This study describes an optimized induction protocol based on a combination of Kuwahara et al.¹² and Döpper et al.²⁷ with detailed instructions. The new method significantly reduces the probability of organoid vesiculation and fusion, increasing the success rate of generating NR. This development holds great promise for disease modeling, drug screening, and cell therapy applications for retinal disorders.

This study was conducted in accordance with the Tenets of the Declaration of Helsinki and approved by the Institutional Ethics Committee of the Chinese PLA General Hospital. The WA09 (H9) ESC line was obtained from the WiCell Research Institute.

1. Culture media and reagent preparation

1. Human ESC culture medium and passage solution
   1. Maintenance medium (MM): Prepare 500 mL of complete MM (Basal Medium + 5x supplement; see Table of Materials) aseptically. Thaw 5x supplement at room temperature (RT) or overnight at 2-8 °C. Prewarm to RT. Stir well before use until the supplement is free of cloudiness.
   2. 1% extracellular matrix (ECM): Use a pre-chilled pipet tip and sterile tubes to dispense 200 µL of ECM (see Table of Materials) per tube on ice. Store the tubes in a -20 °C freezer. Melt ECM on ice and dilute with pre-cooled DMEM/F12 at 1:100.
   3. 0.5 mM pH = 8.0 EDTA: To prepare 500 mL of 0.5 mM EDTA, add 500 µL of 0.5 M EDTA and 0.9 g of NaCl to 500 mL of 1x Dulbecco's phosphate-buffered saline (DPBS) (see Table of Materials). Mix thoroughly and store at 2-8 °C.
2. Retinal differentiation medium
   1. Growth factor-free chemically defined medium (gfCDM): Prepare the gfCDM by combining 45% Iscove's modified Dulbecco's medium-GlutaMAX (IMDM-GlutaMAX), 45% Ham's F12-GlutaMAX (F12-Glutamax), 10% knockout serum replacement (KSR), 1% cholesterol lipid concentrate, and 450 µM thioglycerol (see Table of Materials).
3. Small molecule compounds
   1. Y-27632 2HCl: Add 50 mg of Y-27632 powder to 3.122 mL of dimethyl sulfoxide (DMSO). Dispense and store Y-27632 (see Table of Materials) at a concentration of 50 mM at -80 °C for 2 years, keeping away from light. Dilute the stock 2,500 times (20 µM) for use, equivalent to 0.4 µL of Y-27632 per mL of gfCDM.
   2. IWR1-endo: Add 2.4424 mL of DMSO to dissolve 10 mg of IWR1-endo (see Table of Materials) to obtain a 10 mM stock solution. Dispense aliquots and store them at -80 °C for up to 2 years. Add 0.3 µL of 10 mM IWR1-endo per mL of gfCDM for the 3 µM concentration for induction.
   3. SB431542: To prepare 50 mM SB431542, add 10 mg of SB431542 (see Table of Materials) to 0.5203 mL of DMSO and mix thoroughly. Store at -80 °C in solvent for a maximum of 2 years. For the preparation of SB431542 at a working concentration of 10 µM, add 2 µL of the 50 mM stock solution to 10 mL of gfCDM.
   4. LDN-193189 2HCl: Dissolve 5 mg of LDN-193189 2HCl (see Table of Materials) in 10.4297 mL of DMSO to get a 1 mM of stock solution. Store at -20 °C or -80 °C (as recommended by the manufacturer). Add 0.1 µL of the stock to every 10 mL of gfCDM, resulting in 100 nM of LDN-193189 for induction.
   5. Recombinant Human bone morphogenetic protein 4 (BMP4): Reconstitute at 50-200 µg/mL in 4 mM HCl (see Table of Materials). Store at 2 °C to 8 °C for 1 month or -20 °C for 1 year after reconstitution. Perform induction of NR using 1.5 nM BMP4.
4. Long-term NR culture medium
   1. Retinoic acid (RA): Measure 6 mg of RA powder (see Table of Materials) and add to 3.9941 mL of DMSO. Store in aliquots of 5 mM at −80 °C and use within 3 months. To achieve a working concentration of 0.5 µM, add 10 µL of the master mix to 100 mL of neural retina differentiation medium (NRDM). Add just before use.
      NOTE: Keep away from light during preparation and storage.
   2. Taurine: Add taurine powder (see Table of Materials) weighing 200 mg to 7.9904 mL of DMSO to obtain a 200 mM stock solution. Dispense and store at 2-8 °C. A working concentration of 0.1 mM taurine is achieved by adding 50 µL of the stock solution per 100 mL of NRDM.
   3. NRDM: Compose NRDM with DMEM/F12-GlutaMAX medium, 1% N2 supplement, 10% fetal bovine serum, 0.5 mM RA and 0.1 mM taurine (see Table of Materials). Store at 2-8 °C for up to 2 weeks or 6 months at -20 °C to ensure the activity of the components.
5. Blocking buffer: Dilute 1:9 with DPBS to obtain a 10% working solution for use (see Table of Materials). Store at -20 °C for up to 5 years.
   ​NOTE: Perform all other procedures in a Class II Biosafety Cabinet to ensure sterility, except for weighing. If adding weighed reagents during the preparation process is necessary, use 0.22 µm filters for filtration.

2. Culturing of H9-ESCs

1. Thawing of H9-ESCs
   1. Add 1 mL of 1% ECM to each well of a 6-well plate. Incubate for 1 h in an incubator at 37 °C in a 5% CO₂ atmosphere.
      NOTE: Avoid the addition of ECM along the well walls, because it may stick to the surface.
   2. Take an H9-ESC cryovial stock from the liquid nitrogen tank and quickly shake it in 37 °C water for 30 s.
      NOTE: Do not allow the vial to thaw completely.
   3. Remove the vial and carefully sterilize it using 75% disinfectant alcohol spray. Add the thawed H9-ESC from the cryovial to a 15 mL tube containing 9 mL MM with 10 µM Y-27632.
   4. Centrifuge the tube at 190 × g for 5 min at RT. Carefully remove most of the supernatant with a 1 mL pipette, leaving approximately 50 µL of supernatant to avoid cell loss.
   5. Add 1 mL of MM containing 10 µM of Y27632 to the cell sediment and gently resuspend the cell sediment with a 1 mL pipette by pipetting up and down 5-10 times.
   6. Remove the ECM coating after 1 h of incubation. Add 2 mL of pre-warmed MM containing 10 µM Y-27632 to each well.
   7. Dispense 0.5 mL of cell suspension per well. Gently shake the plate laterally to ensure even distribution of cells.
   8. Incubate the plate at 37 °C under 5% CO₂ for at least 24 h without touching it.
   9. Change the medium daily. When clone density reaches 70% and above, passaging is required.
2. Passaging of H9-ESCs
   1. Prepare the ECM-coated plate as described above (step 2.1.1) and add 2 mL of MM per well.
   2. Remove the spent medium from the 6-well plates. Wash each well twice with 1 mL of DPBS.
   3. Wash each well twice by slowly adding 1 mL of EDTA, and incubating with 1 mL of EDTA for 4-7 min at RT.
      NOTE: During the incubation, examine the 6-well plate for 3 min under a microscope. Proceed immediately to the next step if most cells are close to detaching from the dish. Avoid long time incubation to reduce negative impact on the cells.
   4. Discard EDTA and add 1 mL of MM to abort digestion.
   5. Gently tap the well plate until the majority of the cells are detached.
   6. Carefully transfer the cell suspension into a 15 mL centrifuge tube with a 5 mL pipette. Gently resuspend the H9-ESC colonies up and down 3-5 times to mix them with a Pasteur pitette. Transfer 20-100 µL of cell suspension in a new ECM-coated 6-well culture dish and shake it back and forth.
   7. When the cell density is observed as sufficient under a microscope, incubate the plate at 37 °C under 5% CO₂ for at least 24 h without touching it.
   8. Change the media daily. At 70% clone density and above, passage is required.

3. Generation of human NRs

NOTE: Once the colonies achieve approximately 70% confluence, they can be directed towards differentiation into retinal organoids (ROs) using the procedural steps outlined in Figure 1.

1. Day 0 - Embryoid body (EB) formation
   1. After washing the cells with 2 mL of DPBS, add 0.5 mL of CDS (see Table of Materials) containing 20 µM Y27632 to the well. Incubate for 3 min at 37 °C in a humidified 5% CO₂ incubator.
      NOTE: Check under a microscope. Reduce the time of incubation as much as possible to avoid harmful effects on the cells.
   2. Abort digestion by adding 3 mL of MM containing 20 µM Y27632. Centrifuge at 190 × g for 5 min and discard supernatant.
   3. Resuspend the cell pellet in 1 mL of gfCDM containing 20 µM Y27632, and count the cells. Add the corresponding volume for 1.2 × 10⁶ cells (1.2 × 10⁴ cells per well) to 10 mL of gfCDM, which is pre-incorporated with 20 µM Y-27632, 3 µM IWR1-endo, 10 µM SB431542, and 100 nM LDN-193189 (see Table of Materials).
   4. Add 100 µL of cell suspension to each well of 96 V-bottomed conical wells. Place the plate in a humidified 5% CO₂ incubator until day 6.
      NOTE: Do not move the dishes for at least 24 h to enhance the adherence of EBs.
2. Day 6 - Retina induction
   1. On day 6, add 10 mL of gfCDM containing 55 ng/mL BMP4 to allow a complete change of culture medium on the EBs. Return the plate to the incubator.
      NOTE: Pipette toward the walls of the 96 V-bottomed conical well to avoid air bubbles. Avoid taking up any organoids when changing the medium.
   2. Carry out half-medium change on day 9, day 12 and day 15. Replace half of the medium with fresh gfCDM to gradually dilute the BMP4.
3. Day 18 - Long term NR culture
   1. On day 18, carefully transfer the formed EBs to 15 mL centrifuge tubes using 5 mL Pasteur pipettes and gently rinse again with NRDM. Transfer the EBs to low-adsorption 6-well or 24-well plates (see Table of Materials). Return the plate to the incubator.
   2. Replace the medium with fresh NRDM every 3 days.
      ​NOTE: Switch off the light during the medium change because the RA is light-sensitive. Remove badly differentiated organoids, and separate adherent organoids under a microscope.

4. Analysis of human NRs

1. Mounting of the ROs
   1. Select different generations of H9-ESCs to induce three batches of ROs.
      NOTE: Three plates of the number of forming NRs for each of the three assessed methods (Kuwahara et al.¹², Döpper et al.²⁷, and the modified method in the present study) are calculated to assess the induction success rate. Induction is considered successful if light microscopy reveals the formation of neuroepithelial-like structures on day 30.
2. Immunofluorescent staining of ROs
   1. Transfer 3-5 ROs to 1.5 mL tubes. Pipette off excess medium and wash the ROs once with 1 mL of DPBS at RT. Allow the ROs to sink and carefully remove the DPBS. Add 1 mL of 4% paraformaldehyde (see Table of Materials) per 1.5 mL tube and fix at 4 °C.
      NOTE: ROs on day 6 and day 18, fix 2 h. Fix for 12 h on day 30 and 14 h on day 60.
   2. After fixation, dehydrate using gradient alcohol. Leave the ROs to stand for 15 min each in 50%, 60%, and 70% alcohol, and subsequently for 10 min each in 80%, 90%, 95%, 100%, and 100% alcohol. Then, add a 1:1 mix of 100% alcohol and xylene for 10 min, followed by xylene twice for 10 min each.
   3. Place the ROs in metal molds filled with melted paraplast (see Table of Materials) for 40 min, and then quench the molds on ice to fix the ROs. Place embedding boxes into the molds and freeze them at -20 °C overnight. Subsequently, carefully detach the embedding boxes. Finally, store them at RT.
   4. Create continuous sections (5 µm thickness) with a paraffin slicer. Fix slices on adhesion microscope slides, dry, and store at RT.
   5. Perform dewaxing and rehydration before antigen repair^12,27.
   6. Add 10 mL of 20x pH 6.0 Citrate Antigen Retrieval Solution (see Table of Materials) to 190 mL of ddH₂O (double distilled H₂O). Heat in a microwave on high mode for 4 min until boiling. After adding paraffin sections, heat the sections on low mode for 20 min to complete antigen repair.
   7. Allow paraffin sections to cool naturally in a ventilated area and then put them into a humid chamber.
   8. Use a pap pen (see Table of Materials) to outline the sections. Incubate with 0.2% Triton X-100 at RT for 30 min to break the membranes, followed by washing the slides three times with TPBS, for 5 min each.
   9. Block with 10% donkey serum diluted in DPBS at RT for 1 h in a humid chamber with 10 µL per staining area.
   10. Add primary antibodies diluted in 10% donkey serum. Incubate overnight at 4 °C in a humidity chamber. Wash samples thrice with TPBS for 10 min each to remove the unbound antibody.
       NOTE: Primary antibodies are as follows: anti-PAX6 (1: 250), anti-SOX2 (1: 200), anti-KI67 (1: 200), anti-CHX10 (1: 200), anti-β tubulin III (1: 250) and anti-NESTIN (1: 200) (see Table of Materials).
   11. Incubate with secondary antibodies diluted in DPBS for 1 h at RT in a humidified chamber. Repeat the wash step with TPBS three times for 10 min each.
       NOTE: Keep in the dark to prevent quenching of the fluorescent secondary antibody. Secondary antibodies are Alexa Fluor 488 conjugated with donkey anti-mouse IgG and Alexa Fluor 568 conjugated with donkey anti-rabbit IgG at a dilution of 1:400 (see Table of Materials).
   12. Incubate with DPBS containing DAPI (1: 500; see Table of Materials) for 10 min at RT in the dark. Wash three times with TPBS for 10 min each.
   13. Drop an appropriate amount of anti-fluorescent sealer (see Table of Materials) onto the slide and cover with cover slips.
   14. Visualize using a flow-like tissue cell quantitative analyzer (see Table of Materials), or similar.
   15. Store slides at -20 °C in the dark after microscopic analysis.

A graphical overview of the modified protocol is shown in Figure 1. H9-ESCs were used to generate ROs when the cells were grown to a density of 70%-80%. Single-cell suspensions of H9-ESCs in 96 V-bottomed conical wells aggregated on day 1 and formed well-circumscribed round EBs by day 6. As the culture time increased, the volume of EBs gradually increased. On day 30, neuroepithelial-like structures were clearly formed and thickened during long-term NR differentiation.

In addition, the modified method was compared with two other methods (Kuwahara et al.¹² and Döpper et al.²⁷) to evaluate its repeatability and efficiency (Supplementary Figure 1). This study also observed that the morphology of EBs on day 1 following the modified method was slightly worse than in the other two methods. However, the neuroepithelial structures that formed using the modified method were more continuous on day 18 (Figure 2). On day 30, the early NRs obtained using the modified method were similar in shape and size and showed a regular round shape (Figure 3). Compared to the other two methods, the NRs formed by the modified method had a lower incidence of vesiculation and adhesion. Furthermore, based on the H9-ESCs, the success rate of generating NRs using the modified method increased to 87.39% from the 26.9% and 69.24% for the Kuwahara et al.'s and Döpper et al.'s methods^12,27, respectively (Figure 3M).

Immunofluorescence staining was performed on paraffin-embedded sections of the NRs to evaluate the expression of markers related to retinal development (Figure 4). The results showed the expression of human genes associated with the retina in NRs derived from H9-ESCs using a modified method. The first cell subtype to appear is the retinal ganglion cell, which mainly accumulated on the basal side of the NRs. TUJ1 staining was used to identify the axons of the retinal ganglion cells (Figure 4I-L). Moreover, markers of retinal progenitor cells were detected in both the inner (PAX6⁺) and outer (CHX10⁺) layers (Figure 4A-H). Cells positive for the proliferation marker KI67 were distributed in the outer layer (Figure 4A-D).

Figure 1: Schematic overview. The schematic overview of the modified culture protocol shows the addition of different factors at specific time points and the representative images of the neural retina development. SFEBq: serum-free floating culture of embryoid body-like aggregates with quick reaggregation; KSR: knockout serum replacement; gfCDM: growth factor-free chemically defined medium; NRDM: neural retina differentiation medium; BMP4: bone morphogenetic protein 4.Scale bars: 200 µm. Please click here to view a larger version of this figure.

Figure 2: The representative brightfield images of retinal organoids derived from three different protocols in the early stage. On day 1, only the modified method did not from the EBs (A,E,I). On day 6, EBs formed in all three methods (B,F,J). On day 18, early neural retinas were formed, and different morphology was found among the three induction methods (C,G,K). On day 21, neural retina generated by the modified method showed continuous neuroepithelial structure (D,H,L). Scale bars: 200 µm. Please click here to view a larger version of this figure.

Figure 3: Mature 3D neural retina in suspension culture at days 30, 45, 60. On day 30, neural retina formed in all three methods (A,E,I). Arrows indicate cystic structures in Kuwahara et al.'s method (B). White dotted boxes represent the region of organoids adhesion and fusion to each other (D,F,H). Retinal organoids generated from the modified protocol are similar in size and morphology compared to the other two methods (C,G,J,K,L). Scale bars: 100 µm (white); 200 µm (black). Statistical chart of induction success rate by the three methods using H9-ESCs (M). One-way analysis of variance was used to test the significance (error bars shows standard error, n = 864, **P < 0.01, ***P < 0.001). Please click here to view a larger version of this figure.

Figure 4: Characterization of neural retina in the modified protocol. Immunostaining images of neural retina at day 6, day 18, day 30 and day 60. Green stains are for KI67 (proliferating cell), PAX6 (retinal progenitor cell), and NESTIN (neural stem cell). Red stains are for CHX10 (retinal progenitor cell), SOX2 (neural progenitor cell), and TUJ1 (retinal ganglion cell). Scale bars: 100 µm (A,B,E,F,I,J); 200 µm (C,D,G,H,K,L). Please click here to view a larger version of this figure.

Supplementary Figure 1: Comparison of flow charts for the three retinal organoid induction protocols. SFEBq: serum-free floating culture of embryoid body-like aggregates with quick reaggregation; KSR: knockout serum replacement; gfCDM: growth factor-free chemically defined medium; NRDM: neural retina differentiation medium; BMP4: bone morphogenetic protein 4. Please click here to download this File.

Supplementary Figure 2: The morphology of embryoid body (EB) induced by Kuwaharaet al.'s method. The gradual appearance of dead cell clumps was observed around EBs after the addition of BMP4 on day 6 (A-D, black arrows). Large morphological differences were observed, after day 6 (E-L), and some EBs were even completely inactivated (H,L). ESC: embryonic stem cell, PBMC: peripheral blood mononuclear cell, IPSC: induced pluripotent stem cell. Scale bars: 200 µm. Please click here to download this File.

Human ROs can spatially and temporally recapitulate the development of the fetal retina, and early ROs exhibit a high degree of similarity to the fetal retina at equivalent stages of development¹⁵. In terms of tissue morphology and molecular expression, human RO closely mirror the actual growth status of the retinal tissue, providing tremendous and unprecedented opportunities in the fields of disease modeling, drug screening, and regenerative medicine. Currently, several different methods have been established to generate ROs from human PSCs in vitro, and continuous modifications and optimizations are underway to further improve efficiency⁹. The Sasai laboratory reported for the first time the generation of PSC-derived ROs from human ESCs¹¹. Human ESCs were first prepared into single-cell suspensions, and then equal numbers of cells were seeded in 96 V-bottomed conical wells, with rapid aggregation to form circular-like EBs. The external addition of key cell signaling pathways promoted the formation of optic vesicles in EBs, which subsequently matured into laminated ROs in suspension culture. The NR contains one glial cell type and six different neuronal cell types, representing many aspects of retinal development, such as cell morphogenesis, neuron differentiation, and apical-basal polarity⁹. Although only 10% of the aggregates formed a double-walled optic cup structure, this was a major milestone in the evolution of producing more advanced ROs¹¹.

Subsequently, Zhong et al. introduced a new alternative, in which hiPSCs were first grown close to confluences, followed by chemical or mechanical disintegration into floating small aggregates¹³. Then, the small aggregates formed EBs in suspension culture, which detached separately from the bottom of the plate, further differentiating in the neural direction. Zhong et al. demonstrated a completely laminated 3D retina with relatively well-developed photoreceptor outer segments that responded to light stimulation. This method required less external regulation of cell signaling pathways and was primarily performed in a self-directed manner⁹. The third technological revolution was the introduction of the embedding method by Lowe group²⁵. This involved embedding hESC aggregates in ECM to form a single-lumen epithelial structure. After enzymatic dispersion, the aggregates were maintained in a suspension culture to form mature ROs. Although all three methods can successfully induce ROs with comparable structure and function, large-scale applications have the disadvantages of being time-consuming and laborious because of their high heterogeneity and low success rate²³.

In addition to the above methods, Kuwahara et al. proposed a new induction method, namely, the key factor gradient decline method¹². They found that NR could be generated by the addition of BMP4 at diminishing concentration from day 6. This induction system used fewer external factors to form ROs with increased uniformity of size and morphology. However, studies have reported that the formed EBs were prone to cystic lesions, and the induction success rate of this method was approximately 40%^12,28. Döpper et al. modified this protocol to improve the success rate by combining a commercial medium with three small molecule inhibitors at the early stage of induction to stabilize the EBs²⁷. In contrast, cells of the neural epithelium of the retina more easily adhere to each other, necessitating separate culture of each RO during long-term suspension culture. Döpper et al.'s method increases the complications and costs of experimental operations and is not suitable for large-scale induction²⁷. The optimized RO production protocol improved induction efficiency. The current modified method was compared with the methods of Kuwahara et al. and Döpper et al.

This study showed that both Kuwahara et al.'s and Döpper et al.'s methods formed EBs with smooth edges on day 1, whereas the present modified method required a relatively long time to form EBs, approximately until day 6. The volume of the EBs obtained using Kuwahara et al.'s method was larger than that obtained using Döpper et al.'s method, and the EBs obtained using Döpper et al.'s method showed increased fluidity under liquid flow. Notably, dead cell clumps gradually appeared around the EBs after the addition of BMP4 on day 6 when using Kuwahara et al.'s method (Supplementary Figure 2). By day 18, significant morphological differences were observed in the EBs of Kuwahara et al.'s method, and some EBs were completely inactivated (Supplementary Figure 2). In Döpper et al.'s method, the majority of EBs were rounded and well circumscribed, and exhibited only a small number of scattered dead cells around them. The modified method formed tight and well-structured EBs with slight morphological differences, and a few dead cells were observed around them.

Moreover, this study also found that some EBs generated vesicles by Kuwahara et al. in the early stage following transfer to 6-well plates, resulting in a failure to form ROs. Adhesion and fusion of ROs occurred around day 40, which significantly decreased the utility of the experiments. The culture system of Döpper et al.'s method is more complicated, and requires more operational steps and higher cost than Kuwahara et al.'s method. The majority of ROs by Döpper et al. also inevitably developed adhesions and fusions with each other. The modified method in this study successfully generated 3D NRs in vitro that expressed the human retina-related markers CHX10, KI67, PAX6, SOX2, TUJ1, and NESTIN. And most of ROs by the modified method were consistent in size and morphology, and very few vesicles or adhesions developed during the late stages of retinal differentiation.

Compared to the other two methods, the new culture system had a simpler operational steps and lower costs. The critical step in the modified method is to add three small molecule inhibitors into media and ensure that the plate is not moved for the first six days to stabilize the structure of EBs. The movement of plates or any cell manipulation in the early stage might affect the formation of EBs. Compared with the other two methods, the modified method did not change the medium on day 1 and also reduced the use of small molecule inhibitors on day 15. In short, the operation steps of the modified method are simplified to a certain extent, and the experimental cost is saved with less use of media and reagents. However, the modified method still cannot solve the common problems of current organoid culture system. In addition, ROs induction efficiency largely depends on the quality and differentiation ability of hPSCs. In this study, only the efficiency of NR generation using H9-ESC was completely calculated. We successfully induced ROs in different hPSCs lines with the new method in the present study, and achieved the same induction efficiency, including H9, H1 and PBMC-iPSC. There is no statistical difference in induction efficiency among the three cell lines. In the future, more studies need to explore the induction efficiency of different hPSCs using this method.

In conclusion, the optimized retinal induction protocol is simple and inexpensive, has high repeatability and efficiency, offers promising personalized models of retinal diseases, and provides an abundant cell source for cell therapy, drug screening, and gene therapy testing.