Molecular Vision 2012; 18:920-936
Received 19 July 2011 | Accepted 8 April 2012 | Published 12 April 2012
1Neurobiology-Neurodegeneration & Repair Laboratory (N-NRL), National Eye Institute, National Institutes of Health, Bethesda, MD; 2Laboratory of Molecular Biology and NIH Stem Cell Unit, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD
Correspondence to: Igor O. Nasonkin, N-NRL/NEI/NIH, 6 Center Drive, MSC0610, Building 6, Room 341, Bethesda, MD, 20892; Phone: (301) 435-6149 or (617) 388-4104; FAX: (301) 480-1769; email: firstname.lastname@example.org
Purpose: To examine the potential of NIH-maintained human embryonic stem cell (hESC) lines TE03 and UC06 to differentiate into retinal progenitor cells (hESC-RPCs) using the noggin/Dkk-1/IGF-1/FGF9 protocol. An additional goal is to examine the in vivo dynamics of maturation and retinal integration of subretinal and epiretinal (vitreous space) hESC-RPC grafts without immunosuppression.
Methods: hESCs were neuralized in vitro with noggin for 2 weeks and expanded to derive neuroepithelial cells (hESC-neural precursors, NPs). Wnt (Integration 1 and wingless) blocking morphogens Dickkopf-1 (Dkk-1) and Insulin-like growth factor 1 (IGF-1) were used to direct NPs to a rostral neural fate, and fibroblast growth factor 9 (FGF9)/fibroblast growth factor-basic (bFGF) were added to bias the differentiation of developing anterior neuroectoderm cells to neural retina (NR) rather than retinal pigment epithelium (RPE). Cells were dissociated and grafted into the subretinal and epiretinal space of young adult (4–6-week-old) mice (C57BL/6J x129/Sv mixed background). Remaining cells were replated for (i) immunocytochemical analysis and (ii) used for quantitative reverse transcription polymerase chain reaction (qRT–PCR) analysis. Mice were sacrificed 3 weeks or 3 months after grafting, and the grafts were examined by histology and immunohistochemistry for survival of hESC-RPCs, presence of mature neuronal and retinal markers, and the dynamics of in vivo maturation and integration into the host retina.
Results: At the time of grafting, hESC-RPCs exhibited immature neural/neuronal immunophenotypes represented by nestin and neuronal class III β-tubulin, with about half of the cells positive for cell proliferation marker Kiel University -raised antibody number 67 (Ki67), and no recoverin-positive (recoverin [+]) cells. The grafted cells expressed eye field markers paired box 6 (PAX6), retina and anterior neural fold homeobox (RAX), sine oculis homeobox homolog 6 (SIX6), LIM homeobox 2 (LHX2), early NR markers (Ceh-10 homeodomain containing homolog [CHX10], achaete-scute complex homolog 1 [MASH1], mouse atonal homolog 5 [MATH5], neurogenic differentiation 1 [NEUROD1]), and some retinal cell fate markers (brain-specific homeobox/POU domain transcription factor 3B [BRN3B], prospero homeobox 1 [PROX1], and recoverin). The cells in the subretinal grafts matured to predominantly recoverin [+] phenotype by 3 months and survived in a xenogenic environment without immunosuppression as long as the blood–retinal barrier was not breached by the transplantation procedure. The epiretinal grafts survived but did not express markers of mature retinal cells. Retinal integration into the retinal ganglion cell (RGC) layer and the inner nuclear layer (INL) was efficient from the epiretinal but not subretinal grafts. The subretinal grafts showed limited ability to structurally integrate into the host retina and only in cases when NR was damaged during grafting. Only limited synaptogenesis and no tumorigenicity was observed in grafts.
Conclusions: Our studies show that (i) immunosuppression is not mandatory to xenogenic graft survival in the retina, (ii) the subretinal but not the epiretinal niche can promote maturation of hESC-RPCs to photoreceptors, and (iii) the hESC-RPCs from epiretinal but not subretinal grafts can efficiently integrate into the RGC layer and INL. The latter could be of value for long-lasting neuroprotection of retina in some degenerative conditions and glaucoma. Overall, our results provide new insights into the technical aspects associated with cell-based therapy in the retina.
Photoreceptor death in retinal and macular degenerative diseases is a leading cause of inherited vision loss in developed countries. Novel therapeutic strategies have recently emerged, from mechanical to cell based, to repair neural circuits affected by photoreceptor (PR) cell loss . Trophic factor delivery to extend the life of dying PRs has been pursued in experimental animals [2-5] and in some instances in the clinic . Gene therapy approaches have been applied successfully in one type of Leber congenital amaurosis and remain viable when etiology of disease is understood and the size of a gene is not prohibitive for packaging capacity of the viral vector. Retinal implants [7,8] utilize a high-tech mechanical device placed on the retina to capture photons and transmit the electric signals to ganglion cells. Such a device is designed to replace lost PRs and has been used in the clinic with promising outcomes [9,10]. The concept of transplanting an immature retinal sheet into the subretinal space goes back to 1946 . While seemingly unattainable, the approach has shown some promising outcomes , with the ability of such grafts to survive long-term, preserve layers, establish synaptic connectivity in the host retina, and evoke activity in the visual cortex [13,14].
Transplantation of human embryonic stem cell (hESC)- or induced pluripotent stem cell (iPS)-derived retinal progenitors or retinal neurons is a relatively recent direction for retinal therapies . hESCs or iPSs can be directed to retinal fate with variable efficiency [16-21]. Compared to the retinal sheet transplantation strategy where the donor retina and its preservation constitute major limitations, the stem cell approach is based on using hESCs or iPSs, which provide an unlimited source of cells. Furthermore, there is optimism stemming from research on mouse ESCs that such human protocols might be improved by engineering the development of the whole retina in a dish .
The transplantation of retinal cells into mammalian retina produces variable outcomes and success. A greater understanding of biology and improvements in methodology are required before such protocols may be introduced into the clinic [19,23-27]. Some of the key obstacles in transplantation studies include immunological compatibility of graft and host [27,28], the outer limiting membrane (OLM) being a barrier for retinal integration [25,26,29], and glial cells/glial scar at transplantation site preventing efficient integration [19,26,30,31]. The retinal stem cell-based approach requires that transplanted cells migrate into the retinal layer(s) affected by genetic lesion, undergo terminal maturation, acquire the appropriate cell fate, and establish needed synaptic interactions. This is different from other systems, such as transplantation of insulin-producing β-cells , skin cells , or blood cells , which require the newly grafted cells to primarily acquire the proper postmitotic cell fate. Although biology of specific synaptic connectivity during retinal development is still poorly understood, promising reports indicate the feasibility of this direction, and such an approach, still largely heuristic in nature, may at least partially alleviate blindness in experimental animals with PR degeneration [29,35-37]. The progress in cell-sorting techniques [38,39] based on cell-surface antigens rather than fluorescent markers gives further hope of generating a defined cell population for cell replacement.
Transplantation is traditionally accompanied by immunosuppression, which has detrimental side effects, such as tumorigenesis [40,41], and may contribute to regenerative processes in the central nervous system (CNS) [42,43], thus masking the therapeutic effect exerted by neural graft. The retina and brain are reported to be partially immunoprivileged sites [31,44,45]. However, survival of neural grafts in subretinal space with and without immunosuppression still varies [5,27,46-50], requiring a more systematic examination. Interestingly, the transplanted neural progenitors themselves may exert an immunomodulatory effect on the host CNS . Allogeneic mouse retinal grafts can also undergo apoptosis in degenerating mouse retina .
This study was initiated to examine the survival and integration of hESC-derived neural progenitors that were transplanted into normal adult mouse retina with no immune suppression. We show that xenogenic human grafts comprising of postmitotic hESC-RPCs carrying PR markers can survive in adult mammalian retina for up to 12 weeks with no signs of deterioration. We also report that hESC-RPCs can integrate from the epiretinal grafts into host’s RGC layer and inner nuclear layer (INL) but not PR layer. The cells from the subretinal grafts, however, show limited integration into the PR layer and only when retina was damaged during transplantation. We also noted the instructing role of the subretinal but not epiretinal niche in promoting further maturation of grafted cells to PRs. Taken together, our data may help in refining protocols of hESC-derived retinal cell transplantation.
Young adult C57BL/6Jx129/SvJ wild-type mice (4–6-weeks old; the Jackson Laboratory, Bar Harbor, Maine) were used for transplantation experiments. All animals were housed and treated in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and in accordance with the standards issued by the NIH animal facilities, NIH approval #ASP 08–610.
hESCs were maintained and supplied by the NIH Stem Cell Unit. TE03 (Technion-Israel Institute of Technology, Haifa, Israel) and UC06 (University of California San Francisco, San Francisco, CA) were separately differentiated and used for grafting.
Neurobasal medium with B27/N2 supplementation (all from Invitrogen, Grand Island, NY) and plastic Petri dishes (Corning Inc., Corning, NY) covered with 0.1% gelatin (Sigma-Aldrich, St. Louis, MO) were used for cell culture during differentiation. Noggin, Dkk-1, IGF-1, FGF9 were from R&D Systems, Minneapolis, MN, and bFGF was from Sigma-Aldrich, St. Louis, MO.
The Microm HM550 cryostat (Thermo Scientific, Rockville, MD) was used to produce 16-μm serial sections of mouse eyes. Microscope slides were purchased from Fisher Scientific (Pittsburg, PA). Glass coverslips were purchased from Brain Research Laboratories (Newton, MA).
hESC lines were grown on a mouse embryonic fibroblast feeder layer (CF1 strain, The Jackson Laboratory, Bar Harbor, Maine) according to the protocols described (NIH research). The colonies (day 4–5 after passage) were dislodged and placed on gelatin-coated Petri dishes and cultured at high density in Neurobasal medium supplemented with 1x B27 without retinoic acid, 1x N2, 1x penicillin–streptomycin antibiotic mix, L-glutamine (1%), Minimal Essential Medium nonessential amino acid solution (1%; all from Invitrogen), BSA fraction V (0.1%), β-mercaptoethanol (0.1 mM; both from Sigma-Aldrich), and recombinant noggin (100 ng/ml; R&D Systems) but without bFGF to induce neuralization . At 14 days of differentiation, the cultures were supplemented with bFGF (10 ng/ml) , and 50% of the media was renewed every other day. At day 28, neural rosettes were excised mechanically and replated as large clusters of neuroepithelial cells (hESC-NPs)  on gelatin/laminin-coated plates and cultured at high density (95%–100% confluency). Wnt-blocking morphogens Dkk-1 and IGF-1 (10 ng/ml each) were applied for 1 week immediately after replating to anteriorize cells to a rostral neural fate . Cells were then cultured further with the addition of FGF9 and bFGF (both at 10 ng/ml) as well as noggin until grafting to bias cells to an NR rather than an RPE cell fate [13-15]. At day 50, hESC-RPCs were dissociated with cell dissociation buffer (Invitrogen) and trypsin-like enzyme (Invitrogen) and suspended in Neurobasal medium at ~50×103 cells/μl for transplantation.
Immediately after transplantation, some of the remaining cells were replated to evaluate the viability and differentiated state of transplanted cells. Antibodies against human nuclei (HNu), human nestin, recoverin, Kiel University-raised antibody number 67 (Ki67), doublecortin (DCX), and neuronal class III β-tubulin (Tuj1) were used for this analysis. Immunohistochemistry was done as described previously . Total RNA was prepared from (i) undifferentiated hESC cells and (ii) cells at day 50 (transplantation) using RNeasy Mini kit (Qiagen, Valencia, CA) according to the manufacturer’s protocol; briefly cells were first lysed and then homogenized, the lysates were then loaded onto the RNeasy silica Mini spin columns, and after RNA was bound to silica gel, all contaminants were washed away, and pure concentrated RNA was eluted in water. RNA was then converted to cDNA with Superscript II (Invitrogen, Carlsbad, CA), and used for quantitative reverse transcription polymerase chain reaction (qRT–PCR) analysis on an ABI 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA) with SYBR Green Master Mix (Applied Biosystems). Oligonucleotide primers (Table 1) specific to (i) pluripotent hESCs (octamer-binding transcription factor 3/4 [OCT3/4], NANOG, sex-determining region Y gene-related high mobility group box 2 [SOX2]), (ii) markers of the anterior neuroectoderm (forkhead box protein G1 [FOXG1], [SIX3], sine oculis homeobox homolog 6 (Drosophila) [SIX6], [LHX2), (iii) markers of the eye field (PAX6, SIX3, SIX6, RAX [RX]), (iv) retinal progenitors (CHX10, MASH1, NEUROD1), RPE (microphthalmia-associated transcription factor, MITF), PRs (recoverin, cone-rod homeobox gene [CRX], neural retina-specific leucine zipper, [NRL]), RGCs (MATH5, BRN3B, insulin gene enhancer protein [ISL1]), and horizontal neurons (PROX1) were used for qRT–PCR analysis, which was performed in triplicate at both time points (“undifferentiated hESCs” and “day 50, grafting”). The qRT–PCR data were analyzed using the ΔΔCt method (as described in Livak KJ, Schmittgen TD. Methods 2001 paper), with geometric averaging of β-actin (ACTB) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the endogenous controls (outlined in Vandesompele J. et al., Genome Biol 2002). Briefly, the standard approach of DNA quantification by real-time qRT–PCR is based on plotting measured fluorescence (in our case SYBR Green I cyanine dye incorporated into DNA during PCR) against the number of PCR cycles on a logarithmic scale. During the exponential phase of qRT–PCR, when reagents are not limited, the amount of cDNA (target) is assumed to be doubling every cycle. First, ΔCt analysis is done, which takes the Ct (cycle number, or cycle threshold) value for the gene of interest, divided by Ct of a housekeeping gene in the same sample, at the point when the signal just becomes detectable above the background and the amplification is in exponential phase; the log2 difference is then generated. The more abundant the mRNA for gene X is, the quicker this point is reached, thus giving earlier Ct values and allowing to quantitatively evaluate the expression of gene X. In our case Ct values were generated automatically using SDS2.3 software (Life Technologies, Carlsbad, CA). Second, ΔΔCt analysis is done, where ΔΔCt equals ΔCt [sample] (Ct value for “day 50” sample normalized to the endogenous housekeeping gene) minus ΔCt [reference] (Ct value for “undifferentiated hESCs also normalized to the endogenous housekeeping gene). Collectively, ΔΔCt method is a normalization procedure, which allows comparison of gene expression levels in different RNA samples by taking into account the differences in quality and total amount of RNA in samples. Expression levels at day 50 (grafting) were presented as log2 values of the expression level differences compared to that found in undifferentiated hESCs. The analysis was done in technical triplicates, with one biological replicate.
Transplantation equipment included a nano-injector (World Precision Instruments, Sarasota, FL), pulled glass micropipettes (Drummond Scientific Company, Broomall, PA), 29G insulin syringe (Becton Dickinson & CO, Franklin Lakes, NJ), and a dissection microscope (SZ61; Olympus, Center Valley, PA). General anesthesia was used during transplantation and included a mixture of ketamine (87 mg/kg) and xylazine (10 mg/kg) administered intraperitoneally at 0.1 ml/g bodyweight. Once properly sedated, the animals were placed under the dissection microscope; the eyes were covered with a thin film of mineral oil to prevent drying, and a small incision was made in the cornea using a sharp insulin syringe. Using the nano-injector, the blunt-ended tip of a micropipette filled with cells was guided through the incision in the cornea and advanced trans-retinally until it met resistance due to rigid choroid/scleral tissue. The needle was then slightly withdrawn, and as it was very slowly being pulled out to create space for grafted cells, hESC-RPC suspension (≤1.5 μl, total of about 50,000 cells) was slowly deposited into the subretinal space. This transplantation methodology, due to a mouse NR being so thin, inevitably left about 20% of grafted cells epiretinally, adjacent to the RGC layer. Both eyes were injected for each animal. For the TE03 hESC line, 26 subretinal grafts (13 animals) were generated; six animals (12 eyes) were analyzed at 3 weeks and seven animals (14 eyes) were analyzed at 3 months after grafting. For the UC06 hESC line, 14 subretinal grafts (seven animals) were generated; three animals (six eyes) were analyzed at 3 weeks and four animals (eight eyes) were analyzed at 3 months after grafting.
Enucleation of the eye was done with a fine-point microdissection forceps and fine-point microdissection scissors (Electron Microscopy Sciences, Ft. Washington, PA). Eyes were fixed in paraformaldehyde (4%, Sigma-Aldrich) for 5 min, rinsed with 1x PBS (KD Medical, Columbia, MD; NaCl 90 g/l; Na2HPO4, anhydrous, 7.10 g/l; KH2PO4, 2.3g/l; UltraPure Water), cryoprotected in 20% and then 30% sucrose, and then snap frozen in OCT embedding material (Tissue-Tek, Torrance, CA) and serially sectioned at 10 μm.
For cresyl violet (CV) staining, serial sections were sequentially washed with PBS and deionized water, stained with CV for about 1 min, dehydrated with increasing concentrations of ethanol, mounted with DPX solution (Sigma-Aldrich), and examined with a light microscope for the presence of hESC-RPC grafts.
Immunohistochemistry (IHC) staining was performed using primary antibodies (Table 2) for human nuclei (HNu), human nestin, DCX, Tuj1, recoverin, rhodopsin, glial fibrillary acidic protein (GFAP), human synaptophysin, Ki67, and ionized calcium-binding adaptor molecule-1 (Iba-I).
Eye sections demonstrating the presence of grafted cells by CV (Figure 1H,I) were sequentially incubated with 0.1% Triton X-100/PBS (PBS-T) at room temperature for 30 min, followed by 1 h incubation in blocking solution (5% pre-immune serum and 0.1% PBS-T) at room temperature, and then incubated with primary antibodies diluted in blocking solution at 4 °C overnight. HNu antibody was used to identify grafted human cells [18,21]. Ki67 antibody was used to examine the mitotic activity of cells . Human nestin antibody was used to identify multipotential human neural precursor cells . Tuj1 antibody was used to identify neuronal cells. DCX antibody was used as a marker for neuroblasts and young neurons . Recoverin and rhodopsin antibodies were used as PR markers . Human synaptophysin antibody was used to identify the presynaptic part of human boutons established by maturing hESC-RPCs . GFAP antibody was used as a marker for activated Müller glia cells . Iba-1 was used to identify microglia cells . DAPI staining was used to identify nuclei of any cell type. Following overnight incubation with primary antibodies, sections were washed three times with 0.1% PBS-T and then incubated with the corresponding secondary antibodies (Alexa Fluor 594 goat antimouse, Alexa Fluor 488 goat antirabbit) at room temperature for 45 min. The slides were washed three times with 0.1% PBS-T solution, incubated with 4', 6-diamidino-2-phenylindole (DAPI) solution (1 μg/ml) for 10 min, and then washed again with 0.1% PBS-T solution. For negative controls, slides were treated similarly except that primary antibodies were omitted. The specimens were mounted with ProLong Gold Antifade medium (Invitrogen) and examined using an Olympus (Center Valley, PA) epifluorescent microscope IX51 with a Spot (Sterling Heights, MI) CCD Camera RT3 and Leica (Buffalo Grove, IL) SP2 confocal microscope. For high-resolution confocal microscopy, z-series of images (with a z-step of 0.2 μm, 15–20 optical sections) were collected using a 63x1.32 numerical aperture oil immersion objective (Leica SP2). Consecutive optical planes (z series) of selected fields were analyzed to evaluate distribution and co-localization of fluorescent signals, with subsequent virtual resectioning at the x and y axes.
Data on human RPC grafts at 3 weeks and 3 months were obtained from serial sections and evaluated by the StatView program (Abacus Corporation, Baltimore, MD). The difference in Tuj1 and recoverin expression between TE03 and UC06 grafts was minimal at 3 weeks and 3 months. Thus, results were grouped for two hESC lines for each time point and plotted as a mean of the percentages of HNu – positive ([HNu [+]) human cells carrying Tuj1 or recoverin in grafts, with corresponding standard error of the mean (SEM). Comparison of the statistical significance between expression of Tuj1 and recoverin in the subretinal space versus the epiretinal (vitreous) space was calculated with an unpaired Student t test (with p<0.05 considered statistically significant) after converting the percentage values to arc sin values .
We used noggin in the absence of bFGF mitogen for 2 weeks to neuralize hESCs, as described  (Figure 1A). The detailed protocol is outlined in Methods, in vitro differentiation section of this paper. At day 28, the plates with differentiating hESC colonies were 95% confluent, and about one-third of each plate area consisted of neural rosettes (Figure 1B). These rosettes were isolated mechanically as described  (briefly, excised with a fine fire-polished and sealed pulled glass pipette), replated on gelatin/laminin-coated plates, and induced to a rostral neural tube cell fate by Wnt blocking morphogen Dkk-1 and IGF-1 for 1 week. Following retinal induction, the cells (hESC-RPCs) were cultured with FGF9 and bFGF until transplantation. Immediately after transplantation, the remaining cells were replated and evaluated by immunocytochemistry (ICC) the following day (Figure 1F).
We noted an efficient neuralization of hESCs by day 28 and downregulation of pluripotency markers and upregulation of neural and retinal markers by IHC and/or qRT–PCR by day 50 of our differentiation protocol (Figure 1A). About half (52.7%) of the cells were mitotically active as judged by Ki67 positivity (not shown). The majority (64.2%) were human nestin - positive (human nestin [+]; Figure 1D,F), and 39.8% were Tuj1 [+] (Figure 1E,F; averaged for both hESC lines). No recoverin [+] or rhodopsin [+] cells were detected at this stage by ICC. Less than 1% of cells were human nestin [-] and Tuj1 [-]. Pluripotent hESCs immediately before the differentiation protocol and hESC-RPCs at day 50 (grafting) were used for total RNA preparation and qRT–PCR. Pluripotency markers NANOG, OCT3/4, and SOX2 were downregulated at day 50, while the eye field and NR progenitor markers, such as RX, SIX6, PAX6, CHX10, NEUROD1, early PR marker recoverin, and pan-RGC marker BRN3B, showed substantial upregulation. The RPE-specific isoform of MITF showed only a slight upregulation (Figure 1G).
Serial CV staining of grafts at 3 weeks after transplantation showed surviving transplanted cells clustered around the transplantation site (Figure 1H). At 3 months more cells were found spreading within the subretinal space (Figure 1I, also see Figure 2G). Successful subretinal grafts (Figure 1H,I) were observed in about 25% of transplanted eyes (n=3–4 grafts/hESC line). Many surviving grafts (n=14 examined) were detected in the epiretinal area. In subretinal grafts, there was a distinct border between the graft and the outer nuclear layer (ONL) separated by the OLM (Figure 2A,B,G, arrowheads). However, in cases where the host retina had been damaged during transplantation, we observed some HNu [+] Tuj1 [+] cells, and by 3 months we also observed HNu [+] recoverin [+] neurons integrating into the ONL (Figure 2F,J). The survival of xenogenic human grafts was best when the host RPE/choroid was not damaged, as evaluated by CV staining (as in Figure 1H,I). Sections that displayed damage to the RPE/choroid (Figure 3A-C, arrows) displayed few or no surviving HNu [+] cells and strong GFAP activation  and microglial cell accumulation  in and around the grafted area.
Some grafts demonstrated slower cell degradation evident by the release of human nuclear proteins into the subretinal space, weak nuclei HNu antibody staining, and HNu [+] immunoreactivity outside the grafted cells. Such grafts also had strong activation of GFAP around, but not inside, the grafted area (Figure 4). Iba-1 immunoreactivity was prominent in grafts that did not survive (Figure 5).
Grouped data for both hESC lines showed a reduction of immature neuronal marker Tuj1 in 3-month subretinal grafts (57.2% Tuj1 [+] hESC-RPCs [n=6], Figure 2E-G) compared to that at 3 weeks (75.7% Tuj1 [+] hESC-RPCs [n=7], Figure 2B, also see the plotted graph in Figure 2D). Further maturation of subretinally located hESC-RPCs was evident as only 1.3% hESC-RPCs were recoverin [+] at 3 weeks (n=7; Figure 2C and Figure 6) whereas at 3 months 67.5% were recoverin [+] (n=6; Figure 2I,J, also see the plotted graph in Figure 7H). Approximately 15% of grafted cells were mitotically active at 3 weeks (n=7), but only a few HNu [+] cells were stained with proliferation marker Ki67 by 3 months (less than 0.01%, data not shown). No tumor formation was observed in grafts. Human-specific synaptophysin [+] sparse human boutons resembling boutons en passant were found both on Tuj1 [+] axons emanating from the grafted HNu [+] cells (Figure 2K, inset) and on host PRs. In a few grafts (at 3 months only), cells were positive for rhodopsin (Figure 2L).
Substantial differences were found in maturation of hESC-RPCs in subretinal versus epiretinal grafts (Figure 7). Subretinal grafts demonstrated a little decrease of Tuj1 immunostaining (from about 75.7%, [n=7] at 3 weeks to 57.2% [n=6] at 3 months). However, the difference was not statistically significant at p<0.05 when combined for both TE03 and UC06 cells. Three-week-old epiretinal grafts had less than 8% of Tuj1 [+] cells. By 3 months, only about 1% of cells in the epiretinal grafts were HNu [+] Tuj1 [+], and these were mostly in small clusters (Figure 7A). However, cells from epiretinal but not subretinal grafts were able to easily integrate into the host’s RGC and INL layers even when host retina was not damaged (Figure 2H, Figure 7A,B,E). Cells from the subretinal grafts were detected in the ONL (and rarely in the INL) only when the host retina was damaged (Figure 2F and Figure 7C,D,F). IHC with recoverin and HNu antibodies demonstrated a sharp increase in the number of HNu [+] Rec [+] cells in subretinal grafts from about 1% at 3 weeks [n=7] to about 67.5% at 3 months [n=6] (combined for both TE03 and UC06). Cells in the epiretinal grafts displayed a low presence of recoverin [+] human cells at 3 weeks (less than 2%), and no such cells were present by 3 months. Note the complete absence of HNu [+] recoverin [+] cells in the RGC/INL and clusters of HNu [+] recoverin [+] cells in the subretinal space (Figure 7E). qRT–PCR analysis corroborates this data and shows that early progenitor/PR markers (RCVRN [recoverin], MASH1 and NEUROD1) are upregulated at the time of grafting. Evidently, such cells could undergo further maturation in the subretinal but not epiretinal niche.
Stem cell-mediated cell replacement therapy for retina has advanced rapidly in the past several years and has the possibility of becoming a treatment method for some retinal degeneration (RD) conditions [35,36,56]. Apart from reproducibility of the data from different hESC lines, many issues require further evaluation; these include OLM barrier [23,26,57,58], immunorejection of graft by a host [27,59] (excellent discussion in ), and the formation of glial scar containing extracellular matrix and Müller glia endfeet, preventing further cell integration [58,61]. In addition, the host retinal niche and preservation of retinal architecture of the recipient seem to contribute to the complexity of any graft’s survival and functional integration .
We considered it important to investigate two separate recurrent questions frequently reported in retinal cell transplantation papers: the survival of the retinal grafts in a non-immunocompatible recipient and the population of retinal layers with grafted hESC-RPCs. We approached this by first selecting normal (non-RD) young adult mouse eyes as recipients of hESC-RPC grafts to avoid the influence of a degenerating and rapidly changing neural niche on the survival of the graft [24,62-64]. Such reports, although debated, suggest that an injured or degenerative neural environment might adversely affect the survival of human stem cell-derived grafts. We also chose not to apply immunosuppression, as retina is considered an immunoprivileged site due to the blood-retinal barrier (BRB). In addition, survival of xenogenic human grafts in retina has been reported . To account for the expected differences in graft survival, we correlated the survival of transplanted cells with the overall integrity of the RPE/choroid tissue, which comprises the BRB . Lastly, we compared the dynamics of cell integration into the host’s retina from the subretinal and epiretinal space to circumvent the OLM barrier. The advantage of such an approach is that in any given transplantation case the grafting niche remains the only difference, which may be informative for data interpretation. Overall, we find that both hESC lines UC06 and TE03 (cultured for 50+ passages) can differentiate to mature retinal phenotypes using the noggin/Dkk-1/IGF-1/bFGF/FGF9 protocol. After 3 months in a subretinal environment, transplanted cells demonstrated the ability to acquire mature PR-specific immunophenotypes (e.g., recoverin and rhodopsin staining) and no tumorigenicity was detected in all examined grafts. Importantly, we observed that the survival of xenogenic grafts with no immunosuppression correlates with the integrity of the RPE/choroid structure (BRB) but not the NR. Whenever the histology showed no damage to the RPE/choroid, the graft survived and thrived for up to 12 weeks with no immunosuppression and no signs of deterioration. The damage to the host’s NR alone and/or strong activation of GFAP by reactive Müller glia of the host (Figure 2 and Figure 3) did not affect graft survival. In cases when the RPE/choroid showed signs of substantial damage by a blunt needle guided by the nano-injector, xenogenic hESC-RPC grafts did not survive, displayed lysed human cells, were filled with host’s Iba-1 [+] microglia, and were GFAP [+]. Therefore, we conclude that the xenogenic grafts may survive and thrive in the subretinal space when the BRB is intact. Consequently, systemic immunosuppression may not be necessary for graft survival when nonautologous PR progenitors are transplanted into retina.
Our results showed limited integration of subretinally grafted hESC-RPCs into the host’s retina and only in cases when the ONL had some structural damage. However, no HNu [+] cells (except one case) were found in INL or RGC layers, likely due to intact OLM present in the wild-type retina, consistent with other reports [19,23,57]. In contrast, integration of hESC-RPCs into the host INL and especially the RGC layers was efficient from the epiretinal grafts, irrespective of whether the retina had any structural damage. Some HNu [+] cells were co-localized with host RGCs and also expressed RGC marker Tuj1 . qRT–PCR analysis of cells at the time of grafting showed that hESC-RPCs upregulated RGC markers (such as MATH5 and BRN3B) and the horizontal neuronal marker (e.g., PROX1). Thus, hESCs could potentially generate RGCs and horizontal cells.
A limited number of human synaptophysin [+] boutons en passant could be detected in the INL and RGC layer, indicating initiation of synaptogenesis. Due to the lack of a barrier for cell penetration from the epiretinal side, such grafts may be used for long-term trophic support of degenerating retina , including the trans-synaptic transport of neurotrophins , as well as for potential RGC and INL cell-replacement strategies. Although the migration of cells into the ONL from subretinal grafts was clearly impeded, we suggest that in RD conditions this migration could be helped by a porous OLM  as well as guided by tropism of grafted progenitor cells to the sites affected by degeneration [5,70]. It is also possible that the maturation state of hESC-RPCs affects integration as some studies have reported integration of postmitotic progenitors and even mature PRs into normal retina [35,71]. Although immunosuppression may not be crucial for xenogenic graft survival, it may be beneficial for retinal integration in a clinical setting when nonautologous (i.e., stem cell-bank-derived) hESC-RPCs are transplanted subretinally. For example, removal of glial barrier in GFAP−/− and vimentin −/− mice provided a permissive environment for retinal integration of transplanted neurons . Such a glial barrier, induced by the host, may be partially alleviated by immunosuppression and chondroitinase ABC .
We also noted that the subretinal but not the epiretinal niche can provide further cues for hESC-RPC maturation to PRs, resulting in a sharp gain of mature PR marker recoverin, a neuronal calcium-binding protein found almost exclusively in PRs . However, the epiretinal grafts demonstrated no cell maturation and retained the original, mostly nestin [+] immunophenotype. This is consistent with a previous observation  indicating that paracrine morphogens in the host retina and/or RPE can promote further maturation of hESC-RPCs.
As the cell population at the time of grafting showed almost 100% neuralization with noggin and over 67% of cells in grafts were positive for PR marker recoverin by 3 months, the overall efficiency of PR-fate specification from both hESC lines appears to be comparable to that reported . Only a small number of cells neuralized by noggin may be expected to remain non-neural after 4 weeks in culture . Since only neural rosettes were collected for further induction with Dkk-1 and IGF-1, the number of non-neural cells in such cultures should be minimal by day 50 (grafting), thus reducing tumorigenicity. There are several important distinctions resulting in faster derivation of recoverin/rhodopsin immunophenotypes in cultures reported earlier . These differences potentially originate from somewhat longer exposure to Dkk-1 and IGF-1, culturing on Matrigel rather than defined gelatin/laminin coating, and likely different culturing densities, which may profoundly influence the dynamics of neuronal cell fate acquisition and maturation [72-74]. We also chose to maintain both bFGF and FGF9 in neural cultures, which earlier received anteriorizing Dkk-1 and IGF-1 induction, as both bFGF  and FGF9  reportedly bias early retinal cells to an NR rather than an RPE cell fate.
The influence of FGF9 on the NR versus RPE cell fate is especially interesting as it is unexplored in retinal differentiation protocols. Fgf9 is expressed in the distal part of the developing optic vesicle in the mouse that is destined to become a NR and was reported to induce activation of Ras by receptor tyrosine kinase in early optic neuroepithelium . Ectopic expression of Fgf9 in the proximal region of the optic vesicle destined to become RPE promotes conversion of the RPE cell fate to an NR cell fate in early retinal development by suppressing the expression of RPE marker Mitf and induction of NR-specific markers Rx, Chx10, and Atoh7 (Math5) . As a result of such ectopic expression, a duplicated NR has been produced. Notably, the original NR and duplicated NR differentiated and laminated symmetrically but with a mirror-image polarity. The same study delineated the likely downstream target of FGF9 signaling, promoting the acquisition of the NR cell fate: the RAS-mediated RAF-MEK-mitogen-activated protein kinase pathway. Specifically, the transient expression of a constitutively active human Ras oncogene by tyrosinase-related protein2 (TRP2) promoter in mouse transgenic embryos also converted the developing RPE to a second NR. Because the retinal development in both types of transgenic mice was overall normal, it was concluded by Zhao et al.  that FGF9 signaling was needed to define the boundary between the retina and the RPE. Collectively, transient FGF9 signaling, likely through RAS signaling, was sufficient to promote NR cell fate at the expense of RPE, which was one of the goals of our differentiation protocol. Other factors, such as ectopic Pax6 expression or null mutation of Chx10, are known to shift the cell fate in the developing retina from RPE to NR and vise versa, respectively. However, such signaling requires genetic manipulations in hESCs compared to easy delivery of FGF9 (and bFGF) morphogens during the differentiation protocol.
FGF9 belongs to a different subfamily of FGF factors compared to bFGF (FGF2) and can inhibit the canonical Wnt pathway via upregulation of Dkk-1, a canonical Wnt antagonist, and regulate the transcription of Hedgehog targets patched homolog 1 (Ptch1) and glioma-associated zinc finger 1 (Gli1) independently of the Hedgehog ligand . Both effects may promote NR differentiation [16,78]. Additional investigations are necessary to clearly delineate the role of FGF9 in NR differentiation.
In summary, we show that (i) xenogenic human hESC-RPC grafts from both hESC lines survive in the subretinal space without immunosuppression when little structural damage occurs to the RPE/choroid; (ii) gradual maturation of hESC-RPCs in subretinal but not epiretinal grafts occurs over a period of 3 months, indicating that the subretinal but not the epiretinal (vitreous) niche provides further differentiation cues for retinal cell fate maturation; (iii) substantial migration and integration of hESC-RPCs into the RGC and INL layers from epiretinal grafts occurs, even when the host retina lacked signs of damage. Our data provide new insights into differentiation and integration of grafted cells and may advance the protocols for cell therapies of retinal degenerative diseases.
This research was supported by intramural programs of the National Eye Institute (NEI) and National Institute of Neurologic Disorders and Stroke (NINDS). We are grateful to Dr. Ginger Tansey (NEI) for help with experimental animals, Dr. Robert Fariss for help with confocal microscopy and Dr. Jacob Nellissery for help with selecting the antibodies. We thank the members of N-NRL laboratory for discussion of our results, and especially Dr. James Friedman for carefully reading the manuscript.