Our results also appear to be in line

Our results also appear to be in line with earlier observations that ABCG2 expressing cells are overall smaller than cells positive for CK3 (de Paiva et al., 2006; Kim et al., 2004). As to the effect of oxygen, it is interesting to note that both the ABCG2+ and CK3+ were in general larger in sub-ambient oxygen concentrations, and that ABCG2+ was largest at 15% oxygen. This leaves an intriguing question whether ambient air concentration renders ABCG2 phenotype hitherto unknown properties that are distinct from the physiological state. Future studies will be necessary to resolve in detail the differences between the profiles of ABCG2+ cells having experienced either hypoxic or normoxic environment. An important outcome of our investigation is that low oxygen levels maintain the LESC phenotype. The cells expanded in 2% and 5% oxygen exhibited inhibition of rifampicin and rather slow growth, high clonogenicity, high expression of ABCG2, low expression of CK3, and high co-expression of ABCG2 and p63α. It is notable that with increasing oxygen concentration, the cultures changed in a way resembling limbus compartments beyond the basal region. Increasing the oxygen concentration to 10% resulted in faster growth, lower clonogenicity, decrease of ABCG2, and increase of CK3 expression. These are characteristics of TACs. Finally, in 15% oxygen, the cultures were marked by a slow growth and low clonogenicity, and there was minimal presence of ABCG2 on the backdrop of a very high CK3 expression. This phenotype resembles TDCs and PMCs that can be found at and close to the ocular surface. By controlling the culture atmosphere, we were thus able to reproduce the phenotypes as they occur along the limbal differentiation path (Fig. 6). This finding indicates that oxygen is a major regulatory factor that is involved in the maintenance of homeostasis in the limbal niche. It is noteworthy, that the rate of terminal differentiation peaked at 15% oxygen. In some other stem cells systems, it has also been found that terminal differentiation peaks at moderately reduced oxygen conditions (Chen et al., 2006; Pilgaard et al., 2009). However whether or not moderate hypoxia promotes or inhibits differentiation appear to be cell type- and lineage-specific. Interestingly, the correlation between oxygen tension and rate of differentiation was only apparent for the lower oxygen tensions. Culturing the cells at the even higher oxygen concentration of 20% led to cellular responses, which did not fit the observed trends for the moderate and low levels of oxygen concentrations. We do not have a definite explanation for this phenomenon. However, as culturing primary cells at higher-than-physiological oxygen tensions, has been shown to affect the cell characteristics of lymphocytes (Atkuri et al., 2007), neurons (Tiede et al., 2011), and fibroblasts (Tamm et al., 1998), it is quite likely that the hyperoxic conditions of 20% oxygen compared to the physiological lower conditions from which the cells were isolated, affected their growth and differentiation characteristics. In this paper we demonstrate that when culturing LECs on a feeder layer, exposure to hypoxia enhances the colony forming efficiency. We know that colonies are only formed, if feeder cells are both viable, metabolically active, and in physical contact with LECs. Moreover, the inactivated feeder cells support serial subculture, allow colony formation of epithelial cells, inhibit growth of contaminating fibroblasts, and maintain human telomerase reverse transcriptase expression and Sp1/Sp3 activity, and are thus thought to function as a surrogate niche with an unknown molecular function (Miyashita et al., 2008). As cellular metabolic activity requires oxygen, and as all cells in the culture vessel are exposed to the same gaseous environment, it is highly plausible that both LECs and supporting feeder cells are influenced by the reduced oxygen availability. However, it is impossible to discern if the increased colony forming efficiency of the LECs in hypoxia is mainly due to an effect on LECs or mainly on the supporting feeder cells. Nonetheless, the data obtained from the feeder-free EpiLife system, indicate a strong direct effect of hypoxia on the limbal epithelial cells.