Derivation of Neurons with Functional Properties from Adult Limbal Epithelium: Implications in Autologous Cell Therapy for Photoreceptor Degeneration



The limbal epithelium (LE), a circular and narrow epithelium that separates cornea from conjunctiva, harbors stem cells/progenitors in its basal layer that regenerate cornea. We have previously demonstrated that cells in the basal LE, when removed from their niche and cultured in reduced bond morphogenetic protein signaling, acquire properties of neural progenitors. Here, we demonstrate that LE-derived neural progenitors generate neurons with functional properties and can be directly differentiated along rod photoreceptor lineage in vitro and in vivo. These observations posit the LE as a potential source of neural progenitors for autologous cell therapy to treat photoreceptor degeneration in age-related macular degeneration and retinitis pigmentosa.

Disclosure of potential conflicts of interest is found at the end of this article.


Neurodegenerative diseases and brain injury are major problems affecting millions of people, usually leading to permanent disability. With the discovery and characterization of neural stem cells/progenitors, there is a hope of reversing or slowing degenerative changes using ex vivo cell therapy. However, there are outstanding issues that currently constitute barriers to practical and successful clinical use of neural stem cells/progenitors. These include whether the grafted cells will be accepted or rejected and whether cells will be easily available in sufficient quantity to sustain clinical use. To address these practical issues we are examining the renewable heterologous stem cells/progenitors that regenerate adult cornea as a potential source of neural progenitors towards an autologous stem cell therapy concept [1].

Cornea is a renewable tissue and, like other surface epithelium, is regenerated throughout life. The source of cells that participates in its regeneration is limbus, a circular narrow band of epithelium that separates the cornea from the conjunctiva. Several lines of evidence suggest that cells in the basal layer of the limbal epithelium (LE) are stem cells/progenitors [2, 3]. First, like adult stem cells that divide infrequently, basal cells are slow-cycling with high proliferating potential [4, [5]–6]. Second, the basal cells lack the expression of Keratin 3 [7]. Third, basal cells display side population (SP) cell phenotype, the general phenotype of stem cells, [8, [9]–10] and fourth, limbal epithelial cells can regenerate corneal epithelium [11].

Both neural and limbal stem cells/progenitors are derived from embryonic ectoderm. Based on the default model of neural induction, which proposes that embryonic ectoderm cells can differentiate into neural cells autonomously unless exposed to bone morphogenetic protein (BMP) signaling [12], we hypothesized that removal of stem cells/progenitors from the LE niche will expose their vestigial neural properties. Testing of this hypothesis showed that cells in the basal layer in the adult rat LE, when subjected to neurosphere culture assay, acquired neural progenitor properties under the regulation of BMP signaling and differentiated along neuronal and glial lineages, in vitro and in vivo [13]. Similar observations were made with adult human limbal stem cells/progenitors [14]. Here, we demonstrate that adult LE-derived neural progenitors give rise to neurons with functional properties. In addition, we demonstrate that these neural progenitors can differentiate along photoreceptor lineage under the influence of neonatal retinal cells, suggesting their use in autologous cell therapy for photoreceptor degeneration.

Materials and Methods

LE Cell Culture

LE cell culture was done as previously described [13]. Briefly, eyes of adult Sprague-Dawley rats (Sasco, Wilmington, MA, or GFP/Sox2-GFP/Nrl-GFP mice were enucleated in Hank's balanced salt solution. The entire cornea was removed by circular incision below the limbus. The limbal region was carefully dissected away from the peripheral cornea and cut into small pieces. Tissues were incubated first in 0.05% trypsin (Sigma, St. Louis, and then in 78 U/ml of collagenase (Sigma) and 38 U/ml of hyaluronidase (Sigma) at 37°C for 27 and 30 minutes, respectively. Tissues were dissociated into single cells by trituration. The mode of dissection ensured that LE was not contaminated with cells from the ciliary epithelium or peripheral retina, which are known to harbor neural stem cells/progenitors [15, [16]–17]. Dissociated cells were cultured in DMEM:F12 supplemented with 1× N2 supplement (Gibco), 20 ng/ml of EGF (R&D, Minneapolis,, 10 ng/ml of FGF2 (R&D) and 100 ng/ml of Noggin (R&D Systems, Inc., Minneapolis, at a density of 105 cells/cm2. After a week in culture, resulting neurospheres of cells were plated on glass coverslips coated with poly-D-lysine and laminin. BrdU (10 μM) was added to the last 8–12 hours of culture to tag proliferating cells. Medium was changed every other day. Neurospheres were not detected when stromal cells, which are enriched with neural crest and mesoderm derivatives [18], were similarly cultured, thus ruling out extra-ocular tissue-derivatives as contaminants [13] (supplemental online Fig. 1). Neurospheres generated by LE, derived from pigmented mice, consisted exclusively of non-pigmented cells, thus ruling out contamination with ciliary epithelium neural stem cells, which are pigmented and generate neurospheres consisting of pigmented cells (Fig. 1H, 1I) [13, 16]. To promote pan neuronal differentiation, neurospheres were co-cultured with embryonic day 18 (E18) hippocampal cells using a Millicel CM insert (pore size, 0.4 μm; Millipore, Billerica, MA, in the medium defined above except that it lacked mitogens and was supplemented with 1% FBS, 1 ng/ml of BDNF (R&D) and 1 μM of all trans retinoic acid (RA) (Sigma). Neurospheres were plated in 24-well tissue culture plates and allowed to settle. Inserts containing the hippocampal cells were slowly and carefully lowered into wells to avoid any spill. Medium was changed without removing the insert. Controls consisted of neurospheres cultured in the mitogen-free medium supplemented with serum and factors as above. To promote differentiation along retinal lineage, neurospheres were co-cultured with postnatal day 1 (PN1) retinal cells as described above. Following differentiation, cells were collected for RT-PCR analysis or fixed with 4% paraformaldehyde for 15 minutes at 4°C for immunocytochemical analysis.

Figure Figure 1..

LE cells acquire neural properties in vitro. A section through the anterior aspect of the rat eye display areas of cornea, limbus and conjunctiva (A). A subset of LE cells in the presence of Noggin and FGF2 generate neurospheres (B) containing cells that incorporate BrdU and express Nestin (inset). Cells from LE-derived neurospheres when subjected to Hoechst dye efflux assay demonstrate SP cell phenotype (C). A subset of BrdU+ cells in the LE-derived neurospheres expresses neural progenitor markers Nestin (D, E) and Sox2 (F, G). The graph demonstrates the relative proportion of Nestin+ and Sox2+ proliferating cells (H). RT-PCR analysis of LE cells cultured in the increasing concentration of Noggin (+FGF2) demonstrate the concentration-dependent increase and decrease in transcripts corresponding to Sox2 and p63, respectively, compared to those cultured in the absence of Noggin (+FGF2) (I). LE cells from the Sox2-GFP mice generate neurospheres containing GFP+ cells, suggesting the activation of Sox2 promoter in the presence of Noggin (+FGF2) (J, K). Insets (J', K') demonstrate cells cultured in the absence of Noggin (+FGF2). Temporal RT-PCR analysis of LE cells, cultured in the presence of Noggin and FGF2 for 5 days, revealed that the expression of transcripts corresponding to epithelium-specific genes, α-enolase and p63 decreased and that of neural progenitor specific genes, Nestin and Musashi increased with the time in culture (L). The graph represents temporal changes in gene expression (M). Data are expressed as mean ± SEM from triplicate culture of three different experiments.

Immunofluorescence Analysis

Immunofluorescence analyses were carried out as previously described [13]. Briefly, cells or cryosections were incubated with specific antibodies Map2 (Chemicon), Nfl (Sigma), β-tubulin (Covance), Sox2 (Chemicon), Nestin (DSHB), Pax6 (Babco), Rx [19], Opsin [20], rhodopsin kinase (Affinity Bioreagents), Chx10 (Exalpha Biologicals), PKC (Sigma) and Syntaxin one [20] in blocking serum overnight at 4°C. Following incubation in species-specific IgG antibodies conjugated to CY3, incorporation of BrdU was detected by using anti-BrdU (Sigma). Double immunocytochemical staining was carried out by simultaneously incubating cells with antibodies against specific antigens at 4°C overnight followed by simultaneous incubation in species-specific IgG coupled to CY3 and FITC. Images were captured by a CCD cool digital camera (Princeton Instruments) using image-capturing software (OpenLab).

Electron Microscopy

Neurospheres were cultured on 25 mm round Thermanox cover slips (Fisher Scientific) and were rinsed at least three times with 0.1 M sodium phosphate buffer. Tissue was fixed in 2% glutaraldehyde, 2% paraformaldehyde, 0.5% Acrolein in 0.1 M Sorensen's phosphate buffer (pH 7.2) for 1 hour, washed 2–3 times in buffer to remove all traces of glutaraldehyde and then dehydrated through graded concentrations of ethanols. The tissue was critical point dried and mounted on Al stubs. The mounted tissue was coated with gold palladium using a Sputter Coater (Hummer VI, Anatech) and placed in the vacuum chamber of the FEI Quanta 200 scanning electron microscope (SEM; Hillsboro, OR) and visualized. Representative samples were obtained using the digital image acquisition of the microscope.

RT-PCR Analysis

Isolation of total RNA from cells and synthesis of cDNA were carried out as previously described [13]. cDNA (2 μl) was amplified by using the forward and reverse gene-specific primers (Table 1) using step cycles (denaturing for 30 seconds at 94°C; annealing for 40 seconds at primer-specific annealing temperature, extension for 45 seconds at 72°C for 30 cycles). PCR products were resolved on 1.5% agarose gel.

Table Table 1.. Gene-specific oligonucleotide primers for RT-PCR
original image

Hoechst Dye Efflux Assay

Hoechst dye efflux assay was carried out as previously reported [21]. Briefly, dissociated neurosphere cells were resuspended in Iscove's modified Dulbecco's medium (IMDM) at a concentration of 106 cells/ml and incubated at 4°C overnight followed by staining with Hoechst 33,342 (4 μg/ml) at 37°C for 60 minutes and sorted on a FACStar Plus (BD Biosciences, San Diego, cell sorter. Hoechst dye was excited at 350 nm, and fluorescence was measured at two wavelengths using a 485 BP22 (485 nm bandpass filter) and a 675 EFLP (675 nm long-pass edge filter) optical filter (Omega Optical, Brattleboro, VT, The SP region was defined on the flow cytometer on the basis of its fluorescence emission in both blue and red wavelengths. Dead cells and debris were excluded by establishing a live gate on the flow cytometer using forward versus side scatter. The sorted SP cells were processed for immunocytochemical and RT-PCR analyses.

Microarray Analysis

Total RNA was isolated from LE-derived neurospheres cultured with PN1 retinal cells across a semi-permeable membrane as described above or in the presence of FBS. The RNA samples were processed for microarray analyses using Amino Allyl Message Amp kit (Ambion, Austin, TX, In short, 0.75 μg of total RNA was reversed transcribed and then subjected to in vitro transcription to generate cRNA. Resultant RNA (5 μg) was coupled to either CY5 or CY3 and purified by gel-exclusion chromatography. The CY3 and CY5 labeled probes were mixed and hybridized for 16 hours at 42°C using Ambion hybridization buffer II. Prior to use in the experiments the microarray slides were prehybridized with 4× sodium chloride and sodium citrate, 0.5% SDS, 5% bovine serum albumin at 42°C for 1 hour. Following hybridization, slides were washed according to suggested protocols by Ambion and scanned using an Axon 4,000b scanner and images gridded using GenePix software (Molecular Devices, Sunnyvale, CA, Genes were identified as differentially expressed when all triplicate spots corresponding to a specific gene exhibited twofold up and down differential expression, and confirmed by three independent RT-PCR analyses.


Retinal transplantation was carried out as previously described [22]. Briefly, animals were anesthetized with an intraperitoneal injection of 20–25 μl of a 1:1 dilution of ketamine (60 mg/ml) and xylazine (8 ng/ml), and a 30-gauge needle was inserted into the eye near the equator. The needle was retracted, and a glass micropipette connected to a 10 μl Hamilton syringe was inserted through the wound to deliver 50,000–100,000 GFP-expressing LE-derived neural progenitors in a 1–2 μl volume. Animals were sacrificed 1–2 weeks after transplantation. Eyes were enucleated, fixed in 4% paraformaldeyde and cryprotected in 30% before embedding for sectioning. Cryosections (12 μM) were screened for incorporated GFP+ cells and subjected to immunohistochemical analysis.

Electrophysiological Analysis and Ca2+ Imaging

Electrophysiological analysis and Ca2+ imaging was carried out as previously described [21]. Briefly, cells were plated on coverslips, placed in a chamber, and perfused on the stage of an upright, fixed-stage microscope (for electrophysiology, model BHWI; Olympus, Lake Success, NY,; for imaging experiments, model E600FN; Nikon, Melville, NY, with an oxygenated solution containing NaCl, 140 mM; KCl, 5 mM; CaCl2, 2 mM; MgCl2, 1 mM; HEPES, 10 mM; glucose, 10 mM (pH 7.4). Experiments were performed at room temperature. For whole-cell recording, patch pipettes were pulled on a vertical puller (model PB-7; Narishige, Tokyo, Japan) from borosilicate glass pipettes 1.2 mm outer diameter, 0.95 mm inner diameter; with internal filament (World Precision Instruments, Sarasota, FL, and had tips of 1–2 μm outer diameter with tip resistances of 6–12 MΩ. Pipettes were filled with a bathing solution containing KCH3SO4, 98 mM; KCl, 44 mM; NaCl, 3 mM; HEPES, 5 mM; EGTA, 3 mM; MgCl2, 3 mM; CaCl2, 1 mM; glucose, 2 mM; Mg-adenosine triphosphate (ATP), 1 mM; guanosine triphosphate (GTP), 1 mM; and reduced glutathione, 1 mM (pH 7.2). For Ca2+ imaging studies, cells were incubated for 45 minutes in Fura-2/AM (10 μM; Molecular Probes, Eugene, OR) plus nonionic detergent (10 μM; Pluronic F127; Biotium, Hayward, CA). After loading, cells were rinsed (1 mL/minutes) for 30 minutes by perfusion with the bathing solution. Test solutions were applied by bath perfusion. Cells were viewed through a 60× water-immersion objective (1.0 numerical aperture [NA]; Nikon), and fluorescence was stimulated with a 150-W xenon lamp attached to the microscope's epifluorescence port by a liquid light guide (Sutter Instruments, San Raphael, CA), and excitation wavelengths (340 or 380 nm) were changed with a filter wheel (Lambda 10–2; Sutter Instruments). Fluorescence changes were monitored with a cooled charge-coupled device (CCD) camera (SensiCam; Cooke Corp., Auburn Hills, MI, using Imaging Workbench software; Axon, Inc., Burlingame, CA). Images were acquired every 5–10 seconds, using 2 × 2 binning with 0.5–1-second acquisition times.

Statistical Analysis

Values are expressed as mean fold change ± SEM from three different experiments. Statistical analysis was done using Student's t-test to determine the significance between different conditions.


LE-Derived Neural Stem Cells/Progenitors Differentiate into Functional Neurons

Before examining the potential of LE-derived neural progenitors to differentiate into specific neuronal types, we determined the fidelity of acquiring neural potential. Cell dissociates from the basal LE (Fig. 1A), cultured in the presence of Noggin, and generated neurospheres as previously described [13] (Fig. 1B). Cells in neurospheres were examined for progenitor properties as follows: First, we subjected cell dissociates from neurospheres to Hoechst dye-efflux assay. The assay revealed the presence of side population (SP) cells in neurospheres, suggesting that cells in LE-derived neurospheres possess the SP cell phenotype that is characteristic of stem cells/progenitors [8] (Fig. 1C). Second, we examined the phenotype of cells in neurospheres. Immunocytochemical analysis of BrdU-exposed neurospheres revealed BrdU+ cells, expressing neural progenitor markers Nestin [Figure 1B (inset), 1D, and 1E] and Sox2 (Fig. 1F, 1G). The proportions of BrdU+Nestin+ and BrdU+Sox2+ cells in neurospheres were 24.12 ± 1.75% and 59.72 ± 5.5%, respectively (Fig. 1H). Together, these observations suggested that a subset of cells have acquired neural progenitor properties. Third, we examined whether or not the neural progenitor properties are regulated by BMP signaling. RT-PCR analysis of neurospheres, cultured in increasing concentrations of Noggin, demonstrated a dose-dependent increase and decrease in levels of transcripts corresponding to Sox2 and LE markers, p63 [23], respectively (Fig. 1I). Next, we carried out neurospheres assay on LE cells obtained from Sox2-GFP mice in which the expression of green fluorescent protein (GFP) is driven by Sox2 promoter [24]. We observed more GFP+ cells in neurospheres cultured in the presence of Noggin (Fig. 1J, 1K) than in controls (Fig. 1J, 1K, inset). Together these observations suggested that the acquisition of neural properties, that is, expression of Sox2 is facilitated in the presence of Noggin. Lastly, we examined the temporal expression patterns of putative limbal (α-enolase and p63) and neural (Nestin and Musashi) stem cells/progenitor markers in the presence of Noggin. Although α-enolase and p63 are expressed elsewhere besides LE, in the context of LE they are associated with cells in the basal layer, which harbors stem cells [2, 3, 23, 25]. Therefore, along with other properties such as self-renewal, their expression characterizes LE stem cells [13]. We observed that levels of transcripts corresponding to α-enolase and P63 decreased while those encoding Nestin and Musashi increased with the time (Fig. 1L, 1M). Together, these observations suggested that LE cells, when removed from the niche and exposed to Noggin, acquired neural properties in a regulated manner.

To promote neuronal differentiation that included the display of physiological properties, we shifted neurospheres from proliferating to differentiating conditions, which consisted of co-culturing them in the presence of neonatal hippocampal or retinal cells, across a semi-permeable membrane, in a culture medium lacking mitogens (Fig. 2A). This approach is based on observations that cells from embryonic/neonatal central nervous system elaborate factors that promote differentiation of cells that display physiological properties similar to that of functional neurons [26, 27, [28]–29]. Five days later we examined cells for morphological features and the expression of multiple neuronal markers. Scanning electron microscope (SEM) analysis revealed that cells in co-culture conditions were relatively smaller than those in control conditions and displayed neuronal morphology with long extensions (Fig. 2B, 2C). The extensions contained varicosities observed in the SEM analysis of neurons [30]. Cells in control conditions, besides being larger, were spherical with microcilia like structures, characteristic of epithelial cells [31]. Immunohistochemical analyses of the neurospheres in co-culture revealed cells expressing multiple neuronal markers, β-tubulin (Fig. 2D, 2E), neurofilament (Fig. 2F, 2G) and Map2 (Fig. 2H, 2I). The proportion of cells expressing these markers was ∼20% (Fig. 2J). These observations, based on morphological and biochemical/molecular criteria, suggested that these cells had acquired neuronal features.

Figure Figure 2..

Cells in LE-derived neurospheres display neuronal properties. LE-derived neurospheres were cultured in the presence of neonatal hippocampal cells, FBS+BDNF+RA for 5–7 days (A). Controls included neurospheres cultured in FBS+BDNF+RA. Scanning EM analysis of neurospheres revealed a neuronal morphology of cells in co-culture conditions (C), compared to those in control conditions, the majority of which were larger, spherical and ciliated like epithelial cells (B). Arrowheads point to varicosities in the neuronal processes. Immunocytochemical analysis of neurospheres in co-culture conditions, pre-exposed to BrdU to tag proliferating cells, revealed BrdU+ expressing immunoreactivities corresponding to neuronal markers, β-tubulin (D, E), Nfl (F, G) and Map2 (H, I). The graph demonstrates the relative proportion of cells expressing respective neuronal markers (J). Data are expressed as mean ± SEM from triplicate culture of three different experiments (J).

Next, we examined whether or not these cells have attributes of functional neurons by whole cell patch recording. The resting membrane potential of cells in co-culture conditions averaged −54.4 + 2.2 mV (n = 50). In contrast, cells in control conditions exhibited a significantly more depolarized resting membrane potential averaging −34.5 + 3.6 mV (n = 11). When currents were evoked by a series of voltage steps (150 ms, −90 mV to +30 mv), applied from a steady holding potential of −70 mV, cells in co-culture conditions revealed the presence of fast inward currents that were activated above −50 mV and likely due to voltage-gated Na2+ channels, and sustained outward currents likely due to outwardly rectifying K+ channels (Fig. 3A, 3C). In contrast, cells in control conditions had less pronounced outwardly rectifying currents and did not display voltage dependent Na2+ currents (Fig. 3B, 3C). To examine the physiological differentiation further, we tested for the presence of functional ionotropic glutamate receptors. Calcium imaging with Fura2 showed rapid increase in intracellular calcium in a subset of cells exposed to kainate (KA, 30 μM), indicating the expression of ionotropic KA/AMPA glutamate receptors, a neuronal feature (Fig. 3D, 3E). The upper left and upper right images in Figure 3D show pseudocolor images of the Fura-2 ratio obtained before and during application of KA. The graph in Figure 3E plots the ratios from a series of such images measured within an elliptical region encompassing two neighboring cell bodies (Fig. 3D). KA evoked detectable calcium increases in 19/58 cells examined from eight cultures. The calcium influx evoked by KA is at least partly due to activation of voltage-gated calcium channels since depolarization of neurons with high K+ stimulated similar Ca2+ increases in 25/58 cells. Fura-2 ratio images obtained before and during application of high K+ are illustrated in the bottom two panels of Figure 3D and the ratio changes produced by high K+ application in this same region of interest are plotted in Figure 3E. Consistent with the presence of KA/AMPA receptors, application of 20 μM AMPA to the bath solution also elicited inward currents in 5/13 voltage clamped cells (Fig. 3F). NMDA (100 μM), applied in the absence of Mg2+ and presence of glycine (100 μM), failed to evoke clear Ca2+ responses or inward currents (data not shown). These observations suggested that cells in co-culture conditions displayed both voltage-and ligand-gated currents characteristic of neurons. Lastly, we examined whether or not like neurons cells in neurospheres have the ability to establish inter-neuronal connections by light and electron microscopy. Immunohistochemical analysis revealed that these cells extend processes towards one another and express immunoreactivities corresponding to synaptic vesicle protein, synatophysin at the points of contacts (Fig. 3G). Electron microscope analysis of these cells revealed electron density at the point of intercellular contact, suggestive of axo-somatic synapse (Fig. 3H, 3I) [32]. Together, these observations suggest that LE-derived neural progenitors can generate neurons with functional properties.

Figure Figure 3..

Cells in LE-derived neurospheres display electrophysiological properties typical of functional neurons. LE-derived neurospheres were cultured as described is the preceding figure. Membrane currents evoked in a cell in LE-derived neurospheres in co-culture conditions (A). Currents were evoked by a series of voltage steps (150 ms, −90 mV to +30 mV) applied from a steady holding potential of −70 mV. A transient inward current characteristic of voltage-dependent sodium currents was observed at potentials above −50 mV (inset in C). Pronounced outward currents were observed at potentials above −30 mV. Membrane currents evoked in a cell in LE-derived neurospheres cultured in control conditions showed less pronounced outward currents and no transient sodium currents (B). Current/voltage relationships for the steady state currents in a cell in control conditions (open circles), steady state currents (filled circles) and initial transient current (filled triangles; the peak inward current measured during the first 15 ms) in a cell in co-culture conditions (C). Cells in LE-derived neurospheres in co-culture conditions were examined for their ability to mobilize intracellular calcium in response to non-NMDA agonists. Bath application of kainic acid (30 μM) stimulated an increase in intracellular calcium measured with Fura-2 in two closely adjacent cells (D and E). Depolarization with high K+ (E, 140 mM) also stimulated an increase in intracellular calcium consistent with the presence of voltage-gated calcium channels (E). Images obtained in control and test conditions are shown with an ellipse indicating the measurement region encompassing two neighboring cell bodies (D). Graph of the 340/380 ratio changes measured within the ellipse are shown in E. (F): shows an inward current evoked by bath application of AMPA (20 μM) to a cell in LE-derived neurospheres voltage clamped at −70 mV. Immunocytochemical analysis of LE-derived neural progenitors in co-culture conditions demonstrates expression of immunoreactivities corresponding to synaptophysin at the inter-cellular contacts (G, arrows). EM analysis of cells in LE-derived progenitors in co-culture conditions reveals the electron density of a putative synapse between neighboring cells (H and I). Abbreviation: Undiff, undifferentiated.

LE-Derived Neural Progenitors Differentiate into Retinal Neurons

We were interested in knowing if LE-derived neural progenitors can differentiate into specific neuronal types. We examined their ability to differentiate into retinal neurons. We co-cultured BrdU-tagged neurospheres with PN1 rat retinal cells across the semi-permeable membrane and examined the ability of cells to differentiate into late born retinal neurons, the rod photoreceptors (Fig. 4A). It has been previously demonstrated that PN1 cells elaborate activities that can influence the differentiation of homologous [33, [34]–35] and heterologous [36, 37] stem cells/progenitors into rod photoreceptors. We observed that the presence of PN1 retinal cells led to the induction of photoreceptor regulatory gene, Nrl [38] and Crx [39, 40], accompanied by those encoding photoreceptor-specific markers, opsin, rhodopsin kinase, arrestin and IRBP (Fig. 4B). Immunocytochemical examination of neurospheres in co-culture conditions revealed the presence of BrdU-tagged cells expressing immunoreactivities corresponding to opsin (Fig. 4C, 4D, 4E, 4F) and rhodopsin kinase (Fig. 4G, 4H), the proportion of which was significantly higher (∼4 fold; p < .001) as compared to those in control conditions (Fig. 4K). Next, we co-cultured neurospheres derived from LE of GFP mice with PN1 retinal cells. We observed a subset of GFP+ cells expressing immunreactivities corresponding to rhodopsin kinase (Fig. 4L, 4M) suggesting that the differentiation of LE-derived progenitors along rod photoreceptor lineage is not species-specific. To ascertain that the normal mechanism of rod differentiation is involved, we co-cultured neurospheres derived from the LE of Nrl-GFP mice [41] with PN1 retinal cells. We observed cells expressing GFP, suggesting the activation of Nrl promoter, which is necessary for LE-derived neural progenitors to acquire rod phenotype [38, 41] (Fig. 4N, 4O). Finally, to examine their retinal potential in vivo, GFP+ neural progenitors, derived from the LE of GFP mice, were transplanted intravitreally in PN7 rat eyes containing mechanical damage to the retina. The outcomes of transplantation were examined a week later. Immunofluorescence screening of retinal sections demonstrated incorporation of GFP+ cells in different lamina of the retina. The majority (∼85%) of incorporated cells were observed in the inner retina as previously reported [42]. On average, 15% of incorporated cells were observed in the outer nuclear layer, and immunohistochemical analysis revealed that a subset of these cells expressed immunoreactivities corresponding opsin, suggesting their lamina-specific differentiation along photoreceptor lineage (Fig. 4P, 4Q, 4R, 4S, 4T). Together, these observations suggested that LE-derived neural progenitors possess the potential for retinal differentiation in vitro and in vivo.

Figure Figure 4..

Cells in LE-derived neurospheres differentiate along rod photoreceptor lineage. LE-derived neurospheres were co-cultured in the presence of PN1 retinal cells and FBS+BDNF+RA for 5–7 days (A). Controls included LE-derived neurospheres cultured in FBS+BDNF+RA. LE-derived neural progenitors in co-culture conditions (B, lane 3), expressed transcripts corresponding to regulators (Nrl, Crx) and markers (opsin, rhodopsin kinase, arrestin, IRBP). These transcripts, were either undetectable or detected at much lower levels (opsin, rhodopsin kinase and arrestin) in controls; freshly dissociated LE cells (B, lane one) and LE-derived neurospheres cultured in the presence of FBS+BDNF+RA (B, lane 2). Immunocytochemical analysis of BrdU-exposed LE-derived neurospheres in co-culture conditions revealed the presence of BrdU-tagged cells expressing rod photoreceptor-specific immunoreactivities, opsin (C–F) and rhodopsin kinase (G–J). The graph represents the proportion of cells expressing photoreceptor-specific markers in co-culture and control conditions (K). Neurospheres, obtained from the LE cells of GFP mice and similarly co-cultured, displayed the presence of GFP+ and rhodopsin kinase+ cells suggesting that the acquisition of rod photoreceptor phenotype is not species-specific (L, M). Neurospheres, obtained from the LE cells of GFP-Nrl mice and similarly co-cultured, displayed the presence of GFP+ cells suggesting that the acquisition of rod photoreceptor phenotype involved the activation of Nrl(N, O). GFP-expressing cells in LE-derived neurospheres were transplanted intravitreally in PN7 rats eyes and the outcomes were examined one weeks later. A subset of GFP-expressing cells expressed immunoreactivities corresponding to opsin (P–T). Abbreviations: GFP, green fluorescent protein; PN1, postnatal day 1.

However, the efficiency of directed differentiation was less than 20% in vitro. To get an insight into signaling pathways that may allow us to increase the efficiency of photoreceptor differentiation, we examined the transcription profile of LE-derived neural progenitors in co-culture and control conditions, using custom-made microarray chips that contained sequences corresponding to 1,000 known genes, encoding components of disparate signaling pathways and regulators and markers of retinal progenitors and specific retinal cell types. Genes, whose expression was activated in LE-derived neural progenitors under the influence of PN1 retinal cells included photoreceptor regulatory genes (i.e., Crx and NeuroD) and photoreceptor-specific genes (i.e., Arrestin and Opsin) (Fig. 5A). The expression of genes expressed in LE (i.e., α-enolase) were observed to be decreased under the influence of PN1 retinal cells. Interestingly, the expression of genes encoding the activator of Shh pathway (i.e., Shh and Gli) and negative regulators of Wnt pathway (i.e., Dkk1, sFRP2 and sFRP3) was enhanced, suggesting that one of the mechanisms by which PN1 rat retinal cells induce photoreceptor differentiation may include the activation of Shh pathway and inhibition of Wnt pathway. To test this notion further, we co-cultured LE-derived neural progenitors with PN1 cells in the presence of Shh and Dkk1 to further enhance and inhibit Shh and Wnt signaling, respectively. We observed that the proportion of cells expressing rhodopsin kinase immunoreactivities increased significantly (∼2-fold; p < .001) when LE-derived neural progenitors were co-cultured with PN1 cells in the presence of Shh and DKK1, compared to controls (PN1 cells minus Shh and Dkk1) (Fig. 5B, 5C, 5D). The increase in the proportion of rhodopsin kinase positive cells was accompanied by increase in the levels of rhodopsin kinase and opsin transcripts in the former (Fig. 5E). The specificity of the influence on Wnt and Shh pathways was demonstrated by decrease and increase in the levels of transcripts corresponding to Lef1 and Gli, respectively. In addition, we observed that changes in Wnt and Shh signaling led to a decrease and an increase in levels of transcripts corresponding to Cyclin D1 and P27Kip1, respectively, suggesting that these signaling pathways influence photopreceptor differentiation by promoting cell cycle exit. Together, these observations confirmed the involvement of Shh and Wnt signaling in rod photoreceptor differentiation, and therefore, identify an approach to increase the efficiency of photoreceptor differentiation.

Figure Figure 5..

Differentiation of LE-derived neurospheres along rod photoreceptor lineage is regulated by Wnt and Shh signaling. Expression of genes, whose levels appeared modulated by PN1 retinal cells in LE-derived neurospheres in a focused transcription profiling, was confirmed by RT-PCR analysis (A). Transcripts corresponding to LE (a) and photoreceptor regulators and markers (b) decreased and increased, respectively, in co culture condition (lane 2), compared to controls (lane 1). Transcripts corresponding to Shh pathway (c) and negative regulators of the Wnt pathway (e) increased and decreased, respectively, in co culture condition (lane 2), compared to controls (lane 1). LE-derived neurospheres were co-cultured in the presence of Shh and Dkk1. Immunocytochemical analysis revealed an increase in the proportion of rhodopsin kinase (RK)-positive cells in LE-derived neurospheres co cultured in the presence of Shh and Dkk, compared to controls (B–D). RT-PCR analysis revealed an increase in the levels of transcripts corresponding to opsin and RK in Shh and Dkk1 containing culture (lane 2), compared to controls (lane 1) (E). Levels of transcripts corresponding to Gli/P27Kip1 and Lef1/CyclinD1 increased and decreased, respectively, in Shh and Dkk1 containing culture (lane 2), compared to controls (lane 1). Data are expressed as mean ± SEM from triplicate culture of three different experiments (D).


Corneal epithelium, like other surface epithelium, is renewed throughout the life. The regeneration is supported by stem cells/progenitors in the basal layer of the surrounding LE. We have previously demonstrated that these cells, when removed from their limbal niche, generated cells with self-renewing and neural potential [13]. We observed that these properties were characteristic of LE cells and not of those obtained from the stroma. In addition, the freshly dissociated LE cells did not express retinal (i.e., Rx and Chx10) and neural (i.e., Nestin and Musashi) progenitor markers, demonstrating that the observed properties were not due to the contamination with neural progenitors of retinal or extra-retinal origin. Our observations suggested that the expression of neural properties and potential by LE stem cells/progenitors in vitro is a highly regulated phenomenon, involving BMP and Notch signaling. For example, Noggin-mediated inhibition of BMP signaling promoted neural properties, the expression of which was preceded by the activation of Notch signaling [13]. These observations and the fact that in vitro expanded LE cells, obtained from healthy eyes, have been successfully used to treat corneal diseases and injury [43, [44]–45] make LE a practical source of neural progenitors for autologous stem cell therapy for neurodegenerative diseases. The study presented here demonstrates an incremental progress towards such an approach. The first and the most critical step towards the utilization of cells obtained from heterologous sources for cell therapy is the ability of the derived neural progenitors to differentiate into neurons not only by apparent features such as the expression of a set of specific markers but display of functional maturity. We provide proof, based on electrophysiological, calcium imaging and ultrastructural analyses that LE-derived neural progenitors can differentiate into neurons with functional properties. Signaling by neurons in the retina, like other parts of the CNS, relies heavily on fast glutamatergic transmission involving KA/AMPA receptors and thus every class of retinal neuron besides photoreceptors possess KA/AMPA receptors. By contrast, cells of non-neuronal lineage rarely exhibit these receptors. Fast sodium currents are commonly found in retinal ganglion cells and amacrine cells but have also been reported in some bipolar cells and cultured human photoreceptors [46, [47]–48]. Fast sodium currents are also commonly encountered in cells derived from retinal neural stem cells [1, 35]. Responses to KA and the presence of fast sodium currents in LE-derived neurons are thus consistent with the electrophysiological phenotype of many but not all classes of retinal neurons. Such properties and the indications that these cells form synapses in vitro predict the likelihood of their functional integration into host circuitry upon transplantation.

The next important step was to demonstrate that the LE-derived neural progenitors could be directed to differentiate into specific neuronal types for potentially replacing them in specific degenerative diseases. The degeneration of photoreceptors is at the heart of devastating blinding diseases such as AMD and RP. We observed that the LE-derived neural progenitors could be directed to differentiate along rod photoreceptor lineage in response to specific culture conditions. Their differentiation into cells expressing photoreceptor-specific markers is enhanced by their exposure to neonatal cells, which elaborate inductive activities that can promote photoreceptor differentiation in heterologous somatic stem cells [36] and ES cells [37]. The inductive influence of neonatal retinal cells can be observed in vivo where transplanted LE-derived neural progenitors incorporate in the outer nuclear layer and express photoreceptor-specific marker, opsin. Whereas the LE-derived neural progenitors show potential for photoreceptor differentiation both in vitro and in vivo, it does so with a low efficiency. Transcriptional profiling of LE-derived neural progenitors under the inductive influence neonatal retinal cells suggests that the process of photoreceptor-specific differentiation may involve Shh and Wnt signaling, therefore, manipulation of these signaling pathways may be used as strategies to enhance the efficiency of directed photoreceptor differentiation. Both Wnt and Shh pathways regulate cell proliferation in the retina; Wnt signaling promotes proliferation of both early and late retinal progenitors [49, 50] and Shh signaling direct their exit from cell cycle thus promoting their differentiation [51, 52]. Therefore, it is likely that the inhibition of Wnt signaling by Dkk1 and accentuation of Shh signaling by Shh act synergistically to direct LE-derived retinal progenitors to exit cell cycle thus facilitating their differentiation into rod photoreceptor under the influence of PN1 retinal cells. The efficiency of in vivo differentiation along photoreceptor lineage is related to the mode of transplantation (intravitreal vs. subretinal), incorporation within the retinal lamina and migration to the outer nuclear layer, where the transplanted cells are exposed to lamina-specific cues. Although, intravitreally transplanted progenitors incorporate in the host retina only ∼15% of incorporated cells on average reach the outer nuclear layer [42]. This could be enhanced by subretinal transplantation, where transplanted cells have to migrate shorter distance to reach the outer nuclear layer. Those that are incorporated in the outer nuclear layer express rod photoreceptor-specific markers but do not seem to show mature phenotype that includes elaboration of outer segments, a characteristic of photoreceptors. The lack of the expression of mature phenotype could be attributed to the short period (7 days) of examination of the transplantation outcomes and /or incomplete commitment of LE-derived neural progenitors along photoreceptor lineage before transplantation. It has been shown that progenitors that are pre-committed along rod lineage differentiate into rod photoreceptors in the host retina and display phenotype maturity [53].


In summary, our study identifies the LE as a putative renewable source of neural progenitors for an autologous cell therapy approach. However, there are several outstanding issues that need to be addressed before considering this approach for clinical application. First is the issue of the expansion of LE-derived neural progenitors to obtain cells in sufficient quantity to sustain clinical use. This issue could be addressed taking advantage of the fact that the rat LE-derived neural progenitors display robust self-renewing potential, could be passaged easily, frozen, and re-expanded [13] (Zhao and Ahmad, unpublished observation). This property has yet to be examined in human limbal epithelial-derived neural progenitors. Second is the issue of efficiency and specificity of directed differentiation. The efficiency of differentiation along photoreceptor lineage is low and we do not know what other cell types are generated under the inductive influence. The former can be addressed by changing the milieu such as by modulating Shh and Wnt signaling to improve the efficiency of photoreceptor differentiation and transplantation of progenitors pre-committed along the rod photoreceptor lineage [53]. The latter can be addressed by a comprehensive analysis of the expression of different cell-type-specific markers. Lastly, although our results demonstrate the engraftment and survival of induced progenitors in host retina, their functional status, remains to be studied, along with the long-term effects of transplantation on function. Our study suggests, based on observations of successful treatment of corneal injury using autologous LE stem cells [44], the possibility of autologous stem cell therapy in which in vitro expanded LE stem cells, obtained from 2 mm of LE tissue obtained from patients' eyes, could be used to rescue degenerative changes in diseased retina, such as in age-related macular degeneration. The concept could be easily adapted for treating other neurodegenerative diseases, such as the Parkinson's disease. In addition, to offering another approach for autologous cell therapy, the LE-derived neural progenitors can be posited as a robust and naïve in vitro model system for drug and gene discovery.


The authors indicate no potential conflicts of interest.


We thank Drs. Anand Swaroop and Mahendra Rao for Nrl-GFP and Sox2-GFP mice, respectively, and Ganapati Hegde for help in microarray analysis. This work was supported by the Lincy Foundation, Pearsons Foundation, Nebraska Tobacco Settlement Biomedical Research Fund and Research to Prevent Blindness.