Br J Ophthalmol 96:1431-1437 doi:10.1136/bjophthalmol-2012-301546
  • Original articles
    • Laboratory science

Characterisation of mouse limbal neurosphere cells: a potential cell source of functional neurons

  1. Andrew Lotery1,2
  1. 1Clinical and Experimental Sciences, Clinical Neurosciences Research Group, Faculty of Medicine, University of Southampton, Southampton General Hospital, Southampton, UK
  2. 2Southampton Eye Unit, Southampton General Hospital, Southampton, UK
  1. Correspondence to Professor Andrew Lotery, Clinical and Experimental Sciences, Clinical Neurosciences Research Group, Faculty of Medicine, University of Southampton, Southampton General Hospital, Southampton SO16 6YD, UK; a.j.lotery{at}
  • Accepted 23 July 2012
  • Published Online First 5 September 2012


Background/aims To characterise the origin, ultrastructure and functional properties of corneal limbal neurospheres (LNS).

Methods Limbal cells were isolated from the corneal limbus of adult mice and cultured in a serum-free sphere forming culture system. LNS were characterised by immunocytochemistry, Reverse-transcription-PCR and electron microscopy. LNS cells were also cocultured with neonatal mouse retinal cells. Phenotype and function were then assessed by immunofluorescence and a calcium influx/efflux assay.

Results LNS cells displayed clonal growth and self-renewal, and expressed a wide range of stem cell and neural lineage markers. The acquisition of neural properties was concordant with expression of neural crest markers including CD34, Sca1, Sox9, Twist1, but not CD45. LNS exhibited similar morphology and microstructure to neurospheres derived from the central nervous system. Following culture in a conducive environment, the derived cells displayed mature neural markers and exhibited electrical excitability.

Conclusions Corneal limbal stromal progenitor cells are a potential and convenient autologous cell source to generate functional neurons.


Cell-based therapies are an attractive approach to treat neurodegenerative diseases.1 Such therapies can, theoretically, result from the use of embryonic, induced pluripotent or adult-derived stem cells. However, deriving such therapies from embryonic/foetal tissues is difficult due to limited resources, ethical issues and risks, such as tumour formation or transplant rejection.2 Induced pluripotent stem cells (iPSC) are an alternative source of stem cells. They can be generated from a variety of somatic cell types via the introduction of transcription factors. However, the possibility of mutagenesis by viral vector integration or tumour formation remains.3 ,4 Therefore, because of these potential complications, adult-derived stem cells remain an attractive strategy for tissue regeneration and engineering.

One potential source of adult-derived stem cells is the corneal limbus. This is a transition zone which is 1.5 mm wide, lying between the conjunctiva/sclera and the transparent cornea. It is a readily accessible area, where the superficial layers are amenable to tissue harvesting. Within the corneal limbus are two populations of stem-like cells. These are putative epithelial stem cells,5 and multipotent stromal stem cells.6 ,7 Neural potential has been reported in both populations of stem/progenitor cells in vitro.6–8 Zhao et al demonstrated that a neurosphere assay (NSA) accompanied with inhibition of bone morphogenetic protein 4 (BMP4) signalling encouraged limbal epithelial progenitor cells to acquire neural properties.8 These neural sphere colonies (neurospheres) could subsequently differentiate into either functional neurons or cells which express photoreceptor-specific markers. Using the same NSA culture system, neural-crest origin neurospheres can be derived from the corneal/limbal stroma.6 ,7

The purpose of this paper is to characterise the neural potential, ultrastructure, origin, as well as functional properties of limbal neurospheres (LNS). Our data demonstrates that LNSs are derived from neural crest limbal stroma, rather than limbal epithelium or bone marrow (BM). We report the first ultrastructural characterisation of LNS. We demonstrate they are similar to neurospheres derived from the central nervous system, and that functional neuron-like cells can be derived from limbal stroma.

Materials and methods

Cell culture

Adult mouse (8-weeks-old, C57BL/6) corneal limbal cells were cultured as previously described.7 ,9 Following enucleation of the eyes and removal of the central cornea, a circular incision was made below the limbus to isolate the corneal limbal region. Tissue was then digested with 0.025% (w/v) trypsin/EDTA at 37°C for 10–12 min, and then in 78 U/ml of collagenase (Sigma–Aldrich, Irvine, Ayrshire, UK) and 38 U/ml of hyaluronidase (Sigma–Aldrich) for 30 min at 37°C.8 Dissociated cells were cultured in DMEM: F12GlutaMAX (Invitrogen, Paisley, Scotland, UK) supplemented with 2% B27 (Invitrogen), 20 ng/ml of epithelial growth factor (EGF, Sigma–Aldrich) and basic fibroblast growth factor (FGF2, Sigma–Aldrich) at a density of 1×104 for sphere forming efficacy, and 1×105 for cell expansion. Cells were cultured both with and without Noggin (100 ng/ml, R&D Systems, Abingdon, UK).8 To promote LNS neural differentiation, LNS cells were plated onto Poly-D-Lysine and laminin (Sigma–Aldrich) coated wells, and cocultured with dissociated postnatal 1–3 mouse retinal cells using Millicel-CM inserts (pore size 0.4 µm; Millipore, Watford, UK) for 1–2 weeks. Differentiation medium was Neurobasal A media (Invitrogen), 2% B27, 0.5 mM L-Glutamine 0.5%–1% fetal bovine serum, 1 µM retinoic acid (Sigma–Aldrich) and 1 ng/ml brain-derived neurotrophic factor (R&D system).


As described in Supporting Information Methods.

Calcium influx imaging

Calcium imaging was conducted as previously described.8 ,10 Cells were incubated for 30 min in Fruo-4 acetoxymethyl (AM) (4 μM, Invitrogen) in Hank's Buffered Salt Solution (HBSS) plus non-ionic detergent (10 μM; Pruronic F-127, Invitrogen) at 37°C. Prior to imaging, excess dye was washed out and cells were equilibrated in HBSS to allow the AM to be cleaved by cytoplasmic esterase. Cells were imaged at 20× magnification, and fluorescence was stimulated with light wavelengths near 488 nm and acquired near 520 nm. Images were acquired every 10–30 s, with 500 ms acquisition times by a cooled camera on an inverted microscope (Leica) and analysed using Volocity software. Intracellular calcium concentration was presented as fluorescence intensity (raw pixel intensities). Mean background intensity was subtracted from each cellular region to minimise the noise and vibration due to staining or fluorescence fading (F=cellular average-background average). All cells were then normalised by their initial intensity, respectively F/F0.10 ,11

Transmission electron microscopy

As described in Supporting Information Methods.

Reverse transcription-polymerase chain reaction

As described in Supporting Information Methods.

Statistical methods

All results are presented as mean±SEM (SE of the mean) unless otherwise stated; n represents the number of replicates. Statistical comparisons were made using a Student's t test, or one-way analysis of variance, with a significance threshold of p<0.05. GraphPad Prism Software (GraphPad) was used for statistical analysis.


Neurospheres from adult mouse corneal limbus display clonal growth and neural potential

The ability to form neurospheres in serum-free medium, in the presence of mitogens, was first described as a method to select and expand neural stem cells.12 We initially investigated whether neurospheres could be derived from the adult mouse corneal limbus. Neurosphere-like cell aggregates were generated in serum-free medium in the presence of EGF and FGF2. The size of spheres produced ranged from 50 µm to 150 µm in diameter (61.1±17.1 µm) after 7 days in vitro. Spheres had defined edges and a smooth surface (figure 1A). We then investigated whether sphere formation was due to clonal cell growth or cell aggregation. The growth curve demonstrated a selective growth characteristic, with approximately 90% of cells eliminated after 3 days in culture (figure 1C,D). Sphere formation began on day 5 and continued with a doubling time of 3–4 days (figure 1C). LNS cells could be passaged for over 3 months in these serum-free conditions (maximum observation time). Primary spheres were redissociated into single cells and plated at an extremely low density (5–10 cells/well in 100 μl medium) into 96-well plates, in order to completely avoid cell aggregation. Approximately 1% of primary sphere cells were capable of generating secondary neurospheres, indicating that LNS formation was due to clonal growth.

Figure 1

Generation of limbal neurospheres (LNS). (A) Sphere colonies were generated with defined edges from adult mouse corneal limbus cells. Scale bar: 100 μm. (B) Effect of extrinsic factor on LNS generation. Cell density: 10 000 cells/ml. Spheres with a diameter of over 50 μm were included for analysis. EGF: Epidermal Growth Factor; FGF2: basic Fibroblast Growth Factor. Results are expressed as mean±SEM (n≥4). Significant difference p<0.001 for EGF vs. each other condition, **p<0.001 EGF+FGF2 compared with FGF one-way analysis of variance followed by the Bonferroni multiple comparison test. (C) A growth curve of adult mouse corneal limbal neurosphere cells (n≥4). (D) Limbal cell numbers sharply decreased after 3 days in neurosphere culture (n=4) p<0.0001 (paired t test).

In the presence of specific mitogens, EGF and/or FGF2, serum-free media was able to support stem/progenitor cell proliferation, with gradual elimination of postmitotic non-proliferating mature cells.13 Noggin has been reported to assist acquisition of neural progenitor properties by limbal epithelial cells in neurosphere culture assays.8 We assessed the effect of EGF, FGF2 and the BMP4 inhibitor (noggin) on LNS generation (figure 1B). Sphere formation was dependent on FGF2 rather than EGF. A combination of both led to a significant increase in the sphere generation efficiency (p<0.01), suggestive of a synergetic effect between EGF and FGF2. A comparison of sphere generation in the presence and absence of noggin revealed no significant difference in efficiency (p>0.05). This implied that inhibition of the BMP pathway did not affect LNS generation.

The derived LNS cells expressed the stem cell marker ABCG2, neural stem cells markers Sox2 and Nestin and early differentiated neural marker Class III β-tubulin (Tuj1) (figure 2A,C–J). The proportions of nestin and TuJ1-positive cells within the LNS were 33±6% and 31±3%, respectively. Nestin-positive cells were distributed at the periphery of LNSs (figure 2J), similar to neurospheres derived from the central nervous system.14 Conversely, the epithelial stem cells marker P63 was found to be absent in LNS cells (figure 2B).

Figure 2

Characterisation of adult mouse limbal neurospheres (LNS). (A–E, H) Expression of ABCG2, Nestin, Sox2, Tuj1, but not P63, was detected in LNS cells. Nuclei were counterstained with DAPI (blue). Scale bar: 13 μm. (F, I) Quantification of Tuj1 and Nestin-positive cells (vs. total cells) (data represents three independent experiments). (G, J) Single optical section of intact LNS from Z-stack confocal images: nestin-positive progenitor cells (green) were located at the periphery of the sphere. Nuclei were counterstained with Sytoxin Orange. (K) Shows the bright field image of the same LNS. Scale bar: 50 μm. This figure is only reproduced in colour in the online version.

LNS cells are neural crest-derived progenitor cells

To eliminate the possibility that generation of LNS was due to contamination by adjacent iris pigment epithelial cells, the same dissociation and culture conditions were applied to iris-derived cells. No sphere formation was detected after 2 weeks in vitro (figure 3B–C). To further determine the origin of LNS cells, gene expression was assessed during LNS formation (figure 3A). Reverse-transcription PCR confirmed expression of neural stem/progenitor markers (Sox2, Nestin, Musashi1, TuJ1 and CD133). No neural lineage markers were detected during the first 3 days in this NSA. Neural potential was acquired following LNS generation and subculture. Conversely, the corneal epithelial lineage markers P63 and K12 gradually decreased and/or disappeared during culture (figure 3A, supplementary figure S1). To determine whether this shift in cell lineage markers was due to cell transformation or selective growth of other local cells, we investigated the expression of neural crest markers including Sox9, CD34, Twist1, Sca1 and the mesenchymal stem cell marker Vimentin (figure 3A). The concomitant increase in neural potential and neural crest marker expression suggests that LNS cells are neural crest-derived limbal stromal cells. LNS cells were negative for CD45, a protein tyrosine phosphatase located in haematopoietic cells. This implied that LNS did not originate from migrated BM cells.7 ,15 The BMP4 inhibitor Noggin, has been reported to assist limbal epithelial cells acquiring neural properties.8 Therefore, we assessed the effect of Noggin on gene expression during LNS generation. The presence or absence of Noggin had no effect on gene expression of epithelial or neural markers. In addition, neural crest markers remained abundant throughout LNS generation.

Figure 3

Origin and ultrastructure of limbal neurospheres. (A) Gene expression during sphere culture in the presence and absence of Noggin. Day 0: Freshly isolated limbal cells after a two-step enzyme digestion. Day 1–3: No spheres detected; Day 5: Primary spheres detected; Day 10: Secondary spheres. (B, C) Homogenous adult mouse iris pigment epithelium and corneal limbus cells. Cells from adult mouse iris and corneal limbus isolated by the same dissociation process and maintained in the same culture conditions. (B) No obvious spheres were formed from adult mouse iris pigment epithelium; cells remained heavily pigmented. (C) Sphere colonies were generated with defined edges from adult mouse corneal limbus cells. Scale bar: 100 µm. (D–F) Transmission electron micrographs of adult mouse corneal limbal spheres. The cell showed dissolved nuclear membrane indicative of cell division (arrow, D). Gap junctions were detected between cells within spheres (arrow, E). Adherence-like junctions were also detected between cells within spheres (arrow, F). Microtubulin (MT), processes (P), rough endoplasmic reticulum (RER). Scale bar: 500 nm.

Ultrastructure of neurospheres derived from adult corneal limbus

Spheres derived from the adult mouse corneal limbus were three-dimensional in structure. Cells within spheres were connected with each other through gap junctions (figure 3E) and adherence-like junctions (figure 3F). Cell division within spheres was apparent (figure 3D), as was the presence of vesicles and abundant rough endoplasmic reticulum (figure 3F), indicative of active cellular mitosis, exocytosis/endocytosis and protein synthesis. Spheres also contained cells with processes at the periphery, while melanosomes were not detected in the sphere cells.

Functional neural-like cell derived from adult LNS cells

Cell inserts with a semipermeable membrane were utilised to avoid cell contamination, while still allowing the passage of diffusible factors (figure 5B). Following withdrawal of mitogens and culture on P-D-L and laminin-coated tissue culture plates, LNS cells formed a monolayer, with cells displaying neural morphologies, as demonstrated in figure 5A. After 7–10 days coculture, 31±10% of differentiated LNS cells were immunopositive for neurofilament 200, a mature neuronal marker. Approximately 8%–10% of cells exhibited strong immunoreactivity to Syntaxin3, a major component of synapses within the retina (figure 5).16 Syntaxin3 protein also has an important role in the growth of neuritis in the brain.17

Figure 5

Neural properties of limbal neurospheres (LNS) cells. LNS cells displayed a neural morphology following one week in coculture with neonatal retinal cells (A). Phase contrast images show cells with apparent neurite-like processes, small nuclei, bipolar shape (A left, D arrows) or multiple dendrites-like processes (A right). The schematic diagram illustrates the coculture system using semipermeable Millicel CM inserts to promote LNS differentiate towards neural-like cells (B). Cells were stained with antibodies directed against Neurofilament 200 (C–D) and Syntaxin3 (E–F). Cell nuclei were counterstained in blue with DAPI. Scale bar: A,C,D: 26 μm; E, F: 13 μm. This figure is only reproduced in colour in the online version.

Electrical excitability is one of the characteristic of neural lineage cells. Therefore, we investigated the calcium influx/efflux of induced limbal cells using the calcium indicator Fluo-4.8 ,10 A significant influx of free calcium was evoked by voltage stimulus, the high concentrations of K+, indicating the existence of voltage-gated calcium channels in the induced limbal cells (figure 4 left). Neonatal mouse retinal cells cultured in vitro also displayed the same reactivity (figure 4 right). Approximately 60% of induced limbal cells demonstrated excitability. A quick and significant influx of calcium was observed upon stimulation. The intracellular free calcium level gradually efflux after the peak, reaching to the prestimulation level at approximately 300 s following stimulation.

Figure 4

Limbal neurospheres cells displayed electrical excitability after coculture with neonatal retinal cells in vitro. Intracellular calcium changes evoked by K+ (100 mM) in cocultured limbal cells and neonatal retinal cells. Fluo-4 was used as a Ca2+ indicator. Both differentiated limbal cells and neonatal retinal cells showed an increase in intracellular calcium in response to a depolarizing stimulus, while control conditions (Hank's Buffered Salt Solution) did not induce calcium flux. Images show free calcium (green fluorescence) of individual cells before and after K+ stimulation. Graphs depict changes in fluorescence intensity (F/F0) of individual cells (n≥12). These results demonstrate that both cells types are excitable. This figure is only reproduced in colour in the online version.


This study demonstrates that stem/progenitor cells with neural potential can be derived from the adult corneal limbus and expanded in vitro. The resulting neural colonies (neurospheres) appear to be neural crest-derived limbal stromal stem/progenitor cells (NC-LSPC), instead of transformed limbal epithelial or BM-derived cells. For the first time, this work demonstrates that functional neural-like cells can be derived from neural crest-derived limbal stromal stem/progenitor cells.

The derivation and expansion of stem/progenitor cells in vitro is critical for cell-based therapy. We show that robust passaging and expansion can be achieved through a neurosphere-forming assay, a bona fide serum-free culture system. From a typical (1×3 mm) limbus, approximately 300 primary neurospheres (diameter ≥50 µm) can be generated after 7 days in vitro. LNS cells can be passaged and give rise to further generations of spheres. Thus, with a very limited amount of limbal tissue, sufficient progenitor cells for further manipulation and/or transdifferentiation can be generated.

It has been reported that LNS cells originate from the limbal epithelial layer,8 ,9 the stromal layer,6 ,7 and also from migrated BM cells.18 In this study, we show that inhibition of BMP4 does not change the gene expression profile or affect the production of LNS. In addition, we show that neural crest markers are upregulated in synchrony with expression of neural lineage markers. While a downregulation of epithelial lineage markers is concurrent with elimination of over 90% cells in the first 3 days of culture. These observations suggest that LNS generation is from neural-crest origin corneal limbal stromal progenitor cells. Several studies have demonstrated that BM-derived cells can migrate to the corneal/limbal stroma, with expression of the haematopoietic stem cell marker (CD34) and leukocyte marker (CD45) in the stromal layer of the cornea and limbus, but not in the epithelial or endothelial layers.7 ,18 Here, we show that limbal-derived neurospheres express CD34, but are negative for CD45. This indicates that LNS are stromal keratocyte progenitor cells. This supports previous studies showing that corneal/limbal stromal keratocytes are multipotent neural crest stem cells capable of differentiating into a neural lineage.7 ,14 ,15

We compared the ultrastructure of LNS with well documented neurospheres derived from the CNS19 and the ciliary body.20 Gap junctions were observed while tight and desmosome junctions, typical of spheres of an epithelial origin, were absent. This suggests that sphere formation is unlikely to be due to reprogramming of epithelial cells to a neural lineage.6 ,7 ,15

Consistent with previous reports, our neural crest-derived LNS cells are immunopositive for a range of neural lineage markers.6 ,7 ,15 ,21 In this study, we confirm that following neural differentiation, a subpopulation of cells display typical neural morphology and expressed mature neural markers (neurofilament 200). A synaptic component (Syntaxin3) was also expressed. Furthermore, we demonstrate, for the first time, that induced LNS cells possess voltage-gated calcium channels capable of electrical excitability.

Our data highlights a promising new candidate progenitor cell source for neural-like cell generation. The corneal limbus is one of the most accessible regions of the human eye. The stromal layer from the peripheral cornea and limbus is clear of the visual axis. It represents 90% of the thickness of the front eye wall. Hence, cells can be readily harvested from this area without compromising ocular function. In addition, LNS are adult cells, and so the risk of tumour formation is significantly reduced. To date, adult corneal/limbal stromal stem cells with neural potential have been successfully derived from a number of species including humans,6 cows,22 rabbits23 and mice.7 ,15 This suggests that the existence of limbal stromal stem/progenitor cells is not species specific. Our data, for the first time, shows derivation of functional neuron-like cells from adult corneal limbal stroma. It expands the concept of plasticity of neural crest-derived stromal progenitor cells. The expression of retinal synaptic component (Syntaxin3) on LNS cells, suggests further research is warranted in order to promote their differentiation towards retinal-specific phenotypes. If successful, this could provide a useful cell source for retinal cell therapies.


We thank David Johnson for instruction and help on confocal imaging; Dr Anton Page for TEM processing and analyses. This work was supported by the National Eye Research Centre, the Gift of Sight Appeal and Rosetrees Trust.


  • Contributors XC designed the experiments, undertook the laboratory work, analysed the data and co-wrote the manuscript. HT supervised and helped with cell culture, immunocytochemistry; data analysis and co-wrote the manuscript. PH supervised and helped with data analysis and interpretation. AL designed and supervised the research and co-wrote the manuscript.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.


Relevant Article

Register for free content

Free sample
This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of BJO.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.