We report the presence of neural crest-derived corneal precursors (COPs) that initiate spheres by clonal expansion from a single cell. COPs expressed the stem cell markers nestin, Notch1, Musashi-1, and ABCG2 and showed the side population cell phenotype. COPs were multipotent with the ability to differentiate into adipocytes, chondrocytes, as well as neural cells, as shown by the expression of β-III-tubulin, glial fibrillary acidic protein, and neurofilament-M. COP spheres prepared from E/nestin-enhanced green fluorescent protein (EGFP) mice showed induction of EGFP expression that was not originally observed in the cornea, indicating activation of the neural-specific nestin second intronic enhancer in culture. COPs were Sca-1+, CD34+, CD45−, and c-kit−. Numerous GFP+ cells were observed in the corneas of mice transplanted with whole bone marrow of transgenic mice ubiquitously expressing GFP; however, no GFP+ COP spheres were initiated from these mice. On the other hand, COP spheres from transgenic mice encoding P0-Cre/Floxed-EGFP as well as Wnt1-Cre/Floxed-EGFP were GFP+, indicating the neural crest origin of COPs, which was confirmed by the expression of the embryonic neural crest markers Twist, Snail, Slug, and Sox9. Taken together, these data indicate the existence of neural crest-derived, multipotent stem cells in the adult cornea.
The cornea is an avascular, structurally unique tissue that functions as the primary refracting medium of the eye. Although anatomically continuous with the vascularized sclera and conjunctiva, all three major components of the cornea function together to maintain optically clear tissue. Therefore, homeostasis of the corneal epithelium, stroma, and endothelium—the cellular components of the cornea—is vital in preserving transparency and optical precision.
Stem cell researchers of the cornea have identified the epithelial stem cell to be located in the vascular rim, or limbus, of the cornea . In contrast, there is little evidence of the existence of stem/progenitor cells for keratocytes [2, 3], the resident cells of the corneal stroma. Keratocytes, mesenchymal cells distinct from keratinocytes of the skin, repopulate the corneal stroma during tissue remodeling after its depletion due to disease, such as herpes simplex virus infection, and trauma [4, 5]. Although the stroma of the cornea develops from the cranial neural crest [6, 7], the origin of keratocytes involved in the turnover of stromal tissue is unknown.
We have previously demonstrated that the neurosphere culture technique, which was originally developed for neural stem cells (NSCs) isolated from the forebrain of mouse , can be adapted to culture mouse cornea stromal cells for more than 15 passages while still maintaining the keratocyte phenotype . A recent report demonstrated that multipotent precursor cells from adult mouse and human dermis, termed skin-derived precursor cells (SKPs), also form spheres and differentiate into neural and mesenchymal cells [10, –12]. We therefore hypothesized that the corneal stroma-derived spheres we have isolated also include putative keratocyte precursor cells similar to SKPs of the skin.
Here, we show the existence of multipotent keratocyte precursor cells (termed COPs, for cornea-derived precursors) in cornea stromal spheres isolated from adult mice. Single cells dissociated from spheres initiated clonal growth of progeny spheres, and a subset of COPs exhibited the side population (SP) cell phenotype. We sought to determine whether COPs were of bone marrow (BM) origin or of neural crest lineage by initiating COP spheres in various transgenic mice.
Materials and Methods
Normal, specific pathogen-free, adult C57BL/6J mice were purchased from CLEA Japan, Inc., Tokyo, http://www.clea-japan.com/index.html). Green fluorescent protein (GFP) transgenic mice (C57BL/6 TgN [act-enhanced GFP (EGFP)] OsbC14-Y01-FM131) were obtained from the Genome Information Research Center (Osaka University, Osaka, http://www.gen-info.osaka-u.ac.jp/welcome_en.html). Transgenic mice expressing Cre recombinase under the control of the Wnt1 promoter/enhancer (Wnt1-Cre mice)  and P0 promoter (P0-Cre mice)  were mated to CAG-CATloxP/loxP-EGFP (CAG-CAT-EGFP) transgenic mice  to obtain Wnt1-Cre/CAG-CAT-EGFP (Wnt1-Cre/Floxed-EGFP) and P0-Cre/CAG-CAT-EGFP (P0-Cre/Floxed-EGFP) transgenic mice, respectively. P0-Cre transgenic mice and CAG-CATloxP/loxP-EGFP transgenic mice were obtained from Dr. Ken-ichi Yamamura and Dr. Jun-ichi Miyazaki, respectively. All animal procedures were performed in accordance with institutional guidelines.
Cells were dissociated from adult C57BL6/J mice and then cultured as described previously . All animals were handled in full accordance with the ARVO (Association for Research in Vision and Ophthalmology) Statement for the Use of Animals in Ophthalmic and Vision Research. In brief, corneal stromal discs were cut into small pieces and digested in 0.05% trypsin (Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) for 30 minutes at 37°C, followed by 78 U/ml collagenase (Sigma-Aldrich) and 38 U/ml hyaluronidase (Sigma-Aldrich) treatment for 30 minutes at 37°C. Stromal cells were mechanically dissociated into single cells and cultured in Dulbecco's modified Eagle's medium (DMEM)/F-12 (1:1) supplemented with 20 ng/ml epidermal growth factor (EGF) (Sigma-Aldrich), 10 ng/ml fibroblastic growth factor 2 (FGF2) (Sigma-Aldrich), B27 supplement (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), and 103 U/ml leukemia inhibitory factor (Chemicon International, Temecula, CA, http://www.chemicon.com) at a density of 1 × 105 cells per milliliter at 37°C, 5% CO2.
For clonal sphere expansion, COPs were initiated from corneas of wild-type C57BL/6 strain and transgenic strain expressing GFP ubiquitously . Cells dissociated from COPs were plated on six-well dishes at a cell density of 5 × 103 cells per milliliter and cultured for 6–7 days in DMEM/F-12 containing 0.8% methylcellulose with EGF, FGF2, and B27 supplement. The use of methyl cellulose in the clone culture of NSCs (neural spheres) is an established method reported by several groups [17, , , , , , –24].
To examine the expression of nestin in COPs, cells were prepared from transgenic mice carrying EGFP (Clontech, Mountain View, CA, http://www.clontech.com) under the control of the second intronic enhancer of the nestin gene, which acts selectively in neural stem/precursor cells (E/nestin-EGFP) . To confirm the neural crest origin of COPs, corneal stromal cells were prepared from six corneas of neonatal (13 days) and three corneas of adult (8 weeks) P0-Cre/Floxed-EGFP mice, as well as from one cornea of an adult (10 weeks) Wnt1-Cre/Floxed-EGFP mouse and cultured as described above.
In Vitro Differentiation
To examine neural differentiation, COPs were dissociated into single cells and suspended at a cell density of 10 cells per milliliter. One-hundred microliters of the cell suspension was divided into 48-well culture plates, and only clonal spheres from single cells were subcultured and expanded. Clonal COPs were plated and cultured on poly(l-ornithine)/laminin-coated Lab-Tek chamber slides (Nalge Nunc International, Rochester, NY, http://www.nalgenunc.com). For α-smooth muscle actin (α-SMA) expression, clonal cells were also plated on Lab-Tek chamber slides in transforming growth factor (TGF)-β-containing medium. For adipogenic or chondrogenic differentiation, dissociated COPs were cultured in differentiation-inducing medium (Cambrex Bio Science Walkersville, Inc., Walkersville, MD, http://www.cambrex.com) according to instructions provided by the manufacturer. To visualize adipogenic differentiation, cells were stained with oil red O (Sigma-Aldrich). Chondrogenic differentiation was observed by the expression of the specific markers collagen II and aggrecan analyzed by immunocytochemistry of cell pellets (see above). Results were expressed as mean ± SD.
Cultured cells and frozen-tissue sections were fixed with 4% paraformaldehyde (PFA) for 10 minutes at room temperature and then stained with the following antibodies: anti-Bcrp1 (1:250; R&D Systems, Minneapolis, http://www.rndsystems.com), anti-α-SMA (1:200; NeoMarkers, Fremont, CA, http://www.labvision.com), anti-collagen type II (1:40; Chemicon International), anti-aggrecan (1:40; Chemicon International), anti-Musashi-1 (Msi1) (1:500, clone 14H1) , anti-class III β-tubulin (1:100, R&D Systems), anti-glial fibrillary acidic protein (GFAP) (1:200; Chemicon International), and anti-neurofilament-M (NF-M) (1:500; Abcam, Cambridge, U.K., http://www.abcam.com). Immunohistochemistry for GFP was performed using an anti-GFP antibody (1:500; MBL, Nagoya, Japan, http://www.mbl.co.jp) in 10-μm frozen sections from eyes fixed in 2% PFA overnight at 4°C. Immunoreactivity of primary antibodies was visualized using secondary antibodies conjugated with fluorescein isothiocyanate or cyanine 3 (Jackson ImmunoResearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com).
Reverse Transcription-Polymerase Chain Reaction
COPs and cells freshly dissociated from mouse corneal stroma were collected and immediately frozen in liquid N2. cDNA was synthesized using a commercially available kit (Life Sciences, Inc., St. Petersburg, FL, http://www.lifesci.com) from total RNA prepared using RNeasy kit (Qiagen, Hilden, Germany, http://www1.qiagen.com). Primers used for Abcg2, Notch1, nestin, Msi1, Twist, Snail, Slug, Sox9, and Gapd are shown in Table 1 (supplemental online data). Polymerase chain reaction (PCR) was performed using GeneAmp 9700 (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com).
Table Table 1.. Polymerase chain reaction primers
For Hoechst dye efflux assays, single cells dissociated from COP spheres were incubated with Hoechst 33342 dye (Dojindo Laboratories, Kumamoto, Japan, http://www.dojindo.com) for 60 minutes at 37°C in the presence or absence of 50 μM reserpine (Daiichi Pharmaceutical, Tokyo, http://www.daiichius.com). SP cells were gated using FACS Vantage (BD Biosciences Immunocytometry Systems, San Jose, CA, http://www.bdbiosciences.com) as described previously . Surface marker expression was also analyzed by flow cytometry using antibodies for CD45, CD34, Sca-1, c-kit, and CD133 (eBioscience, San Diego, http://www.ebioscience.com). Isotype-matched immunoglobulin G was used as negative control.
Whole BM (WBM) cells (1 × 106 cells) were prepared from GFP-transgenic mice  and transplanted into the retro-orbital space of C57BL6/J recipient mice treated with a lethal dose (10.3 Gy) of irradiation. Eight weeks after transplantation, recipient mice were sacrificed, and corneal stromal cells were prepared for sphere culture. Cells from the transplanted animals and nonirradiated animals were then mixed and cultured as described above to assess WBM-derived cell contribution to COP sphere formation.
COPs Initiate Clonal Sphere Formation
Mouse corneal stromal-derived spheres were first prepared and cultured as described previously . To determine whether spheres arise from single putative COPs or from aggregates of floating cells, we first performed the clonal sphere-forming assay . As shown in Figure 1, homogeneous GFP-positive or -negative spheres were found 6 days after plating. More than 70% of spheres were homogenous; however, nonclonal spheres composed of GFP-positive and -negative cells were also observed. The same observation was made by Kawase et al. , who reported that SKPs may aggregate at an initial cell density of 1 × 103 cells per milliliter during sphere cultures. Clonal sphere formation was observed for several passages (P5), suggesting that COPs possess high “self-renewing” potential.
COPs Differentiate into Neural and Mesenchymal Lineage Cells
COPs differentiate into keratocytes when cultured on plastic (Fig. 2A), into fibroblasts when cultured with serum, and into myofibroblasts under TGF-β stimulation (Fig. 2B) . To determine whether these cells possess the ability to differentiate into other cells of mesenchymal lineage, COPs were cultured in various differentiation-inducing media. After 10 days of culture in medium containing insulin, approximately 7.9% (mean, n = 2) of the cells differentiated into oil red O-positive lipid-producing adipocytes (Fig. 2C). In addition, cell pellets were formed when cells were cultured in chondrogenic-inducing medium containing TGF-β3 (n = 9) (Fig. 2D). Immunofluorescent staining showed expression of the chondrocyte markers, type II collagen and aggrecan  in the pellets (Fig. 2E, 2F).
The NSC marker Msi1, an RNA-binding protein involved in the maintenance of NSCs and activation of Notch signaling [26, 29, 30], was expressed in COP spheres (Fig. 3A, 3C). COP spheres also expressed the NSC markers nestin [25, 31] and Notch1 (Fig. 3A); the latter is a gene involved in the self-renewal of various types of tissue stem cells, including NSCs . Because Nestin is an intermediate filament expressed by several cell types , COP spheres were prepared from E/nestin-EGFP transgenic mice, which carry the EGFP transgene under the control of a NSC-selective enhancer . As expected, EGFP-positive spheres were formed (Fig. 3B) from these mice, which originally did not show EGFP-related fluorescence in the cornea.
Neural differentiation of COPs was shown by the expression of class III β-tubulin, GFAP, and NF-M in cells cultured on poly(l-ornithine)-coated slides in differentiation-inducing medium (Fig. 3E, 3F). Approximately 1.4% of cells stained with anti-NF-M antibody (n = 3), 36.9% ± 17.7% expressed GFAP (n = 3), and 32.8% ± 15.8% expressed β-III-tubulin (n = 3).
COP Spheres Are Rich in SP Cells
Several studies have shown that the ability to exclude Hoechst dye is a property of stem cells commonly referred to as SP cells , which are distinguished from the “main population” by flow cytometric analysis. The SP cell phenotype is defined by the dye exclusion ability of an ABC transporter, ABCG2, which is inhibited by ABC-transporter inhibitors such as reserpine. We found that COP spheres expressed ABCG2 when examined by reverse transcription (RT)-PCR and immunocytochemistry (Fig. 4A, 4B). Reserpine-sensitive SP cells were detected in dissociated sphere cells, representing 3.3% ± 1.2% (n = 8) of viable cells analyzed by flow cytometry (Fig. 4C).
We also analyzed several stem cell-related surface markers by flow cytometry. COP spheres expressed CD34 (Fig. 5A, 5B), a cell surface marker reported in rodent epithelial stem cells in the bulge area [35, 36], skeletal muscle stem cells [37, 38], and corneal stromal cells [9, 39]. In addition, expression of stem cell antigen-1 (Sca-1), a cell surface protein expressed in BM-derived hematopoietic stem cells (BM-HSCs) , mammary epithelial stem cells , a subpopulation of BM stromal cells , skeletal muscle stem cells , and SKP spheres , was found in 56.1% ± 19.2% (n = 7) of viable cells (Fig. 5). The expression of CD133, found in different types of primitive cells such as BM-HSCs, NSCs, and SKPs [43, , –46], was not observed (data not shown). Another cell surface marker, c-kit (CD117), the receptor for stem cell factor and a marker of BM-HSCs , was also not detected by flow cytometric analysis (Fig. 5) and RT-PCR (not shown).
COPs Are Neural Crest Lineage Cells
Although we found CD34+ cells in COP spheres, Sosnova et al.  reported that all CD34+ cells in mouse corneal stroma are CD45+ BM-derived cells. In addition, the ability of BM-derived mesenchymal stem cells (BM-MSCs) to differentiate into multiple cell types has been reported [49, 50]. However, we found that COPs did not express CD45 (0.2% ± 0.2%, n = 6; Fig. 5A), indicating a nonhematopoietic origin for these cells. We further prepared COPs from mice transplanted with GFP+ WBM cells. GFP+ cells were not found in COP spheres prepared from the recipient mice 8 weeks after transplantation (Fig. 6C), although numerous GFP+ cells were observed in the recipient cornea (Fig. 6A, 6B). GFP+ cells in sphere culture preparations were found attached to the bottom of the culture dish, and immunofluorescent staining showed that the GFP+ cells were CD45+ and some also expressed CD34 (Fig. 6D). CD34 was therefore expressed in both BM-derived GFP+ cells as well as GFP-COPs, indicating that WBM-derived cells are not likely to contribute to COP sphere-initiating cells.
Given that cranial neural crest-derived mesenchymal cells contribute to corneal stroma development, we next investigated whether COPs were of neural crest origin [6, 7]. COP spheres were prepared from Wnt1-Cre/Floxed-EGFP and P0-Cre/Floxed-EGFP transgenic mice in which neural crest-derived cells are tagged by EGFP expression [14, 51]. As expected, COP spheres prepared from both transgenic mice were GFP+ (Fig. 7D, 7E). To visualize GFP+ neural crest-derived cells in the cornea, sections of Wnt1-Cre/Floxed-EGFP mouse were immunostained using anti-GFP antibody. Expression of GFP was detected in stromal keratocytes, although the expression level was low in vivo (Fig. 7B). Strong immunoreactivity was detected in the corneal endothelium (Fig. 7A), which are also neural crest-derived [52, 53]. We also examined embryonic neural crest-associated genes by RT-PCR analysis. Twist, Slug, Snail, and Sox9 were expressed in COPs (Fig. 7F). These data confirm that COPs are neural crest-derived stem cells that are not recruited from the BM.
The expansion of stem cells in vitro while maintaining properties of progenitor cells is critical from the standpoint of using stem cells for research as well as medical purposes. Culture conditions for several adult somatic stem cells, including BM-HSCs and NSCs, have been well-established. The sphere culture technique, which was originally developed for culturing NSCs as neurosphere from the central nervous system (CNS), was recently applied to isolate sphere-initiating cells from adult tissues other than CNS [2, 10, –12, 17, 51, 54]. COPs have been subcultured for more than 13 months (more than 18 passages, corresponding to more than 90 population doublings) to date. As we discovered, not only do these cells have the ability to differentiate into keratocytes, fibroblasts, and myofibroblasts as observed in primary stromal keratocytes , COPs can also be induced to differentiate into adipocytes, chondrocytes, and neural cells.
Clonal spheres in this study were initiated using methylcellulose, which is an established method to clone hematopoietic cells and, more recently, embryonic stem cells and NCSs [17, , , , , , –24]. Cells within a sphere arising from a single cell were not necessarily homogeneous, which may be due to the position within a sphere or to autocrine and paracrine mechanisms. Dark cells in spheres (Fig. 1) were not GFP-negative but simply had low fluorescence under the conditions of our photography, which were set with exposure settings that do not cause saturation of fluorescent levels. This is vital because long exposure times can give the misleading impression of strong fluorescence in 100% of the cells, which is not the case.
We have also demonstrated that COPs include a high ratio of SP cells with Hoechst dye exclusion activity, which are regarded as a general property of progenitor-candidate cells. Although a higher percentage of cells seem to be ABCG2-positive by immunocytochemical analysis compared with flow cytometry, not all ABCG2-positive cells are drawn into the SP gate, which was defined by the inhibition of functional ABC transporters. Indeed, reserpine-sensitive SP-like cells were found outside the SP gate, which may have been dividing cells exhibiting higher fluorescent intensity. The results of ABCG2 expression in COPs and the high fraction of SP cells suggest that the Hoechst dye exclusion assay may be used to further characterize COPs. A recent study by Du et al. also demonstrated the presence of SP cells in the human peripheral corneal stroma, which were shown to express neural and cartilage markers in addition to keratocyte markers . We have confirmed that COP SP cells re-formed spheres after cell sorting (data not shown).
Other stem cell-related markers, including nestin, Notch1, and Msi1, were also expressed in COP spheres. Although the upregulation of Nestin is often used as evidence of a NSC phenotype, expression of this intermediate filament protein in non-neuronal cells has also been reported . We therefore prepared COP spheres from transgenic mice carrying EGFP under the control of a neural-selective enhancer of the nestin gene [57, –59]. Given that no fluorescence was observed in corneas of these mice, expression of nestin in corneal stromal cells revealed by RT-PCR analysis is probably due to non-neuronal expression. However, the fluorescence observed in COPs prepared from E/nestin-EGFP transgenic mice suggests that the neural stem/progenitor cell-specific enhancer was activated. Interestingly, we also found that expression of Msi1 was upregulated only in COPs but not in the corneal stroma of the original mice. Recent reports demonstrated that Msi1 is expressed by epithelial stem cells in intestine [60, 61] and mammary gland , making Msi1 a candidate marker of adult stem cells in a variety of tissue sources.
There are only a limited number of reports describing putative progenitor cells for corneal keratocytes [2, 3]. Stromal keratocytes develop from mesenchymal cells originating in the cranial neural crest [6, 7]. A recent study demonstrated that late embryonic keratocytes maintain plasticity to differentiate into other neural crest-derived tissue when transplanted into embryos . On the other hand, several reports have shown that BM-derived cells migrate to the corneal stroma [64, 65]. Recently, Sosnova et al.  reported that keratocytes do not express CD34 in the mouse corneal stroma and that all CD34+ cells coexpressed CD45 and were therefore BM-derived. However, Espana et al.  reported CD34 expression in cultured human keratocytes. We found CD34+CD45− cells in COP spheres (Fig. 5), which were distinct from the CD34+CD45+ adhesive cells isolated from corneas of GFP+ WBM transplanted mice (Fig. 6). Given that GFP+ COP spheres were not observed in GFP+ WBM transplanted mice, COPs appear to be non-BM progenitors that express CD34, at least during sphere cultures. Furthermore, COPs prepared from Wnt1-Cre/Floxed-EGFP and P0-Cre/Floxed-EGFP mice were EGFP+ (Fig. 7), strongly suggesting that these cells prepared from the cornea are of neural crest origin. Anti-GFP immunostaining also revealed neural crest-derived cells in the corneal stroma of Wnt1-Cre/Floxed-EGFP mice, with weaker levels of GFP expression in the stroma (Fig. 7B) compared with the endothelium (Fig. 7A). The weak expression of GFP in the stroma is probably due to the thin dendritic morphology of keratocytes, as well as the fact that stromal keratocytes are quiescent in vivo [66, –68].
Further investigations are required to determine whether COPs are unique cells that reside in the corneal stroma or whether they represent a lineage of NSCs common with SKPs that migrate to the cornea. Although there is still controversy as to the identity of SKPs , the similarity of COPs with SKPs also has several clinical implications for the possible use of dermal cells for reconstructing the corneal stroma. If abundant dermal SKPs can be induced to differentiate into keratocytes, the development of corneal equivalents using autologous tissue may become a reality. Further studies to isolate COPs from humans for regenerative purposes are under way.
The authors indicate no potential conflicts of interest.
We thank Kimie Kato for technical assistance, Hiroko Kouike for expert assistance with flow cytometric analysis, Fumito Morito for cell cultures, and Prof. Masaru Okabe (Genome Information Research Center) for providing the GFP-transgenic mice (C57BL/6 TgN [act-EGFP]OsbC14-Y01-FM131). This study was partly supported by a grant from the Advanced and Innovational Research Program in Life Sciences from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to T.K. and H.O.), a grant from Core Research for Evolutional Science and Technology (CREST) of Japan Science and Technology Corporation (to H.O.), and a grant-in-aid to Keio University from the 21st Century Center of Excellence (COE) program.