Spheroid transplantable and functional retinal pigment epithelium derived from non-colony dissociated human induced pluripotent stem cells

Background: Retinal pigment epithelium (RPE) cells derived from human induced pluripotent stem cells (hiPSCs) exhibit great promise in treating retinal degenerative diseases. Here, we would explore the feasibility of non-colony dissociated hiPSCs to differentiate into functional RPE cells (hiPSC-RPE), and offer an alternative transplantation method based on cell spheroids. Methods: hiPSC-RPE cells were identied using reverse transcription-polymerase chain reaction (RT-PCR), immunouorescence assay, Western blotting, and ow cytometry assay. The functions of hiPSC-RPE cells in vitro and in vivo were assessed by uorescein leakage test, transepithelial electrical resistance (TEER) assay, atomic force microscopy observation, POS phagocytosis assay, frozen tissue sections, live/dead assay, SA-β-Gal staining, and immunocytochemistry. Results: hiPSC-RPE cells positively expressed biomarkers of RPE cells but not iPSCs, such as CRALBP (97.4%), EMMPRIN (93.8%), Oct4 (2.1%), and Sox2 (2.0%). hiPSC-RPE cells displayed RPE-like characteristics including barrier function, phagocytic activity, and polarized membrane. The cells derived from hiPSC-RPE spheroids positively expressed Nestin and exhibited reduced SA-β-Gal staining. hiPSC-RPE cell spheroids could form monolayer on decellularized corneal matrixes (DCM). After one month of subretinal transplantation, hiPSC-RPE cell spheroids could survive and maintain segmental sheet growth in sodium iodate (NaIO 3 ) induced RPE-degenerated chinchilla rabbits. Conclusion: This study suggested that non-colony dissociated hiPSCs were effectively differentiated into functional RPE cells, and hiPSC-RPE cell spheroids maintained segmental sheet growth in the subretinal of RPE degenerate chinchilla rabbits in vivo, which may lay the foundation for cell spheroid transplantation as an alternative method for RPE

serious adverse events were noted after one year of follow-up [29]. In addition, Westenskow et al. reported that the iPS-RPE grafts remained viable and did not induce any obvious pathological changes in rats [30]. One month of transplantation of hESC-derived retinal progenitors (RPs) into the sub-retinal space of sodium iodate (NaIO 3 ) induced RPE degeneration adult rabbits, transplanted retinal cells survived, migrated into retinal layers, and restored a small but signi cant b-wave of ERG testing [31]. Moriguchi et al. explored the mechanism causing degeneration of RPE in mice after an intravenous injection of NaIO 3 [32]. So, in our study, we would also use NaIO 3 treatment to establish RPE degeneration model in chinchilla rabbits for cell transplantation in vivo.
In this study, we would examine the feasibility of non-colony dissociated hiPSCs instead of clonal hiPSCs to differentiate into transplantable and functional RPE cells. Then, we seeded hiPSC-RPE cell spheroids onto decellularized corneal matrixes (DCM) for simulation experiments in vitro. Lastly, we injected hiPSC-RPE cell spheroids into the subretinal space of chinchilla rabbits with RPE degeneration induced by NaIO 3 treatment in vivo. Our study will lay the foundation for the use of hiPSC-RPE cell spheroids as an alternative method for cell transplantation to therapy of RPE degenerative diseases.

Material And Methods
Culture of hiPSCs hiPSCs were a gift from the Guangzhou Institutes of Biomedicine and Health, Chinese Academy of Sciences. These hiPSCs were reprogrammed from human umbilical cord mesenchymal cells using methods described in a previous report [33]. hiPSCs were cultured in 1% Matrigel (BD Biosciences, USA) coated dishes at 37 °C, 5% CO 2 and refreshed daily with mTeSR1 medium (StemCell Technologies Inc., Canada). They were passaged once every 6 days with 0.25% EDTA (Sigma, USA) and were then seeded into 6-well dishes at a ratio of 1:6. They were supplemented with 10 μM Y-27632 (Sigma) on the rst day of passaging. For generating dissociated hiPSCs, the harvested clonal hiPSCs were pipetted approximately 30-50 times and then were ltered through a 40 μm cell strainer (BD). The dissociated hiPSCs (1×10 5 cells/cm 2 ) were plated into 6-well dishes. Half-exchange mTeSR1 medium (HM) was used for further culture of non-colony dissociated hiPSCs [16].

Isolation and culture of adult RPE cells
Human eyes were obtained from six male donors without ocular disease after informed consent at the mean age of 45±5 years old. The study was approved by the Human Research and Ethical Committee of Jinan University, and the procurement and use of human tissues were in compliance with the Declaration of Helsinki. Human adult RPE cells were isolated and cultured as described in a previous report [34].
Brie y, RPE cells were isolated from the posterior section of the eyeballs using 0.25% EDTA-trypsin (Gibco, USA) and harvested by centrifugation at 400×g for 5 min. hRPE cells were cultured in RPE medium that consisted of high-glucose Dulbecco's modified Eagle's medium (HG-DMEM, Gibco), 10% FBS (Gibco), 100 U/mL penicillin and 100 mg/mL streptomycin (P/S, Gibco) at 37 °C and 5% CO 2 . After reaching 100% con uence, RPE cells were passaged and seeded in 6-well plates with a 1% Matrigel coating (Sigma). The medium containing 100 ng/mL Activin A (R&D Systems, USA) was changed every 3 days.

Gene expression analysis
Total RNAs from the cells was isolated using a Tissue RNA Miniprep Kit (Biomega, China). These RNAs were reversely transcribed into cDNA using the Superscript II kit (Invitrogen), and cDNAs were synthesized and used for reverse transcription-polymerase chain reaction (RT-PCR) to identify gene expressions such as GADPH, Nanog, Oct4, Sox2, Klf4, RPE-65, CRALBP, EMMPRIN, Otx2, c-Myc, Lin28, Nestin, and Pax6. RT-PCR products were examined after electrophoresis on 2% agarose gels. Gels were scanned for further analysis. For quanti cation of gene transcripts, cDNAs were rst denatured at 95 °C for 3 min, followed by 40 cycles of 95 °C for 10 s and 58 °C for 30 s. The relative expressions of the genes such as Nanog, Oct4, Sox2, Klf4, c-Myc, Lin28, Nestin, and Pax6 were normalized against GAPDH, and quanti cation was performed using the comparative Ct (2 -DDCt ) method. The primer sequences are shown in Table 1.

Flow cytometry
Samples including hiPSC, hiPSC-RPE, and hRPE cells were xed using 4% paraformaldehyde and permeabilized by 0.1% Triton X-100 (Sigma). Then, they were incubated with isotype control or primary antibodies as shown in Table 2 at 4 °C for 30 min. Primary and isotype control antibodies such as Oct4, Sox2, RPE-65, CRALBP, Mitf, and EMMPRIN were labeled with uorophore-conjugated secondary antibodies at 4 °C for 30 min. The labeled samples were detected by a ow cytometry analyzer (BD, USA).

Fluorescein leakage test
A uorescein leakage test (FLT) was used to assess the barrier function of the epithelial cells as described in a previous report [36]. First, 200 μL of cells (1×10 5 cells/cm 2 ) were seeded in 6-transwell inserts (Millipore, USA) and incubated at 37 °C and 5% CO 2 . After reaching 100% con uence, cells were xed with 4% paraformaldehyde, and 500 μL of Na-uorescein solution (10 μg/mL, Sigma) was added into each insert before incubation for the designated times (0, 4, 12, 24, and 36 h) to allow solution leakage into the bottom wells. The values of uorescein leakage into the wells were measured using a microplate spectrophotometer (VSERSA Max, USA) at 485 nm excitation and 530 nm emission wavelengths. The uorescein leakage through the inserts without cells was set as 100% (maximum leakage). The values of uorescein leakage test (FLT) were calculated as the amount of uorescein leakage from hRPE and hiPSC-RPE cells compared to the corresponding maximum leakage.
Transepithelial electrical resistance (TEER) assay A TEER assay was used to assess the dynamic barrier function of the epithelioid cells [37]. Cells were seeded into 24-transwell inserts at 1×10 5 cells/cm 2 . After reaching 100% con uence on day 7, the dynamic barrier of the cells was determined through measuring TEER across the cell monolayer using Millicell-ERS-2 (Millipore, Temecula, USA). The value of TEER was calculated per the following equation: TEER (Ω cm 2 ) = (R total -R insert w/o Matrigel ) ×A R total is the resistance measured (Ω), R insert w/o Matrigel (Ω) is the resistance of the insert without 1% Matrigel coating, and A is the membrane area (cm 2 ) of the insert.

Atomic force microscopy observation
Atomic force microscopy (AFM) was used to observe the ultrastructure of cells as described in a previous report [38]. Cells were xed with 4% paraformaldehyde for 10 min and dried at room temperature before imaging. The curvature radius of the AFM tips was 10 nm. The spring constant was 20-50 N/m with a resonance frequency of 278-317 kHZ. The scanning speed was kept at 0.5 Hz. The ultrastructure of the cells was measured in contact mode. The data analysis was performed using Nanoscope Analysis Software (Thermo Microscopes Proscan Image Processing Software Version 2.1, USA).

POS phagocytosis assay
The photoreceptor outer segment (POS) was isolated as described in a previous report [39]. Brie y, the retinas of porcine eyeballs were collected and agitated in KCl buffer (0.5 mM CaCl 2 , 1 mM MgCl 2 , 0.3 M KCl, and 10 mM HEPES) with 48% w/v sucrose at pH 7.0, and then were centrifuged at 5000×g for 5 min.
The supernatant containing the POS was ltered using sterile gauze, diluted 1:1 with KCl buffer without sucrose, and centrifuged at 4000×g for 10 min. The isolated POS were then resuspended in 1 mL of PBS and were labeled with FITC (Sigma-Aldrich) at room temperature for 1 h. The labeled FITC-POS were then rinsed and resuspended using HG-DMEM medium with 5% sucrose. Cells were incubated with FITC-POS at 37 °C and 5% CO 2 for 2 h. Last, immuno uorescence was performed using Cy3-conjugated mouse monoclonal antibody ZO-1 and DAPI, and then they were examined under an inverted uorescence microscope (OLYMPUS, Japan).
Z-stack confocal microscopy polarized membrane observation Z-stack confocal microscopy was used to observe the polarized membrane of the cells. First, 200 μL of cells (1×10 5 cells/cm 2 )were seeded into 6-well dishes and incubated at 37 °C and 5% CO 2 for 5 days.
Immuno uorescence was conducted with Cy3-conjugated mouse monoclonal antibody ZO-1 and DAPI and the stained cells were examined under a confocal microscope (LSM 510 META; Zeiss, Thornwood, USA).
Production of agarose micro multiwell dishes and hiPSC-RPE cell spheroids A volume of 500 µL liquid solution of 2% (g/mL) agarose was pipetted into an commercially available silicone micro-mold with eighty-one wells (Micro Tissues Inc., CA, USA). After solidi cation, the microwell agarose mold was removed using sterilized forceps. The agarose molds were placed in 6-well dishes.

Preparation of decellularized corneal matrix (DCM)
The lamellar corneal matrix (100 µm thickness) was excised from porcine eyeballs using a microkeratome (Kangming, China) and was rinsed 3 times with PBS. The excised lamellar corneal matrix was treated with 0.25% EDTA-trypsin (Invitrogen) at 37 °C for 30 min, and then was xed with 4% paraformaldehyde for 1 day at 4 °C. It was then treated with 0.8% SDS (Sigma) solution at -80 °C for 30 min and then was transferred to a 37 °C shaking table (350 rpm) for 1 h. It was then rinsed 3 times with PBS and preserved in 100% glycerol at 4 °C as decellularized corneal matrix (DCM). Before the seeding of cells, the DCM was washed 3 times with PBS containing P/S solution and sterilized under ultraviolet light for 2 h.

Seeding the cell spheroids on the DCM
To test whether cell spheroids in the biomimetic microenvironment could grow well, an in vitro simulation experiment of seeding cell spheroids onto DCM was conducted as described previously [26]. Brie y, cell spheroids were seeded on the DCM. The medium containing 10 μM Y-27632 was changed every 3 days. Viable cell staining with Calcein AM was used for better observation of the cell spheroids on the DCM under a uorescent microscope. The adherent growing area of the cell spheroid periphery stained by a Live-Dead Cell Staining Kit (Biotium, USA) was measured using ImageJ on days 7 and 14, respectively.

Frozen tissue sections
Tissue samples rstly were xed with 4% paraformaldehyde for 15 min, and then were mounted using tissue freezing medium (SAKURA Tissue-Tek, USA) and placed at -80 °C refrigerator for at least 2 h until frozen. The frozen tissues were sectioned at a thickness of 15 μm using a cryo-microtome (Thermo Fisher, USA) as described in a previous report [40]. Sections were placed on one side of microscope slides (SAKURA Tissue-Tek, USA). Some sections were incubated with DAPI for 15 min, washed with PBS for 3 times, and examined under an inverted uorescence microscope (OLYMPUS, Japan). Other sections were used for hematoxylin-eosin (H&E) staining and were imaged using an inverted microscope (Nikon, Japan).

Live/dead assay of Calcein AM and EthD-III double staining
Calcein AM and EthD-III double staining (Molecular Probes, USA) was performed as described in a previous report [16]. Brie y, a standard working solution containing 2 µM Calcein-AM and 4 µM EthD-III was prepared. Cells were incubated with the standard working solution at room temperature for 40 min, and were then imaged under an inverted uorescence microscope.
SA-β-Gal stainingassay SA-β-Gal staining was performed using a Cellular Senescence Assay Kit (Beyotime Biotechnology, China) following the manufacturer's instructions. We used spheroid or monolayer adherent cells on day 7 and 14 to conduct SA-β-Gal staining. Brie y, after reaching 100% con uence, the cells were xed with 4% paraformaldehyde for 15 min at room temperature and were then incubated in a staining solution overnight at 37 °C. On the next day, the stained cells were washed with PBS and observed under an inverted microscope. A blue color indicated the presence of SA-β-Gal. The intensity of the SA-β-Gal staining was calculated using ImageJ software.

Tagging hiPSC-RPE cells with PKH26
The standard protocol was performed as described on the PKH26 Product Information Sheet (MINI2, Sigma). Brie y, a suspension containing 2×10 7 cells was centrifuged (400×g, 5 min) and washed once using fresh medium without serum. After centrifuging, the cells were resuspended in 1 mL of Diluent C.
Dye Solution (4×10 -6 M) was prepared by adding 4 μL of PKH26 ethanolic dye solution into 1 mL of Diluent C. Then, 1 mL of Dye Solution was rapidly added to the cell suspension. The nal concentration after mixing was 2×10 -6 M PKH26 with 1×10 5 cells/cm 2 . The mixing suspension was incubated with periodic mixing at room temperature for 5 min. The staining was stopped by adding an equal volume (2 mL) of serum. Then, the suspension was centrifuged at 400×g for 10 min and washed 3 times. Finally, cells tagged with PKH26 were used for injection.

RPE degeneration chinchillarabbitmodel
Rabbits were weighed on an electronic scale and then injected with 1% NaIO 3 (40 mg/kg, Sigma) via the ear marginal vein. After one week, the rabbits were injected with 1% NaIO 3 (40 mg/kg) again. The successful lesioned models were assessed by ERG assay using the Roland Retiport (RETI-Port/Scan 21) system, which showed no distinctly functional ERG response compared with normal rabbits under similar excitation. Then, these successful models would be used for following cell transplantation.

Preliminary test of hiPSC-RPE cell spheroids in vivo
Animal experiments were approved by the Institutional Animal Care and Use Committee of Jinan University, and animal procedures were conducted following the guidelines of the US National Institutes of Health. Twelve six-month-old chinchilla rabbits with a weight of 1-2 kg were raised in a 12 h dark/light cycle, temperature at 23 ± 2 °C, and relative humidity of 45% to 55%. Water and food were changed every day. These chinchilla rabbits were randomly separated into four groups: control group (n=3, 6 eyes), which received no treatment; Na 2 IO 3 group (n=3, 6 eyes), which received Na 2 IO 3 treatment; Na 2 IO 3 +PBS group (n=6, 6 right eyes), which received Na 2 IO 3 treatment and PBS injection; and the Na 2 IO 3 +hiPSC-RPE group (n=6, 6 left eyes), which received Na 2 IO 3 treatment and hiPSC-RPE cell injection. The leakage or bystander effects were excluded through the inspection of different people. Water containing 210 mg/L cyclosporin A (Sigma) and prednisone were given to these rabbits throughout the experiment to reduce allograft rejection. For PBS or hiPSC-PRE cell injection, the model rabbits were anesthetized with pentobarbital sodium (25 mg/kg, Sigma) and chlorpromazine (5 mg/kg, Sigma). The pupil was dilated using tropicamide (Alcon, Canada) and the eye lid was kept open using a lid speculum. Cell transplantation was performed under a surgical microscope (Ocular Instruments, China) [41]. For subretinal injection, the peritomy was made 2.0 mm posterior to the limbus in the superotemporal quadrant of each eyeball. A sideport knife Beaver blade (BD) was used to make a longitudinal triangular scleral incision starting 2 mm away from the limbus at about the 5° axis toward the choroid until minimal blood re ux appeared. At this point an additional tract through the choroid toward the RPE layer was created using a 30-gauge needle. The hiPSC-RPE cell spheroids labeled with PKH26 were suspended in PBS containing 10 μM Y-27632, and then 10 µL of cell spheroids (approximately 1×10 5 cells) at a density of 20 cell spheroids/µL were slowly injected through the scleral tunnel using a 50 µL Hamilton blunt syringe with a 30-gauge needle (BD), then the syringe was immediately pulled back.

Electroretinographic (ERG) Analysis
ERG was recorded using an Espion ERG Diagnosys machine (Diagnosys, Littleton, MA, USA) [42]. In brief, pigment rabbits were anesthetized with pentobarbital sodium (25 mg/kg, Sigma) and chlorpromazine (5 mg/kg, Sigma) after overnight dark adaption. Eyeballs were anesthetized with oxybuprocaine (Alcon, Ontario, Canada) and pupils were dilated with compound tropicamide (Alcon, Ontario, Canada). Flash ERG waves were measured using a gold wire corneal electrode, a forehead reference electrode, and a ground electrode near the tail. Scotopic, rod-mediated responses were obtained from dark-adapted animals at the following increasing light intensities: 0.1, 1 and 10 cd. s/m 2 . Photopic, cone-mediated responses were performed following 10-minute light adaptation on the background light intensity of 30 cd. s /m 2 . Recordings were obtained at the light intensity of 10 cd. s/m 2 . The ERG a-wave amplitudes were measured from the baseline to the negative peak and the b-wave was measured from the trough of the a-wave to the peak of the positive wave, or when the a-wave was not present, from baseline to the peak of the rst positive wave.

Immunocytochemistry
Rabbits were sacri ced with an overdose of sodium pentobarbital. Their eyeballs were removed, punched using a 23-gauge needle (BD), and xed in 4% paraformaldehyde overnight at 4 °C. Then, these eyeballs were dehydrated with a graded series of ethanol and xylene and subsequently were embedded using para n wax [43]. The para n sections (5 mm) were cut and dewaxed in water. After antigen repair, these sections were xed with 4% paraformaldehyde for 15 min and washed 3 times with PBS. They were permeabilized with 0.1% Triton X-100 for 15 min at room temperature and were blocked with 3% (w/v) BSA (Sigma) at room temperature for 1 h. Then, they were incubated with primary antibodies as shown in Table 2 for 2 h at room temperature followed by FITC-conjugated secondary antibody (1:1000, concentration of 1mg/mL, Bioword) for 1 h at room temperature. These sections were rinsed 3 times with PBS and incubated with DAPI (10 μg/mL) for 15 min. Then, they were examined under an inverted uorescence microscope.

Statistical analysis
All of the data are presented as the mean ± SEM of at least three separate experiments, and statistical signi cance was evaluated using one-way ANOVA followed by Tukey's multiple comparison tests. Student's unpaired t-test was used to compare two different groups. P< 0.05 was considered statistically signi cant.

Result
Comparison of non-colonydissociatedhiPSCsand clonal hiPSCs hiPSCs showed clonal growth, and the clones gradually became larger over time. The dissociated hiPSCs in non-colony culture could adherently grow well. The dissociated hiPSCs displayed a homogenized cellular state. Immuno uorescence assays demonstrated that both clonal hiPSCs and dissociated hiPSCs were positive for the expression of Nanog, Oct4, Sox2, SSEA4, and TRA-1-60 proteins (Fig. S1A).
RT-PCR also showed that the genes of Nanog, Oct4, Sox2, c-Myc, Klf4, Lin28, and Nestin were expressed but the Pax6 gene was not expressed in dissociated hiPSCs and clonal hiPSCs (Fig. S1B). The results of qPCR showed that the mRNA expression levels of Nanog, Sox2, Oct4, c-Myc, Klf4, Lin28, and Nestin had no signi cant differences between dissociated hiPSCs and clonal hiPSCs (Fig. S1C). The chromosomal stabilities of the dissociated hiPSCs (P97) and clonal hiPSCs (P89) were detected by karyotype analysis and showed that both maintained a normal 46XX chromosome complement (Fig. S2). These results demonstrated that dissociated hiPSCs and clonal hiPSCs can maintain their undifferentiated characteristics and chromosomal stability.

Differentiation and identi cation of hiPSC-RPE cells
Because a prerequisite for hiPSC differentiation is shutdown of the self-renewal machinery, dissociated hiPSCs were pretreated in E7 medium without FGF2 for 2 days to encourage spontaneous differentiation. FGF2 withdrawal from the culture medium may promote neuro-ectoderm induction [44], and RPE cells belong to neuro-ectoderm lineages. Therefore, the dissociated hiPSCs were switched into Proneural medium with the sequential addition of de ned factors at speci c time points. The differentiated cells were mechanically enriched and passaged for further experiments. A schematic illustration of the process is displayed in Fig. 1A.
Meanwhile, western blotting showed that hiPSC-RPE cells also expressed RPE speci c proteins such as RPE-65, CRALBP, EMMRPIN, Mitf, Otx2, and Tyrosinase but not Nanog, Oct4, Sox2, or Klf4. This pattern of marker protein expressions was opposite to that of hiPSCs but was similar to that seen in hRPE cells ( Fig.  2A) (Fig. 2B).
AFM was used to investigate the topography and de ection error morphologies in 2D and 3D of hiPSC, hiPSC-RPE, and hRPE cells on the cell surface and in the cell nucleus. AFM images of hiPSCs in the cells displayed a volcano-like distribution. The cell surface of the hiPSCs was relative rough, and the cell nucleus showed an oval shape. AFM image of hiPSC-RPE in the cells revealed a pancake-like con guration, which was similar to that of the hRPE. The cell surface of the hiPSC-RPE was relatively smooth, and the cell nucleus displayed a cobblestone-like appearance. The nucleo-cytoplasmic ratio of the hiPSCs was higher than that of hiPSC-RPE and hRPE ( Fig. 2C and S3A). Statistical analysis showed that cell length, width, and height parameters of hiPSC-RPE cells were similar to those of hRPE cells ( # P>0.05) but were different from those of hiPSCs ( * P<0.05) (Fig. S3B). At the same time, the cell nucleus length, width, and height parameters of hiPSC-RPE cells were similar to those of hRPE cells ( # P>0.05) but were different from those of hiPSCs ( * P<0.05) (Fig. S3C).
Taken together, these results illustrated that our protocol is able to e ciently differentiate hiPSCs into RPE-like cells and that the morphology of hiPSC-RPE cells was similar to that of hRPE cells but was different from that of hiPSCs.

Maintenance of native RPE function in hiPSC-RPE cells
One of the most important properties of RPE cells is the barrier function created by tight junctions. Tight junctions encircle each cell to form the occluding seal monolayers that retard diffusion across the paracellular space [45]. Cytoplasmic anchor protein ZO-1 is often used to establish tight junctions. Zstack confocal microscopy revealed that ZO-1 showed strong expression on the apical side of both hRPE and hiPSC-RPE cells (Fig. 3A). This polarized expression of ZO-1 is in accordance with previous reports [34,46]. An essential function of RPE cells is phagocytosis of outer segments shed from photoreceptors.
To test whether hiPSC-RPE cells had phagocytosis function, both hRPE and hiPSC-RPE cells were incubated with FITC-POS, xed, and then stained for ZO-1. The results showed that FITC-POS particles were speci cally engulfed by cells, and extensively distributed into the cytoplasm of hiPSC-RPE and hRPE cells (Fig. 3B), indicating that these cells were capable of phagocytosis. Fluorescein leakage test (FLT) results showed that hiPSC-RPE cells had similar FLT values with hRPE cells at 4, 12, 24, and 36 h (Fig.  3C).
In addition, a transepithelial electrical resistance (TEER) assay was conducted to assess the barrier behavior of the hRPE and hiPSC-RPE cells. The data showed that the value of TEER in hRPE is 129.31.4 Ω cm2, and in hiPSC-RPE it is 127.11.1 Ω cm 2 , but there was no signi cant difference between them ( # P>0.05) (Fig. 3D).
Taken together, these results demonstrated that hiPSC-RPE cells had RPE-like functions such as forming a barrier, phagocytic activity, and a polarized membrane.

Seeding of hiPSC-RPE cell spheroids in vitro
To determine whether hiPSC-RPE cell spheroids remained viable after in vivo transplantation, we performed in vitro simulation experiments in which hiPSC-RPE cell spheroids were seeded on DCM. The porcine lamellar corneal matrix with the thickness of 100 µm and the diameter of 9.29±0.24 mm was produced using a microkeratome (Fig. S4A). The frozen DCM and normal cornea (control) were cut to obtain tissue sections for DAPI and H&E staining. The results showed that the DCM had almost no cells in it, but the control cornea had many cells in the epithelial and stroma layers (Fig. S4B).
3D agarose molds containing eighty-one wells were produced using the commercially available silicone concave microwells. These molds were used to generate hiPSC-RPE cell spheroids (Fig. S4C). hiPSC-RPE cell spheroids were harvested and seeded onto the DCM. Inverted light microscopy imaging showed that proliferating epithelial-like hiPSC-RPE cells could migrate from the periphery of adherent cell spheroids. On day 14, hiPSC-RPE cell spheroids had disappeared and instead formed an epithelioid-like monolayer that completely covered the DCM (Fig. 4A). A live/dead assay was performed to assess the viability of hiPSC-RPE cells cultured on the DCM. On day 0, cells in hiPSC-RPE cell spheroids were viable, showing green uorescence by Calcein AM staining. Red uorescence indicated dead cells by EthD-III staining.
After 7 days, the majority of cells were stained by green uorescence, and a limited number of dead cells were located mainly in the center of hiPSC-RPE cell spheroids, showing red uorescence. On day 14, cell spheroids had disappeared, and the remaining monolayer cells were only stained by green uorescence (Fig. 4B). Meanwhile, double immuno uorescence staining showed that the monolayer hiPSC-RPE cells cultured on the DCM expressed Occludin, Mitf, ZO-1, and RPE-65 (Fig. 4C).
In addition, hiPSC-RPE cells were stained by SA-β-Gal to determine the proportion of senescent cells. A blue color indicates the presence of SA-β-Gal staining. Positive staining of SA-β-Gal was detected in spheroids or conventional cultured hiPSC-RPE cells on days 7 and 14 (Fig. 4D). Statistical analysis of SAβ-Gal staining revealed that the percentages of positive staining cells in spheroid or conventional culture hiPSC-RPE cells were 8.01±0.44% and 22.47±0.75% on day 7, respectively. The senescent cells in the spheroid culture hiPSC-RPE cells were lower at 14.46±0.32% than those in the conventional culture hiPSC-RPE cells ( ** P<0.01) (Fig. 4E). After 14 days, the cell spheroids had disappeared, and there were 15.23±0.61% senescent cells in the spheroid culture hiPSC-RPE cells and 68.68±1.34% in the conventional culture hiPSC-RPE cells. There was a 53.38±0.95% increase in senescence in the conventional culture hiPSC-RPE cells compared with the spheroid culture hiPSC-RPE cells ( ** P<0.01) (Fig.  4E). Furthermore, western blotting showed that hiPSC-RPE cells in a spheroid culture could express Nestin, which was not expressed in hiPSC-RPE cells in conventional culture (Fig. 4F); Nestin is a neural progenitor cell marker.
In our study, we reported a method to produce hiPSC-RPE cell spheroids using agarose multiwell dishes. hiPSC-RPE cell spheroids are able to maintain an integrated monolayer epithelioid morphology as well as viability after being injected onto DCM. Spheroid culture could enhance the viability of hiPSC-RPE cells, which will lay the foundation for hiPSC-RPE cell spheroid transplantation in vivo.

Preliminary test of hiPSC-RPE cell spheroidsin vivo
To evaluate the states of transplanted cells in vivo, hiPSC-RPE cell spheroids were labeled with lipophilic red dye (PKH26), which can display red uorescence (Fig. S5A). The classical external approach for RPE transplantation was conducted to inject hiPSC-RPE cell spheroids into the upper temporal region of the equator in rabbit eyeballs (Fig. S5B). A schematic of the external approach cell transplantation is shown in Fig. S5C.
ERG provides a gross measure of light-mediated trans-retinal function. Improved retinal function was documented by scotopic ERG recordings using the Roland Retiport (RETI-Port/Scan 21) system (Fig. 5A).
The results showed that photopic (light-adapted at 0.1, 1, 10 cd. s/m 2 ) responses were reduced in the group that received NaIO 3 in comparison with control. After one month of sub-retinal transplantation, a small functional ERG response was observed in NaIO 3 +hiPSC-RPE group compared with NaIO 3 +PBS group, which was unresponsive under similar excitation (Fig. 5B). Additionally, mean b-wave amplitudes of different groups in response to white ashes of increasing intensity after dark adaptation were statistically analyzed. The results showed that control group had distinct b-wave amplitudes, which were higher than those of other groups, but both NaIO 3 and NaIO 3 +PBS groups almost had no b-wave amplitudes, and had no differences with each other. In addition, NaIO 3 +hiPSC-RPE group could acquire partial b-wave amplitudes (Fig. 5C).
One month after transplantation, H&E staining showed that control group had a complete RPE layer, but the RPE layers in NaIO 3 and NaIO 3 +PBS groups were almost totally degenerated, and a segmental sheet growth RPE layer was observed in NaIO 3 +hiPSC-RPE group (Fig. 6A). The immunocytochemistry assay showed that a speci c marker of human cells (AHNA) could not be detected in control, NaIO 3 , and NaIO 3 +PBS groups, but was positive in NaIO 3 +hiPSC-RPE group. Meanwhile, cells tagged with PKH26 could not be observed in in control, NaIO 3 , and NaIO 3 +PBS groups, but could be detected in NaIO 3 +hiPSC-RPE group (Fig. 6B).
These results demonstrated that hiPSC-RPE cell spheroids injected into the subretina of RPE degenerated chinchilla rabbits could survive for at least one month, maintain segmental sheet growth, and partially recover the visual function.

Discussion
The retina is a sensor and processor of visual information. Retinal damage can cause permanent loss of vision. However, recent studies in regenerative medicine have aroused hope for rescuing visual function.
One of the strategies for treatment of retinal diseases is the transplantation of retinal cells, especially RPE. Partial recovery of visual function by replacement of RPE cells has been reported [47][48][49].
Due to the limited resources of donor cells, RPE cells derived from stem cells, especially ESCs and iPSCs, are required. In the eld of ophthalmology, great progress has been made in research on iPSC differentiation. Currently, iPSCs can be successfully differentiated into corneal epithelial cells (CECs) [50,51], eye trabecular reticular cells [52], rod cells [53], retinal progenitor cells [54,55], RPE [56], and so on [23]. Brandl C et al. documented that adult dermal broblasts can serve as a valuable resource for iPSC-RPE with characteristics highly reminiscent of RPE cells [57].
Usually, the process of differentiation of iPSCs into RPE is time consuming and has a low e ciency. The survival rate of transplanted iPSC-RPE cells in vivo is not high [10,58,59]. Recently, Buchholz et al.
reported that iPSCs induced with a series of small molecules such as IGF-1, Noggin, DKK-1, Nicotinamide, bFGF, Activin A, SU5402, and VIP could be rapidly differentiated into RPE phenotype cells after only 14 days [60]. In our study, we modi ed the differentiation method proposed by Buchholz et al. Non-colony dissociated hiPSCs instead of clonal hiPSCs were pretreated with E7 iPS culture medium without bFGF for 2 days to inhibit proliferation and promote differentiation [59]. RPE-like cells began to appear on day 8 and formed a cell monolayer after removal of the dead cells through changing the medium on day 14.
Our results showed that the percentage of CRALBP + cells among the hiPSC-RPE could reach 97.4%.
Usually, an RPE subpopulation with divergent morphology can be partially separated by selective trypsinization of calcium-dependent adhesion [61]. In our study, RPE-like cells were digested using 0.25% EDTA instead of trypsin to reduce any damage to their calcium-dependent adhesion. Generally, RPE cells tend to undergo morphological changes post passaging due to epithelial-mesenchymal transition (EMT) [62][63][64] . The tight junction of cells is vital to maintain epithelial morphology and restrain EMT occurrence [65]. In this study, hiPSC-RPE cells could retain their epithelial morphology and barrier function as shown by Matrigel and Activin-A (MA) treatment based on our previous study [37]. Their morphological features showed that hiPSC-RPE cells were similar to hRPE cells but not hiPSCs. In addition, hiPSC-RPE cells could express ZO-1 in relation to tight junctions. hiPSC-RPE cells had similar functions to hRPE cells in term of barrier and phagocytosis activity. Taken together, hiPSC-RPE cells in our study possessed normal morphology and functioned like RPE cells. Kokkinaki et al. also demonstrated that iPS-RPE could exhibit ion transport, membrane potential, and gene expression patterns similar to native RPEs [66].
Recent advances in cell culture research have shown that 3D cultures can bridge the gap between cell culture ex vivo and live tissue in vivo. Cells cultured in the 3D environment could generate many differences in cell behaviors and characteristics compared with that in the conventional 2D environment [67]. Cell-to-cell and cell-to-matrix interactions are maintained, and the stemness of progenitor cells can be enhanced in the 3D spheroid microenvironment [68]. 3D spheroid cultures can form distinct ECMs and establish new ECM interactions to improve cell viability and in uence cell fate [69][70][71]. Bayoussef et al. documented that spheroid muscle cells exhibited higher proliferation capacity [25].
There are many approaches to generating cell spheroids including centrifugation, pellet culture, hanging drop culture, spinner or rotary dynamic culture systems, and so on [72,73]. Hanging drop or pipetting cells into multiwell plates are time consuming and laborious approaches. The generation of cell spheroids in a suspension culture in non-adhesive culture vessels results in a lack of homogeneity [72]. Here, we reported another method to produce hiPSC-RPE cell spheroids using commercially available agarose multiwell dishes. Cell spheroids with the uniform size could be quickly formed at the bottom of the nonadhesive agarose microwells. The agarose multiwell dishes easily allowed for the simultaneous generation of hundreds of homogenous hiPSC-RPE cell spheroids.
The progenies derived from cell spheroids showed only limited staining for the senescence marker SA-β-Gal, had a regular morphology, and grew at a higher density compared with passaged CECs [74]. Shortterm spheroid formation of adipose-derived stem cells (ASCs) before monolayer culture could enhance their properties of stemness, angiogenesis, and chemotaxis. ASCs derived-cell spheroids exhibited higher expansion e ciency with less senescence including SA-β-Gal staining and p21 expression [75]. In our study, SA-β-Gal staining assays showed that spheroid cultures could reduce cell senescence in hiPSC-RPE, and hiPSC-RPE cell spheroids could positively express Nestin, which is a major intermediate lament protein of embryonic central nervous system progenitor cells. Expression of "progenitors" marker Nestin revealed a distinct molecular reprogramming of the cells into a stem cell state [76].
The polarized monolayer RPE cells have a basal side adherent to the Bruch's membrane (BM), which consists of 2-4 μm acellular extracellular matrix (ECM) complexes among which collagen forms a major component. The collagenous layer is composed of collagen type I, III, and V brils. The collagen grid is embedded in a mass of interacting biomolecules, such as glycosaminoglycans (chondroitin sulfate, dermatan sulfate, and hyaluronic acid) and components of the coagulation and complement system [77].
To better observe the monolayer growth of hiPSC-RPE cell spheroids in vitro, we used DCM as a BM substitute. DCM has a similar structure to BM but the former is more convenient to manipulate and observe. DCM is also comprised of aligned arrays of hydrated type I/V collagen brils, glycosaminoglycan, keratin sulfate, and dermatan sulfate [78]. An experiment of cell spheroids seeded on the DCM in vitro were used to assess the state of cells, which might offer some useful experiences for transplantation in vivo.
Huang et al. reported that cells from dermal papilla (DP) spheroids were able to maintain their structural integrity and cellular viability after an in vitro simulation experiment of the injectability of DP spheroids on culture plates [51]. As a natural ECM biomaterial, DCM is easier to obtain and is more suitable for observing cell attachment and migration [26]. Our results showed that hiPSC-RPE cells could migrate from the periphery of adherent cell spheroids and grow into the con uent monolayer on the DCM in vitro over 14 days when incubated with the ROCK inhibitor Y-27632. hiPSC-RPE cell spheroids maintained high cell viability for transplantation. Y-27632 can enhance cell adhesion, motility, and proliferation of human CEC spheroids [79]. Y-27632 treatment also allowed prostate colony cells to re-plate e ciently, increased the cloning e ciency of prostate stem cells in a prostate sphere assay, and suppressed cell dissociation induced apoptosis. Our previous study also con rmed that Y-27632 could decrease cell death inside spheroids, enhance cell proliferation in the spheroid periphery, and promote the monolayer growth of injectable B-CEC spheroids.
There are internal and external approaches for subretinal RPE transplantation. Dorey et al. reported that the internal technique was more precise while the external approach consistently resulted in a higher percentage of transplanted cells on Bruch's membrane [80]. Moreover, subretinal injection of hiPSC-RPE cells by the external approach is relatively easier and safer compared with the internal approach. Thus, we chose subretinal injection of hiPSC-RPE cells through an external approach in vivo study. One month after injection of hiPSC-RPE cell spheroids into the subretinal space of RPE degeneration chinchilla rabbits, a segmental sheet of RPE was observed in the cell transplantation eyeballs. Other studies have shown that mono-and multilayer RPEs could be formed after injection of RPE micro-aggregates into the subretinal space, and apoptotic cells were rarely seen [81]. In contrast to transplantation of dissociated cells, micro-aggregate transplantation developed in small patches in the outer segment of the neural retina, and these transplanted cells could survive and maintain their outer segments over the long-term [82]. Since the subretinal space is an immune-privileged site, the post-transplantation hiPS-RPE spheroids could survive at least one month without apparently systemic immunosuppression in our model.
It is important to explore how the anatomical preservation of photoreceptor to bipolar cell synapses correlated with the activation of retina circuits to photoreceptors as re ected by the ERG b wave [83]. In rats that received sub-retinal cell transplantation, there was clear preservation of scotopic ERG b-waves [84]. In our study, the ERG results showed that the transplantation of hiPSC-RPE spheroids could partially recover the ERG responses and have mean b-wave amplitudes in RPE degenerated chinchilla rabbits.