Retina Organoid Transplants Develop Photoreceptors and Improve Visual Function in RCS Rats With RPE Dysfunction

Purpose To study if human embryonic stem cell–derived photoreceptors could survive and function without the support of retinal pigment epithelium (RPE) after transplantation into Royal College of Surgeons rats, a rat model of retinal degeneration caused by RPE dysfunction. Methods CSC14 human embryonic stem cells were differentiated into primordial eye structures called retinal organoids. Retinal organoids were analyzed by quantitative PCR and immunofluorescence and compared with human fetal retina. Retinal organoid sheets (30–70 day of differentiation) were transplanted into immunodeficient RCS rats, aged 44 to 56 days. The development of transplant organoids in vivo in relation to the host was examined by optical coherence tomography. Visual function was assessed by optokinetic testing, electroretinogram, and superior colliculus electrophysiologic recording. Cryostat sections were analyzed for various retinal, synaptic, and donor markers. Results Retinal organoids showed similar gene expression to human fetal retina transplanted rats demonstrated significant improvement in visual function compared with RCS nonsurgery and sham surgery controls by ERGs at 2 months after surgery (but not later), optokinetic testing (up to 6 months after surgery) and electrophysiologic superior colliculus recordings (6–8 months after surgery). The transplanted organoids survived more than 7 months; developed photoreceptors with inner and outer segments, and other retinal cells; and were well-integrated within the host. Conclusions This study, to our knowledge, is the first to show that transplanted photoreceptors survive and function even with host's dysfunctional RPE. Our findings suggest that transplantation of organoid sheets from stem cells may be a promising approach/therapeutic for blinding diseases.

difficult to scale to a clinically relevant level and may not be available for clinical trials owing to ethical concerns.
Stem cells have been considered a promising source for cell therapy. 14 Pluripotent stem cells, such as human embryonic stem cells (hESCs) and induced pluripotent stem cells, can be differentiated into three-dimensional retinal organoids with similar properties as fetal retina. [15][16][17][18][19] Although stem cell-derived transplants improved visual function, transplants only formed rosettes (spherical arrangement of retinal layers, with photoreceptors in the center and separated from RPE). [20][21][22] Furthermore, it has been shown that both photoreceptors and RPE were necessary for optimal transplant photoreceptor organization. 10,23,24 Many retinal diseases include both RPE dysfunction and photoreceptor degeneration. To restore vision in those patients, both RPE and photoreceptors should be replaced. However, this goal is difficult to achieve with stem cell-derived tissues because RPE and retina organoids have different media requirements. Thus, one interesting question will be whether RPE is needed for the photoreceptors in the transplant to function normally. If the transplanted photoreceptors can survive and function in the presence of dysfunctional host RPE, they may restore vision not only in patients with photoreceptor degeneration, but also in patients with RPE dysfunction.
The RCS rat is a model of photoreceptor degeneration owing to RPE dysfunction 2 , caused by a deletion in the Mer tyrosine kinase (MerTK) receptor that abolishes RPE cell phagocytosis. 3,25 Most photoreceptors are lost at 2 to 3 months of age, resulting in severely impaired retinal function. [26][27][28] RPE cells, [29][30][31] and other cell types, [32][33][34] can rescue photoreceptors when transplanted at early stages (P21- 28), but not at later stages of degeneration (P42). 25 Therefore, we investigated whether transplanted photoreceptors can survive and function in a rat model of RPE dysfunction by transplanting retinal organoid sheets at a later stage (P44-56). Furthermore, nude immunodeficient RCS rats were used, which do not reject xenografts owing to lack of T cells. 35

Cell Culture
CSC-14 hESCs (AIVITA Biomedical Inc., Irvine, CA; National Institutes of Health [NIH] registered line 0284) were grown and expanded using a chemically defined and xeno-free custom formulated media (Irvine Scientific, Irvine, CA) supplemented with low levels of bovine fibroblast growth factor and activin-A. Cells were grown on thin Matrigel (Corning, Corning, NY) and passaged every 3-4 days at 1:6 to 1:10 splits using Collagenase IV digestion. Retinal organoids were generated using a protocol previously described. 19,22

Quantitative Polymerase Chain Reaction (qPCR)
A custom qPCR array (Qiagen, Germantown, MD; Cat# CAPH13339; 84 retinal genes and 5 housekeeping genes) was used to analyze the expression pattern in retinal organoids and control human fetal retinal tissues. Fetal control tissues (n = 7) ranged from day 105 to 145 and were obtained from Human Stem Cell Research Oversight Committee-approved suppliers. Differentiated retinal organoids were analyzed at day 37 to 70 (n = 8). Most samples analyzed were postprocessed pieces remaining from organoids used to dissect retinal sheets for transplantation. The genes analyzed are listed in Table 1. RNA was isolated using Trizol reagent (Qiagen), DNase I digested (Thermo Fisher, Waltham, MA), and phenol:chloroform extraction (Thermo Fisher). cDNA was generated using RT 2 cDNA synthesis kit (Qiagen). Amplification was performed using RT 2 Sybr Green with ROX qPCR master mix (Qiagen), with the following cycling conditions: 95°C (10 minutes); followed by 40 cycles of 95°C (1 minute), and 60°C (30 seconds). Cycle threshold (Ct) values were determined using Viia7 RUO software (ThermoFisher). Delta Ct values were calculated using RPL7 as the housekeeping gene. The mean Delta Ct value per gene was determined and scatterplots of mean delta Ct values for human fetal retina vs retinal organoids was plotted. Both the "hclust" R-program algorithm to analyze qPCR data and code used to create the gene array scatterplots were downloaded from The R Foundation (https://www.r-project.org/).

Experimental Animals
For all experimental procedures, animals were treated in accordance with the NIH guidelines for the care and use of laboratory animals, the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and under a protocol approved by the Institutional Animal Care and Use Committee (IACUC #2006-2698, and AUP-18-145). RCS nude rat transplant recipients were generated by crossing pigmented dystrophic RCS rats (RCS-p+) with Hsd:RH-Foxn1rnu (mutation in the foxn1 gene; no T cells) rats. 35 Recipient rats have a mutant MerTK gene and a T-cell deficiency resulting in immunocompromised and retinal degenerate rats. To prevent infections of nude rats, rats were inspected daily, and cages changed under a laminar flow hood. Eyes of nude rats were cleaned every 2 weeks under isoflurane anesthesia.

Retinal Sheet Preparation
Retinal organoids (differentiation days 30-70, average 47 days) were selected based on transparency and morphological criteria (a hollow spherical shape with a laminated structure) under a phase contrast and stereoscope. The structures were prepared into retinal rectangular sheets (0.7-1.3 × 0.6 mm) for transplantation. RPE "blobs" were dissected away. A second, smaller piece (0.2-0.3 × 0.6 mm) was also cut to be used as a "spacer" piece against the mandrel (to prevent the main piece from sticking to the mandrel).

Transplantation
Recipient rats (P44-56, either sex) were randomized into agematched nonsurgery (n = 13), sham (n = 16), and transplant (n = 33) cohorts. The group size for transplanted animals was set higher because some transplanted animals were used only for histology (n = 19) and not for functional tests to investigate transplant development. Two rats were eliminated after the first or second optical coherence tomography (OCT) examination because of faulty surgeries or corneal ulcers. Two rats could not be used for the final analysis because they died from anesthesia after OCT.

Spectral Domain OCT (SD-OCT) Imaging
SD-OCT imaging was used to document and monitor the transplant as it developed in the host retina. The general protocol was described previously. 22,36 The SD-OCT images of the retina were obtained using a Bioptigen Envisu R2200 Spectral Domain Ophthalmic Imaging System (Bioptigen, Research Triangle Park, NC) after anesthesia with ketamine/xylazine and pupil dilation with atropine. Scans of a 2.6 × 2.6 mm area were taken to include the optic disk. If necessary, additional scans were taken to include areas in the further retinal periphery. The 488 × 488 × 5 (# B-scans/#A-scans/B-scan averaging value) scans were used to obtain fundus images; then the same area was scanned by 800 × 20 × 80 or 700 × 70 × 25 scans to obtain B-scans. Transplanted rats (n = 33) were imaged every 1 to 2 months, starting 2 weeks after surgery, up to 9.5 months of age (8.5 months after surgery). Rats with transplant misplacement into the vitreous or excessive surgical trauma such as optic nerve or corneal damage were excluded from further analysis after the first or second examination (n = 2). The last scan was scheduled as close as possible to the terminal experiment (superior colliculus [SC] recording). Sham (n = 10) and control nonsurgery RCS nude rats (n = 10) were imaged at approximately similar ages. Using 488 × 488 × 5 scans, B-scan images were used to outline the transplant edges on corresponding fundus images. Then Image-J was used to calculate the size (area) of the transplants. In B-scans, the transplant showed the same reflectivity as the host retina. B-scan images were carefully examined to exclude high reflections caused by scars.
Pupils were dilated using 1% atropine eye drops (Akorn Pharmaceuticals). Contact lens electrodes were placed on the cornea of both eyes, with reference and ground electrodes placed subcutaneously. An optically clear ophthalmic gel was used to maintain hydration and conductivity between the cornea and recording electrodes. Scotopic testing was conducted with flash stimuli intensities ranging from 1 to 25,000 millicandela followed by photopic testing (light adaptation of 10 minutes before the photopic test that records flash stimuli responses of 10-25,000 millicandela).

Optokinetic Test (OKT)
At the age of 1.5, and 4 to 8 months, corresponding with 1 week before surgery and monthly at 2 to 6 months after surgery, the visual acuity of RCS nude rats (transplanted, sham surgery, and nonsurgery age matched controls) was measured by recording videos of optomotor responses to a virtual cylinder with alternating black and white vertical stripes (Optomotry, Cerebral Mechanics Inc., Alberta, Canada) as described previously. 22,36 Some rats missed testing time points either because they were treated for infections or had been recorded in the SC. Rats were dark-adapted for at least 1 hour before testing. Optomotor responses were recorded at 6 different spatial frequencies (0.05-0.45 cd/m 2 ) for 1 minute per frequency by testers blinded to the experimental condition. Both the left and right eyes were tested by alternating the direction of the moving stripes. Two independent tests were performed at each time point, with at least 1 hour in between, with one test going from lowest to highest frequency, and the other from highest to lowest frequency. The best visual acuity of the two tests was used for analysis. All tests were video recorded and evaluated off-line by two independent observers blinded to the experimental conditions. Any discrepancies between the two observers resulted in a reanalysis of videos by a third observer, and data discussion before giving a final score, and before decoding the experimental condition.

Superior Colliculus (SC) Electrophysiology
Responses to light flashes were recorded from the surface of the exposed SC as previously described. 22,36 During recording, the tester was blinded to the group allocation of the animal. After overnight dark adaptation, responses from transplanted RCS nude rats (n = 14) were recorded between 5.9 and 9.7 months after surgery (age 7.5-11.3 months) and compared with responses from agematched, nontransplanted RCS nude (n = 13), and sham rats (n = 16). A tungsten microelectrode (0.5 M impedance; MicroProbe, Inc., Carlsbad, CA) was used to record multiunit electrical responses from 50 to 55 locations on the superficial layer of the SC, approximately 200 to 400 μm apart (ADInstruments, Inc. Colorado Springs, CO). Light stimuli (20 ms duration) were delivered approximately 10 times at 10-second intervals at an intensity of 0.58 to -6.13 log cd/m 2 . When responses were found, the intensity of the light stimuli was decreased until there was no response to determine the threshold. Responses to the strongest light stimuli (stimulus level 0.58 log cd/m 2 ) were quantified and formed into a map over the area of the SC. Any up or down deflection higher than the background recording in the 100 ms before stimulation was considered a response (spike). All spikes occurring at 30 to 210 ms after the onset of the photic stimulus were counted. The sum was averaged across stimulus presentations. Analyses of the responses (spike counts and locations) were performed using a custom MATLAB program (Mathworks, Natick, MA). 13,22

Histology and Immunofluorescence
After injection of anesthetic overdose, rats were perfusion fixed with cold 4% paraformaldehyde in 0.1 M Na-phosphate buffer at 1.8 to 10.0 months after surgery (transplanted rats, n = 33; sham surgeries, n = 16; age-matched controls (AMC), n = 13). Eye cups were dissected along the dorsoventral axis, infiltrated overnight in 30% sucrose before embedding in Tissue-Tek O.C.T. compound and frozen using -60°C isopentane on dry ice. Retina organoids were frozen in O.C.T. using a similar procedure. Serial 10 μm cryostat sections were cut and stored at -20°C. Every fifth slide was stained using hematoxylin and eosin and analyzed for the presence of donor tissue in the subretinal space of the RCS host. Hematoxylin and eosin-stained slides were imaged on an Olympus BXH10 using an Infinity 3-1U camera. For immunofluorescence and diaminobenzidine (DAB) analyses, cryostat sections underwent antigen retrieval at 70°C with Histo-VT One (Nacalai USA Inc., San Diego, CA) and blocked for at least 30 minutes in 10% donkey serum (IFA) or 20% horse serum (DAB). Primary antibodies are listed in Table 2. Primaries were left on sections overnight at 4°C at specified concentrations. After several phosphate-buffered saline (PBS) washes, slides were incubated for at least 30 minutes at room temperature in fluorescent secondary antibodies, Alexa Fluor 488 donkey anti-rabbit IgG (H+L), Rhodamine X donkey anti-mouse IgG (H+L) and AF647 donkey anti-goat IgG (H+L) or biotinylated conjugated secondary antibodies (dilution of 1:200-1:400) ( Jackson Immunoresearch, West Grove, PA). Fluorescent sections were coverslipped using Vectashield mounting media (Vector Labs, Burlingame, CA) with 5 μg/mL 4,6-diamidino-2-phenylindole. The Stem Cell 121 (SC121) or Ku80-stained sections were incubated with an ABC kit (Vector Labs, Burlingame, CA) and developed with DAB for up to 4 minutes and according to manufacturer's instructions. Fluorescence was imaged using a Zeiss LSM700 confocal microscope (Zeiss, Oberkochen, Germany) taking tiled stacks of 5 to 8 micron thickness at 40× (selected images). Zen 2012 software (Zeiss, Oberkochen, Germany) was used to extract confocal images. Three-dimensional images were extracted separately for each channel and combined in Adobe Photoshop CS6 software (San Jose, CA). Volocity (×64) software (Perkin-Elmer, Waltham, MA) was used to obtain higher magnification three-dimensional opacity rendered images that could be rotated for better viewing of three-dimensional structures including cell bodies and processes, in addition to co-localization analysis. In addition, Imaris software (Oxford Instruments) was used to analyze colocalization (see Fig. 7).

Experimental Design and Statistical Analysis
Rats (either sex) were randomized into age-matched nonsurgery, sham, and transplant cohorts. The group sizes are similar to or exceed those reported in previous publications and that generally used in the field. Experimenters were blinded to the rat's experimental group. For all statistical analyses, the significance level was calculated in Graphpad Prism software (Graphpad Software LLC, La Jolla, CA)   with paired and unpaired two-tailed t tests using mean ± SEM. The level of significance was set at 0.05.

Code Availability Statement
The "hclust" R-program algorithm to analyze qPCR data was downloaded from The R Foundation and the code used to create the gene array scatterplots. The Matlab code that created the SC heatmaps from raw spike counts is a custom code. Both codes are available upon request.

Development and Characterization of Retinal Organoids
CSC14 hESC developed into a refractive annular structure with a flattened area and a curved ridge of multilayered cells often surrounded by a RPE monolayer after 21 to 28 days of differentiation (d21-28) (modified procedure after 19 ), as shown in our previous study. 22 These structures were subsequently cut out between d30 and d70, and placed in suspension culture. Cut-out structures were designated as a "retinal organoid." To confirm the success of our retinal organoids (developing into retinal progenitors), qPCR gene analysis (an array of 84 genes, see Table 1) was used to compare expression patterns between human fetal retina and stem cell derived retinal organoids (8 samples, d37-70 of differentiation). The to-be-transplanted retinal organoids acquired a gene expression pattern similar to human fetal retina (d105-145 of gestation) and contained all the important cell populations needed for successful integration within the host's microenvironment (Fig. 1a). Organoids showed lamination in their distribution of cell types. At d44 and d51, the ganglion cell marker Brn3b (Fig. 1b) and synaptophysin (a marker for presynaptic vesicles, Fig. 1c) was apparent in developing ganglion cells on the inner side of the organoids. Synaptophysin was also found in developing photoreceptor progenitors on the outside of the organoids (Fig. 1c). The localization of different retinal cell markers changed and could shift between apical and basal regions of the organoid during development. MAP2 (microtubule-associated protein 2) was found in putative ganglion cells and other neurons in the inner layer (Figs. 1d, 1e). CHX10 (VSX2, Visual System Homeobox 2) was found in retinal progenitor cells in the outer layer (Figs. 1d, 1e). OTX2 (Orthodenticle Homeobox 2, a transcription factor important for RPE, photoreceptor and bipolar cell development 38 ) was found both in photoreceptor progenitors and in the putative inner retinal cell layers (Fig. 1d). Organoids also showed immunoreactivity to Recoverin (a marker for photoreceptors and cone bipolar cells), both in the developing outer nuclear layer and the inner retina (Fig. 1e) and CRX (data not shown). These findings suggest that with our "retinal determination" protocol, the CSC14 hESC differentiated into retinal organoids comparable with human fetal retina.

In Vivo Development of Retinal Organoid Transplants Monitored by OCT
To monitor the development of retinal organoids after the transplantation, the rats were imaged with OCT monthly, starting at 2 weeks after transplantation. Figure 2a shows retinal layering by SD-OCT in the normal NIH nude retina. Age-matched control RCS rats showed reduced outer nuclear layer thickness in OCT imaging (Figs. 2b1-b4), 35 similar to age-matched sham surgeries (example of sham surgery with nozzle trace in subretinal space shown in Figs. 2c1-c4). The transplants (average size 0.9 ± 0.03 mm 2 ) were deposited subretinally and covered an average area of 3.6 mm 2 at the last OCT scan (Table 3). Transplants continued to differentiate and developed photoreceptor rosettes several months post transplantation (Figs. 2d1-d4). These data show that   Table 5.
retinal organoids could survive for a long time and mature after transplantation to nude RCS rats.

Bipolar Cells
The development of transplants (identified by Ku80, a marker for human nuclei) at different days after surgery (60, 152 and 217 days after surgery ) is shown in Figure 4. PKCα-immunoreactive rod bipolar cells were found in both the host retina and transplants. Transplant PKCα immunoreactivity increased at later times after surgery (Figs. 4c-f), and transplant rod bipolar cells were oriented radially in the transplant inner nuclear layers. In some areas, host rod bipolar cells extended processes into transplants (Figs. 4d and 4f). In some transplants forming rosettes, transplant rod bipolar cells were very close to host rod bipolar cells and integrated with the host (Figs. 4d and 4f). Transplant processes in the host inner plexiform layer (IPL) intermingled with host rod bipolar cell processes (Figs. 4d, 4f). Synaptophysin, a marker for presynaptic vesicles, exhibited strong staining in both host and transplant, indicating the good integration of host and donor tissues (Figs. 4c-f). Ku80 staining demonstrated that increasing numbers of donor cells migrated into and integrated with  Table 3). Extensive synaptic connectivity (synaptophysin) was found between donor and host cells in the host IPL. Scale bars = 50 μm. (a, c); 40 μm (b, d, f) are enlargement of (a, c, and e). For abbreviations, see Table 5. the host retina. Subsequently, the synaptophysin staining became stronger and the migration of donor cells and integration of host and donor cells became more extensive, indicating the continuing development of the donor tissues after transplantation.

Photoreceptors
Donor cells differentiated into photoreceptors after transplantation (Figs. 5, 6, and Supplementary Figure S1). Examples of rhodopsin-immunoreactive rod photoreceptors in the transplant and remaining photoreceptors in the host are shown in Figure 5 and Supplementary Figure S1c. In contrast with Rho S334ter-3 RD nude rats, nude RCS rats retained their photoreceptors for a longer time. Supplementary Figure S1 shows still remaining host photoreceptor outer segments at 2 months after surgery (age 4 months) ( Supplementary Fig. S1a-S1c), which had mostly disappeared at 5 months after surgery (age 7 months) ( Supplementary Fig. S1d-S1f). At the age of SC recording (7.5-10.0 months), host photoreceptors had lost their outer segments (Figs. 5a [left], 5d). Rhodopsin staining showed that transplant rods extended outer segments into the rosette centers (Figs. 5a-c; Supplementary Fig. S1c, S1f).  Fig. S1). This rat (transplant #4, see Table 3) had good responses in the SC. Combination of rhodopsin (rods, red), Ku80 (human nuclei, green) and DAPI (nuclei, blue). (a) Overview. The transplant has developed photoreceptors in rosettes with outer segments (red). There seems to be some host rod photoreceptor rescue close to the transplant (left side in a). Donor nuclei are migrating into the host retina in some places. (b, c) enlargements of (a). Note that there is also a rosette of remaining host photoreceptors (only blue DAPI stain) in (c) (white arrow). (d) Host retina outside transplant, with outer segment debris on the left, and few scattered degenerated host rod cell bodies on the right. Scale bars = 50 μm. DAPI = 4',6-Diamidino-2-phenylindole. See Table 5 for definitions of labels.

Microglia
Iba1-immunoreactive microglia was observed in the retina of all three groups (age-matched controls, sham and transplant, Fig. 8). Microglia cells were also found within the transplanted tissue after transplantation. The number of Iba1 immunoreactive cells was higher in transplanted retinas (n = 10) than in sham (n = 8; P = 0.15) and age-matched control groups (n = 7; P = 0.56), but the above difference was not statistically significant (P > 0.05) (Fig. 8a). Interestingly, some Iba1+ microglia cells were observed in the center of the rosettes (Fig. 8b). Examples of microglial cells in sham surgery and age-matched controls are shown in Figures 8c and 8d, respectively.

Müller Cells
Immunohistochemical staining for anti-cellular retinaldehyde binding protein, specific for Müller cells and RPE,  50 μm (a-f), 4 μm (g, h). See Table 5 for definitions of labels. . The transplant group showed a slightly higher number of microglial cells than age-matched controls and sham groups, but the difference is not significant. (b) Iba1-staining in section of transplant #5 (257 dps, age 308 days). Arrows in enlargement point to microglial cells inside rosettes. (c) Representative section of sham surgery rat (210 dps, age 260 days). (d) Representative section of age-matched control (age 308 days). Scale bars = 200 μm (overviews); 50 μm (enlargements). glial fibrillary acidic protein (a marker for reactive glia) and glutamine synthetase (Müller cells) showed that the retina organoid transplants contained Müller cells that were radially oriented in rosettes (Figs. 9a-c; Supplementary Fig. S4). Müller cells in the host were more reactive than in the transplant as shown by glial fibrillary acidic protein staining (Fig. 9b). The transplants extended many processes into the host IPL that did not stain for cellular retinaldehyde binding protein, specific for Müller cells and RPE ( Fig. 9c; Supplementary Fig. S4). RPE underneath the transplant appeared to be disrupted at times and generally expressed less RPEspecific markers, such as CALBP, Ezrin and RPE-65 than outside the transplant (data not shown). Nevertheless, such transplants could still elicit visual responses.

Visual Function Improvement Evaluated by ERG
To determine whether the organoid could improve visual function, nude RCS rats were tested with ERG at 1 to 2 weeks before surgery (age 4-5 weeks) (Fig. 10a), and 2 to 6 months after transplantation (age 4-7 months) (Figs. 10b, 10c). Our data show at 2 months after surgery a significant improvement in ERG responses to scotopic (P = 0.038) and photopic stimuli (P = 0.008; n = 16) in transplanted eyes (left eyes) compared with nonsurgery eyes (right eyes); the improvement in scotopic ERG (transplant vs sham, P = 0.017; transplant vs AMC, P = 0.022) and photopic ERG (transplant vs sham, P = 0.036; transplant vs AMC, P = 0.009) is also significant when comparing transplanted eyes with that of sham  Table 5 for definitions of labels. and AMC (Figs. 10b, 10c). However, at later time points (4 and 6 months after surgery), ERGs were not detectable for all groups, and a similar difference was not observed (Fig. 10c). ERGs of age-matched control and sham surgeries decreased dramatically with age and showed no difference between both eyes. No ERG responses were detected in any of the RCS rats after 4 months after surgery (6-to 8-month-old RCS rats). These ERG data depicted that visual function improved by the retinal organoid transplants at 2 months after surgery.

Visual Function Improvement Evaluated by OKT
OKT of nude RCS rats showed that the visual acuity of age-matched RCS controls and sham RCS rats significantly decreased after 4 months after surgery (corresponding with an age of 6 months). In contrast, in the transplanted RCS rats, the visual acuity of eyes with hESC-derived retinal organoid transplants showed significant improvement from 2 (transplant n = 16, sham n = 6, AMC n = 5), 3 (transplant n = 15, sham n = 12, AMC n = 10), 4 (transplant n = 25, sham n = 6, AMC n = 10), 5 (transplant n = 15, sham n = 6, AMC n = 8), and 6 (transplant n = 13, sham n = 6, AMC n = 12), months after surgery. There was no difference between age-matched controls and sham control groups in visual acuity (P = 0.865, unpaired t test). In addition, there was no significant difference between the left eye and right eye of sham surgery and age-matched control rats (P = 0.735, paired t test). The nonsurgery eyes of the transplanted rats also showed a similar trend in visual acuity loss. However, a significant improvement in the visual acuity was observed in transplanted eyes compared with nontransplanted eyes (paired t test). This difference was more apparent at later time points (P = 0.001, P = 0.037, P < 0.0001, P < 0.0001, and P < 0.0001 at 2, 3, 4, 5, and 6 months after surgery, respectively; Fig. 11). Furthermore, the visual acuity of the transplanted eyes was significantly higher compared with the nonsurgery control and sham surgery rats (transplant vs sham, P = 0.01; transplant vs AMC, P = 0.007). All these values were lower than those before surgery (1 month old) as compared with normal (nondegenerate) NIH or left eye nude rats with a normal retina (0.42 ± 0.03 cycles/degree). 36

Visual Function Improvement Evaluated by SC Electrophysiologic Recording
Electrophysiologic recording in the SC was also used to test the visual function in the rats ( Fig. 12; Supplementary  Fig. S5). At the age of transplantation (1.5 months), there was still some residual visual activity in a small area of the SC which was disappeared at the age of 3 to 5 months in nonsurgery RCS rats ( Supplementary Fig. S5, Table 4). Strong visual responses (spike activity) were recorded in the SC from transplanted RCS rats (n = 7/14) (Figs. 12c, 12d) at the age of 7.5 to 11.3 months (5.9-9.7 months after surgery). Some of these rats (n = 2) even showed responses to very low levels of light stimulation at the scotopic level; the best light threshold was -1.91 log cd/m 2 (Table 4). No responses could be recorded in the SC of control nonsurgery agematched control rats (n = 13) and sham surgery rats (n = 16) (Figs. 12a, 12b) even at the strongest light stimulation tested FIGURE 11. OKT results. OKT responses of right and left eyes were the same at the age of 1 month (approximately 1-2 weeks before surgery). For eyes of nonsurgery age matched controls and sham, and nonsurgery fellow eyes (right eyes) of transplant group, OKT responses significantly decreased with age. The transplanted eyes showed significant improvements compared with controls which were maintained up to 6 months after surgery, corresponding to the age of 8 months.   Tables 3 and 4). Further statistical tests (R 2 ) in the seven transplant rats with responses to light showed that max spike count, response threshold and percentage of area with response strongly correlated with each other (Figs. 12e-12g). Five of the seven transplanted rats with responses in the SC showed improvements in OKT responses at the last test, whereas only three of the seven rats with responses in the SC showed ERG improvements at 2 months after surgery (Table 3). This finding demonstrates that retinal organoid transplants contribute to the visual functional improvement seen in the nude RCS rats.

DISCUSSION
In the current investigation, consistent with a previous report, 22 we demonstrate that hESC-retina organoid sheets exhibit high expression levels of several key genes for eye-field development, such as RAX, LHX2, Six3, Six6, and Chx10. The similarity between retina organoids and human fetal retina demonstrated the immature development of retinal progenitors in the retina organoids and its potential use as a therapeutic candidate. The present report demonstrates that after the transplantation into the RCS nude rats at advanced disease stage, retina organoid sheets survive and develop into mature photoreceptors and other retinal cells. Visual function tests (ERG, OKT, and SC electrophysiology) show visual improvement after transplantation. For the functional photoreceptors, recycling of visual pigments by isomerization and oxidation, from all-trans retinol to 11-cis retinol, and to 11-cis retinal, is essential. 39 The first essential step of isomerization to form 11cis retinol is catalyzed by RPE65 in RPE for rod visual pigments (review 40 ) and in Müller cells for cone visual pigments. 41,42 Further oxidation is required to restore 11-cis retinal, which is then catalyzed by retinol dehydrogenases, distributed in the retina and RPE in a number of isoforms with overlapping activities. 39,43,44 Another function of RPE is phagocytosis. Photoreceptor cells continuously generate new outer segments from their base while simultaneously releasing outer segments to be phagocytosed by the RPE monolayer. 45,46 In the RCS rat, the genetic mutation in the MerTK gene causes defective phagocytosis of RPE, resulting in photoreceptor degeneration. Theoretically, replacement of both RPE and photoreceptors is needed to restore vision in RCS rats. However, our data show that retinal organoids transplanted without RPE could improve vision. These findings support the theory that RPE phagocytosis is not needed for vision improvement by the retinal organoid transplant; and that other cells, such as Müller cells 42,[47][48][49] and microglia might compensate for the RPE and provide the support required by photoreceptors. 50 Interestingly, some Iba1+ microglia were found in the center of rosettes (location of photoreceptor outer segments) similar to what has been observed in fetal rat and hESC-retina organoid transplants (unpublished observations). The number of Iba1+ microglia was not significantly increased in the transplant group compared with sham and age-matched control groups, suggesting that the surgery did not cause an inflammatory response. The presence of Iba1+ microglia cells and Müller cells in the transplants may compensate for the lost phagocytic function of the RPE. 42,51 However, we do need to be aware that RCS rat is a model of RPE dysfunction (loss of phagocytosis) instead of an RPE degeneration model. Phagocytosis is one of the main functions of RPE. Without it, photoreceptors will die. A recent study showed that surgical removal of debris in subretinal space of RCS rat could rescue the remaining photoreceptors. 52 Further studies are required to test if the photoreceptors in the transplants produce fewer outer segments or if their outer segments are removed by other cells. Another possibility may be that stem cell-derived photoreceptors may have better capability to survive.
Our previous work 22 showed that similar retina organoid (photoreceptor progenitor) sheet transplants improved vision in the RD Rho S334ter-3 nude rats. Because the Rho S334ter-3 rat loses all photoreceptors but still has functional RPE, it is understandable that the transplant could replace damaged photoreceptors and improve visual function. Our current finding that retina organoid sheet transplants also improved vision in the RCS nude rats suggests that transplants can function in the presence of dysfunctional RPE. However, the success rate of visual responses in the SC found in the RD nude rats (9/13) after transplantation was higher than that observed in the RCS nude rats (7/14). Although Müller cells or microglia can support photoreceptors, they may not compensate for the quality of support provided by RPE cells. This finding suggests that cotransplanting retina organoid and RPE may be desirable for a more robust restoration of the visual activity in retinal diseases caused by RPE dysfunction or degeneration.
After transplantation of fluorescently labeled dissociated cells, several studies have demonstrated the possibility of cytoplasmic material exchange between the transplanted cells and the host photoreceptors, instead of direct integration of transplanted photoreceptor precursors. This causes the rescue of host photoreceptors. [53][54][55][56][57][58] Interestingly, only photoreceptor-photoreceptor or Müller-photoreceptor interactions showed the exchange of cytoplasmic material. The donor cells do not exchange intracellular content with other cells (horizontal, bipolar, Müller, or amacrine cells). Furthermore, the cytoplasmic exchange occurred only when the host outer limiting membrane was broken and an outer nuclear layer was present. In our study, there was still a thin outer nuclear layer present at the time of transplantation. Thus, there could be more cytoplasmic transfer in the RCS rat than in the Rho S334ter-3 rat. However, Ku80 staining (a donor cell nuclear marker) also confirmed that donor cells migrated into and integrated with the host retina, and that the photoreceptors in the body of the transplant are of donor origin. Therefore, our data suggest that both integration and cytoplasmic transfer of donor cells with the host retina occurred in the current condition, and result in visual improvement.
After transplantation, the donor tissues survived for a long time (more than 9 months). The donor cells continued to grow and developed into mature retinal cells, including rod and cone photoreceptors, bipolar, Müller, horizontal, and amacrine cells, except ganglion cells. The human cytoplasmic marker SC121 demonstrated donor cell processes' extension into the host retina. Extensive synaptic connectivity was found inside donor tissue and between donor and host cells, indicating synaptic integration between donor and host tissue that might contribute to the observed functional improvement. Interestingly, only a limited area of the SC showed responses to light stimuli after the transplantation, which corresponded with the small area of the host retina containing the transplant. Based on these data, it may be suggested that the transplant could improve vision by replacing the photoreceptors. However, the presence of some surviving host photoreceptors found near transplants in the RCS host retina is suggestive that the transplants also induced some rescue effect by release of neurotrophic factors.
Our present investigation showed significantly improved ERG responses to scotopic and photopic stimuli in transplanted eyes compared with controls at 2 months after surgery, but the ERG responses were not detectable in all groups at later timepoints. At 2 months after surgery, ERG was still detectable in all groups. At this time point, the transplant was not mature enough to replace the function of host photoreceptors; thus, the improvement in transplanted rats seems to be a trophic effect. This is also indicated by the fact that only three of the seven rats with SC responses showed improved ERG responses. Full-field ERGs give an overall assessment of the visual response of the retina, but are not sensitive enough in most cases to detect visual function elicited from a relatively small part of the retina. 59 Because rescued photoreceptors are limited to the transplant area (up to 3.8 mm 2 after full development), corresponding with less than 10% of the retina, conventional full-field ERG may not be capable of identifying visual responses elicited by the transplants alone. In similar experiments (data not shown), we have observed that SC responses but not ERG are recordable in 1-to 3-month-old RD Rho S334ter-3 nude rats, confirming that SC recording is more sensitive than fullfield ERG (see also 36 ). Recording from a retinal wholemount in vitro would be sensitive enough to detect responses, 21 but that recording would have required sacrificing the animal. Other possibilities are multifocal or focal ERGs; however, that process requires a different setup.
OKT data showed a significant decrease after 4 months after surgery (age 6 months) in sham and age-matched control RCS rats, whereas the transplants significantly improved visual acuity from 2 to 4 months after surgery to the end of study period (up to 6 months after surgery). Compared with our previous work in RD Rho S334ter-3 nude rats, 22 the vision improvement owing to retinal organoid transplants was more prominent in the current RCS rat study. This finding may be due to an additional rescue effect on host photoreceptors in combination with photoreceptor replacement. In contrast, almost all host photoreceptors were lost at the time of transplantation in the RD Rho S334ter-3 nude rats.
In summary, our findings demonstrate that retinal progenitor sheets derived from hESC transplanted at an advanced disease stage improve the visual function in nude RCS rats by both cell replacement and cell rescue. This study provides proof of concept that retinal progenitor sheet transplantation may become a possible future treatment not only for RD but also for RPE dysfunction/degeneration diseases. RPE and retinal organoid (mainly photoreceptor progenitors) co-grafting may possibly result in better vision restoration in RD patients.