The mouse corneal epithelium is a continuously renewing 5–6 cell thick protective layer …

• Abstract
• Background
• Results
• Discussion
• Conclusion
• Methods
• Abbreviations
• References
• Table 1
• Figures

Qualitative characterisation of X-inactivation mosaic patterns in the corneal epithelium

Examination of mosaic corneal epithelia of 3–52 week old female XLacZ^+/- mosaic mice showed that the initial pattern of randomly orientated patches seen at 3 weeks was replaced by radial stripes by postnatal week 10 (Fig. 2A–D). This radial striped pattern was also present at all subsequent adult stages examined up to and including postnatal week 52 (Fig. 2E–I). The transition from patches to stripes was gradual and not completed until after 10 weeks (Fig. 2D,E), which is consistent with our earlier study [4]. This presumably reflects activation of LSCs between 3 and 5 weeks and the time required for subsequent replacement of the original mosaic patchwork pattern in both the basal and overlying suprabasal layers. The mature striping patterns (Fig. 2E–K) did not extend outwards into the conjunctival epithelium (Fig. 3), supporting earlier evidence that the conjunctiva is not maintained by LSCs [4,29].

Occasional abnormal mosaic patterns (Fig. 2J) may have been produced by wound healing (discussed below). Some β-gal-positive stripes were paler than normal (arrows in Fig. 2K) and histological sections showed that these pale stripes were composed entirely of pale blue cells in all of the epithelial cell layers (Fig. 2L–M), implying that pale stripes were caused by reduced β-gal expression rather than overlapping, misaligned β-gal-positive and β-gal-negative layers. This suggests that the source of the variation is clonal, presumably reflecting an epigenetic change in the level of β-gal expression in LSCs.

In adult mosaic corneas, the stripes often met at a central whorl (either clockwise or anticlockwise; Fig. 2G,I) or (less frequently) met at a midline to form a more bilaterally symmetrical pattern (Fig. 2H). Excluding eyes in which the pattern was unclear, 45.2% of 135 eyes from mice aged 15–52 weeks had clockwise whorls, 39.3% had anticlockwise whorls, and 15.6% had midline patterns (Fig. 4A). The proportion of unclear patterns decreased from 50% (37/74) at 15–20 weeks to 12.5% (14/112) at 26–52 weeks, implying that midline, clockwise and anti-clockwise patterns become more clearly delineated and easier to resolve after 20 weeks. The relative frequencies of clockwise and anticlockwise whorls were unaffected by age (Fig. 4A). For 56 pairs of eyes with clear patterns, the clockwise (51 eyes) anti-clockwise (45) and midline (16) patterns occurred independently in left and right eyes (Fig. 4B).

Wounds in mosaic corneal epithelia heal clonally

In addition to XLacZ^+/- mosaics, the suitability of transgenic mice from lines Y001deltaDRR and Y223 [30] (both displaying mosaic corneal GFP expression) were evaluated for wound healing studies. These animals carry a yeast artificial chromosome (YAC) containing the human PAX6 locus [31] into which a GFP reporter gene has been inserted at the PAX6 ATG start codon, placing it under the control of the PAX6 regulatory elements. This also eliminates production of PAX6 protein from the YAC ensuring that wild type Pax6 levels in these mice are unaffected. Both lines demonstrated patterns of GFP-positive and GFP-negative radial corneal stripes (Fig. 5) qualitatively similar to those observed in XLacZ^+/- mosaics (Fig. 2). The reason for the mosaic transgene expression is not clear but it probably involves stochastic transgene inactivation early in development so that only a proportion of adult limbal stem cells express GFP. Line Y001 contains a single copy of the YAC with a 10–20 kb truncation, while Line Y223 contains six copies and in both cases tissue specific GFP expression patterns correspond to the endogenous pattern of Pax6 expression [30]. Line Y001deltaDRR was produced by crossing Y001 with the CAGGS-Cre line [32] resulting in germline deletion of the downstream regulatory region (DRR) and leading to severely reduced GFP expression in the lens compared to line Y223 [30]. Therefore, following initial evaluation, line Y001deltaDRR (PAX6-GFP) was used for the wound healing study because corneal stripes were clearer in eyes with weaker GFP expression in the lens. Histological sections showed that the GFP-positive stripes were localised to the corneal epithelium not the stroma (Fig. 5D) and were similar to the β-gal-positive stripes in the XLacZ^+/- mosaics (Fig. 2).

Two approaches were used to study the dynamics of epithelial migration during ex-vivo wound healing more closely. First, the extent of cell mixing was monitored in XLacZ^+/- mosaic corneal epithelia following 24 hours wound healing. 1 mm diameter central wounds closed completely during this period (as demonstrated by absence of fluorescence when fluorescein was applied to the tissue [21]). X-gal staining revealed that striping patterns extended from the edge of the wound to its centre with little lateral mixing between stripes (Fig. 6A–D). This provides the first demonstration that striping patterns in X-inactivation mosaic eyes are restored as a central wound heals in whole eye-organ culture, indicating that corneal epithelial wounds heal in a clonal manner with no significant lateral mixing between stripes. For the second approach, which we term 'temporal mosaic analysis', time-lapse confocal microscopy was used to follow wound healing in mosaic corneas of Y001deltaDRR (PAX6-GFP) transgenic reporter mice [30]. Using this approach, radial GFP stripes and the wound margins were clearly visible in both central and peripheral wounds throughout the healing period without the use of fluorescein (Fig. 6).

Time-lapse recording of the radial GFP stripes allowed visualisation of the movement of GFP-positive cells towards the centre of wounds during the initial 24 h of the wound healing response. GFP-positive stripes extended towards the centre of the previously wounded region and stripes remained contiguous throughout the wound healing response (Fig. 6; for videos see additional files 1 and 2). While central wounds healed by centripetal movement of cells, healing of peripheral wounds involved both centripetal and centrifugal cell movement (relative to the centre of the cornea). This demonstration that cells can move centrifugally (towards the limbus) during wound-healing shows that population pressure from centripetally streaming cells produced by LSCs at the periphery of the cornea is not the main driving force. During healing of central wounds it appears that the original stripe pattern is effectively restored, as cells move inwards from the wound edge to meet close to the original point of convergence, although there could be minor differences. In contrast, during healing of peripheral wounds the stripes form a second point of convergence so that the new pattern of stripes is quite different from the original (Fig. 6H,L).

Wounds in mosaic corneal epithelia sometimes heal asymmetrically

The mean rate of migration calculated from the time taken for the wound shown in Fig 6E–H to completely heal was 0.44 μm min^-1. However, tracking of individual stripes revealed that the wound actually closed asymmetrically (Fig. 7). Because image quality was relatively low and it was difficult to track some stripes, the total distance moved was calculated as the overall distance measured in a straight line between the first and last time points and as the aggregate of individual measurements between each time point (methods A and B respectively, in Table 1). This demonstrated that stripes from different places around the wound extended at different rates and is confirmed by the variation in the mean rates of movement for the leading edge of each of the six stripes (Table 1). Using method A to measure the distance moved, the rates of movement varied from 0.29 μm min^-1 (stripe 6) to 0.80 μm min^-1 (stripe 3) for the 6 different stripes measured. Method B gave very similar results (0.31 μm min^-1 for stripe 6 and 0.82 μm min^-1 for stripe 3). To our knowledge this is the first report of differential rates of cell movement at the leading edge of a circular wound in the corneal epithelium and highlights the advantages of using a mosaic analysis.

Quantitative clonal analysis implies a reduction in the number of active LSC clones with age

Corrected stripe numbers were calculated for stripe patterns in 186 adult XLacZ^+/- mosaic corneal epithelia (see Methods) and used to compare LSC function at different ages up to 52 weeks and in different regions around the circumference. This showed that the corrected stripe number declined with age, up to but not beyond 39 weeks (Fig. 8A; Additional file 3). There was, however, no significant age-related change in the percentage of β-gal-positive cells in the corneal epithelium (ANOVA P = 0.8194; Additional file 3), reflecting the fact that X-inactivation is a single stable event in the early embryo. The decline in corrected stripe number implies there is an age-related decline in the number of active LSC clones and, therefore, an age-related decline in LSC function during this period.

Distribution of active LSC clones around the circumference

Several studies based on analysis of holoclones in humans [33] and BrdU label-retaining cells in mice [34,35] suggest that limbal stem cells may be unevenly distributed around the limbal circumference. There is similar evidence that the proposed TAC marker α9β1 integrin is unevenly expressed around the circumference of the corneal epithelium [36]. Therefore, we compared the distribution of corrected stripes in two pairs of overlapping regions, each representing half of the circumference (nasal versus temporal and superior versus inferior) of a subset of XLacZ^+/- mosaic eyes. This showed that there was no significant difference among the four regions in either percentage of β-gal-positive cells (ANOVA, P = 0.7270) or the corrected stripe numbers (ANOVA, P = 0.5392; Fig. 8B; Additional file 4). Thus, our analysis showed no evidence for an uneven distribution of LSC clones around the limbal circumference.