Effects of Aberrant Pax6 Gene Dosage on Mouse Corneal Pathophysiology and Corneal Epithelial Homeostasis

Background Altered dosage of the transcription factor PAX6 causes multiple human eye pathophysiologies. PAX6 +/− heterozygotes suffer from aniridia and aniridia-related keratopathy (ARK), a corneal deterioration that probably involves a limbal epithelial stem cell (LESC) deficiency. Heterozygous Pax6+/Sey-Neu (Pax6+/−) mice recapitulate the human disease and are a good model of ARK. Corneal pathologies also occur in other mouse Pax6 mutants and in PAX77Tg/− transgenics, which over-express Pax6 and model human PAX6 duplication. Methodology/Principal Findings We used electron microscopy to investigate ocular defects in Pax6+/− heterozygotes (low Pax6 levels) and PAX77Tg/− transgenics (high Pax6 levels). As well as the well-documented epithelial defects, aberrant Pax6 dosage had profound effects on the corneal stroma and endothelium in both genotypes, including cellular vacuolation, similar to that reported for human macular corneal dystrophy. We used mosaic expression of an X-linked LacZ transgene in X-inactivation mosaic female (XLacZTg/−) mice to investigate corneal epithelial maintenance by LESC clones in Pax6+/− and PAX77Tg/− mosaic mice. PAX77Tg/− mosaics, over-expressing Pax6, produced normal corneal epithelial radial striped patterns (despite other corneal defects), suggesting that centripetal cell movement was unaffected. Moderately disrupted patterns in Pax6+/− mosaics were corrected by introducing the PAX77 transgene (in Pax6+/−, PAX77Tg/− mosaics). Pax6Leca4/+, XLacZTg/− mosaic mice (heterozygous for the Pax6Leca4 missense mutation) showed more severely disrupted mosaic patterns. Corrected corneal epithelial stripe numbers (an indirect estimate of active LESC clone numbers) declined with age (between 15 and 30 weeks) in wild-type XLacZTg/− mosaics. In contrast, corrected stripe numbers were already low at 15 weeks in Pax6+/− and PAX77Tg/− mosaic corneas, suggesting Pax6 under- and over-expression both affect LESC clones. Conclusions/Significance Pax6+/− and PAX77Tg/− genotypes have only relatively minor effects on LESC clone numbers but cause more severe corneal endothelial and stromal defects. This should prompt further investigations of the pathophysiology underlying human aniridia and ARK.


Introduction
The Pax6 gene encodes the Pax6 transcription factor with key regulatory roles in eye development [1][2][3][4][5].Abnormal expression results in a spectrum of ocular pathophysiologies, some of which are directly linked to the protein level [6][7][8][9].Some corneal abnormalities associated with Pax6 mutations occur during development and others result from inadequate tissue maintenance.
It is widely accepted that, during adult corneal epithelial homeostasis, cell production is initiated by limbal epithelial stem cells (LESCs) in the limbus at the corneoscleral junction [10][11][12][13][14]. LESCs proliferate to produce transient (or transit) amplifying cells (TACs) that migrate centripetally, dividing a few times before terminally differentiating.As TACs differentiate they lose contact with the basal epithelium, move apically, and are desquamated from the surface layer [15,16].Epithelial abnormalities could be caused by defects in LESCs or epithelial cell proliferation, movement or loss.
Centripetal movement in the mouse corneal epithelium has been demonstrated directly in several experimental systems [17,18] and indirectly by the postnatal switch from a randomly orientated mosaic pattern to radial stripes in X-inactivation mosaics [19][20][21] and lentivirus-labelled lineages [22].Radial stripes emerging from the periphery and extending towards the central cornea from ,5 weeks are thought to represent clones of centripetally migrating epithelial cells produced after LESC activation.Numerical analysis of these striping patterns provides an indirect estimate of the number of coherent clones of LESCs maintaining the corneal epithelium [19][20][21].
Pax6 +/2 mice have corneal and other anterior segment abnormalities [37][38][39][40][41][42][43][44][45][46] and some have been characterised by electron microscopy (EM) [38,47,48].The postnatal corneal deterioration in Pax6 +/2 mice is equivalent to that seen in human aniridia-related keratopathy (ARK), which has been attributed to LESC deficiency.This is based entirely on indirect evidence such as the presence of goblet cells [49] and clinical results for limbal transplants [50] because there are currently no suitable LESC markers.Quantitative analysis of mosaic corneal epithelial patterns in mouse X-inactivation mosaics and chimeras also suggests that Pax6 +/2 mice have fewer active clones of LESCs than normal [51].However, the stripe pattern is disrupted, implying that cell movement is abnormal so Pax6 +/2 corneal deterioration probably involves additional factors.
The present study had three aims.(1) To investigate the effects of altered Pax6 dose on the corneal stroma and endothelium in Pax6 +/2 and PAX77 Tg/2 mice using EM, to complement previous studies of effects on the corneal epithelium.(2) Given that our previous clonal analysis of X-inactivation mosaics identified abnormalities of corneal epithelial maintenance in Pax6 +/2 mice [51], to investigate whether Pax6 over-expression causes similar abnormalities in PAX77 Tg/2 mice.(3) To investigate whether this effect can be rescued by combining the PAX77 transgene with a Pax6 +/2 genotype.We identified previously unreported corneal endothelial and stromal abnormalities in both genotypes by EM.Furthermore, our qualitative analysis of X-inactivation mosaics implied that cell movement was normal during corneal epithelial maintenance in PAX77 Tg/2 mice (unlike Pax6 +/2 heterozygotes) and our quantification suggested that in younger mice LESC clone numbers were reduced in both Pax6 +/2 and PAX77 Tg/2 genotypes.The PAX77 transgene normalised both the qualitative and quantitative defects in Pax6 +/2 corneas.

Results
We used scanning electron microscopy (SEM) and transmission electron microscopy (TEM) to compare the structure of corneas from WT, Pax6 +/2 and PAX77 Tg/2 mice.We also analysed mosaic patterns in the corneal epithelia of WT, Pax6 +/2 and PAX77 Tg/2 X-inactivation mosaics to compare effects of Pax6 doses on corneal epithelial cell movement (from mosaic patterns) and LESC clone numbers (from quantitative analysis of corrected stripe numbers).
PAX77 Tg/2 corneal epithelial cells have abnormally large microvilli and less pronounced cell junctions On the CBA/Ca background Pax6 +/2 and PAX77 Tg/2 eyes were smaller than WT (Fig. 1A-C) and PAX77 Tg/2 mice had microcorneas and a pronounced ring around the corneoscleral junction (Fig. 1C).SEM of the superficial corneal epithelial cells showed that, the WT corneal epithelium consisted of large polygonal cells with tightly opposed cell junctions (Fig. 1D) and numerous microvilli (Fig. 1G).Despite reported increased sloughing of Pax6 +/2 corneal epithelial cells [38], the epithelia of the Pax6 +/2 specimens analysed by SEM appeared similar to WT (Fig. 1E,H).However, the PAX77 Tg/2 epithelial cells had a more irregular surface (Fig. 1F,I), indistinct cell junctions (Fig. 1F) and larger microvilli (Fig. 1I).Other corneal epithelial abnormalities have been well described, so further EM work focused on the corneal stroma and endothelium.Previously unreported abnormalities are discussed below and summarised in Table 1.
Pax6 +/2 and PAX77 Tg/2 corneal endothelial cells are severely abnormal SEM examination revealed serious abnormalities in both Pax6 +/2 and PAX77 Tg/2 corneal endothelia.The WT corneal endothelial cells were hexagonal (mean diameter, 18.762.35mm; Fig. 2A,D) whereas Pax6 +/2 cells were larger (23.7563 mm; Fig. 2B,E) and slightly irregularly shaped.The PAX77 Tg/2 endothelium was highly irregular with indistinct cell borders that were only visible at higher magnification (Fig. 2C,F).The WT endothelial cells had either no vacuoles or very small vacuoles (Fig. 2D,G), whereas the Pax6 +/2 and PAX77 Tg/2 endothelial surfaces appeared highly irregular by SEM (Fig. 2E,F) and TEM revealed large intracellular vacuoles across the entire endothelium in each case (Fig. 2H,I).In PAX77 Tg/2 corneal endothelia the vacuoles seemed smaller towards the central cornea (data not shown).These results imply that Pax6 over-and underexpression both produced significant endothelial defects but over-expression produced a more severe phenotype.
In sections of stained XLacZ Tg/2 corneas from WT, Pax6 +/2 and PAX77 Tg/2 animals, clones of b-gal positive cells were aligned vertically with little overlap of b-gal positive and b-gal negative cells (Fig. 5).We, therefore, treated the stripes as 2-dimensional patterns and reduced them to a 1-dimensional count for quantification (see Materials and Methods).
Over-expression of Pax6 in PAX77 Tg/2 mosaic corneas results in fewer corneal stripes Although stripe patterns in PAX77 Tg/2 , XLacZ Tg/2 mosaics appeared normal, a quantitative analysis was undertaken to identify any subtle differences that might suggest that stem cell clones were affected.The observed number of radial stripes in the corneal epithelium was converted to a corrected stripe number to compare LESC clone numbers between different groups (see Materials and Methods).A preliminary experiment, using PAX77 Tg/2 mice on an outbred CD-1 genetic background showed that the corrected stripe number per eye was significantly lower in PAX77 Tg/2 , XLacZ Tg/2 mosaics than WT, XLacZ Tg/2 controls at 15 weeks (Table 3).This difference remained significant after correcting for the smaller circumferences of PAX77 Tg/2 corneas, suggesting that PAX77 Tg/2 , XLacZ Tg/2 corneas were maintained by fewer active LESC clones than normal.
Although the corrected stripe number does not provide a direct estimate of LESC numbers, it can be used to compare LESC clones between different groups of mice.The corrected stripe number is an estimate of the number of corneal epithelial clones, each of which will be produced by an active coherent clone of LESCs in the limbus.A difference in corrected stripe number, therefore, predicts a difference in the number of active LESC clones and this could reflect a difference in active LESC numbers and/or a change in the LESC distribution (number of LESCs per clone).
The preliminary results were confirmed and extended in a second experiment, using PAX77 Tg/2 mice on an inbred CBA/Ca genetic background and analysing corneas at both 15 and 30 weeks.The mean corrected stripe number per eye (or per mm circumference) was significantly lower in PAX77 Tg/2 , XLacZ Tg/2 mosaics than WT, XLacZ Tg/2 controls at 15 weeks (Fig. 6).It declined between 15 and 30 weeks in WT, XLacZ Tg/2 controls, as reported previously [19,21], but no reduction occurred in the PAX77 Tg/2 , XLacZ Tg/2 mosaics, so by 30 days the PAX77 Tg/2 corrected stripe number was not significantly different from controls (Fig. 6C,D).This suggests that PAX77 Tg/2 , XLacZ Tg/2 corneas were maintained by fewer active LESC clones than normal at 15 weeks (as in the preliminary experiment, Table 3) but, unlike WT, this did not decline further between 15 and 30 weeks.

Discussion
This study identified new corneal abnormalities, particularly in the stroma and endothelium, in both Pax6 +/2 and PAX77 Tg/2 mice, which respectively under-and over-express Pax6.Analysis of mosaic corneal patterns showed cell movement during corneal epithelial maintenance was affected in Pax6 +/2 but not PAX77 Tg/2 mice and implied that LESC clones were affected in both Pax6 +/2 and PAX77 Tg/2 mice at 15 weeks.
Morphological abnormalities of Pax6 +/2 and PAX77 Tg/2 corneas The abnormally large microvilli and indistinct cell junctions of PAX77 Tg/2 corneal epithelial cells are consistent with fragility tests suggesting cell adhesion may be affected [56].Significant new morphological abnormalities were identified in the corneal stroma and endothelium of both Pax6 +/2 and PAX77 Tg/2 mice including intracellular vacuoles.There is some evidence that Pax6 is expressed weakly and transiently in the mouse fetal corneal stroma [47,57] and chimera experiments imply that Pax6
Excessive hydration may cause corneal stromal haze or corneal oedema, as in some human corneal stromal dystrophies.Similar corneal endothelial and stromal keratocyte vacuolation also occurs in human macular corneal dystrophy [58,59].If the corneal endothelium and/or stroma of PAX6 +/2 human aniridia patients are affected like the Pax6 +/2 mice described here, this might have important clinical implications and could underlie some of the abnormal phenotypes associated with aniridia-related keratopathy.

Effects of Pax6 on corneal epithelial cell movement and maintenance
The normal radial striped corneal epithelial pattern of WT Xinactivation mosaics was disrupted in Pax6 +/2 mosaics, implying that corneal epithelial cell movement is abnormal [51] but it is unclear whether this is caused by some intrinsic abnormality or a response to chronic wounding of a thin and fragile cornea.Although maintenance of the PAX77 Tg/2 corneal epithelium is not entirely normal (Table 2), PAX77 Tg/2 mosaics had completely normal radial striped patterns, implying that centripetal cell movement is unaffected by the higher Pax6 dose.In contrast, the pattern in corneas of Pax6 Leca4/+ mosaics resembled the randomly orientated pattern of patches seen in young WT corneas before stripes emerge.This suggests that movement of cells from the limbus is severely reduced and that the Pax6 Leca4/+ corneal epithelium may be maintained by proliferation from within the epithelium, perhaps because of an extreme LESC deficiency.This possibility has previously been suggested to explain a similar phenotype in Dstn corn1/corn1 homozygotes [60].

Evidence for effects of Pax6 dose on limbal stem cell clones
Our quantitative analysis of striped patterns in Pax6 +/2 and PAX77 Tg/2 X-inactivation mosaics showed that, at 15 weeks, the corrected stripe numbers were lower than in WT mosaics, even after correcting for differences in corneal circumference, implying that at this age there were fewer clones of active LESCs maintaining the corneal epithelium.However, no such difference was seen at 30 weeks because, by this age, the stripe number had declined in WT mosaics but the stripe numbers did not decline between 15 and 30 weeks in Pax6 +/2 and PAX77 Tg/2 mosaics.
The decline in corrected stripe number in WT XLacZ mosaics implies that there is an age-related decline in active LESC clones.This may reflect a decline in LESC numbers or activity but it could also be explained by a drift in LESC clone distributions.If LESCs can divide either symmetrically to produce one LESC and one TAC or asymmetrically to produce either two LESCs or two TACs then the pattern of active stem cell clones may follow a pattern of neutral drift as suggested for some other stem cell systems, including spermatogonial stem cells [61] and intestinal crypts [62,63].Over time, stem cells may be lost and replaced by their neighbours.On this basis, clones of stem cells will expand and contract stochastically and some clones will be lost (e.g. if a bgal negative LESC clone, that is flanked by two b-gal positive clones, is lost the b-gal negative stripe will be lost and the two flanking b-gal positive stripes will merge into one larger one).
The lower corrected stripe number in Pax6 +/2 and PAX77 Tg/2 XLacZ mosaics compared to WT mosaics at 15 weeks can be explained in several ways.It is possible that there are initially fewer LESC clones in Pax6 +/2 and PAX77 Tg/2 eyes, either because fewer LESCs are specified, or activated or because the LESCs are grouped into fewer larger clones.Alternatively, LESC clone numbers may initially be similar in all groups (before 15 weeks) but the decline may begin earlier or be more rapid in Pax6 +/2 and PAX77 Tg/2 mice, so by 15 weeks the LESC clone number is similar to that in a 30 week WT mouse.It is not clear why the Pax6 +/2 and PAX77 Tg/2 LESC clone numbers do not continue to decline after 15 weeks, but it has been suggested that this might be related to a minimum required for corneal epithelial maintenance [21].Regardless of the explanation it is clear that, at 15 weeks, LESC clones numbers and/or distributions are different in Pax6 +/2 and PAX77 Tg/2 mice.However, as this difference is no longer detectable at 30 weeks the difference is short-lived and may have little biological significance.
The presence of goblet cells in the human corneal epithelium is often cited as evidence of LESC deficiency [49,64].However, the quantitative analysis of Pax6 +/2 and PAX77 Tg/2 mosaics indicates LESC clones are similarly affected in both genotype at 15 weeks (Fig. 8B,C) but only Pax6 +/2 mice have goblet cells in the corneal epithelium (Table 2).This difference highlights the need for reliable LESC markers.
The PAX77 transgene compensates for defects of corneal maintenance in Pax6 +/2 X-inactivation mosaics On a Pax6 +/2 background, the PAX77 transgene rescued the abnormal stripe patterns that normally occur in Pax6 +/2 heterozygotes and are attributed to low Pax6 levels.Quantitative analysis showed that the PAX77 transgene also normalised the putative deficiency in active stem cell clones (reduced stripe number) that occurs in Pax6 +/2 heterozygotes at 15 weeks.These results imply that restoring the Pax6 dose to a more normal level corrects abnormalities of corneal cell maintenance as well as the developmental ocular defects demonstrated previously [8].
It is widely believed that stem cell deficiency causes most corneal abnormalities in ARK.However, our quantitative analyses of mosaic patterns suggest that Pax6 +/2 and PAX77 Tg/2 mice have only relatively modest reductions in LESC clone numbers.In contrast, both Pax6 +/2 and PAX77 Tg/2 mice have severe corneal Table 3.Preliminary comparison of corrected stripe number in corneal epithelia of wild-type, XLacZ Tg/2 and PAX77 Tg/2 , XLacZ Tg/2 X-inactivation mosaics at 15 weeks.endothelial and stromal defects.This should prompt further investigations of the pathophysiology underlying ARK.

Materials and Methods
Consumables were purchased from Sigma (Poole, UK) and procedures carried out at room temperature, unless stated otherwise.

Ethics statement
All animal work was approved by a University of Edinburgh internal ethics committee and was performed in accordance with institutional guidelines under license by the UK Home Office (project license number PPL 60/3635).

Animals and genetic crosses
Mice were maintained in animal facilities of the College of Medicine and Veterinary Medicine, University of Edinburgh.Heterozygous Pax6 +/Sey-Neu mice (abbreviated to Pax6 +/2 ) and wild-type (WT, Pax6 +/+ ) littermates were produced from Pax6 +/+ female 6 Pax6 +/2 male crosses on a CBA/Ca genetic background and genotyped by PCR as described previously [3].Heterozygous Pax6 Leca4/+ mice, on a mixed genetic background, were provided by Prof. Ian Jackson and Dr Sally Cross (MRC, Human Genetics Unit, Edinburgh) and maintained as a closed, random-bred colony by crossing Pax6 Leca4/+ and WT mice within the colony.Outbred CD-1 mice carrying the PAX77 transgene which expresses 5-7 copies of the human PAX6 gene [8] were provided by Professor Veronica van Heyningen and Dr Dirk A. Kleinjan (MRC Human Genetics Unit, Edinburgh) and the transgene was transferred to the inbred CBA/Ca strain by genetic crosses as reported previously [55].In the present study we designated mice hemizygous for the PAX77 transgene as PAX77 Tg/2 (because the use of 'Tg' to designate presence of the transgene is less ambiguous than '+' used in our previous 'PAX77 +/2 ' notation [55]) and we designate non-transgenic littermates as PAX77 2/2 .The founder colony is designated CD1-PAX77 Tg and was maintained by CD-1 6 CD1-PAX77 Tg/2 crosses.The derived congenic stock is  designated CBA-PAX77 Tg and was maintained by CBA/Ca 6 CBA-PAX77 Tg/2 crosses.Hemizygous PAX77 Tg/2 mice and WT, PAX77 2/2 littermates were genotyped by PCR as described previously [55].No homozygous PAX77 Tg/Tg transgenic mice were used in this study.
H253 strain mice [65], ubiquitously expressing the Tg(Hmgcr-lacZ)H253Sest, X-linked nLacZ transgene (abbreviated to XLacZ), were obtained from the MRC Mammalian Genetics Unit, Harwell, UK, as strain FTH, and maintained on a genetic background that was predominantly a mixture of C57BL/6 and CBA/Ca inbred strains.Males and females hemizygous for this Xlinked transgene are designated respectively XLacZ Tg/Y and XLacZ Tg/2 ; female homozygotes are designated XLacZ Tg/Tg .Xinactivation mosaics hemizygous for the Pax6 Sey-Neu null mutation and the Pax6 Leca4 missense mutation, plus WT littermate controls, were produced from Pax6 +/2 female 6 XLacZ Tg/Y male and Pax6 Leca4/+ female 6 XLacZ Tg/Y male crosses respectively.
For scanning electron microscopy (SEM), samples were transferred to hexamethyldisilazane (HMDS) for 40 minutes and air-dried.The samples were mounted on aluminium stubs and sputter coated with gold using an Edwards S150A sputter coater then examined on a JEOL JSM 5600 scanning electron microscope.
For transmission electron microscopy (TEM), samples were transferred to propylene oxide twice for 20 minutes each time.They were placed in a solution containing 50% propylene oxide and 50% Araldite resin (Agar Scientific, UK) overnight, after which they were transferred to 100% resin and infiltrated overnight under agitation.The samples were embedded in moulds containing fresh resin and polymerised at 60uC for 24-36 hours.Ultra-thin sections (50-70 nanometres thick) were cut on a Reichert Ultracut E microtome, collected on naked copper grids and counterstained for 1 hour each with 1% vanadyl sulphate and phosphotungstic acid and then 15 minutes with Reynolds' lead citrate prior to examination on a JEOL JEM 1010 transmission electron microscope.

Clonal analysis of X-inactivation striping patterns
X-gal staining of XLacZ Tg/2 eyes and the acquisition of images have been described previously [19,21].Striping patterns were analysed automatically as described in Mort [66].Photographs of eyes were taken so that the entire cornea was visible and were then cropped to the edge of the corneal epithelium and analysed using ImageJ, a freeware software package designed by Wayne Rasband for the National Institute of Health (NIH), USA (http:// rsb.info.nih.gov/ij/).The observed number of radial stripes in the corneal epithelium was corrected for the probability that stripes would contain multiple adjacent b-gal-positive corneal epithelial clones.This involved dividing the observed mean width by the function 1/(1-p), where p is the proportion of b-gal-positive cells around the circumference as described previously [19][20][21].The corrected stripe number provides an estimate of the total number of active corneal epithelial coherent clones (both b-gal positive and b-gal negative) per circumference.This is useful for comparing numbers of active clones of stem cells between different groups but because the number of stem cells per coherent clone may vary it is not a direct estimate of the number of active stem cells.For the preliminary experiment mosaic corneal patterns were analysed manually at 15 weeks using Adobe Photoshop software as described previously [19].For later mosaic analyses performed at both 15 and 30 weeks, the ImageJ plugin 'Clonal Tools' [66] was used in batch mode to analyse all the images automatically.Where correction for the actual circumference was required this was calculated by dividing the number of corrected stripes by the circumference of each eye measured using ImageJ.

Histology
Whole eyes dissected at 15 and 30 weeks after birth were fixed and stained for b-gal reporter activity using X-gal as described previously [19,21].X-gal stained eyes were embedded in paraffin wax and 7 mm sections were cut on a microtome, mounted on standard microscope slides and counterstained with eosin and neutral red as described previously [21].

Statistics
2-way ANOVAs and Tukey's HSD post-hoc tests were calculated using R statistical software (http://www.r-project.org/).Student's t-tests were calculated using Microsoft Excel.

Figure 6 .
Figure 6.Quantitative comparisons of PAX77 Tg/2 XLacZ Tg/2 and PAX77 2/2 XLacZ Tg/2 eyes at 15 and 30-weeks.(A) WT Eye mass increased significantly between 15 and 30 weeks (2-way ANOVA P, 0.0001, results of relevant post-hoc tests are shown in the figure).(B) Corneal circumference differed significantly between WT and PAX77 Tg/2 at 15 and 30 weeks but the increase in circumference between 15 and 30 weeks was only significant for WT (2-way ANOVA P,0.0001, results of relevant post-hoc tests are shown in the figure).(C) The mean corrected stripe number was significantly higher in the 15-week WT corneas than the 30-week WT or 15-week PAX77 Tg/2 groups, there was a significant decline in stripe number between 15 and 30 weeks in the WT but not the PAX77 Tg/2 group (2-way ANOVA P,0.0001, results of relevant post-hoc tests are shown in the figure).(D) The mean corrected stripe number was significantly higher in the 15-week WT corneas than the 30-week WT or 15-week PAX77 Tg/2 groups, there was a significant decline in stripe number between 15 and 30 weeks in the WT but not the PAX77 Tg/2 group (2-way ANOVA P,0.0001, results of relevant post-hoc tests are shown in the figure).For each comparison there were 14-36 eyes per group: 22 15-week WT, 36 30-week WT, 14 15-week PAX77 Tg/2 and 20 30week PAX77 Tg/2 .Significant P-values for Tukey's HSD post-hoc tests are shown: ns = not significant; *P,0.05;**P,0.01;***P,0.001;****P,0.0001.For all post-hoc tests see Tables S1, S2, S3, S4. doi:10.1371/journal.pone.0028895.g006

Table 2 .
Comparison of maintenance of corneal epithelium in adult Pax6 +/2 and PAX77 Tg/2 mice with low and high doses of Pax6 respectively.