|Home | About | Journals | Submit | Contact Us | Français|
Retinal degenerative diseases are a major cause of untreatable blindness. Stem cell therapy to replace lost photoreceptors represents a feasible future treatment. We previously demonstrated that postmitotic photoreceptor precursors expressing an NrlGFP transgene integrate into the diseased retina and restore some light sensitivity. As genetic modification of precursor cells derived from stem cell cultures is not desirable for therapy, we have tested cell selection strategies using fluorochrome-conjugated antibodies recognizing cell surface antigens to sort photoreceptor precursors. Microarray analysis of postnatal NrlGFP-expressing precursors identified four candidate genes encoding cell surface antigens (Nt5e, Prom1, Podxl, and Cd24a). To test the feasibility of using donor cells isolated using cell surface markers for retinal therapy, cells selected from developing retinae by fluorescence-activated cell sorting based on Cd24a expression (using CD24 antibody) and/or Nt5e expression (using CD73 antibody) were transplanted into the wild-type or Crb1rd8/rd8 or Prph2rd2/rd2 mouse eye. The CD73/CD24-sorted cells migrated into the outer nuclear layer, acquired the morphology of mature photoreceptors and expressed outer segment markers. They showed an 18-fold higher integration efficiency than that of unsorted cells and 2.3-fold higher than cells sorted based on a single genetic marker, NrlGFP, expression. These proof-of-principle studies show that transplantation competent photoreceptor precursor cells can be efficiently isolated from a heterogeneous mix of cells using cell surface antigens without loss of viability for the purpose of retinal stem cell therapy. Refinement of the selection of donor photoreceptor precursor cells can increase the number of integrated photoreceptor cells, which is a prerequisite for the restoration of sight.
Retinal degenerations are a heterogeneous group of eye diseases that result in the permanent loss of vision and affect millions of individuals worldwide . Although the molecular mechanisms underlying these conditions vary, they share a common endpoint: the irreversible death of the photoreceptor cells. No effective treatment is currently available to restore lost photoreceptors and visual function and most therapeutic interventions can at best only slow down the disease progression. However, stem cell-based therapies aiming to replace damaged photoreceptor cells thereby restoring visual function, or for paracrine support of remaining cells, represent a newly emerging therapeutic strategy with significant potential clinical impact [2–5].
A wide range of different cell and tissue types have previously been considered and tested for restoring visual function by transplantation studies with varying levels of success [6–10]. We have demonstrated that rod photoreceptor precursor cells can be successfully transplanted into the normal and diseased mouse retina [10–12]. Photoreceptor precursor cells isolated from developing retinae at a time concurrent with the peak of rod genesis can integrate into the adult host outer nuclear layer (ONL) after transplantation, even though this is a nonneurogenic environment. Furthermore, the transplanted precursor cells complete their developmental program and mature into functional rod photoreceptors in their new environment. The discovery that transplanted photoreceptor cells can repopulate the retina offers a potential approach for human retinal repair [10, 12, 13]. In the rhodopsin knockout mouse, a model of retinitis pigmentosa, rod precursor cell transplantation resulted in restoration of a light response . Importantly, integrated photoreceptor cells were derived exclusively from postmitotic photoreceptor precursors and not from stem cells or other proliferating retinal progenitors [10, 13]. Transplantation of a mixed population of retinal cells generated in vitro from differentiated human embryonic stem cells (hESCs) also gave rise to new rod photoreceptors after transplantation into the mouse retina providing further evidence that the adult retina presents a feasible target for stem cell-based therapy .
Transplantation competent postmitotic precursor cells for human therapy would have to be derived from fetal tissue in the second trimester, which is ethically problematic and logistically difficult. Alternative promising sources of donor cells are differentiated hESCs or induced pluripotent stem cells (iPSCs). Several groups have recently demonstrated that retinal cells, including photoreceptors, can be derived from stem cell cultures after application of differentiation protocols mimicking retinal development in vivo [14–16]. Retinal progenitor cells isolated from the human retina have also been expanded and differentiated in vitro [17, 18]. However, all these cultures are heterogeneous containing a range of different cell types at different developmental stages and may include proliferating cells that could lead to the formation of tumors. Moreover, these cultures do not show optimal levels of integration after transplantation. Therefore, one critical long-term challenge facing the development of retinal stem cell therapy is the ability to obtain unmodified homogeneous populations of correctly staged donor cells from in vitro sources. While genetic tagging of donor cells using photoreceptor specific transgenes for isolation/enrichment is possible, in a research setting, it is undesirable for human therapy. The method of choice for cell isolation for transplantation in other systems is cell sorting using combinations of conjugated monoclonal antibodies, which bind specifically to surface antigens on the cell type of interest. This strategy permits the isolation of discrete subsets of cells from complex cultures without damaging the desired cells. This approach has already been successfully deployed in immunology and cancer biology [19–21], but it is untested for the isolation of retinal cells that could be effective for transplantation to repair damaged retina.
Retinal cell populations, photoreceptor precursors in particular, remain poorly characterized in terms of cell surface marker expression profiles. Here, we use microarray technology and transcriptome analysis of rod precursor cells to identify retinal cell surface markers that effectively isolate transplantation competent photoreceptor precursor cells. Using a combination of conjugated monoclonal antibodies directed against such cell surface markers, we were able to isolate and purify photoreceptor precursor cells from a mixed population of retinal cells, which were then able to efficiently integrate and differentiate in large numbers into the healthy and degenerating host retina. This is the first report of effective transplantation of photoreceptor cells isolated by means of nongenetic tagging, a prerequisite for the development of stem cell therapy for retinal disease.
All mice were kept in University College London animal facilities and experiments were conducted in accordance with the Animals (Scientific Procedures) Act 1986 and the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. The NrlGFP transgenic mouse line and the CrxGFP line express green fluorescent protein (GFP) in developing, and mature photoreceptors and the CbaGFP transgenic line ubiquitously express GFP [10, 22, 23]. Details of matings and dissections are provided as detailed in the Supporting Information Methods.
Microdissected neural retinae from NrlGFP animals at postnatal day 4(P4) were dissociated using papain according to manufacturer’s recommendation (PSD kit; Worthington, Lakewood, NJ, www.worthington-biochem.com). NrlGFP+ rod photoreceptors were separated from GFP− retinal cells using a Beckman Coulter MoFloTM XPD (Beckman Coulter, High Wycombe, United Kingdom, www.beckmancoulter.com) cell sorter collecting both populations in 3 ml collection medium (50% fetal bovine serum, 50% Dulbecco’s Modified Eagle Medium/F12 Glutamax). Three independent biological replicates were collected for GFP+ and GFP− cells, each replicate from the retinae of two P4 NrlGFP animals. Total RNA was extracted from cell pellets using the RNeasy Mini kit (Qiagen, Crawley, United Kingdom, www.qiagen.com). RNA yield was quantified on a nanodrop spectophotometer (Thermos Scientific, Waltham, MA, www.thermofisher.com). Samples quality was verified using the Agilent Bioanalyzer (Agilent Technology Inc., Edinburgh, UK, www.home.agilent.com). Approximately 50–100 ng of total RNA from sorted retinal cells was used for target generation and subsequent hybridization to a (Affymetrix, Santa Clara, CA, www.affymetrix.com) GeneChip Mouse Gene ST 1.0. Data from a previously performed Affymetrix microarray analysis  were also used from the GEO data base and analyzed to identify genes with changing expression profiles in photoreceptor precursors. This dataset analyzed RNA isolated from NrlGFP expressing rod cells from the wild-type retina and from the Nrl−/− retina at several developmental time points (embryonic day 16 [E16], P2, P6, P10, and 4 weeks). Algorithms used for analysis of microarray data are detailed in the Supporting Information Methods.
It is described in Supporting Information Methods.
Neural retinae of NrlGFP eyes were isolated by microdissection and dissociated into a single cell suspension using papain treatment, according to the manufacturer’s instructions (Worthington Biochemical, Lorne Laboratories, U.K.). Developing eyes from a range of developmental stages (E15.5, E17.5, P4, and P8) were dissected and dissociated. After dissociation cells were resuspended in blocking solution containing 1% BSA (w/v) in phosphate-buffered saline and then incubated for 1 hour on ice. Subsequently, the conjugated primary antibody or IgG isotype control was added and cells were incubated on ice for additional 45 minutes. The following conjugated monoclonal antibodies were used for fluorescence-activated cell sorting (FACS) analysis: eFluor 450 conjugated CD73 (clone TY/11.8, eBioscience, Hatfield, United Kingdom, http://eu.ebioscience.com); Phycoerythrin-Cy7 conjugated CD24 (clone M1/69, BD Bioscience, Oxford, United Kindgom, www.bdbiosciences.com); Allophycoerythrin (APC)-conjugated Prominin-1 (CD133, clone 13A4, eBioscience); APC-conjugated Podocalyxin (RD systems, Abingdon, United Kingdom, www.rndsystems.com); Podocalyxin was tested but did not label postnatal retinal cells. Antibody specificity for these monoclonal antibodies including Western blot analysis has been previously demonstrated [24–29]. The same clones were used for flow cytometry and immunohistochemistry analyses. All FACS antibodies were used at manufacturer’s recommendations. The cells were then centrifuged at 200g for 5 minutes at 4°C and resuspended in PBS and kept on ice until analysis. FACS analysis was carried out on a BD Bioscience LSR II flowcytometer using FlowJo software (Tree Star, Ashland, OR, www.treestar.com). FACS gates were determined according to background staining displayed by IgG isotype controls. At least 10,000 events of live cells were analyzed.
Transplantations were carried out using postnatal cells from the NrlGFP mouse line or CbaGFP mouse line isolated as described above. For cell transplantation via cell surface markers, cells were blocked for 1 hour on ice in 1% bovine serum albumin (BSA) in PBS. After blocking the conjugated monoclonal antibodies or respective isotype controls for CD73 (Alexafluor647, Biolegend, Cambridge, United Kingdom, www.biolegend.com) alone or CD24 (PECy7, BD Bio-science) and CD73 in combination were added according to manufacturer’s recommendations. Stained retinal cells were isolated by FAC-sorting (FACS, Beckman Coulter MoFloTM XPD). FACS-gates were determined in each experiment using samples stained with either single or combined isotype controls. Non-antibody-labelled cells used for transplantation were processed identically to labeled cells except they were ungated. For transplantation experiments in which rod precursors were selected via NrlGFP transgene expression, stage matched GFP− wild-type pups were used to set FACS-gates. Purity of sorted cells was usually more than 90%. Cell viability was more than 90% based on 4′,6-diamidino-2-phenylindole (DAPI) staining. The sorted cells were resuspended at 200,000 cells per millilitre in injection buffer (Earle’s Balanced Salt Solutions, DNaseI) after centrifugation at 124g for 10 minutes using a Heraeus Labfuge 400R (Thermos, U.K.) and injected via subretinal injection into recipient mice as described previously [10, 12] and detailed in the Supporting Information Methods.
It is described in Supporting Information Methods.
It is described in Supporting Information Methods.
As described in Supporting Information Methods.
Previously, we have shown that transplanted photoreceptor precursor cells isolated from retinae at P1–P7 and expressing an NrlGFP transgene are effective for retinal repair [10, 12]. The NrlGFP transgene labels developing and mature rod photoreceptors [22, 30]. Transcriptome analysis of NrlGFP expressing rod precursors was performed to identify markers of postnatal photoreceptor precursor cells. GFP+ and GFP− cells from P4 NrlGFP retinae were isolated by FACS for RNA isolation and their transcriptomes compared using an Affymetrix microarray platform to identify markers enriched in photoreceptor precursors. Analysis of the data set using Bioconductor software revealed 294 genes, whose expression levels were at least fourfold higher in the GFP+ versus GFP− cells and these were classified in 10 clusters (Fig. 1). As expected, the largest cluster with the highest enrichment score related to vision and visual perception (n = 134 genes). The second largest cluster concerned nucleotide binding (n = 29). Supporting Information Table 1 shows the top 40 genes that showed significantly higher expression (greater than ninefold higher expression in the P4 photoreceptor precursors when compared with other retinal cells; range 9.2–127-fold). More than half of these genes cause human retinal disease involving photoreceptor degeneration including the Nrl transcription factor gene (24.3-fold increased expression in the GFP+ population). These data indicate that the P4 immature photoreceptor precursors highly express genes important for photoreceptor function.
As a complementary analysis to identify genes showing changing temporal expression profiles in the postnatal period, which may be useful for selecting stage-specific postnatal rod precursor cells, we used Affymetrix microarray data deposited in Gene Expression Omnibus (GEO, www.ncbi.nlm.nih.gov/geo) from a study comparing isolated NrlGFP expressing cells from Nrl−/− retina, with NrlGFP expressing wild-type cells . This study did not analyze the GFP− cell population and so did not allow the identification of genes upregulated in the P4 photoreceptor precursors when compared with other retinal cells. A total of 2067 genes were identified that showed more than twofold temporally varying expression profiles during development (E16 to P28). Box plots were generated from microarray analysis for all genes identified. Of the top 40 genes identified in the P4 Nrl GFP+ versus GFP− cells, 21 were identified from the analysis of the GEO dataset and 19 were unique to our array (Supporting Information Table 1 and Fig. 1). To identify genes encoding cell surface markers that could potentially be used to isolate precursor cells the DAVID tool, Functional Annotation Chart was used to search the gene lists for genes encoding membrane proteins with extracellular domains or plasma-membrane-associated proteins.
Four genes were identified that fulfilled the criteria of (a) expression in photoreceptor precursors, (b) dynamic developmental expression profiles, (c) available antibodies for flow analysis, and (d) gene ontology (GO) annotations and existing literature indicating they encoded proteins possessing extracellular domains. These were Nt5e, which encodes the 5′ ectonucleotidase, present in the visual perception and nucleotide binding clusters; Cd24a, a proposed marker of immature neurons (both glycosylphosphatidylinositol (GPI) anchored proteins, in which the majority of the proteins are extracellularly located); Prom1 encoding Prominin1/CD133 from the visual perception cluster and Podxl encoding the podocalyxin protein (both transmembrane proteins with extracellular domains) (Supporting Information Fig. 1).
We examined the temporal gene expression profiles of the candidate genes encoding cell surface markers in developing retinae, using quantitative real-time polymerase chain reaction (qRT-PCR) of RNA. Nt5e, Prom1, and Podxl showed lower expression during embryonic development relative to increased expression levels after birth (Fig. 2). All three markers maintained higher gene expression levels in the adult retina (P21). Cd24a showed an opposite expression profile, displaying stronger expression during early retinal development and a decline with progressing maturation in the postnatal period (Fig. 2). qPCR analysis of FAC-sorted P5 NrlGFP cells confirmed that transcripts for Nt5e, Prom1, and Podxl were highly enriched in the GFP+ rod precursor population, displaying a 60-, 22-, and 20-fold enrichment, respectively, in this population when compared with GFP− cells. Cd24a showed slightly higher expression in GFP− cells when compared with GFP+ cells, consistently with the microarray analysis (1.4-fold higher in GFP− compared with GFP+).
We next compared the spatial distribution of the proteins CD73, CD24, and CD133 encoded by the Nt5e, Cd24a, and Prom1 genes, respectively, in tissue sections from the postnatal and adult retina and compared their distribution with NrlGFP and CrxGFP labeling of developing photoreceptor cells (Fig. 3). In the postnatal retina, CD73 protein was located within the outermost edge of the neuroblastic layer, which contains developing photoreceptors expressing GFP. CD133 showed intense staining at the outer aspects of the developing photoreceptor layer and lower level staining within the neuroblastic layer. CD24 was widely distributed throughout the neural retina during this period of development, showing expression within developing photoreceptors as well as high level expression on inner retinal cells. In the adult mouse retina, CD73 was restricted to the photoreceptors (Fig. 3). CD133 showed a punctate distribution at the base of the outer segment of photoreceptors and was absent from other retinal cell types. CD133 immunostaining on dissociated P5 retinal cells from the NrlGFP transgenic mouse line confirmed cell surface labeling of photoreceptor cells (Supporting Information Fig. 2). CD24 protein levels decreased markedly in the mature retina. Immunostaining was weakly present only in the projections of Muller glia cells (Fig. 3C4″′).
To establish the percentage of retinal cells labeled by each candidate cell surface antigen at different developmental time points, we performed FACS analysis using fluorochrome-conjugated antibodies. At E17, only a very small number (2%) of retinal cells is stained with CD73 and this increased markedly in the postnatal period to more than 60% of all cells persisting into adulthood (Fig. 4B, 4F). By contrast, CD24 showed an inverse pattern (Fig. 4C, 4F): more than 80% of cells were labeled at E17 and this decreased in the postnatal period to very low levels in the mature retina of less than 8% consistent with its reported downregulation in maturing neurons in the central nervous system [33, 34]. CD133 stained a significant number of retinal cells during embryonic and postnatal development (>70%) but levels decreased in adulthood (Fig. 4D, 4F).
By gating for each fluorochrome and GFP, the proportion of each cell surface marker population that colabels with NrlGFP was assessed. At E17, very few of the NrlGFP cells colabeled with CD73 (Fig. 4G, left hand panel) indicating that CD73 is not a marker of the very early stages of rod photoreceptor differentiation. By P5, and at later stages, the majority of retinal cells (>60%) were colabeled for CD73 and NrlGFP (Fig. 4G; top right hand quadrants). Conversely, at E17 the vast majority of the CD24+ cells were NrlGFP− (Fig. 4H, left hand panel; top left hand quadrant, 70%). The proportion of retinal cells that colabeled with CD24 and NrlGFP initially increased postnatally so that by P5 more than 40% of all retinal cell colabeled and then subsequently declined such that less than 10% of cells colabel after P10 (Fig. 4H; top right hand quadrants). At E17, CD133 labeled approximately 70% of all cells, but the majority of these were NrlGFP− (Fig. 4I; left hand panel; top left hand quadrant). In the postnatal period, CD133 labeled both GFP+ cells and GFP− cells with approximately 50% of all cells colabeling for GFP and CD133 at P5 and more than 60% at P10 decline thereafter (Fig. 4I).
CD73 and CD24 were evaluated as a candidate surface marker combination that could define postnatal stage photoreceptor precursors as the qPCR and protein analysis indicated that these markers were coexpressed in developing photoreceptors at P4–P7 but not at earlier or later stages. We first tested whether a combination of CD73 and CD24 labeling would select postnatal photoreceptor precursor cells. We found that more than 30% of retinal cells colabeled with CD73 and CD24 at P5, whereas negligible levels colabel at embryonic and adult stages (Fig. 4J). These data indicate that photoreceptors downregulate CD24 expression as they transition to a more mature state. The majority of CD73 and CD24 colabeled cells expressed the NrlGFP transgene (98.4 ± 0.3%; Supporting Information Table 2). Thus, this marker combination selects a stage-specific subpopulation of developing NrlGFP expressing cells; 43.3 ± 5.1% of NrlGFP precursors label with CD73/CD24 at P5 (Supporting Information Table 2). NrlGFP is an established marker of postmitotic cells [10, 22], therefore the CD73 and CD24 colabeled cells are postmitotic.
We next conducted proof-of-principle experiments to address whether the cell surface marker combination identified above, CD73 and CD24, can be used to isolate photoreceptor precursors that are integration competent following transplantation. Immunostaining on dissociated P5 retinal cells from the NrlGFP transgenic mouse line was first conducted to assess cell surface labeling of photoreceptor precursor cells; CD73 was restricted to the cell surface of NrlGFP+ rod precursor cells, while CD24 was present on the cell surface of both photoreceptor and nonphotoreceptors cells (Fig. 5F). Therefore, CD73/CD24 double positive cells were isolated using FAC-sorting (from either NrlGFP or CbaGFP P4 retina), ignoring GFP expression from the transgene for cell selection. P4 CD73/CD24 sorted cells were transplanted into the subretinal space of adult wild-type recipients and compared with cells sorted using CD73 alone or sorted using NrlGFP expression. At 3 weeks post-transplantation, large numbers of NrlGFP or CbaGFP expressing cells had migrated and correctly integrated into the recipient’s ONL following transplantation of each donor population (Fig. 5). Integrated cells displayed the typical features of mature rod photoreceptors, including condensed nuclear morphology, inner/outer segment, and a rod spherule in the outer plexiform layer. The density of integrated cells within the host ONL varied, ranging from single cells to dense clusters of cells. In some cases, the host ONL showed a thickening where new photoreceptors had integrated. The migration of photoreceptor cells was restricted to the ONL (using donor cells from either NrlGFP or CbaGFP P4 retina) and GFP+ cells were not observed in other retinal cell layers (Fig. 5A–5D). All subsequent experiments were performed using donor cells from NrlGFP P4 retina, as this genetic marker was useful for the identification of integrated rods following transplantation of donor populations sorted using cell surface antigens alone.
To further verify transplanted photoreceptor identity and normal maturation following donor cell selection using cell surface antigens we performed immunostaining with several markers of mature photoreceptors (Fig. 6). Integrated NrlGFP+ cells labeled with recoverin, RetGC, and the rod specific marker phosducin (A′–C′). In addition, punctate CD133 labeling was observed at the base of the outer photoreceptor segment of integrated cells (D′). Rod spherules of the newly integrated photoreceptors were situated in close proximity to rod bipolar cells labeled using a PKCalpha antibody (E′). Taken together, these data suggest that rod precursors isolated by means of the cell surface markers and transplanted subretinally into the normal adult mouse retina give rise to mature rod photoreceptors in the host ONL that appear to properly integrate within the host retina. No adverse effects of antibody binding to cell surface antigens for photoreceptor differentiation were noted.
Staining with conjugated monoclonal antibodies to specific cell surface antigens/receptors may have a detrimental effect on cell viability or reduce integration efficiency of rod precursors. Therefore, we directly compared integration efficiency of P4 rod precursors isolated using cell surface antigens to those isolated by NrlGFP fluorescence or unsorted P4 retinal cells from the NrlGFP mouse line. This comparative analysis also allowed us to assess whether the cell surface antigens label a cell population that has enhanced integration properties. Serial sections of transplanted host retinae were prepared and integrated NrlGFP+ cells were counted (Fig. 5E). Significantly, transplantation of P4 rod precursors isolated based on their coexpression of CD73 and CD24 resulted in a twofold increase (median = 10,899, range 544–32,826, n = 10) in the number of integrated photoreceptors compared with precursors isolated by NrlGFP expression (median 4,649, range 1,953–8,206, n = 12), and approximately 18-fold higher than unsorted cells (median 606, range 131–4,032, n = 15; Fig. 5E). Transplantation of P4 rod precursors sorted by CD73 expression alone showed the same efficiency (median 13,134, range 2,840-31,894, n = 10) as CD73/CD24 double positive donor cells and was also more efficient than NrlGFP alone (Fig. 5E).
In summary, these data show that the cell surface marker CD73 both alone and in combination with CD24 can be used to isolate a population of donor cells that can be effectively transplanted into the wild-type retina and this is achieved without reliance on a genetic modification of the donor cell population.
Finally, we tested whether transplantation into a degenerating environment could also be achieved using donor cells selected on the basis of cell surface markers. The diseased retina is particularly challenging owing to the physiological and structural changes that occur during disease progression. Prph2rd2/rd2 (also known as Rds) mutant mice are a model of Leber’s congenital amaurosis (LCA) and are homozygous for a null mutation in the Prph2 gene [35, 36]. In this model, photoreceptors lack outer segment disc structures, contain low amounts of the photo pigment rhodopsin, and degenerate slowly, while leaving the inner retinal cell layers mostly intact. Mutations in the human CRB1 gene cause severe retinitis pigmentosa and LCA. In Crb1rd8/rd8 mutant mice, which contain a single base pair deletion in the Crb1 gene, the outer limiting membrane is compromised and the retina degenerates [37, 38]. Transplanted P4 rod precursors isolated by CD73/CD24 selection integrated in large numbers into ONL of the retinae of both Rds (median 12,855, range 1,387–25,742, n = 4) and the Crb1 (median 1008, range 876-3284, n = 4) mice and differentiated to exhibit the morphological features of mature photoreceptors (Fig. 7).
Potential sources of photoreceptor precursors for future clinical application include ESC-derived retinal cultures [14–16] or progenitor cell cultures from developing retinae [17, 18]. To test whether the cell surface markers are expressed in ESC-derived retinal cultures, we performed RT-PCR on mouse ESCs differentiated using the Takahashi protocol which has previously been shown to achieve successful photoreceptor cell differentiation [16, 39]. This protocol selects RxGFP positive cells from embryoid body cultures on day 10 before addition of growth factors to induce retinal differentiation. Nt5e, Cd24a, and Prom1 expression were detected by RT-PCR in retinal differentiation cultures of days 20 and 26, along with the photoreceptor markers Nrl and Crx. Nt5e, Cd24a, and Prom1 were also expressed in the adult brain and undifferentiated ESCs suggesting that protocols without RxGFP selection may require additional markers to exclude nonretinal cells. As there are differences in marker expression patterns between mouse and humans , we also tested for expression of the human homologues of Cd24a and Nt5e in the developing human neural retina and undifferentiated hESCs by RT-PCR. Prominin 1 (PROM1) expression in the human retina has been previously reported [41, 42]. CD24 but not CD73 was detected in the undifferentiated hESCs, whereas both CD73 and CD24 mRNA together with CRX and NRL expression were detected in the 13 week human neural retina (Supporting Information Fig. 3).
Human color and high acuity vision relies mainly on cone photoreceptors concentrated in the fovea of the central retina. Rod photoreceptors mediate dim light vision and outnumber cones 30:1. In the majority of inherited retinal degenerations rod photoreceptors are first affected and loss of cone photoreceptors, eventually resulting in severe visual impairment is secondary and is due to a trophic dependence on the surrounding rods [43, 44]. There is an urgent need to develop new therapies for human retinal disease as they cause such a high proportion of global blindness. One in 2000 people have an inherited retinal degeneration caused by mutation in one of more than 200 different genes and more than 3.2 million people are blind due to acquired retinal degenerations including age related macular degeneration [1, 45]. Stem cell therapy is one of several therapeutic strategies under investigation. Others include gene therapy to replace single defective genes before photoreceptors degenerate [46–48] or to add light sensitive genes to surviving inner retinal cells to provide light sensitivity  and electronic implants to detect light .
The aim of retinal stem cell therapy is to transplant stem cells or stem cell-derived cells into the diseased retina either to replace photoreceptor cells lost through the degenerative process or to delay or prevent the loss of further cells. In its purest form, the goal of retinal stem cell therapy is to repopulate the damaged retina with new functional photoreceptors that make connections with the remaining inner retinal circuitry and provide visual function equivalent to that of normal photoreceptors.
Recently, we and others have provided evidence that photoreceptor precursor transplantation via subretinal injection presents a feasible strategy for the repair of retinal degeneration [3, 4, 10, 11, 13, 51]. We have shown that transplanted rods make connections and confer low light sensitivity and that cone transplantation is also possible. In future studies, it will be essential to perform assays to demonstrate the efficacy of photoreceptor transplantation in long-term transplants and to provide in vivo assays of visual improvements and synaptic connections. So far, the most promising type of donor cell used in animal models has been the postmitotic yet immature photoreceptor precursor cell derived from the developing retina . To develop strategies to translate this progress toward a clinical therapy, it is essential to develop methods to successfully isolate and purify optimal stage precursor cells from stem cell cultures. The most practical sources of precursor cells are considered to be either ESC or iPSCs [14–16, 52], which offer the potential to supply large numbers of cells, although other cell sources are also being investigated.
Isolation of donor cells for transplantation in the research setting up to this point has depended either on the use of photoreceptor specific transgenes such as NrlGFP and CrxGFP or has used nonselected mixed cell populations [4, 10, 13]. Regardless of the type of stem cell used to synthesize precursor populations, the use of transgenic labeling is not desirable for the future translation of this approach to human therapy. There is an essential requirement to provide purified donor cell populations; mixed cell populations have inherent risks associated with tumorogenesis from the unregulated proliferation of stem or progenitors cells once transplanted into the host as well as showing low transplant efficiency. Several reports have shown that unselected ESC transplants lead to tumor formation [53, 54]. All retinal differentiation protocols reported to date have relatively low photoreceptor yields and contain different stages and types of cells [14–16, 52]. Here, we have explored the possible use of cell surface markers as a means of isolating photoreceptor precursor cells at stages of development when they are most effective at migrating, integrating, and differentiating in an adult host retinal environment. This approach is established in other system most notably in the isolation of hematopoietic cells for bone marrow transplants, where combinations of several cell surface marker antibodies are commonly used for FACS and/or magnetic cell selection such as using MACS technology [55–57]. However, adult blood cells are naturally in suspension and the populations isolated are dividing stem cells able to turnover the cell surface markers and replenish their state. By contrast, the cells that need to be isolated from differentiated stem cell cultures for retinal stem cell therapy are postmitotic developing cells , which are required to maintain the ability to integrate, make connections, and mature after transplantation.
We selected four markers, based on analysis of the transcriptome of postnatal photoreceptor precursor cells, that were either highly upregulated in postnatal photoreceptor precursors compared with other retinal cells, or that showed a developmental expression profile that labeled immature but not mature cells. Previous studies of early endoderm development have also successfully made use of transcriptome analysis to identify cell surface markers to isolate genetically unmanipulated ESCs . We chose markers that we predicted could be useful for isolating transplantation competent cells based on their labeling of postnatal cells expressing an NrlGFP transgene expressed in photoreceptor cells. Among those CD73, a GPI anchored ecto-5′ nucleotidase whose expression commences around the peak of rod birthing is specific to photoreceptor cells in the retina but also labels them at developmental stages that we have determined are not optimal for integration . CD24, also anchored to the cell surface via a GPI connection, on the other hand, is only expressed in immature retinal neurons and is turned off in the mature state providing a marker for selecting immature photoreceptors. CD73/CD24 double labeling of retinal cells delineates a transient state spanning the postnatal period from which donor cells have previously been shown to have high integration efficiency [10, 13]. CD133, the pentaspan transmembrane glycoprotein, also known as PROM1 was shown to label postnatal photoreceptor precursors, but also labels NrlGFP negative cells and other studies have shown its expression in somatic stem and progenitor cells including those derived from the nervous and hematopoietic systems and in various developing epithelia and differentiated cells as well the retina [39, 32, 59–61]. CD133 represents an additional cell surface marker of donor cells, which in combination with other markers may be useful for refining future cell isolation protocols from differentiated stem cell cultures. The fourth marker identified Podocalyxin, a transmembrane sialoglycoprotein, did not successfully label photoreceptors using available antibodies. This protein is more commonly studied in the kidney where it is thought to be the major constituent of the glycocalyx of podocytes . Expression patterns of CD133, CD73, and CD24 were consistent with previous reports of expression in photoreceptors and immature neurons [31–33, 41, 63]. Notably, CD73 expression was previously shown on rod cells in mice and primates; in mice, CD73 positive cells expressed rhodopsin but not the cone marker s-opsin and the common marmoset monkey showed correlation of the expression pattern of rhodopsin and CD73 . Moreover, the effects of ectopic expression in retinal cells of Nrl and Crx, both expressed in postmitotic photoreceptor precursors, suggested that CD73 is genetically downstream of Crx in the rod cell differentiation lineage . We selected the two cell surface markers CD24 and CD73 that based on their opposing gradients of expression during development define a population of photoreceptor precursor cells during the transplantation competent period of postnatal development. In proof-of-principle experiments, we demonstrated that our donor cell selection approach using antibodies to cell surface antigens was feasible; a significantly higher number of integrated rod photoreceptors were achieved using the new selection methods via cell surface antigens (up to 30,000 cells per retina) compared with isolation via GFP fluorescence from a rod expressed NrlGFP transgene or without selection of cells. This is likely due to the defined developmental stage of the photoreceptor precursors selected using the cell surface markers compared with NrlGFP labeling; NrlGFP labels more early-stage precursors than CD73 (Fig. 4F). Addition of further markers and refinement of marker combinations is likely to further improve the ability to select optimal stage cells for effective transplantation from mixed populations of developing retinal cells.
This study has demonstrated that CD24/CD73 selection from a mixed retinal cell population effectively isolates a postmitotic photoreceptor population that is highly transplantation competent. To ensure the removal of any residual non-retinal cells from stem cell cultures, such as ESCs or iPSCs subjected to retinal differentiation protocols, more complex multistep protocols including negative selection and additional cell surface markers will need to be developed in future studies. A number of cell surface markers of undifferentiated mouse ESCs have been identified that could potentially be used for negative selection, such as SSEA-1 (stage-specific embryonic antigen 1), and CD8 which is downregulated as ESCs differentiate [64, 65]. Positive selection approaches to select retinal progenitor cells before retinal differentiation protocols and precursor cell selection may also prove effective; c-kit and SSEA-1 label retinal progenitor cells as well as other cell types . Development of cell-surface marker signatures for human cell sources of transplantable retinal precursor cells will require more investigation; notably, a recent study identified expression of CD133 and CD24 in undifferentiated hESCs, whereas CD73 was only observed in neural induction cultures of hESCs . Future studies will need to test multiple combinations of markers and will require the optimization of the timing of differentiation protocols, optimization to ensure viability of in vitro differentiated stem cells and scale-up of stem cell cultures to generate and isolate large numbers of optimal stage cells sufficient to perform rigorous testing in transplantation experiments. It is imperative that high levels of integrated cells are achieved to conduct reliable testing of visual function provided by transplanted cells. In future work, it will be important to establish whether the same or different cell surface markers can be used to isolate human cells and to better understand the properties of cell surface molecules that enable transplantation competence. In summary, this study indicates that selection using cell surface markers could be an effective approach for the development of retinal stem therapy.
Our proof-of-principle study shows that isolation of photoreceptor precursors via antibodies specific to cell surface markers is a feasible approach because it does not inhibit cell integration or negatively influence survival after transplantation. When transplanted into mouse models of human retinal degeneration, CD73/CD24 selected cells were able to integrate into the remaining ONL of the host animal. Moreover, P4 rod precursor cells isolated by CD73/CD24 expression integrated in greater numbers into the wild-type adult retina than the cells isolated by expression of an NrlGFP transgene indicating that the use of surface markers may allow optimization of cell populations for transplantation. Based on this work, we propose that a future strategy using combinations of cell surface markers including those identified here may be applicable for the isolation of human precursor cells effective for photoreceptor cell replacement from in vitro stem cell-derived cell populations.
We thank Anand Swaroop for providing the NrlGFP mouse line and Connie Cepko for providing the CrxGFP mouse line; Krzysztof Palczewski for the RetGC1 (Gucy2e) antibody; Rebecca Rolfe for assistance with polymerase chain reaction analysis, Angie Wade for statistical advice, UCL Genomics for microarray analysis, the UCL Institute of Child Health flow cytometry and confocal facilities for technical support. We also thank the guidance from Masayo Takahashi to establish protocols for retinal differentiation of embryonic stem cells (ESCs) and Yoshiki Sasai for the EB5 RxGFP ESC line. This work was supported by the Medical Research Council UK (G03000341 and G0901550); Fight for Sight; Child Health Research Appeal Trust; the Macula Vision Research Foundation; the Wellcome Trust (082217); the Ulverscroft Foundation; National Institute for Health Research (NIHR) Biomedical Research Centre for Ophthalmology at Moorfields Eye Hospital and UCL Institute of Ophthalmology; NIHR Biomedical Research Centre for Pediatric Research at Great Ormond Street Hospital for Children and UCL Institute of Child Health. The human embryonic and fetal material was provided by the Joint Medical Research Council UK (grant# G0700089)/Wellcome Trust (grant # GR082557) Human Developmental Biology Resource (http://hdbr.org) and the human ESC line WA09 was from the WiCell Research Institute provided by JA Thomson, University of Wisconsin, Madison. R.A.P. is a Royal Society University Research Fellow and J.C.S. is funded by the Great Ormond Street Hospital Children’s Charity.
Disclosure of Potential Conflicts of Interest The authors indicate no potential conflicts of interest.
See www.StemCells.com for supporting information available online.