Long-term homeostasis and wound healing in an in vitro epithelial stem cell niche model

Cultures of epithelial cells are limited by the proliferative capacity of primary cells and cell senescence. Herein we show that primary human epithelial cell sheets cultured without dermal equivalents maintained homeostasis in vitro for at least 1 year. Transparency of these sheets enabled live observation of pigmented melanocytes and Fluorescent Ubiquitination-based Cell Cycle Indicator (FUCCI) labeled epithelial cells during wound healing. Cell turn over and KRT15 expression pattern stabilized within 3 months, when KRT15 bright clusters often associated with niche-like melanocytes became apparent. EdU labels were retained in a subset of epithelial cells and melanocytes after 6 months chasing, suggesting their slow cell cycling property. FUCCI-labeling demonstrated robust cell migration and proliferation following wounding. Transparency and long-term (1 year) homeostasis of this model will be a powerful tool for the study of wound healing and cell linage tracing.

weeks, however, long-term observation was not performed 27 . On the other hand, epithelial cell sheets used for regenerative medicine are generally cultured without DEQ. Transplanted epithelium can reconstruct damaged ocular surface over a long term, especially when autologous cells sources are used 28 . Although cultured epithelial cell sheets may maintain homeostasis in vivo after transplantation, it is difficult to perform detailed examination of these cells due to ethical issues.
In order to obtain an ideal in vitro model of the human epithelial stem cell niche, we previously reported that replacing epidermal growth factor (EGF) with fibroblast growth factor 7 (FGF7 or keratinocyte growth factor; KGF) combined with the rho kinase inhibitor Y27632 can extend the culture life of a confluent epithelial cell sheet for up to 3 months (hereafter termed as KY sheet) 29 . Herein, we further show that KY sheets can maintain homeostasis for over 1 year, and can undergo wound healing demonstrated by live fluorescence imaging. The unique transparent property of KY sheets was indispensable for such imaging techniques. In addition, we report changes in cell turnover and the expression pattern of the epithelial stem cell marker during the one-year culture period.

Results
Continuous turnover of primary KY sheets for 1 year in vitro. At first, we set to confirm how long KY sheets could be maintained in vitro. We initially set the upper limit of cultivation period at 1 year, and to our surprise, KY sheets were maintained for over 1 year (Fig. 1). Cell morphology in KY sheets was easy to observe by inverted microscopy, whereas cells in OTC were hard to observe except for mesenchymal cells invading into the acellular layer of DEQ during early OTC ( Fig. 1a; Supplementary Fig. S1a). Although air-lifting diminished vacuolar structures in suprabasal cells and enhanced differentiation marker KRT3 expression ( Fig. 1b; Supplementary Fig. S1b), we cultured KY sheets under submerged conditions in order to simplify the culture method and to reduce variable factors. Basal cell morphology was similar for 1 year (Fig. 1a,b,c; Supplementary Fig. S1c). Even though the culture medium was replaced daily, desquamated cells were observed every day (Fig. 1c, arrows). Feeder layer of human bone marrow derived mesenchymal stem cells also survived throughout the culture period, although cells autonomously embedded into secreted extracellular matrix after 3 months ( Supplementary Fig. S1c). Exchanging feeder cells during culture did not seem to influence the condition of KY sheets ( Supplementary Fig. S1d).
To confirm whether the epithelium structure changed during culture, basal cell density and the number of desquamated cells were counted at several time points (Fig. 1d, Table 1). Basal cell density was stable throughout the culture period. In contrast, number of daily desquamating cells collected from the supernatant increased from 1 to 3 months, and stabilized thereafter. These results suggest that the total number of epithelial cells was stable throughout the culture period, whereas the flow of cell mass required 2 to 3 months to stabilize. The number of daily desquamated cells was not related to donor age or gender, seeding cell density, tissue storage period (days from donor death to cell preparation) or passage number of feeder cells ( Supplementary Fig. S1e).
Since we previously experienced the loss of corneal specific markers during the establishment of a murine cell line 30,31 , we confirmed the expression of corneal and limbal markers after 1-year culture (Fig. 1e). Cultures at 1 month were used as control. Immunohistochemistry (IHC) showed that PAX6 was observed in all cells. KRT3 was detected in several suprabasal cells and KRT12 was observed in suprabasal cells and several basal cells. In addition to these corneal lineage markers, tight junction protein 1 (TJP1/ZO1) was observed between superficial cells (Fig. 1e, arrows). HRP permeability assay 32 showed that KY sheets maintained a functional barrier, as HRP did not intrude into cell sheets without impairing the cell barrier by treatment with 0.02% benzalkonium chloride (BAC) 33 (Supplementary Fig. S1f). These data suggest that KY sheets maintain cell barrier function. Staining of the epithelial stem cell marker TP63 was observed in almost all cell nuclei. Epithelial stem cell marker KRT15 22,[34][35][36] was expressed uniformly in basal cells at 1 month, but was limited to small cell clusters at 1 year. These results show that expression of limbal markers was stable during the culture, except for KRT15.
Turnover rate signifies the ratio of shedding cells to total cell number. Shedding cells were monitored by counting desquamating cells daily, and total cell number was estimated as the product of (culture area) × (basal cell density) × (the ratio of total cell number to basal cell number). To obtain the ratio of total cell number to basal cell number, we counted cells in cryo-sectioned samples. In addition, we measured the thickness of the cell layer during culture. We found that the number of cell layers and the ratio of total cell number did not change throughout the culture period (Fig. 1f, Table 1). Turnover rate increased from 1 to 3 months (Fig. 1f, Table 1), with the lowest turnover at 1 month compared to 2 to 6 months (p < 0.05, One way ANOVA followed with Sheffe's F test). Senescence-associated β -galactosidase (SA-β -gal) 37 was not detected in both early and late cultures ( Supplementary Fig. S1g).
To confirm whether KY sheet retained proliferative ability after long-term culture, we performed colony forming assays by using cells dissociated from sheets cultured for 1 year (Fig. 1g). Growing colonies formed at an efficiency of 9.0% ± 3.8% (n = 3), showing that proliferative potential of KY sheets was maintained even after 1-year culture.
The existence of KRT15-bright cell clusters in KY sheets. Since KRT15 expression was different between 1 month and 1 year, we performed whole mount staining (WMS) to visualize KRT15 expression patterns ( Fig. 2a). At 1 month, almost all basal cells expressed KRT15, although expression levels seemed to differ among cells. In several regions, densely packed clusters of KRT15 bright cells were observed. At 2 months, KRT15 expression varied by cell lot; the pattern in several samples resembling that at 1 month, while other lots resembled patterns observed at 3 months. After 3 months, KRT15 dim areas were apparent, which allowed densely packed KRT15-bright cell clusters to be distinguished easily. KRT15-bright cell clusters were observed during the rest of the culture period at 6 months and 1 year. Percentage of KRT15 bright cells was 34.2% ± 4.1% at 1 mo (n = 4), 13.2% ± 8.2% at 3 mo (n = 8), and 11.3% ± 3.0% at 6 mo (n = 4), respectively, and was statistically significant between 1 and 3 months (p < 0.01, One way ANOVA followed with Sheffe's F test). To observe the association of cell proliferation with KRT15 expression, several wells were treated with the thymidine analogue EdU 1 day prior to fixation, followed by WMS for EdU and KRT15 (Fig. 2b). Some KRT15 bright cells incorporated EdU (Fig. 2b, arrow), indicating that KRT15 bright cells were not growth-arrested cells. Although EdU uptake seemed to be rare in KRT15 bright cells after 3 months, the number of KRT15 bright cells itself was reduced. EdU staining in KRT15 bright cells and KRT15 dim cells were not significantly different at both 1 month (p = 0.44, n = 4, paired t test) and 3 months (p = 0.49, n = 8, paired t test), respectively.
To investigate the existence of slow cycling cells, we performed label-retaining assay (Fig. 2c). Semi-confluent KY sheets were serially labeled with EdU for 3 days, and chased for 6 months without EdU. Immediately after labeling, 58.8% ± 9.6% cells were positive for EdU, and EdU positive cells significantly decreased to 8.8% ± 3.1% at 6 months (n = 4, p = 0.003, paired t test). KRT15 bright cells appear to contain more LRCs compared to KRT15 dim cells, however the difference was not significant (n = 4, p = 0.09, paired t test). LRCs were spread around KRT15 bright cell clusters, and in some areas a linear progression from KRT15 bright basal cells to KRT15 dim superficial cells was observed (Fig. 2c, lower right panel). This result suggests a cell linage from slow cycling KRT15 bright basal cells to differentiated cells.

The association of KRT15-bright cell clusters and dendritic melanocytes in KY sheets.
Transparency of KY sheets enabled fluorescent observation of live cells, resulting in the discovery of dendritic non-epithelial cells in CMV-GFP labeled culture (Fig. 3a). Similarly, when we used densely pigmented limbus as a cell source, we observed pigmented epithelial cells associating with pigmented melanocyte-like cells (Fig. 3b). These observations suggest that limbal melanocytes are preserved and functional even after long-term primary culture. One-year cultured KY sheets were immunostained with the melanocyte marker PMEL (Fig. 3c) and MELANA/MART-1 (Fig. 3d). PMEL positive cells were located immediately above the KRT15 bright cell clusters, although limbal melanocytes in vivo locate within the basal cell layer. We speculate that this difference is due to the absence of limbal stroma in our culture. Dendrites of MELANA positive cells enwrapped KRT15 bright cell clusters (Fig. 3d). However, KRT15 bright cell clusters without melanocytes were observed, and melanocytes were also located in KRT15 dim areas implying that contact of KRT15 bright cells with melanocytes was not essential for mutual survival, although they tended to associate with each other. In LRC experiments, several melanocytes incorporated EdU after 3 days labeling and retained label after 6 months chasing (Fig. 3e, arrow), suggesting that melanocytes in KY sheets underwent cell cycle slowly.
Wound healing ability of KY sheets. Since both continuous cell turnover and wound healing are required for epithelial homeostasis, we next examined the wound healing ability of 3-month KY sheets in a steady state. We made circular φ 4 mm wounds by peeling the epithelium ( Supplementary Fig. S2). Epithelial cells migrated into the wound on the next day covering 10.0 ± 0.8 cm 2 of the wound (Fig. 4a, mean ± S.D., n = 4). Complete epithelization of φ 4 mm wounds (11.6 ± 0.9 cm 2 ) was achieved within 2 days. To monitor cell proliferation during wound healing, wounded cultures were administrated with EdU from wound day 0 to day 1 and from day 1 to day 2, respectively ( Fig. 4b). EdU positive cells were sparse at day 1. In contrast, intense labeling of EdU was observed at day 2, expanding from the middle to periphery of the wound. Both KRT15 bright cells and KRT15 dim cells migrated into epithelized area. Although EdU was mostly located in KRT15 dim cells, several KRT15 bright clusters were also positive for EdU. To observe proliferation for longer periods, KY sheets were labeled with lentiviral vectors carrying Fucci at culture day 1, and wounded after 3 months culture, followed with serial observation (Fig. 4c). In Fucci cells, green fluorescence of mAG-hGeminin (1/110) indicates cells in the S, G 2 , and M phases, whereas red fluorescence of mKO2-hCdt1 (30/120) indicates cells in G 1 phase 38 . Before wounding, Fucci red cells were predominantly observed, whereas green fluorescence was rarely observed. After wounding, Fucci green cells were observed from day 2 after wounding, but decreased at day 4 in the periphery as well as in the center of the wound by day 8. Fucci red cells were observed throughout the wound healing process. These results show that transient proliferation occurred for several days during wound healing, and also demonstrate that the transparency of KY sheets allows fluorescence observation in living cells.

Discussion
We successfully demonstrated how primary culture of human limbal epithelial cell sheets reached a steady state within 3 months, and showed continuous cell turnover and wound healing ability for at least 1 year. These results indicate the homeostasis of cultured epithelium in vitro, and strongly suggest the maintenance of stem cells in niche-like structures. Long-term culture of epithelial cells has been known for over two decades, with serial passage of epidermal keratinocytes for over 200 days (17-18 passages, 140 cell population doublings) 39 and for 2-3 months (14 passages, 80-100 cell population doublings) in limbal epithelial cells 19 . In these serial passage experiments, cells are subcultured before they reach confluence, and confluent cells lost viability and decreased in number during prolonged culture 40 . In contrast, KY sheets maintained total cell number for 1 year. As shown in our previous report, this difference was due to replacing EGF with KGF and Y27632 29 . Administration of EGF from the seeding medium decreases CFE 39 . EGF stimulates proliferation, but also stimulates cell motility and diminishes KRT3 expression in limbal epithelial cells 41 . In our previous study, cells cultured with EGF from initial seeding did not survive over 3 months 29 . KGF/FGF7 stimulates the proliferation of keratinocytes 42 and limbal epithelial cells without increasing cell motility 41 , and enhances the expression of p63 via the p38 pathway in limbal epithelial cells 43 . EGF seems to have a superior effect on epithelial cells compared to KGF, since EGF stimulates epithelial cell migration even in the presence of KGF 41 . Y27632 is known to inhibit keratinocyte differentiation in cell suspension 44 , immortalize keratinocytes 45 , and increase CFE of limbal epithelial cells 29,46 . Y27632 cannot preserve KRT15 expression in EGF-treated cultures, while Y27632 increased CFE in KGF supplemented-cultures 29 . From these observations, removing EGF while supplementing KGF and Y27632 may be crucial for the maintenance of primary cultures over long term.
Recently, several methods were developed to induce corneal epithelial linage cells from human embryonic stem cells (ESCs)/induced pluripotent stem cells (iPSCs) [47][48][49][50][51][52][53][54][55] . However, since ESC and iPSC are pluripotent, other cell types may be induced together with corneal epithelial cells. Purification and detailed characterization is required to prove that the induced cells are truly of corneal epithelial linage 51 . One of the greatest advantages of our KY sheets is the simplicity of the technique. Compared to ESC/iPSC cultures, our method only requires few steps to isolate cells from the limbus and co-culture with feeder cells under submerged conditions. The limbal origin of cells also ensures that the cells are of corneal epithelial lineage.
Transparency of KY sheets is useful for cell linage analysis and investigating cell signaling in specific cells, when combined with proper fluorescent reporters. Despite its simplicity, LRC rate was 8.8% after 6 months chasing in KY sheet, compared to 0.9% after 6 weeks chasing in OTC 26 , suggesting that cells in KY sheets had a slower cell cycle compared to OTC. In addition, KY sheets contained colony-forming cells even after 1 year culture, proving that KY sheets maintained stem/progenitor cells in vitro throughout culture. Storage period of donor tissue did not influence the quality of KY sheet states, which may allow flexibility of experiment schedules. Our culture protocol can also be used for other experiments requiring human corneal epithelial cells. For example, culture inserts can be treated with different substrates to study the effects of biomechanical stress, or stiffness of extracellular matrix on cell homeostasis 56 .
Compartmentalization of stem cells following confluency was reported in organotypic culture of primary human epidermal keratinocytes 57 . Similarly, we found densely packed cell clusters expressing epithelial stem cell marker KRT15 [34][35][36] in KY sheets. KRT15 negative basal cells were positive for KRT12, showing characteristic features of corneal epithelium, which is differentiated from limbal epithelium. Thus the steady state KY sheet may be used as an ocular surface model including both corneal epithelium and limbal epithelium. The mechanism by which KRT15 was maintained in densely packed clusters during culture is unclear. One feasible explanation is that KRT15 bright clusters were protected by the niche, while KRT15 bright single cells were not, thus driving KRT15 bright single cells to differentiate and become lost during culture. There is also the possibility that KRT15 bright clusters were formed from cell aggregates at the time of seeding. Further studies using fluorescent probes driven by human KRT15 promoters 58 may elucidate the mechanism cluster formation.
Wound healing process of corneal epithelium consists of an initial phase and a closure phase (Reviewed in ref. 3). The closure phase starts with the cell migration without mitosis, followed by the cell proliferation and differentiation. Similarly, KY sheets showed the cell migration without mitosis at day 1 after wounding, followed by proliferation at day 2 after wounding, indicating the usefulness of KY sheets as wound healing model. Interaction between epithelial cells and mesenchymal cells also occurs during the wound healing (Reviewed in refs 3,59). Since KY cultures contained a feeder layer consisting of mesenchymal cells, crosstalk between epithelial cells and mesenchymal cells via soluble factors may have occurred during the wound healing assay. A limitation of our model is that unlike stem cells in vivo that are presumably anchored to the stromal niche, our cells in vitro are not in contact with the feeder cells. This may explain the migration of KRT15 bright clusters during wound healing.
Melanocytes are a niche candidate to support limbal epithelial stem cells 10,60,61 , and dendritic melanocytes surrounding several KRT15 bright cell clusters were observed in KY sheet. However, KRT15 bright clusters not associated with melanocytes were also observed, suggesting that other niche factors may maintain KRT15 bright clusters. Since melanocytes were a serendipitous observation, conditions that allow these cells to be maintained in culture need to be elucidated. It is possible that the re-association of epithelial stem/progenitor cells and melanocytes may have occurred immediately following dissociation. While not within the scope of this study, the transparent nature of KY sheets should allow the study of interaction between human epithelial cell and melanocytes in vitro.
In conclusion, we showed the homeostasis of primary cultured human limbal epithelial cells for 1 year in vivo without the use of dermal equivalent, maintained by continuous cell turnover with the maintenance of KRT15 bright cell clusters often associating with melanocytes and wound healing ability. This protocol will be a powerful tool in the study of stem cells, wound healing and epithelial/melanocyte interaction, and cell linage analysis using fluorescent reporters. While our protocol was not intended for clinical use, long-term maintenance of cells sheets with a proven stem cell population may also open doors to a new generation of regenerative medicine techniques.
Human limbal epithelial cells were isolated from U.S. eyebank eyes (Sightlife, Seattle, WA). Donor age was ranged from 20 to 75, 60.6 ± 10.8 years old (n = 71, male = 39, female = 32). Eyes were shipped in cold preservation medium (Optisol, Bausch & Lomb, Rochester, NY) in a cornea viewing chamber (Bausch & Lomb), and stored at 4 °C. Mean time from death to cell isolation was 11.9 ± 3.6 days. After the removal of excess tissue, limbal epithelium was separated from stroma by Dispase II treatment (final 4U/mL, Roche, Basel, Switzerland) at 37 °C for 1 hour, followed with the dissociation by pipetting to obtain the mixture of single cells and cell aggregates. In 3 experiments, epithelium was further dissociated by enzyme treatments (TrypLE Express, Gibco) for 30 min at 37 °C, followed with passing through 40 μ m nylon mesh (Cell strainer, Falcon, Corning Incorporated, Corning, NY) to remove large cell aggregates. Dissociated cells were seeded in plastic cell culture inserts (3450, Corning), subsequently co-cultured with feeder cells in the bottom of paired well. Since harvested cell number varied among donors (2.8 ± 1.8 × 10 5 ), isolated cells were evenly separated to 3 inserts. Average seeding cell density was 9.1 ± 7.3 × 10 4 cells/insert.
Organotypic culture. DEQ was constructed by using MASCs and a commercially available collagen type I kit (acidic collagen type I from porcine tendon, KP-7000, Nitta-gelatin Inc., Tokyo, Japan). At first, 1 mL of acellular collagen layer was reconstructed on cell culture insert. After gelation, 3 mL of cellular layer was cast on the acellular layer 64 . Each DEQ contained 5 × 10 5 of MASCs and 0.7 mg/mL of collagen type I 65 . DEQs were fed with MEM-α containing 10% FBS at day 3 and day 6. Human limbal epithelial cells were cultured in the 25 cm 2 flask with feeder layer of MMC treated NIH/3T3, and semi-confluent epithelial cells (P0-P1) were passaged on 1wk-old DEQ at a density of 5 × 10 5 epithelial cells/DEQ. OTCs were cultured as submerged for 4 days 65 , followed with air lift culture for additional 1 month.
EdU pulse labeling, serial labeling and the chasing of LRCs. For the pulse labeling of proliferating cells, EdU (10 μ M, Thermo Fisher Scientific, Waltham, MA) were administrated to the culture 24 hours before fixation. For LRC assay, EdU (1 μ M) were administrated to semi-confluent cells (day 6) for serial 3 days. One of triplicated inserts was fixed immediately after serial labeling, and remaining inserts were subsequently cultured for an additional 171 days without EdU. EdU was detected by Alexa flour-488 conjugated azide (Click-iT kit, Thermo Fisher), followed by immunostaining.
Measurement of layer numbers, ratio of total cell number to basal cell number, and basal cell density. DAPI images of IHC and WMS under objective power x20 were used for cell counting. The number of cell nuclei was counted by using cell counter in Image J 1.45 s software (NIH). Four to seven different lots were Scientific RepoRts | 7:43557 | DOI: 10.1038/srep43557 used in each time point. Number of layers was counted as the nuclei number in the vertical direction from basal cells to superficial cells in IHC images. Ratio of total cell number to basal cell number was calculated by using total nuclei number in basal layer and total nuclei number in all suprabasal layers in each IHC image; total cell number was calculated as the sum of basal cell number and suprabasal cell number. Basal cell density was calculated by the number of nuclei in each WMS image field (1.48 × 10 −1 mm 2 at Objx20).
Calculation of daily desquamated cell number and turnover rate. Supernatants from each cell culture insert were collected by P1000 micropipette. Volume of supernatant was also measured by P1000. Number of desquamated cells was counted by hemocytometer at day 7 (n = 4), day 14 (n = 4), day 21 (n = 6), day 27-29 (n = 19), day 41-43 (n = 16), day 59-61 (n = 21), and every 30 days to day 359-361 (n = 6-22). The average number of desquamating cells during 3 continuous days was recorded as the data for each time point, except for 1-3 weeks. Turnover rate was estimated as (the number of daily desquamated cells)/(the mean total cell number in each insert). Total cell number in each insert was calculated as (basal cell density) × (the ratio of total cells to basal cells) × (the area of cell culture insert, 4.67 cm 2 ).
Colony forming assay. Primary cultures at 1 year were sampled for IHC and WMS without fixation, and remaining cells were treated with cell dissociation buffer (TrypLE express) with Y-27632 at 37 °C for 30 min. After passing through a 40 μ m cell strainer to remove cell aggregates, cells were seeded on NIH/3T3 feeder-prepared 100 mm dishes at a density of 1 × 10 3 cells per dish. Dishes were cultured until colonies became apparent (18 days), and fixed with buffered 10% formalin and stained with Rhodamin B to visualize colonies. Colony formation efficiency was calculated as the percentage of colonies per seeded cell number. Four indicator dishes were used for each lot, and three independent experiments were performed.
Wound healing assay. KY sheets cultured for at least 3 months were used for wound healing assays. Culture plastic dishes were marked with φ 4 mm trephines, and cell culture inserts with KY sheets were moved onto a dish ( Supplementary Fig. 2). A section of the cell sheet above the mark was peeled off with fine forceps (No. DU-5, Dumont, Montignez, Switzerland) under a stereomicroscope. The shape of the wound was marked on the plastic insert by scratching with forceps. One to four wounds were created in each insert. After washing with culture medium twice, wounded epithelia were returned to the feeder prepared wells and co-cultured. Phase contrast images were merged by raster graphics editor (GIMP 2, The GIMP Development Team). To measure the epithelialized area, day 0 line was scratched on the underside of the cell culture insert, and day 1 line observed as the border of epithelial cells were traced by raster graphics editor and the area between the lines was measured by ImageJ.
Statistical analysis. Two groups were analyzed by paired t test. Groups over 3 were analyzed by one way ANOVA followed with Sheffe's F test, performed by excel software and add-in Statcell. Statistical significance was set as p < 0.05.
Methods for HRP permeability assay and SA-β -galactosidase staining are provided in Supplementary Information.