To examine whether transplantation of human mesenchymal stem cells (MSCs) could reconstruct the corneal damage and also whether grafted MSCs could differentiate into corneal epithelial cells, we isolated MSCs from healthy donors. After growth and expansion on amniotic membrane, cells were transplanted into rat corneas 7 days after chemical burns. Reconstruction of the damaged cornea and the rat vision were measured once a week by slit lamp and by an optokinetic head-tracking instrument, respectively. Corneas were then cut out, fixed, and imbedded for immunofluorescent study of the expression of keratin 3 and keratin-pan as epithelial cell markers. Expression of CD45, interleukin 2, and metalloproteinase-2 was also investigated for inflammation and inflammation-related angiogenesis. The data showed that transplantation of MSCs, like limbal epithelial stem cells, successfully reconstructed damaged rat corneal surface. Interestingly, the therapeutic effect of the transplantation may be associated with the inhibition of inflammation and angiogenesis after transplantation of MSCs rather than the epithelial differentiation from MSCs. This study provides the first line of evidence that MSCs can be used for reconstruction of damaged corneas, presenting a new source for autotransplantation in the treatment of corneal disorders.
Corneal damage can be the result of a variety of clinical disorders, including aniridia, Stevens-Johnson syndrome, or a chemical, mechanical, or thermal injury. In severe injury, both the eye limbus and central epithelia can be lost, accompanied by inflammation, neovascularization, and conjunctivalization in the patient's eye. Such patients may subsequently lose their vision. The prognosis of corneal damages in many cases depends on the extent of deficiency in limbal epithelial stem cells (LSCs) [1, 2]. Current treatment relies, therefore, on the inhibition of inflammation, protection, and provision of LSCs for reconstruction of the damaged cornea.
In addition to anti-inflammation drug therapy, several novel methods using bioengineering materials or LSCs recently have been tested to treat corneal disorders. For example, amniotic membrane was used for treating corneal disorders [3–5], but amniotic membrane alone is not sufficient for inhibition of neovascularization and conjunctivalization . As suggested by our previous study, however, LSCs cultured on amniotic membrane can effectively inhibit inflammation and reconstruct the injured corneal surface [6–10]. It appears that LSCs play a major role in this process.
LSCs are located at the corneal limbus of the eye. They can be obtained from tissue donors (for allotransplantation) who are under other clinical conditions for surgery or donate their eyes due to other reasons. Alternatively, LSCs can be isolated from the other eye of the same patient (for autotransplantation). Whether used via allotransplantation or autotransplantation, the availability of LSCs is always the limiting factor for treatment of corneal disorders by cell transplantation. Furthermore, immunorejection induced by allotransplantation underscores the importance of seeking cell sources other than LSCs for cell therapy of corneal disorders. Because mesenchymal stem cells (MSCs) are easy to isolate and have the potential to differentiate into epithelial cells [11–17], we tested whether MSCs can be used to treat corneal disorders.
Materials and Methods
Preparation of Human Amniotic Membrane
Preparation of human amniotic membrane has previously been reported [3–6, 18]. Briefly, human amniotic membrane was obtained at the time of cesarean sectioning and washed with phosphate-buffered saline (PBS) containing penicillin and streptomycin. After separation of epithelium and basement membrane from bulk tissues, amniotic membrane was stored at −70°C in Dulbecco's modified Eagle's medium (DMEM) containing 50% glycerol. Before use, amniotic membrane was pretreated with 0.25% trypsin in 0.02% EDTA for 15 to 30 minutes to remove the amniotic epithelium. The denuded amniotic membrane was ready for use.
Bone marrow cells were obtained from healthy donors. Briefly, 10 ml bone marrow was diluted 1:2 with PBS and loaded over 5 ml Histopaque (density, 1.077; Sigma-Aldrich, St. Louis, http://www.sigma-aldrich.com). Cells were harvested from the interface after centrifugation at 2,000 rpm for 20 minutes and washed with PBS. Cells were resuspended in modified Eagle's medium of alpha (GIBCO BRL, Invitrogen, Grand Island, NY, http://www.invitrogen.com) containing 10% fetal bovine serum (Hyclone, Logan, UT, http://www.hyclone.com), 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine and plated in a flask at a density of 3 × 105 cells/ml. Nonadherent cells were discarded after cultivation for 48 hours. The adherent cells were washed twice and cultured for 10 to 14 days until cell clones were formed. Cells were analyzed by flow cytometry to confirm their identity of MSCs (Fig. 2). Cells were then plated at a density of 1 × 105 cells/cm2 on amniotic membrane and expanded until 90% confluence. MSCs used in the present study were from three healthy donors with informed consent. Permission to use human tissue was granted by Peking University Health Science Center's Ethical Committee.
Measurement of Rat Vision by Head Tracking
The head-tracking method is based on an optokinetic test devised by Cowey and Franzini  and others [20, 21]. This head-tracking instrument consists of a motorized drum 58 cm in diameter and 58 cm in height and a stationary platform in the center of the drum. The drum can rotate either clockwise or anticlockwise. The inside wall of the drum is lined with a card colored in vertical stripes of black and white at a frequency of 0.5 cycles per degree (Fig. 4A1). If a sighted rat was placed on the platform, it would move its head to track the movement of a square wave (Fig. 4A2). When subjected to the test, each rat was placed on the platform and allowed to settle for 30 seconds. The drum was then rotated clockwise and the computer photography recorded the total distance of the rat on head-tracking movement for the first 10 minutes (the red curve recorded clockwise movement and the blue curve recorded anticlockwise movement). For system quality control, the head-tracking measurement was carried out using normal rats (Figs. 4B1, 4C1), rats with one eye blocked (Figs. 4B2, 4C2), and rats with both eyes blocked (Figs. 4B3, 4C3). Based on the total head-tracking responses, 40 of 120 rats were eliminated due to their poor performance, with the total distance less than 500 cm. Eighty rats performed total distances over 500 cm and were therefore selected for the present study. Measurement of head-tracking response before and after operation also served as an indication for corneal surface repairing.
Generation of Damages in Rat Corneas
Male CD(SD)IGS(SPF/VAF) rats (180 to 200 g) were purchased from Peking University Animals Science Research Center. All animals were handled according to the animal protocols approved by the Peking University Institutional Animal Care and Use Committee. After selection by head-tracking measurement, 80 rats were subjected to corneal surface damage in the right eye of each rat. A disc paper saturated with 1 N (mol L−1) NaOH was used to damage rat corneas, but animals developed severe corneal ulcers, hyphema, hypopyon, and even corneal fistula, and, consequently, the stroma of the rat cornea was often severely damaged. Because corneal epithelial stem cells are confined to a 4- to 8-mm-diameter ring on the ocular surface and stroma is essential for repairing of corneal epithelium, to avoid stroma damage, a filter paper ring with an inner diameter of 4 mm and outer diameter of 8 mm was saturated in 1 N NaOH, applied to the eyes for 30 seconds, and then quickly washed with 0.9% NaCl. The residual limbus and central epithelium were removed with a surgical blade. This protocol including the concentration of NaOH was generally used in generation of animal models for corneal damage [22, 23]. Dexamathazone-gentamicin was applied to the operated eye twice a day.
One week after chemical burns, 29 rats with severe hypema, hypopyon, or even corneal fistula were dropped out and 51 rats were selected for the experiment.
Transplantation of Cells
Seven days after rat corneal injury, cells grown on amnion membrane were transplanted. The damaged corneal surfaces were carefully keratectomized under anesthesia. Amnion membrane with or without grown cells was sutured onto the corneal surface. The grafted cells on amnion membrane were then covered with a blank amnion membrane with the basement membrane facing up. After surgery, the rat eyelid was sutured to avoid grasping. Ten days after surgery, the eyelid was opened and dexamathazone-gentamicin was applied to the operated eye twice a day throughout the experiment.
Ocular Surface Evaluation
Rat cornea was examined using either a hand-held light daily or a slit-lamp once a week. A fluorescein sodium solution was used to check the integrity of corneal epithelial cells. The corneal surface was also examined for smoothness, clarity, and neovascularization by slit-lamp examination, in addition to the measurement of vision by head tracking.
Histology and Immunohistochemistry Staining
Four weeks after transplantation, rats were killed with 10% overdose trichloroacetaldehyde monohydrate solution. Corneas were collected, fixed in 4% paraformaldehyde, and embedded in Tissue Tek OCT compound (Electron Microscopy Sciences, Washington, PA, http://www.emsdiasum.com/ems) for sectioning. For hematoxylin and eosin (H&E) staining, slides were refixed, washed, and stained with H&E. For immunofluorescent staining, slides were refixed and rinsed three times in PBS containing 0.5% Triton-X 100 for 15 minutes, incubated with 2% bovine serum albumin in PBS for 60 minutes at 37°C. Primary antibodies were applied to tissues for 2 hours at 37°C. After wash, fluorescein-conjugated antibodies were applied for 30 minutes at 37°C for immunofluorescent microscopy.
Antibodies against human used for flow cytometry were from Becton, Dickinson and Company (BD Pharmingen, San Diego, http://www.bdbiosciences.com). Monoclonal antibodies specific for human keratin 3, human nuclei, and monoclonal antibodies against rat CD45 were from Chemicon (Temecula, CA, http://www.chemicon.com). Anti-rat matrix metalloproteinase-2 (MMP-2) monoclonal antibody was from Neomarkers (Lab Vision Corporation, Fremont, CA, http://www.labvision.com). Monoclonal antibodies specific for human keratin-pan and monoclonal antibodies against rat interleukin 2 were from Sigma-Aldrich. Fluorescein isothiocyanate-conjugated or tetra-methylrhodamine isothiocyanate-conjugated secondary antibodies were also from Sigma-Aldrich.
As shown in a histological study in Figure 1, normal rat cornea surface was smooth and the epithelium was integral (Fig. 1A). By contrast, damaged corneas became thicker; inflammation took place and the epithelium and stroma were exfoliated (Fig. 1B). After transplantation, the damaged cornea surface was repaired in the rat transplanted with either MSCs on amniotic membrane (Fig. 1C) or LSCs on amniotic membrane (Fig. 1D). The repair was not observed in rats transplanted with human fibroblast cells on amniotic membrane (Fig. 1E) or amnion membrane alone (Fig. 1F), as was the case for rats treated with either dexamathazone (Fig. 1G) or gentamicin (Fig. 1H).
In addition to the histological study, the therapeutic effect was further evaluated based on transparency, neovascularization, and epithelial integrity. In all three regards, transplantation of MSCs on amniotic membrane or LSCs on amniotic membrane significantly improved the damaged corneal surface compared with other treatments (Table 1). Further analysis by slit-lamp photographs also showed that the rat corneal surface treated either with MSCs on amniotic membrane (Fig. 3A) or LSCs on amniotic membrane (Fig. 3B) was so smooth and transparent that iris vessels underneath the cornea can be observed. Although neovascularization and inflammation were not detected in the corneal surface of these animals, transplantation of fibroblast cells on amniotic membrane or amnion membrane alone showed severe neovascularization and inflammation (Figs. 3C, 3D); so were the cases for animals treated with dexamathazone (Fig. 3E) or gentamicin (Fig. 3F).
Improvement of Rat Vision After Transplantation of MSCs
To test whether transplantation of MSCs could improve the rat corneal function, rat vision was measured 4 weeks after cell transplantation using an optokinetic head-tracking instrument described in detail in the Materials and Methods section and in Figure 4. Because all experimental rats had one eye operated on and one eye untreated, rat vision was tested with the untreated eye being blocked. As shown in Figure 4D, rats transplanted with either MSCs on amniotic membrane or LSCs on amniotic membrane performed significantly better (p < .05) than other treatments in measurements (Fig. 4D, bars 1 and 2). Compared with normal animals (Fig. 4D, bar 7), no significant vision difference was detected in rats transplanted with MSCs on amniotic membrane and LSCs on amniotic membrane (Fig. 4D, bars 1 and 2), whereas the significant difference was detected between normal rats (Fig. 4D, bar 7) and the rats treated with other methods (Fig. 4D, bars 3 through 6). These data suggest that transplantation with either MSCs or LSCs on amniotic membrane could indeed improve rat vision.
Therapeutic Effect Did Not Come from Epithelial Differentiation of MSCs but May Be a Result of Inhibition of Neovascularization and Inflammation
To investigate the possible mechanism in regard to the observed corneal recovery upon transplantation of MSCs, the corneal tissues were subjected to immunofluorescent analysis to detect the expression of human keratin 3, a specific marker for corneal epithelial cells; human keratin-pan, a marker for epithelial cells; and human nuclear antigen, an indicator for transplanted human cells. In addition, we investigated the expression of CD45 and interleukin 2 in the operated rat eyes to assess whether inflammation took place in the repairing process [24, 25]. Because MMP-2, a proteolytic enzyme, plays a role in the regulation of inflammation-related angiogenesis, we also measured the expression of MMP-2 by immunofluorescence .
Human cells were easily detected in the epithelium of the rat transplanted with either MSCs or LSCs (Figs. 5A, 5B, green). Whereas human keratin 3 (Fig. 5D) and human keratin-pan (Fig. 5F) were detected in the corneal epithelium of the rat transplanted with LSCs on amniotic membrane, neither human keratin 3 (Fig. 5C) nor human keratin-pan (Fig. 5E) was detected in the rat eyes transplanted with human MSCs on amniotic membrane, suggesting that the observed therapeutic effect of MSC grafts may not come from differentiation of MSCs into epithelial cells. We therefore tested whether this therapeutic effect came from the inhibition of inflammation and angiogenesis. Indeed, expression of both CD45 and interleukin 2 were significantly depressed in rat eyes transplanted with either MSCs (Figs. 6A, 6G) or LSCs on amniotic membrane (Figs. 6B, 6H) compared with other treatments (Figs. 6C–6F, 6I–6L). Moreover, MMP-2 was highly expressed in rat corneas transplanted with fibroblasts (Fig. 6O) or amniotic membrane alone (Fig. 6P) or treated with dexamathazone (Fig. 6Q) or gentamicin (Fig. 6R). On the contrary, MMP-2 was not detected in the rat eyes transplanted with MSCs on amnion membrane (Fig. 6M) or with LSCs on amnion membrane (Fig. 6N). All of these data suggest that inhibition of both inflammation and inflammation-related angiogenesis may partially account for the recovery of the damaged rat corneal surface upon transplantation of human MSCs.
In the present study, the therapeutic efficacy of transplantation of human MSCs was evaluated by a systematic comparison with LSCs in a rat model. The data demonstrate that neither fibroblast cells nor amniotic membrane alone had therapeutic effect. Transplantation of human MSCs on amniotic membrane, like LSCs on amniotic membrane, could reconstruct severely damaged rat corneal surface. This study may provide a new method to treat corneal disorders using patients' own MSCs, thereby escaping the allotransplantation-induced immunorejection.
Due to their easy isolation from patients and relatively easy expansion, MSCs have been tried to treat a variety of human diseases based on the idea that MSCs could become functional cells in host tissues. In this experiment, however, the expression of human keratin 3 and keratin-pan were not detected in the rat cornea transplanted with MSCs on amniotic membrane by immunofluorescent staining. Our data raised the question of whether MSCs could differentiate into epithelial cells as reported by others [27, 28].
If transplanted MSCs cannot become epithelial cells, how could they provide therapeutic effect on damaged rat cornea? Based on the immunofluorescent staining of CD45, interleukin 2, and MMP-2, we conclude that inhibition of both inflammation and inflammation-related angiogenesis may partially account for the therapeutic effect of recovery of the damaged rat cornea treated with human MSCs. This explanation is at most only a part of a whole story, because neither dexamathazone nor gentamicin showed obvious therapeutic effect. Some other unknown factors may be involved in the process of repairing damaged ocular surface of the rat cornea. The real reason, however, remains unknown.
One advantage of using patients' own MSCs for treating human diseases is to overcome immune rejection. In our case, however, we did not observe any obvious immune rejection without having to resort to immunosuppressive drugs such as cyclosporine A. It might be due to immune-privileged eyes; alternatively, it may be due to immunosuppressive effect of MSCs. So far, we do not have a definitive answer for this question. Nevertheless, this study demonstrates that transplantation of MSCs could successfully reconstruct damaged corneal surface. It may provide a novel method to use patients' own MSCs in the treatment of corneal disorders.
Table Table 1.. Improvement in the damaged corneal surface after transplantation
Y.M., Y.X., and Z.X. contributed equally to this study. This work is supported by grants from the Chinese National 973 Project (2002CB510100), the 863 Project (2003AA205070), the Ministry of Education 211 project, the Beijing Ministry of Science and Technology (2002-489), and Peking University (grants 985 and 211 to L.L). We thank Aili Lu, Xiaoyan Zhang, Shuling Wang, and Meiyu Li for their help.
The authors indicate no potential conflicts of interest.