Interaction between hedgehog signalling and PAX6 dosage mediates maintenance and regeneration of the corneal epithelium

Mice

Pax6^+/Sey-Neu (Pax6^+/−) mice were maintained on the CBA/Ca genetic background by heterozygous mating and eyes were usually enucleated from Pax6^+/+ and Pax6^+/− 8 to 10 week old littermates. H253 (XLacZ) mice were as described in Tan et al. [19]. The Gli3 Extra toes mice (Gli3^+/XtJ, henceforth Gli3^+/−) were obtained from Prof. David Price (University of Edinburgh, Edinburgh, UK) and maintained on an inbred CBA/Ca background as for Zaki et al. [11]. PI3Kγ^−/− mice [20] were as described in Zhao et al. [21]. Nude mice (CD1-Foxn1^nu/nu) were purchased from Charles River (Ramsgate, UK). All experiments were performed in accordance with institutional guidelines and UK Home Office regulations and followed the ARVO Guidelines for treatment of laboratory animals.

Chemicals and reagents

Shh peptide was from R&D system laboratories (Abingdon, UK). Cyclopamine was purchased from Calbiochem (Nottingham, UK). Anti-Gli-3, was from Santa-Cruz Biotechnology (Santa-Cruz, CA) and was validated by use of blocking peptide to quench the signal. The 5E1 anti-Shh monoclonal antibody was obtained from the Developmental Studies Hybridoma Bank maintained under the auspices of the University of Iowa (Iowa City, IA), and an anti-Shh goat polyclonal antibody from R&D systems. Anti-Dhh was from Abcam (Cambridge, UK). Anti-BrdU was a mouse monoclonal from Millipore (Watford, UK). Cell culture reagents were from Invitrogen (Paisley, UK).

Boyden chamber migration assay

Eyes were incubated in a solution containing Dispase II (15 μg/ml Dispase II; Invitrogen, Paisley, UK; 17 μg/ml sorbitol) at 4 °C for 18 h. The corneal epithelium was peeled from the stroma and incubated in 0.05% trypsin for 5 min at 37 °C. The cells were then washed, dissociated and resuspended in serum-free SHEM (modified SHEM: DMEM F-12 + Glutamax; Invitrogen, Paisley UK), 0.5% DMSO, 0.5 μg/ml hydrocortizone, 0.1 mg/ml cholera toxin A (Sigma, Poole, UK), 50 μg/ml gentamicin (Invitrogen, Paisley, UK), and 1.25 μg/ml amphotericin B (Invitrogen, Paisley, UK).

Boyden chamber assays were performed according to the manufacturer’s protocol for 1 h at 37 °C with 96 well microchemotaxis chambers (Neuroprobe, Warwickshire, UK) and 5 µm pore polycarbonate membranes (Neuroprobe). Dissociated epithelial cells, (1×10⁵/well) were added to the upper chambers and serum-free medium containing Shh peptide, or medium alone was added to the lower chambers. Cells that did not migrate were subsequently removed by gentle scraping of the upper side of the membrane, whereas the remaining cells were fixed and stained using the Diff-Quik staining kit (Techlab, Blacksburg, VA), and the filters were mounted on a glass slide. Cell migration was assessed by counting migrated cells in ten fields of view per well at 40× magnification with light microscopy.

Ex vivo whole eye wound healing

Pax6^+/+ and Pax6^+/− mice were killed and eyes were wounded using a trephine blade 0.8 mm or 1.5 mm in diameter. The corneal epithelium was debrided from the wound with an opthalmological scalpel. Eyes were enucleated and placed into 1 ml of serum free corneal culture medium (CCM, 15 ml Keratinocyte Growth Medium; Cambrex, Cambridge, UK), 19 ml DMEM:F12, Mercaptoethanol, 1 mM Hepes, Gentamycin (Sigma), according to Hazlett et al. [22]. Where described the culture medium was supplemented with 5 μg/ml Shh peptide reconstituted in 0.1% BSA in PBS. The wound healing rate was measured by fluorescein sodium staining. Eyes were photographed and the wound size measured in Adobe Photoshop 7 (Adobe Systems, San Jose, CA). The diameter of the wound during a 24 h experiment was calculated. The initial wound diameter was measured (X₀) and measurements at each time point (X₆, X₁₂, X₂₄) were subtracted from this reference point. The values plotted represent average diameter of circular wounds within the experimental population. Diameters for individual wounds were calculated as an average of two approximately perpendicular measurements across a single wound.

In vivo cyclopamine injection

Nude mice (Foxn1^nu/nu) of approximate mass 25 g received a subcutaneous injection of 0.2 ml of 4 mg/ml cyclopamine (LC Laboratories, Woburn, MA) in 45% (2-hydroxypropyl)-beta-cyclodextrin (Sigma) vehicle. Control mice received vehicle only. After 1 h, all mice received 0.2 ml of 10 mg/ml bromodeoxyuridine (BrdU) intraperitoneally, and were killed 2 h later, followed by fixation of eyes for 2 h in paraformaldehyde and processing for histological analysis as described below.

Histological analysis

Eyes were fixed in 4% paraformaldehyde (PFA) for 2 h, processed to wax and 7 μm sections were cut. Eyes were sectioned in an anterior-posterior plane to include the cornea and lens. Deparaffinised sections were washed in 100% ethanol, incubated with 3% hydrogen peroxide in methanol for 20 min to block endogenous peroxidise activity. Sections were then rehydrated in ethanol gradient and washed in PBS. Antigen retrieval was performed by boiling in 0.01 M citrate buffer, pH 6, for 20 min. Slides were washed with Tris-buffered saline (TBS: 100 mM Tris, 0.9% NaCl, pH 7.6) and blocked with 4% normal serum (species according to secondary antibody) in TBS, 0.3% BSA for 2 h at room temperature. Sections were incubated in primary antibodies overnight at 4 °C (Gli-3 as above, diluted 1:100) and BrdU (diluted 1:400). Slides were washed in TBS and then treated with secondary antibody (biotinylated rabbit anti-mouse IgG diluted 1:200 in blocking buffer. Slides were then treated with avidin-biotin complex (ABC; Dakocytomation UK Ltd., London, UK) for 30 min and washed with TBS. The bound antibody was then visualized by 3,3′-diaminobenzidine (DAB) stain (5.9 ml of 20 mM Tris, pH 7.6, 100 μl of 50 mg/ml DAB, 1 μl of H₂O₂). For negative control, primary antibody was omitted and for Gli3, the antibodies were incubated with their blocking peptides before addition to slides, to demonstrate specificity of signal. X-Gal staining was performed on whole mount tissue following Collinson et al. [23]. Mosaic patterns in corneas of 15-week old XLacZ X-inactivation mosaics were analysed as described previously [8,23] but ImageJ was used for the quantification (see Results section).

In situ hybridization

Digoxygenin-labeled riboprobes were transcribed from linearized Shh BSll-SK^- and Ptch1 cDNA (a kind gift from Dr. Christopher Hayes, Medical Research Council Mouse Genome Centre, Harwell, UK) using the DIG RNA-labeling kit (T3/T7; Roche Diagnostics Ltd., Burgess Hill, UK). In situ hybridization was performed on 7-μm sections hybridized overnight at 65 °C in a chamber humidified with 50% formamide and 5×SSC (Sigma). Slides were washed twice to remove unbound probe with 5×SSC for 2 h at 65 °C then in 0.1× SSC for 15 min at room temperature then in 50 mM Tris-HCl, 150 mM NaCl, pH 7.5. Samples were blocked in 2% blocking reagent in 0.1 M Maleate buffer, pH 7.5, 150 mM NaCl for 30 min and incubated with alkaline phosphatase conjugated anti-DIG antibody, diluted 1:500 for 1 h at room temperature. Sections were washed for 10 min in 100 mM NaCl, 50 mM MgCl₂, 1 mM levamisole, 100 mM Tris-HCl, pH 9.5 then immersed in a buffer containing 3.5 μl/ml 5-bromo-4-chloro-3indolyl-phosphate (BCIP, 9.4 mg/ml) and 5 μl/ml 4-nitroblue tetrazolium chloride (NBT, 18.75 mg/ml) at room temperature in the dark until color developed. Slides were then briefly washed in PBS, re-fixed in 4% PFA for 10 min, washed in PBS and mounted in 80% glycerol.

RT-PCR

Corneas (4 per genotype and/or treatment) were pooled and homogenized. RNA was isolated using the PeqGOLD total RNA isolation kit (Peqlab, Southampton, UK), with reverse transcription using Superscript II (Invitrogen, Paisley, UK) according to the manufacturers’ instructions with oligo-dT and random primers. PCR was performed on cDNA using the primers described in Table 1, with annealing temperature of 58–62 °C (optimized for each primer set). All primers were design to span introns to avoid false positives due to residual genomic DNA.

Western blot

One dimensional denaturing 12% PAGE was used to detect proteins in cell lysate. Prior to gel electrophoresis lysed protein samples were heated at 95 °C for 5 min. Gels were blotted onto nitrocellulose membrane (GE Healthcare, Little Chlafont, UK) in transfer buffer using Bio-Rad Mini-Protean III system (Berkeley, CA) for 1.5 h at 100V at 4 °C. Membranes were then washed in TBS 0.3% Tween followed by pre-blocking in 10% milk in TBS 0.3% Tween for 1 h at room temperature before incubation in primary antibody diluted in 0.5% skimmed milk in TBS 0.05% Tween-20, overnight at 4 °C. Unbound primary antibody was washed off with TBS 0.05% Tween-20 for 3×15 min at room temperature followed by incubation in secondary (horseradish peroxidise-conjugated) antibody for 1 h. Excess of secondary antibody was washed in TBS, 0.05% Tween-20 for 3×15 min at room temperature. Detection was performed using Immobilon™ chemiluminescent HRP substrate (Millipore, Watford, UK) and Amersham ECL Hyperfilm (GE Healthcare, Little Chlafont, UK) according to the manufacturers’ instructions.

Statistical analyses

Statistical analyses were using either Microsoft Excel (Mountain View, CA), or the 'R' statistics package, an open source software package based on the 'S' programming language, using the package 'stats' (R Development Core Team, 2008).

The migratory response of corneal epithelial cells to exogenous Shh requires normal Pax6 dosage

Topical application of Shh has already been reported to improve corneal epithelial wound healing in wild-types [17], but its effect on Pax6 mutants has not been investigated, in spite of its therapeutic potential for corneal disease. We investigated the stimulatory effect of Shh on corneal epithelial wound healing by determining the rate of wound closure in both Pax6^+/+ and Pax6^+/− littermates in serum-free media supplemented with 5.0 μg/ml Shh. For each animal, one eye was treated in Shh and the contralateral eye was wounded concurrently and incubated in medium without Shh (Figure 1A). Exogenous Shh produced a small but significant increase in the rate of wound healing in Pax6^+/+ eyes 24 h post injury (control 17.9±1.4 μm/h, Shh 22.1±1.2 μm/h, n=8, paired t-test: p=0.007; Figure 1B). In contrast, Pax6^+/− eyes showed a reduced rate of healing and no response to exogenous Shh (control 10.33±1.7 μm; Shh 10.99±1.1 μm/h; n=5; p=0.32; Figure 1B). Application of 200 μM cyclopamine in 1.5% DMSO, a treatment that in other systems has the effect of blocking the canonical hedgehog pathway by antagonising the Hedgehog protein co-receptor, Smoothened [24], did not significantly reduced the rate of wound closure after 24 h in wild-types (17.7±2.8 μm/h; n=6). Therefore our data suggest that although exogenous Shh accelerates wound healing in wild-type corneas, endogenous hedgehog signaling is not required for normal wound closure.

Exogenous Shh increases corneal epithelial proliferation irrespective of Pax6 dosage

Hedgehog signaling can act as a mitogen, which in part underlies the oncogenic activity of the signaling pathway. The potential role of exogenous Shh for increasing corneal epithelial proliferation after wounding was investigated. Corneal epithelia were wounded as described above and allowed to heal with or without exogenous Shh as above. Bromodeoxyuridine (BrdU; 10 mg/ml) was added to the medium for 6 h before fixing the tissue to assay proliferation (Figure 1C,D). Exogenous Shh increased the rate of cell proliferation in the basal layer of the wild-type corneal epithelium: 4.07±0.47% of cells were BrdU-positive in controls in serum-free medium, compared to 6.56±0.89% in Shh-treated corneas (n=3 of each genotype; Mann–Whitney U test, p=0.009). Shh also increased proliferation rate in Pax6^+/− corneas after wounding, from 5.01±0.38% to 6.75±0.64% (n=3 of each genotype; Mann–Whitney U test, p=0.006) showing that Pax6^+/− corneal epithelia can respond to Shh application by increasing their proliferation rate. In vivo, injection of adult mice with 32 mg/kg cyclopamine completely inhibited corneal epithelial cell proliferation (Figure 2; n=3 cyclopamine-treated, 3 controls). Hence pharmacological upregulation or downregulation of Hedgehog signaling, which leads to increased or decreased mitosis, does not necessarily translate to a increase or decrease in wound healing rate (as shown above). It was concluded that the effect of Shh application on increasing the early stages of healing in wild-type corneas is not likely to be a function of its mitogenic effects. Although Pax6^+/− corneal epithelia increase their proliferation rate when Shh is applied, this has no effect on their rate of wound healing.

Shh promotes corneal epithelial cell migration

Data above suggested that changes in wound healing rates seen in wild-type corneal epithelia following Shh application were unlikely to be due to an effect on cell proliferation. We hypothesized that Shh directly increases the migratory potential of corneal epithelial cells. This was investigated for dissociated corneal keratinocytes in a Boyden chamber migration assay. It was found that wild-type mouse corneal epithelial cells migrated toward Shh (5 μg/ml), which produced a 1.72 fold increase in migration compared with control serum-free media (n=27 replicates, p<0.001; Figure 3). Epithelial cells also migrated toward other growth factors including EGF (50 ng/ml), which produced a 1.33 fold increase over serum-free control (Figure 3, n=12, p=0.0002). These data were comparable to the effect of serum or FGF (50 ng/ml) that is known to stimulate epithelial cell migration.

This experiment showed that Shh stimulates the migration of wild-type corneal epithelial cells. Pax6^+/− corneal epithelial cells could not be tested in the Boyden chamber for technical reasons (see Discussion).

Activity of PI3Kγ is not required for the increased healing rate after Shh addition

It was shown previously that Shh peptide can induce PI3K-dependent activation of Akt [16,25,26]. PI3Kγ has previously been shown to be necessary for optimal corneal epithelial cell migration [21]. We examined wound healing in PI3Kγ^−/− mice, with and without application of exogenous 5 μg/ml Shh. The results are not directly comparable with our other data because the mice are on a different genetic background (C57BL/6) but wound healing in control serum-free medium was slow (13.4±3.6 μm/h). In contrast, with application of Shh the rate of epithelial wound healing over 12 h rose significantly to 28.8±4.7 μm/h (t-test: p=0.014). Hence although application of Shh activates Akt, PI3Kγ is not required for the accelerated wound healing response to exogenous Shh.

Expression of Shh signaling pathway components in corneal epithelium

Multiple canonical Shh signaling components were detected by RT–PCR in wild-type and Pax6^+/− corneas, before and after injury (Figure 4), suggesting that the failure of Pax6^+/− corneas to increase wound healing rate in response to Shh application was not attributable to failure of pathway gene expression. By RT–PCR we were able to demonstrate mRNA for Dhh in uninjured wild-type and Pax6^+/− adult corneas, but no expression of the genes encoding Shh or Ihh ligands (Figure 4). These data are consistent with an earlier study of Hedgehog family ligands in adult ocular tissues [27] and suggest that Dhh is the predominant ligand. However, Shh expression was induced in wild-type and Pax6^+/− corneas assayed 2 h after injury. Expression of receptors Ptch1 and Smo and the signaling intermediates Su(fu) and GSK3β were detected in all adult corneal preparations (Figure 4). Canonical Shh signaling is mediated by Gli family transcription factors, and we were able to demonstrate expression of Gli2 and Gli3, but not Gli1 in the wild-type cornea by RT–PCR. In contrast to wild-type, Gli1 expression was detected in uninjured and injured Pax6^+/− corneas. Gli1 upregulation is normally taken as an indicative readout of active Hh signaling, and the significance of this observation is not yet known.

Although Shh message was detectable by RT–PCR in wounded corneas, expression must be at very low levels because western blot of corneal epithelia, wounded in vivo, detected no Shh protein using the 5E1 anti-Shh monoclonal antibody (Developmental Studies Hybridoma Bank; Figure 5A). Major bands at around 50 kDa and 22 kDa corresponding to the uncleaved and NH₂-terminal Shh peptides were detected from embryonic notochord and neural tube. The antibody used in previous work (goat anti-mShh; AF464; R&D Systems) [17], was found to be non-specific. Western blotting of wounded corneas with this antibody yielded 16 bands between 64 and 16 kDa, and neither of the major bands corresponded to known sizes of Shh peptides. No Shh protein was detected in the corneal or limbal epithelia using the 5E1 monoclonal antibody (Figure 5B,C) in contrast to strong signal in the embryonic notochord and floor plate (Figure 5D). In situ hybridization showed Ptch1 upregulation in mesenchyme surrounding the notochord (an assay for cells where Hh signaling is active; Figure 5E). At the healing edge of wounded corneal epithelia, Dhh protein was detected but no Shh labeling was visible using the 5E1 monoclonal antibody (Figure 5F,G). There was some evidence for weak Ptch1 upregulation, suggesting active Hh signaling (Figure 5H). It was concluded that there is little endogenous limbal or wound edge localization of Shh protein in adults and that Dhh is probably the major corneal Hedgehog family ligand.

Immunohistochemistry using an antibody directed at the COOH-terminus of Gli3 confirmed that activator form of Gli3 was expressed predominantly in cell nuclei of the corneal epithelium (Figure 6A). Nuclear staining appeared weaker in Pax6^+/− littermates (Figure 6B). The NH₂-terminal product of Gli3 was localized in cytoplasm of Pax6^+/+ epithelial cells but was also weaker in Pax6^+/− littermates (Figure 6C,D).

Gli3 deficiency exacerbates the corneal migration defect resulting from Pax6 heterozygosity

Immuohistochemistry suggested that Gli3 nuclear localization was disrupted or absent in Pax6^+/− corneal epithelia. Previously, we and others have shown that although Gli3^+/− mice have only very subtle eye defects, Pax6^+/− Gli3^+/− double heterozygotes have ocular defects that are much more severe than either single heterozygote phenotype [11]. To determine whether heterozygousity for Gli3 affects corneal epithelial cell migration, both mutations were bred onto the H253 XLacZ background. This reporter mouse carries an X-linked LacZ transgene under the control of a housekeeping promoter. Due to X-inactivation during early embryogenesis, females show mosaic β-galactosidase expression in all tissues [19]. Their corneas display a radial pattern of blue striping that appears to reflect centripetal migration of streams of cells from the limbus during adult life, and has previously been shown to be disrupted in Pax6 heterozygotes [4,8,23].

Hemizygous XLacZ^Y/+ Gli3^+/− male mice were mated to Pax6^+/− females and the resulting female littermates were genotyped and their eyes stained with XGal at 15 weeks of age. At least 10 mice of each genotype were analyzed. The pattern of corneal epithelial staining in Pax6^+/+ Gli3^+/− XLacZ adults was found to be grossly similar to Pax6^+/+ Gli3^+/+ XLacZ littermates (Figure 7A). That of Pax6^+/− Gli3^+/+ XLacZ adults showed irregularities in the radial striping, of variable severity, as described previously. In contrast, patterns of LacZ expression in Pax6^+/− Gli3^+/− XLacZ double heterozygotes were very highly disrupted with a patchwork organization and little evidence in some cases of any radial orientation or migration at all (Figure 7A). Wild-type XLacZ corneal epithelia often display a spiral pattern of cell migration as noted previously [23], but in one Pax6^+/− Gli3^+/− case there was a double spiral at the center (right hand Pax6^+/− Gli3^+/− eye in Figure 7A) which suggested particular instability of cell migration patterns. By counting the mean number of blue and white stripes and correcting for the total proportion of blue corneal area it is possible to estimate the number of coherent clones of limbal epithelial cells (LSCs) that maintain the corneal surface [8,28]. Analysis of the corrected patch number around the circumference of the four genotypes generated from the above cross was performed using the freeware image analysis software package ImageJ with the 'Clonal Tools' plugin described in [28]. The analysis showed a statistically significant trend when the corrected patch number (expressed per millimeter of corneal circumference) was compared by one-way ANOVA (p<0.02). Subsequent post-hoc tests revealed no significant change in the number of corrected patches per millimeter of the corneal circumference in Pax6^+/+ Gli3^+/− adults (Mean±95% confidence interval [C.I.]: 9.66±1.72; n=16) compared to wild-type litter-mates (10.4±1.97; n=17; Tukey HSD: p=0.96). However the corrected patch number in the Pax6^+/− Gli3^+/+ single heterozygotes (6.66±1.44; n=27) was lower than for wild-type or Pax6^+/+ Gli3^+/− single heterozygous littermates although the difference only reached statistical significance for the comparison with wild-types (Tukey HSD: p=0.024; Figure 4B). The lack of clear stripes in Pax6^+/− Gli3^+/− XLacZ double heterozygotes made the analysis less easy to interpret but counting the number of blue and white patches of cells at the corneal circumference produced an estimate of 9.43±1.67 (n=31) corrected patches in the double heterozygotes; this was not statistically significant when compared to the wild-type group, (p=0.87).

The data confirm a functional genetic interaction between Hedgehog signaling and Pax6 dosage in the corneal epithelium. Because the Pax6^+/− Gli3^+/− XLacZ double heterozygotes do not always form stripes a change in the number of active LSCs can not be inferred however it is clear from the data that Gli3^+/− heterozygotes have migratory defects that are only visible on the Pax6^+/− background and are distinct from animals that are heterozygous for Pax6^+/− or Gli3^+/− alone. This is consistent with evidence for a genetic interaction between Pax6 and hedgehog signaling in the retina [11].