Growing evidence suggests that glial cells may have a role as neural precursors in the adult central nervous system. Although it has been shown that Müller cells exhibit progenitor characteristics in the postnatal chick and rat retinae, their progenitor-like role in developed human retina is unknown. We first reported the Müller glial characteristics of the spontaneously immortalized human cell line MIO-M1, but recently we have derived similar cell lines from the neural retina of several adult eye donors. Since immortalization is one of the main properties of stem cells, we investigated whether these cells expressed stem cell markers. Cells were grown as adherent monolayers, responded to epidermal growth factor, and could be expanded indefinitely without growth factors under normal culture conditions. They could be frozen and thawed without losing their characteristics. In the presence of extracellular matrix and fibroblast growth factor-2 or retinoic acid, they acquired neural morphology, formed neurospheres, and expressed neural stem cell markers including βIII tubulin, Sox2, Pax6, Chx10, and Notch 1. They also expressed markers of postmitotic retinal neurons, including peripherin, recoverin, calretinin, S-opsin, and Brn3. When grafted into the subretinal space of dystrophic Royal College of Surgeons rats or neonatal Lister hooded rats, immortalized cells migrated into the retina, where they expressed various markers of retinal neurons. These observations indicate that adult human neural retina harbors a population of cells that express both Müller glial and stem cell markers and suggest that these cells may have potential use for cell-based therapies to restore retinal function.
Disclosure of potential conflicts of interest is found at the end of this article.
Müller cells constitute the main glial population of the retina . They share a lineage with retinal neurons, and both Müller cells and neurons share a common progenitor that is multipotent at all stages of retinal histogenesis . This evidence derives from examination of the progeny of a single mouse retinal progenitor cell transfected with a retrovirus, which generated clones containing up to three types of neurons, whereas others contained a combination of neurons and Müller glia, Müller glia alone, or a single type of neuron . For several decades, it has been known that fish and amphibians are capable of regenerating neural retina , and previous studies have indicated that Müller cells may regenerate chick  and rat retina . More recent findings have shown that, in the adult zebra fish, Müller glia form the retinal stem cell niche and are able to generate retinal stem cells in the regenerating retina . Furthermore, Müller glia from the adult rat retina have been more recently shown to exhibit neural stem cell properties .
Müller cells have a structural role in the retina in addition to providing metabolic support to neurons and blood vessels. They are easily characterized by their phenotypic characteristics in vitro, including the presence of intracellular glycogen granules and the expression of epidermal growth factor receptor (EGF-R), vimentin, cellular retinaldehyde binding protein (CRALBP), and glutamine synthetase [8, 9]. They also express the glutamate transporter GLAST and depolarize in response to l-glutamate without changes in membrane resistance, consistent with the electrogenic uptake of this amino acid . We first described the Müller cell characteristics of the spontaneously immortalized human cell line MIO-M1 , but our recent work has revealed that many similar cell lines can be derived from the adult human retina. Since immortalization is one of the characteristics of stem cells, and based on recent evidence that postnatal Müller cells in zebra fish and rat exhibit progenitor characteristics in experimental models of retinal injury, we investigated whether the immortalized human cell lines that we have obtained from the neural retina of adult human eyes have stem cell characteristics. On this basis, we examined whether MIO-M1 cells and similar cell preparations express markers of neural progenitors, whether they are capable of differentiating into retinal neurons in vitro, and whether they have the potential to migrate and differentiate, a property of neural and retinal stem cells [12, , , –16], when grafted into the subretinal space of the dystrophic Royal College of Surgeons (RCS) rat, a model of retinal degeneration, or the neonatal Lister hooded rat.
Materials and Methods
Isolation of Müller Cells from the Adult Neural Retina
Müller cells were isolated as previously described by us and others [9, 11, 17, 18] from the neural retina of cadaveric donor eyes with no eye disease (age range 18 months to 83 years old). Upon approval of the ethics committee of the local health authority, eyes consented for research were obtained from Moorfields Hospital Eye Bank between 24 and 48 hours postmortem. After removal of the cornea and the lens by holding the optic nerve on the upright position, vitreous and retina were gently dislodged from the eyecup with a pair of small forceps, leaving behind the retinal pigment epithelium (RPE) and choroid. The retina was then carefully cut from the optic nerve and placed in a Petri dish. Using a surgical blade, the neural retina was excised approximately 4 mm away from the ciliary body (supplemental online Fig. A). This was done to avoid contamination with retinal stem cells reported in this region . Since separation of the retina from the ciliary body and the vitreous may potentially dislodge cells from the ciliary body that could contaminate cell preparations obtained from the neural retina, we obtained small fragments of neural retina from two eyes without disturbing the ciliary body. For this purpose, we made four radial cuts through the sclera and the vitreous (the deepest one in the superior pole) to obtain a flattened structure. Then, using a punch cutter 1 cm in diameter, we cut a section of retina attached to vitreous, RPE, and choroid from the posterior section of the eye. Retina was then carefully dissected from these small fragments (supplemental online Fig. B). To isolate the Müller cells, neural retina was rinsed with phosphate-buffered saline (PBS), placed in Trypsin-EDTA (5% trypsin, 2% EDTA; Gibco-BRL, Gaithersburg, MD, http://www.gibcobrl.com), homogenized by vigorous pipetting, and incubated in the same solution for 20 minutes at 37°C. Large tissue debris was then removed by filtration through a stainless steel sieve. Dissociated cells were washed and cultured in tissue culture flasks (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com) coated with 10 μg/cm2 fibronectin (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) in the presence of 40 ng/ml epidermal growth factor (EGF; Sigma) in Dulbecco's modified Eagle's medium (DMEM) containing l-Glutamax I (Gibco-BRL) and 10% fetal calf serum (FCS; Gibco-BRL). The original medium was maintained until the first colonies were formed, after which it was replaced weekly with fresh medium without EGF. After 2–3 weeks, cells were detached by incubation with Trypsin-EDTA (5.0 g/l trypsin, 2.0 g/l EDTA) (Sigma) for 3 minutes at 37°C. Cells were then expanded in culture and examined for their characteristic morphology, electron-microscopic features, and expression of CRALBP (antibody kindly provided by Professor J. Saari, University of Washington), cytokeratin 8/18 (CAM 5.2 antibody from Becton, Dickinson), fibroblast surface protein (antibody from Abcam, Cambridge, U.K., http://www.abcam.com), vimentin (antibody from Dako, Glostrup, Denmark, http://www.dako.com), glial fibrillary acidic protein (GFAP), EGF-R, and glutamine synthetase (antibodies from Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com) using our methods previously described . Cells isolated from whole neural retina or retinal sections obtained by punch cutting through the vitreous and sclera yielded similar cell preparations that exhibited Müller cell characteristics and spontaneous immortalization.
Expression of mRNA Coding for Markers of Neural Retinal Stem Cells and Mature Retinal Neurons by Immortalized Cells
Following trypsinization of Müller cell monolayers, total cellular RNA was isolated from cell pellets using the RNeasy system (Qiagen, Hilden, Germany, http://www1.qiagen.com) according to the manufacturer's instructions. Cells used to isolate RNA were derived from passages 8–45. For the reaction, 1 μg of total RNA was reverse-transcribed in 20-μl reactions consisting of 5 mM MgCl2, 1 mM deoxynucleoside-5′-triphosphate (dNTP), 1 U/μl RNase inhibitor, 0.8 U/μl AMV reverse transcriptase (Roche Diagnostics, Basel, Switzerland, http://www.roche-applied-science.com), and 80 ng/μl oligo(dT)-15 primers (Roche) in 10 mmol/l Tris/HCl buffer containing 50 mM KCl. The mixture was incubated as follows: 10 minutes at 25°C, 60 minutes at 42°C, 5 minutes at 99°C, and 5 minutes at 4°C in a thermal cycler (Eppendorf AG, Hamburg, Germany, http://www.eppendorf.de). Polymerase chain reaction (PCR) amplification was then performed using our published methods  and specific primers for CRALBP, glutamine synthetase, Notch 1, Sox2, Chx10, Pax6, calretinin, S-opsin, and recoverin. Sequences were derived from the human genome sequence, GenBank (Table 1). The amplification was performed in a final volume of 50 μl by addition of 1.5 mM MgCl2, 0.2 mM dNTP, 2.5 U Expand HiFi Taq DNA polymerase (Roche), 0.4 μM primers in 50 mM KCl, and 10 mM Tris/HCl, pH 8.0. The mixture was initially incubated at 94°C for 2 minutes followed by 30–34 cycles under the following conditions: 94°C for 30 seconds, annealing temperature for 30 seconds, 72°C for 1 minute, and one cycle of 72°C for 5 minutes. PCR products were then analyzed by agarose gel electrophoresis (1%) containing 25 ng/ml ethidium bromide (Sigma).
Table Table 1.. Primer sequences used for reverse transcription-polymerase chain reaction analysis
Differentiation of MIO-M1 Cells and Other Cell Lines into Neural Phenotypes
The Müller cell line MIO-M1 and similar immortalized cells, named as MIO-M2, MIO-M3, MIO-M4, etc., were maintained in culture as previously described . In more detail, cells were maintained as monolayers cultured in the presence of DMEM containing l-Glutamax I and 10% FCS. Upon confluence, cells were detached using Trypsin-EDTA (5% trypsin, 2% EDTA) and seeded at a 1:6 dilution of the original flask. For dedifferentiation studies, cells were cultured at a density of 500–800 cells per cm2 in DMEM containing 10% FCS on cell culture slides coated with 50 μg/ml extracellular matrix (ECM) gel or 10 μg/ml fibronectin (Sigma). Fibroblast growth factor-2 (FGF2) or retinoic acid (RA) (Sigma) were then added to the medium to achieve concentrations of 40 ng/ml and 500 nM, respectively. Cells were cultured for 3–7 days at 37°C with replenishment of medium and growth factors every 2 days. To assess the clonality of the cells, a mixture of cells transfected with enhanced green fluorescent protein (EGFP) (see below) was mixed with nontransfected cells at a ratio of 1:1 and the formation of green or nongreen neurospheres examined after 4 days in culture using the same culture conditions as above. Acquisition of neural morphology was examined by phase contrast microscopy using the Leica imaging system DC200 (Heerbrugg, Switzerland, http://www.leica.com) and confocal microscopy (see below). To examine the expression of markers of retinal stem cells, slides were fixed in 4% paraformaldehyde in PBS, pH 7.4, and stained with the following antibodies: rabbit polyclonal anti-Notch 1, sonic hedgehog (Shh), Pax6 (Santa Cruz Biotechnology), Sox2 (Chemicon, Temecula, CA, http://www.chemicon.com), goat polyclonal anti-Chx10 (Santa Cruz Biotechnology), monoclonal anti-nestin and anti-βIII tubulin (Chemicon), and rabbit polyclonal anti-cyclin D (Santa Cruz Biotechnology). Cells were also labeled with fluorescent peanut agglutinin (Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com). Specificity of staining for Sox2, Pax6, Shh, and Chx10 was confirmed by overnight incubation of the corresponding blocking peptides (catalog numbers: sc-17319P, sc-7750P, sc-1194P, and sc-21692P, respectively; Santa Cruz Biotechnology) with the primary antibodies at a ratio of 10:1 by weight. Additional controls included the omission of primary antibodies and the use of IgG isotypes from the same species in which the primary antibodies were raised. To determine the expression of retinal neural phenotypes, cells were stained with monoclonal antibodies to cone/rod peripherin (clone 5, kind gift from Professor R. Molday, University of British Columbia); polyclonal anti-protein kinase C (PKC) (Santa Cruz Biotechnology); polyclonal anti-calretinin identifying ganglion cells and major retinal neurons (Swant, Bellinzona, Switzerland, http://www.swant.com); polyclonal anti-HuD and Brn3 identifying ganglion and amacrine cells (Santa Cruz Biotechnology); and polyclonal anti-160 kDa neurofilament protein identifying major retinal neurons (Chemicon). Binding of primary antibodies was detected with donkey anti-IgG labeled with Alexa Fluor 488 or 546 reacting with the species in which the primary antibody was raised (Molecular Probes, Eugene, OR, http://www.probes.invitrogen.com). Slides were costained with 4,6-diamidino-2-phenylindole (Sigma) to visualize cell nuclei. Fluorescent images were recorded using a confocal microscope (LSM 510; Carl Zeiss, Jena, Germany, http://www.zeiss.com).
To assess the percentage of cells expressing the different markers, following fixation and immunostaining of cells cultured in 96-well plates, positively stained cells were counted using the Cellomics ArrayScan VT1 HCS automated reader (Cellomics, Pittsburgh, http://www.cellomics.com). Using the integrated reader's software, over 5,000 cells from five replicates were analyzed in each group. Results were expressed as the percentage of positive cells.
Western Blot Analysis
Subconfluent cell monolayers of Müller cell lines cultured on either uncoated tissue culture flasks or culture flasks coated with fibronectin (FN) or ECM gel in the presence or absence of RA or FGF2 extracellular matrix proteins were lysed with Laemmli buffer, followed by centrifugation of the lysates at 13,000 rpm for 5 minutes and storage of the supernatants at −85°C until use. Western blotting of cell lysates (1.5 mg/ml) was performed as previously described  using the same antibodies used for immunostaining, secondary antibodies coupled to horseradish peroxidase (Jackson Immunoresearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com), and enhanced chemiluminescence reagent (Amersham Biosciences, Piscataway, NJ, http://www.amersham.com). The images were analyzed using a Fuji image reader LAS-1000 Pro, version 2.1 (Fuji, Bedford, U.K., http://www.fujifilm.com).
The use of animals in this study was in accordance with the Home Office regulations for the care and use of laboratory animals and the U.K. Animals (Scientific Procedures) Act (1986). RCS rats were bred in-house, and Lister hooded rats were purchased from Harlan (Indianapolis, http://www.harlan.com) or Charles River Laboratories (Wilmington, MA, http://www.criver.com). Animals were kept under a 12-hour/12-hour light-dark cycle (light cycle mean illumination: 30 cd/m2). To trace the transplanted cells, spontaneously immortalized Müller cells were transfected with a retroviral vector harboring EGFP protein (Clontech, Palo Alto, CA, http://www.clontech.com) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, http://www.invitrogen.com) according to the manufacturer's instructions and our published methods. Three cell lines (MIO-M1, MIO-M4, and MIO-M7) obtained from whole neural retinae were used for the transplantation studies. Dissociated cells (2 × 104 cells in 2 μl) were grafted into the subretinal space of 3–4-week-old dystrophic RCS rats susceptible to retinal degeneration [21, 22] as previously described . Adult animals were anesthetized using a mixture of medetomine hydrochloride and ketamin reversed with atipamezole hydrochloride. Sham-injected rats received DMEM without cells. All injections were given in a single eye, with the contralateral eyes used as untreated controls. Animals were immunosuppressed with oral cyclosporine A (210 mg/l drinking water) from day 2 before transplantation until termination of the experiment. For transplants into neonatal Lister hooded rats, the dams were maintained on oral cyclosporine and azathioprine immediately after the birth of the rat pups. Each experimental group consisted of 6–8 animals. Under terminal anesthesia and following intracardial perfusion with 4% paraformaldehyde in PBS, eyes were enucleated. After washing in PBS, eyes were cryoprotected in 30% sucrose. Frozen sections 15-μm thick were cut, and the fate of the transplanted cells was investigated by confocal microscopy of retinal sections costained with monoclonal antibody to GFP and antibodies to neural retinal markers as above. Nontransfected donor cells were identified using antibodies to human mitochondria (Chemicon).
Establishment and Characterization of Immortalized Müller Cell Lines from Adult Human Retina
Similar to the protocols used to grow the Müller cell line MIO-M1 , cell colonies formed after 2–3 weeks in culture were dissociated, expanded in the absence of EGF, and examined for Müller characteristics at the time of the first passage and after 20 passages. Cells that continued dividing after twenty passages (approximately 100 divisions) were considered immortalized. To date, out of 15 eye donors, we have generated 12 cell lines with similar characteristics. Six of the cell lines have undergone 20 passages (approximately 100 divisions), four of the cell lines (MIO-M3 to M6) 57 passages (approximately 285 divisions), one cell line (MIO-M2) 102 passages (approximately 510 passages), and the previously published cell line MIO-M1  has undergone 120 passages (approximately 600 divisions). Although Müller glia that became immortalized in vitro are responsive to EGF, the first two cell lines established in our laboratory did not have EGF in culture at any time. However, we found that faster colonies were formed when EGF was added to the primary culture and that, once we dissociate the first colonies for expansion, EGF is not necessary to maintain cell growth. These cell lines have been frozen and thawed several times without losing their characteristics, and all cells examined after 50 passages exhibit the same features of cells tested at passages 1 and 20. We were also able to grow cells with immortalized characteristics from two eyes in which we sectioned the neural retina by punch cutting through the vitreous and sclera without disturbing the eye anatomy. This was done to avoid any potential contamination with cells from the ciliary body. The natures of the immortalized cell preparations obtained from adult neural retina were identified by gene expression and immunostaining for Müller cell markers as well as by their phenotypic characteristics using phase contrast, transmission, and scanning electron microscopy using standard techniques in our laboratory . Subconfluent immortalized cells exhibited bipolar morphology, irregular membrane appearance, and formation of cytoplasmic projections (Fig. 1A), which are well recognized features of Müller cells in culture [9, 11, 17, 18]. They also contained cytoplasmic glycogen granules as observed by transmission electron microscopy (Fig. 1B) and displayed microvillous projections on their apical surface as observed by scanning electron microscopy (Fig. 1C), as previously described for Müller glia . Immunocytochemical staining confirmed that these cells expressed well characterized markers of Müller cells, including CRALBP, EGF-R, vimentin, and glutamine synthetase [8, 9, 17, 24] (Fig. 1D–1G). Only a small proportion of these cells (<5%) expressed GFAP (not shown), which is in accordance with previous reports that GFAP is found at low levels or is completely absent in mammalian Müller cells , and none of the cell preparations expressed cytokeratin 8 or 18, known markers of RPE cells, or fibroblast surface protein. Interestingly, only 1/500–1/600 of the freshly dissociated cells from the retina coexpressed nestin and CRALBP, and only cells that coexpressed CRALBP and nestin in culture became spontaneously immortalized. Those that did not express this molecule acquired a flattened morphology and did not proliferate for more than 4–6 passages (Fig. 1H). Examination of gene expression by reverse transcription (RT)-PCR confirmed that these cells express mRNA coding for glutamine synthetase and CRALBP (Fig. 1I). In addition, Western blot analysis of cell lysates showed that bands of molecular weights 42 kDa, 40 kDa, and 170 kDa corresponding to glutamine synthetase, CRALBP, and EGF-R, respectively (Fig. 1J), were obtained by immunoblotting with the corresponding antibodies. Cells obtained from punch cutting through the sclera and vitreous exhibited similar characteristics to the cells obtained by dislodging the whole vitreous body with the retina. When cultured in the absence of extracellular matrix and growth factors, all cell lines permanently expressed Müller glial markers.
Gene Expression of Retinal Stem Cell Markers by Spontaneously Immortalized Cell Lines
Investigation of the expression of various markers of retinal stem cells was performed by RT-PCR of mRNA extracted from different cell lines cultured on tissue culture flasks in the absence of extracellular matrix and without the addition of growth factors. Expression of mRNA coding for Sox2, Pax6, Chx10, and Notch 1 was observed in all cell lines examined. Figure 2A shows the expression of mRNA by 6 of the 12 cell lines established in our laboratory. Similarly, mRNA coding for markers of postmitotic retinal neurons, including calretinin, recoverin, and S-opsin, was also observed in cells cultured under the above conditions (Fig. 2B). Variations in the levels of expression of these factors were observed among cell lines, and further studies will elucidate the significance of these differences.
Differentiation of Immortalized Cell Lines
We induced differentiation of the different immortalized cell lines using conditions known to promote neural retinal differentiation [19, 26, –28]. Culture of cells at a low density (500–800 cells per cm2) on ECM gel in the presence of FGF2 or RA caused individual cells to form neurospheres, which could be examined for multipotentiality. Confirmation of clonality was achieved by limiting the dilution of cells dissociated from neurospheres so that a single cell placed in a tissue-culture well could multiply until confluence. Cells originating from a single clone were then dissociated and repeatedly cultured at low density (as above) to induce new formation of neurospheres with pluripotent progeny. We observed that, after 3–4 days of culture on ECM gel in the presence of FGF2 or RA, between 10% and 20% of individual Müller cells formed neurospheres (Fig. 3A–3D). Cells present in the neurospheres expressed (a) cyclin D, a cell cycle protein involved in the regulation of proliferation during retinal development ; (b) nestin, an early marker of neuronal differentiation ; and (c) binding of peanut agglutinin, a lectin that binds to glycoconjugates of pluripotent neural stem cells  (Fig. 3B–3D). Following 4 days of culture on fibronectin in the presence of RA, we observed that neurospheres formed by a 50/50 mixture of EGFP transfected and nontransfected cells resulted in the formation of neurospheres mainly of a single type: 50.3% contained only green cells, 46% contained only nongreen cells, and 3.7% contained a mixture of green and nongreen cells. The small proportion of neurospheres containing the two types of cells may be attributed to cell aggregation; nevertheless, the high proportion of single green and nongreen neurospheres confirms the clonality of the cells (Fig. 3E).
After 3–4 days, a large number of cells (>70%) exhibited neural morphology (Fig. 3F–3H), and a variable proportion also expressed markers of postmitotic retinal neurons at 4 and 7 days (see below; Table 2). These included expression of PKC (a marker of bipolar cells), peripherin (a marker of photoreceptor cells), HuD and Brn3 (markers of ganglion cells), 160 kDa neurofilament protein (a marker of ganglion, amacrine, and horizontal cells), and calretinin (a marker of ganglion, amacrine, and horizontal cells) (Fig. 3I–3N) [32, –35]. Cells cultured under the same conditions for 4–7 days also expressed βIII tubulin, a marker of immature neurons (Fig. 3O) . Acquisition of neural markers has been examined in 10 of the 12 spontaneously immortalized human cell lines established in our laboratory, and all have the ability to differentiate into cells expressing markers of retinal neurons.
Table Table 2.. Percentage of cells expressing markers of various cell types upon differentiation on extracellular matrix with addition of growth factors
Influence of Growth Factors and Extracellular Matrix on Müller Cell Differentiation
Depending on the growth factor used for differentiation of cells cultured on ECM gel, there was variability in the proportion of cells expressing markers of retinal progenitors and mature retinal neurons as well as in the pattern of staining for various progenitor and mature retinal markers (Table 2; Fig. 4A). Staining for Pax6, a progenitor and amacrine cell marker , was observed in cells cultured with FGF2 and RA, but very little or no staining was observed in cells cultured in the absence of these factors (Fig. 4A–4C). Staining for Sox2, a transcription factor and progenitor marker , was characteristically associated with cytoplasmic and neurite extensions on cells cultured with FGF2 (Fig. 4D), but nuclear staining was observed in 25%–30% of cells cultured in the presence of RA (Fig. 4E). In the absence of ECM gel and growth factors, the majority of cells expressed cytoplasmic Sox2 (Fig. 3F). Staining for the signaling protein Shh, a key regulator of neural development , was predominantly perinuclear in cells cultured with FGF2, but less dense perinuclear and cytoplasmic staining was observed in cells cultured with RA (Fig. 4G, 4H). Staining for Chx10, a progenitor and amacrine cell marker [40, 41], was mainly observed in the cytoplasm of a small proportion of cells cultured with FGF2, but weak or no staining for this molecule was observed on a very small number of cells cultured in the presence of RA (Fig. 4J, 4K). In the absence of extracellular matrix and growth factors, a very small proportion of cells expressed Shh and Chx10 (Fig. 4I, 4L). Table 2 shows the variations in the proportion of cells staining for the various markers examined. When cultured with FGF2, 75% expressed Sox2 when compared with 59% of cells cultured with RA (p < .01). A higher number of cells expressed Pax6 when cultured with FGF2 (40%) compared with RA (19%) (p < .01), and a similar number of cells expressed Shh when cultured in the presence of FGF2 or RA (40% and 38%, respectively). Although a small proportion of cells expressed Chx10 in the presence of FGF2 (15%), this was significantly higher (p < .05) than in the presence of RA (5%). Variations in the proportion of cells expressing markers of postmitotic retinal neurons were also observed when cells were cultured for 7 days in the presence of different growth factors (Table 2). A significant decrease in the number of cells expressing CRALBP was observed when cells were cultured in the presence of FGF2 or RA (12% and 15%, respectively) when compared with cells cultured in the absence of these factors (74%). HuD was expressed by 37% of cells cultured with FGF2 and 35% of cells cultured with RA. PKC was expressed in a higher proportion of cells when cultured with FGF2 (33%) (p < .01) than in the presence of RA (17%), and, similarly, a higher number of cells expressed peripherin in the presence of FGF2 (29%) (p < .01) than in the presence of RA (14%) (Table 2). The majority of cells cultured in the absence of ECM gel and growth factors also maintained the expression of Sox2 (86%) after 7 days, but very few cells expressed markers of retinal neurons (Table 2). Confirmation of the specificity of staining for Sox2, Pax6, Shh, and HuD was achieved by blocking the activity of these antibodies by overnight incubation with their corresponding blocking peptides. Using this approach, complete depletion of immunoreactivity was achieved, resulting in negative staining for these molecules (not shown). Lack of immunoreactivity was also observed when the primary antibody was omitted or when control isotype antibodies were used. Western blot analysis of cell lysates following culture on ECM gel or FN in the presence or absence of FGF2 or RA showed that bands of molecular weights 48 kDa, 47 kDa, 46 kDa, and 34 kDa corresponding to Pax6, Shh, Sox2, and Chx10, respectively, were obtained by immunoblotting with the corresponding antibodies. In addition, immunoblotting with the anti-Shh antibody yielded bands of approximately 19 kDa (corresponding to the secreted N-terminal domain of Shh) and 95 kDa (which appears to be a Shh dimer) (Fig. 4M). Equal amounts of protein were applied to the blots, which were controlled by immunodetection with anti-β actin antibodies (Fig. 4M).
Retinal Grafting of Immortalized Cells with Müller and Stem Cell Characteristics
We investigated the biological activity of human Müller glia with neural stem cell characteristics by examining their ability to migrate into the retina and to express markers of retinal neurons in vivo, following grafting of dissociated cells into the subretinal space of the dystrophic RCS rat, a model of retinal degeneration  and neonatal Lister hooded rats. Confocal microscopy examination of retinal sections showed that transplanted cells could be observed in the outer and inner nuclear cell layers 9–14 days after grafting. We observed better migration of the Müller cell line MIO-M1 into the neonatal retina of Lister hooded rats than into the dystrophic retina of the RCS rat (Fig. 5A, 5B). Cells that migrated into the photoreceptor cell layer stained for recoverin and rhodopsin (Fig. 5C, 5E), markers of photoreceptor cells , whereas cells that migrated to the inner nuclear layer expressed HuD, a marker of major retinal neurons found in this layer  (Fig. 5D). Cells that had migrated to the ganglion cell layer also stained for calretinin (Fig. 5F), a marker of ganglion cells and amacrine cells  that may be found in this layer. As shown in Figure 6, subretinal injection of the Müller cell lines MIO-M4 and MIO-M7 also yielded similar results to those obtained with the Müller cell line MIO-M1. Following 12 days after subretinal injection of the MIO-M4 cell line, cells costaining for EGFP and recoverin were observed in the outer nuclear cell layer (Fig. 6A). Similarly, MIO-M7 cells that migrated to the ganglion cell layer costained for EGFP and calretinin. Cells with long processes, in some cases resembling axons, were observed with all of the three cell lines investigated (Figs 6C, 6D).
Glial cells have the role of stem cells in the adult and embryonic brain in various species, and it is suggested that differentiation along the glial lineage may be a default state of development reflected in the progression of stem cells along the neuroepithelial to radial glia to astrocyte lineage . Stem cells are thought to have an undifferentiated phenotype, but a population of glia retains neural progenitor ability in the central nervous system of different species . This is illustrated by observations that radial glia, astrocytes, and oligodendrocyte precursors generate neurons [43, 45, , , –49], for which it has been proposed that progression along a glial pathway is related to neuronal production .
The presence and neural regenerative ability of Müller glia in postnatal retina have been shown after cytotoxic damage in chick , zebra fish , and rat [5, 7] retinae, but Müller glia with progenitor characteristics have not been demonstrated in the adult human retina. Although there are obvious limitations to the study of individual cell progenies during human development, the present findings that a small proportion of cells isolated from the human neural retina becomes spontaneously immortalized in vitro and exhibits both Müller glia and neural stem cell characteristics suggest that adult human retina may also harbor these cells.
It is well known that Müller glia express EGF-R [8, 9], and studies have shown that overexpression of EGF-R by neural progenitors in vivo enhances glial cell differentiation . Observations that EGF-R was localized to the cytoplasm and nuclei of immortalized Müller glia are in agreement with those previously reported that intracellular expression of EGF-R strongly correlates with highly proliferating activities of tissues [51, –53]. This is in accordance with our observations that the immortalized cells reported in this study are highly proliferative. EGF responsiveness has been shown to correlate with EGF-R expression [53, 54] and may explain our observations that initial exposure to EGF by cells expressing EGF-R induces rapid proliferation. Expression of cyclin D by cells present in neurospheres is in agreement with observations that this cell cycle regulatory protein is found in the cytoplasm of proliferating cells  and neurons during development .
All the spontaneously immortalized cell lines we have established not only express mRNA for well known markers of Müller glia but also express proteins (as determined by immunostaining and Western blotting) that are characteristically expressed by these cells, including CRALBP, glutamine synthetase, vimentin, and EGF-R [1, 9, 11, 25]. In addition, electrophysiological studies on the first characterized cell line (MIO-M1) showed that they depolarized in response to glutamate without change in membrane resistance, a well known feature of mammalian Müller cells . Other investigations on the cell line MIO-M1 have shown that, in response to proinflammatory factors, these cells produce matrix metalloproteinases 1, 2, and 9 [20, 57] as well as hepatocyte growth factor and vascular endothelial growth factor , which are known features of mature cells. These observations indicate that these cells not only possess features of mature Müller glia but also of neural stem cells when exposed to various extracellular matrix and growth factors in vitro.
As for the requirements for in vitro isolation and establishment of immortalized cell lines with Müller and stem cell characteristics, it is important to emphasize that, in our study, we did not derive the immortal cells from neurospheres grown under conditions normally used to culture retinal stem cells. Unlike previous reports in which retinal stem cells were derived from the ciliary body as sphere colonies in the absence of FCS , the cells reported in this study were grown as monolayers of adherent cells obtained from the neural retina and cultured in the presence of FCS. Using the culture conditions used by other investigators to grow neurospheres from the ciliary body , we were unable to grow these from the neural retina, which is in agreement with that previously reported . Here, we show that factors present in FCS do not induce differentiation of neural retinal cells with stem cell and Müller glia characteristics, but that presence of extracellular matrix and neurogenic factors such as FGF2 and RA induce their differentiation. It is unlikely that the cell lines that we have established are derived from the ciliary body, as the human eye is a large organ in comparison with the eyes of fish and rodents, where Müller glia with stem cell characteristics have been identified. Potential contamination with cells from the ciliary body was avoided by removing the neural retinal tissue at a considerable distance from this region, which is clearly identifiable when dissecting the human eye. That only 10%–20% of the cells formed neurospheres under the culture conditions used to induce expression of neural markers may be explained by the fact that cells were cultured in the presence of FCS, which, by promoting adherence, might inhibit neurosphere formation.
Nuclear and cytoplasmic staining for Sox2 is found in the inner cell mass of the blastocyst , supporting our observations that Sox2 is expressed in the nucleus and cytoplasm of a population of Müller glia. Immunocytochemical studies were confirmed by identification of gene expression using RT-PCR analysis as well as by Western blots of cell lysates, which yielded corresponding molecular weights. Sox2 expression is associated with dividing stem cells and precursors of the central nervous system [38, 59] and has been shown to maintain neurogenesis in the adult mouse brain . It is therefore possible that this factor may maintain the neurogenic properties of the spontaneously immortalized cells from the adult human retina. Shh is implicated in regulating adult neural stem cell proliferation , and it is a mitogenic factor for retinal progenitors , suggesting that it may promote the proliferation of stem cells with Müller characteristics isolated from the adult human retina. Chx10 constitutes an early marker of the developing retina and is required for retinal progenitor cell proliferation as well as formation of bipolar cells [40, 41]. In the mouse and chick, Chx10 is expressed in nearly all retinal progenitors but is absent from all postmitotic cells types except bipolar interneurons and a subset of Müller glia , which is in agreement with our findings that immortal cells expressing Müller glial characteristics express this protein. The Notch 1 signaling pathway is important at several stages of retinal development including the differentiation of retinal ganglion cells and Müller glia [62, 63]. Expression of this molecule by the cell lines reported in this study further indicates that they have neural stem cell characteristics. In our study, it is of interest that human-derived cells with Müller characteristics not only express genes coding for neural stem cells but also genes coding for proteins expressed by mature retinal neurons, such as Calretinin, recoverin, and S-opsin. Although a very small number of cells cultured under nondifferentiating conditions expressed markers of retinal neurons, as judged by immunohistochemical staining, a greater number of cells cultured in the presence of ECM and FGF2 or RA expressed neural retinal markers (Table 2). These observations highlight the proneural ability of the immortalized cells reported in this study. Immunohistochemical staining showed that HuD was expressed at both nuclear and cytoplasmic levels, which is in accordance with what is described by others, that intracellular trafficking of HuD is important for neuronal differentiation [64, 65]. Extracellular matrix proteins have been recognized to play a very important role not only in the maintenance of the stem cell niche [66, 67] but also in the tissue-specific stem cell differentiation . It is therefore of interest that, when cultured in the presence of ECM and growth factors, cells with Müller and stem cell characteristics are induced to express stem cell and neural markers in different proportions (Table 2; Fig. 3). We did not investigate the differentiation of Müller glia into astrocytes. However, under the experimental conditions used, we did not observe cells expressing the astrocyte marker GFAP. Differentiation of retinal stem cells into astrocytes has not been carefully investigated and it may be due to the scarcity of these cells in the neural retina . Further studies using different matrix substrates and growth or differentiation factors may help to clarify this issue.
Although Müller glial cell proliferation (reactive gliosis) has been shown to occur during degenerative retinal processes , neurogenesis occurring in these conditions has not been demonstrated in the adult human retina. It is possible to speculate that if neural progenitors are harbored in the adult retina, these may have the potential to regenerate neurons. However, as there is no evidence that postdevelopmental neurogenesis may occur in the human eye, the possibility arises that this process is suppressed in vivo by unidentified factors present in the adult eye. Investigations into the feasibility of activating neurogenesis mediated by Müller progenitors in vivo may have the potential to develop into treatments for endogenous replacement of dysfunctional neurons.
To date, attempts to regenerate murine retina by transplantation of brain-derived or retinal progenitors have produced mixed results, and better migration and integration of stem cells have been shown when brain- or retinal-derived neural progenitors are transplanted into immature or injured retina [12, 70, 71]. Our study showed that migration and survival of cells with Müller and stem cell characteristics was better in normal neonatal retina than in the degenerating retina of the RCS rat, indicating that developmental cues may be beneficial to cell integration. The migration and acquisition of neural retinal markers in vivo by our immortalized human retinal cells closely resemble that observed with other neural progenitors [13, 15, 16]. At present, there are many limitations to the effective delivery of stem cells to regenerate retina [13, 15, 16], but the fact that cells with Müller glia and stem cell characteristics can be easily grown from human cadaveric retina, cryopreserved, and renewed for long periods of time without losing their phenotypic or genetic characteristics suggests that, once other obstacles in the retina itself can be overcome, these cells may have potential for clinical application and may merit further investigations.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
This work was supported by the Medical Research Council (MRC) (Grant number 67386), the Helen Hamlyn Trust (in memory of Paul Hamlyn), the Wellcome Trust (Grant reference 062290), the Henry Smith Charity, and the Michael and Ilse Katz Foundation, U.K. We thank Dr. James Ellis for help with tissue processing. We are also very grateful to Professor A.R. Venkitaraman from the Hutchison MRC Research Centre for allowing us to use the Cellomics ArrayScan VT1 automated reader. This research has been partly funded by The Department of Health's National Institute for Health Research Biomedical Research Centre at Moorfields Eye Hospital and the UCL Institute of Ophthalmology, U.K. S.S. is supported by the Inlaks Foundation and the Henry Smith Trust.