A dual role of FGF10 in proliferation and coordinated migration of epithelial leading edge cells during mouse eyelid development

During normal development in mammals, the eyelids grow across the eye, fuse and subsequently reopen. In mice, eyelid formation begins on day 11.5 of gestation (E11.5), and from E14 to 16 the eyelids grow, flatten across the eye, progressively meet beginning at the inner and outer canthi and fuse tightly with each other (Harris and McLeod,1982; Li et al.,2001). The eyelashes and the glands lying along the margins of the lids start to differentiate from this common epithelial lamina before the lids reopen at 14 days after birth (Findlater et al., 1993).

The developing eyelids are composed of loose mesenchyme covered by an epithelial sheet, the epidermis (outer surface) and conjunctiva (inner surface) and the periderm, which covers the epidermis(Weiss and Zelickson, 1975). Only the peridermal and epidermal layers are involved in eyelid fusion; the mesenchymal layers of the upper and lower eyelids remain separate(Pei and Rhodin, 1970). A profusion of rounded periderm cells appears, and they pile up at the leading edges of the advancing eyelids during eyelid growth(Harris and McLeod, 1982; Harris and Juriloff, 1986; Juriloff and Harris, 1989). Once contact is made between the apposed eyelids, these cells flatten and form a strip along the fusion line, until they slough off with the rest of the periderm on day 17 of gestation (Harris and Juriloff, 1986; Juriloff and Harris, 1989; Findlater et al.,1993).

Failure of the eyelids to grow across the eye and fuse during the fetal stage in mice leads to a birth defect of open-eyelids at birth. Mutations at several distinct loci have been found to cause such open-eyelids, often as part of a syndrome with other defects. For example, open-eyelids results from spontaneous or gene-knockout mutation at the loci of transforming growth factor α (Tgfa), its receptor epidermal growth factor receptor(Egfr), activin/inhibin βB (Inhbb – Mouse Genome Informatics) fibroblast growth factor receptor type 2b (Fgfr2b), Jun, MEK kinase 1 and some forkhead genes(Luetteke et al., 1993; Mann et al., 1993; Miettinen et al., 1999; Thereadgill et al., 1995; Vassalli et al., 1994; Celli et al., 1998; De Moerlooze et al., 2000; Li et al., 2001; Li et al., 2003; Zenz et al., 2003; Zhang et al., 2003; Kume et al., 1998; Uda et al., 2004). However,the molecular and cellular events occurring during eyelid development and the interactions among the signaling molecules have not been fully elucidated. We previously reported briefly that Fgf10-null mice exhibit open-eyelids at birth with multi-organ developmental defects(Sekine et al., 1999; Ohuchi et al., 2000). Here, we report the expression pattern of Fgf10 during eyelid formation and the phenotype of Fgf10-null eyelids in detail. We found that FGF10 is dually required for proliferation and coordinated migration of epithelial cells during mouse eyelid development by reorganization of the cytoskeleton,through the regulation of activin, TGFα and SHH signaling.

Fgf10-knockout mice were generated on a C57BL/6 × CBA background and genotyped as previously described(Sekine et al., 1999). In the Fgf10-null mice, the phenotype of open-eyelids at birth is seen in homozygotes, while the eyelids of heterozygotes appear normal at birth, as reported previously (Sekine et al.,1999). So far, all examined homozygous mice have exhibited open-eyelids at birth. Noon of the day when the vaginal plug was detected was considered embryonic day 0.5 (E0.5) of development for embryos of the overnight mating. For analysis of eyelid morphogenesis, the heads of Fgf10^–/–, Fgf10^+/–and wild-type mice were collected at the desired stages (n=3∼5 for each genotype and stage). For analysis of the normal expression patterns of Fgf10 and other genes, the heads of Slc:ddy mice (no pigment in the eye) (Japan SLC, Inc.) were used at the desired stages(n=3 for each stage).

For histological analysis, Hematoxylin and Eosin (HE) staining was performed according to the standard procedure.

The Fgf10^+/+, Fgf10^+/– and Fgf10^–/– embryos at around E15 and E16(n=3∼6 for each genotype and stage) were fixed in 4%paraformaldehyde (PFA)/5% glutaraldehyde/PBS (10 mM phosphate-buffered saline)overnight and processed for scanning or transmission electron microscopy observation according to standard procedures.

Normal and Fgf10^–/– mutant embryos at the desired stages were fixed in 4% PFA in PBS overnight and used for in situ hybridization. Digoxigenin-labeled riboprobes for mouse Fgf10, Shh(Sekine et al., 1999), Patched(Ptch1), Ptch2 (Motoyama et al., 1998) and activin βB (provided by Dr Tsuchida,University of Tokushima, Japan) were prepared as described. The mouse Tgfa cDNA (Tgfa; 2578 bp) was isolated by RT-PCR using E18.5 mouse whisker mRNA. The PCR primers were as follows: the sense primer was 5′-tgtgtctgccactctgggtacgtg-3′ and the antisense primer was 5′-aacgcagcagggctgtcatacgtc-3′. Sense probes were used as a control and produced virtually no signals. Whole-mount in situ hybridization was performed as previously described (Tao et al., 2002). Section in situ hybridization was performed on 18-μm thick frozen sections according to standard procedures, or by using a tyramide signal amplification method on 7-μm thick paraffin sections(Yang et al., 1999). The gene expression patterns were compared between littermates, and the in situ hybridization experiment was repeated at least three times for each gene.

Total RNA was isolated from the back skin (of normal E18.5 mice), wild-type and Fgf10-null keratinocytes (at E18.5) and eyelid mesenchyme [of normal mice at E15, treated with Dispase II (Roche) to remove epithelial tissues] by using an RNAqueous Kit (Ambion). To prevent contamination of the genomic DNA, the samples were treated with RNase-free DNase (Promega). The DNase was subsequently inactivated at 70°C for 5 minutes, and the samples were subjected to chloroform/phenol extraction and ethanol precipitation. Reverse transcription (RT) was carried out using 1 μg of total RNA,Superscript II reverse transcriptase (Invitrogen), and gene-specific primers as follows: MA-5′ primer (ATGACCCAGATCATGTTTGAGACC) and MA-3′primer (AGGAGGAGCAATGATCTTGATCTT) for β-actin (645 bp); Fgf10-Se primer (AAGCTCTTGGTCAGGACATGG) and Fgf10-An primer(ATGGGGAGGAAGTGAGCAGA) for Fgf10 (506 bp); and Fgfr2b-Se primer (ACACCGAGAAGATGGAGAAG) and Fgfr2b-An primer(GTTTGGGCAGGACAGTGAG) for Fgfr2b (609 bp). PCR was performed using Ex Taq polymerase HS (Takara Bio, Japan) and 1/10 of the volume of the cDNA reaction mix. In total, 35 cycles were performed at annealing temperatures of 60°C for Fgf10 and 65°C for Fgfr2b. The analysis was repeated twice with samples from two different fetuses for each genotype, and all gave the same results.

To analyze BrdU uptake in embryos, pregnant mice were injected intraperitoneally with BrdU (Roche) at a dose of 100 μg/g body weight and were sacrificed 1 hour later. Immunohistochemical staining for BrdU was performed using a monoclonal antibody (G3G4, 1:100) from Developmental Studies Hybridoma Bank, a M.O.M. kit (Vector) and a NovaRed substrate (Vector)according to the manufacturers' instructions. At E11.5, the BrdU-positive cells were counted in the epithelium (along the line shown in Fig. 4A,B, approximately 200μm long in each eyelid) and in three serial sections (21 μm) for each embryo (wild type, n=5; Fgf10^–/–, n=4). Since the total number of cells was not found to differ between genotypes, the number of BrdU-positive cells did not reflect varying cell density. The epithelial areas counted for BrdU-positive cells at E13.5 are indicated in Fig. 4D,E. The percentage of BrdU-positive cells was calculated (wild type, n=3; Fgf10^–/–, n=3). The epithelial areas counted for BrdU-positive cells at E15 are indicated in Fig. 4I,J. The percentage of BrdU-positive cells was again calculated (wild type, n=4; Fgf10^–/–, n=3). The means and standard errors of the means (s.e.m.) were calculated from the pooled data. Differences were judged significant if P<0.05 (as shown by the asterisks in Fig. 4), as determined by Student's t-test.

A TUNEL assay of apoptotic cells on tissue sections was carried out as recommended by the manufacturer (ApopTag, Intergen). The sections were pretreated with 0.5% Triton X-100, and diaminobenzidine was used as a substrate for horseradish peroxidase.

Immunostaining for γ-tubulin was performed essentially according to the procedure of Thompson et al. (Thompson et al., 2004). Briefly, heads from wild-type and Fgf10-deficient embryos at around E15 were cut at the midline and fixed overnight in 3% formaldehyde in PBS after Triton X-100/PEG treatment(Libusova et al., 2004). Cryosections with a thickness of 18 μm were prepared and processed for immunofluorescence. The anti-γ-tubulin (Clone GTU-88; Sigma, diluted at 1:1000) and anti-vimentin (Clone Vim3B4; Dako, diluted at 1:200) monoclonal antibodies were visualized by using Cy3-conjugated goat anti-mouse IgG(Jackson, diluted at 1:1000). To illuminate the cell boundaries,Bodipy-ceramide (FLC5; Molecular Probes) was used, according to Lele et al.(Lele et al., 2002), at a concentration of 10 μM. The number of γ-tubulin-expressing cells in the basal epidermis of the eyelid tip (approximately 17 cells along the broken line shown in Fig. 7J,K) was calculated.

Staining for F-actin of flat mounts and frozen sections (18 μm) was carried out using eye samples from E15 embryos fixed in 4% paraformaldehyde. The tissues were incubated with Rhodamine-phalloidin (Molecular Probes) and observed with a fluorescence stereomicroscope (Leica) or by laser scanning confocal microscopy (BioRad) according to standard procedures.

Isolation and culturing of primary keratinocytes were performed basically according to Li et al. (Li et al.,2003) with minor modifications. Briefly, E18 embryos were collected, and their skin removed, washed and incubated in Dispase medium[defined keratinocyte-serum free medium (DK-SFM; Invitrogen), 5 U/ml Dispase II, 100 μg/ml streptomycin, 0.25 μg/ml fungizone, 50 μg/ml gentamicin] at 4°C for 18 hours. The dermis was separated from the epidermis, and the epidermis was minced and digested in 0.05% trypsin-EDTA for 10 minutes. A mouse keratinocyte culture medium containing DK-SFM, 10 ng/ml EGF (Sigma) and 10 ng/ml choleratoxin (Wako, Japan) was used. In the case of the E18 embryonic keratinocytes, we did not observe a distinct difference in motility with or without EGF.

To determine the cell motility, wild-type and mutant keratinocytes were seeded onto 6-well culture dishes or chamber slides (Nalge Nunc), grown to confluence and transferred to a growth factor-free medium plus mitomycin C(Sigma) for 2 days. The confluent monolayers were wounded using a disposable Pasteur pipette tip (Iwaki, Japan) and EGF was added again. For immunostaining, the cells were washed twice with PBS and fixed in 3.7%formaldehyde solution in PBS for 10 minutes at room temperature.

Eyelids (with the anterior segment of the eye) at E15 were cultured at the air-fluid interface by placing them on 0.4-μm Milli Cell-CM (PICM 03050,Millipore, Bedford, MA, USA) in 6-well plates containing DK-SFM. The organ cultures were maintained at 37°C under 100% humidity and 95% air-5% carbon dioxide for 1 day. Heparin-coated acrylic beads (H5263, Sigma), 250-300 μm in diameter, were incubated in 0.5 mg/ml recombinant human FGF10 (Peprotech)at 37°C or 40 minutes and then stored at 4°C before being placed on the explant. For control experiments, beads were soaked in PBS according the same protocol. An FGF10-soaked bead or PBS-bead was inserted into the eyelid mesenchyme. The distance between the eyelid margin and the bead was estimated,from microscope observation, to be approximately 300 μm.

The eyelid closure percentage was calculated by measuring the area of the epithelium covering the cornea (Fig. 9M) at the beginning of the culture and after 8 hours, using the NIH Image program(http://rsb.info.nih.gov/nih-image/). The means and s.e.m. values were calculated from the pooled data (FGF10-bead, n=3; PBS-bead, n=2).

The processes of mouse embryonic eyelid development are shown in Fig. 1. They include the following: initiation (ectoderm morphogenesis and groove formation; E11.5),eyelid mesenchymal protrusion (E13.5), protruding epithelial ridge formation at the tip of the eyelid margin (E15) and subsequent extension of the upper and lower eyelid epithelium first (E15∼16) and mesenchymal cells later(E16.5∼17.5) (Li et al.,2001; Stepp,1999). The upper and lower eyelid fusion occurs in the epithelium,and the mesenchymal cells are not fused for subsequent reopening. The once-fused eyelid normally re-opens at 2 weeks after birth. As Fgf10-null mice die at birth because of the absence of lung formation(Sekine et al., 1999), we speculated that FGF10 might be required for certain developmental steps that lead to eyelid closure.

To specify the role of FGF10 in eyelid development, we sought to determine its expression pattern in mouse eyelid primordia. It has been reported that Fgf10 is expressed in the mesenchyme of developing eyelids at E12.5(Li et al., 2001). Its expression domain and profile, however, have not been fully described. Therefore, we re-examined the expression pattern of Fgf10 until eyelid fusion. At E11.5, Fgf10 was expressed in the mesenchyme underneath the epithelium of the emerging eyelid groove(Fig. 2A,B). By E13.5, Fgf10 was expressed in the eyelid mesenchyme just beneath the epithelial tip and in the developing corneal stroma(Fig. 2D-F). At E15, it was expressed around the eye (Fig. 2G), in the eyelid mesenchyme(Fig. 2H,I). By contrast, the major receptor for FGF10 during organogenesis, Fgfr2b, was expressed in the eyelid epithelium (Fig. 2J-L). RT-PCR analysis using E18.5 keratinocyte RNA verified that Fgf10 was not expressed in the epithelium but Fgfr2b was expressed there (Fig. 2M). These expression patterns suggest that Fgf10 may be involved in eyelid development from a very early stage, and that in the absence of FGF10 primary defects may be found in the eyelid epithelium.

The Fgf10-null embryos had no eyelids at birth, but eyelid protrusion was distinctly observed (Fig. 3C,D; see A,B for control). Taken together with the expression pattern of Fgf10 described above, FGF10 may be required for maintenance of eyelid formation, rather than for induction of eyelid anlagen. Other significant differences at this stage involved the developing corneal layers. In Fgf10^–/– eyes, the corneal stroma appeared less organized (Fig. 3Q). The role of FGF10 in corneal development will be described in detail elsewhere.

To determine which processes require FGF10 during eyelid development, we examined the developmental processes of mutant eyelid primordia by comparing them with normal ones. The initiation stages of eyelid development include ectoderm morphogenesis and groove formation. The ectoderm morphogenesis initiates at E11.5 when the flat ectoderm cells above and below the optic vesicle undergo morphogenetic changes to form cube-shaped epithelial cells. The epithelium also starts to form small eyelid grooves above and below the eye, which were obvious at E11.5 in the wild type(Fig. 3E,F). In the mutant embryos, however, the initial morphogenetic change from flat ectoderm cells to cuboid epithelial cells was not observed, as the mutant cells maintained a flat appearance (Fig. 3G,H). As the eyelid grooves deepened by E13.5 in the wild type, mesenchymal cells nearby started proliferating to form primitive eyelids with protruding ridges of epithelium (Fig. 3I-K). However, the mutant eyelid anlagen appeared much smaller, the groove was shallower and the epithelium overlying the eyelid mesenchyme remained flat(n=3; Fig. 3L-N).

To investigate the mechanisms underlying defects of eyelid primordia, we assessed the number of proliferating cells using a BrdU incorporation assay. At E11.5, the mutant epithelium showed a dramatic decrease in proliferation,as only 17.1±1.9 cells incorporated BrdU, while 41.2±2.6 wild-type cells were BrdU positive in the lower eyelid (see Materials and methods; n=4; Fig. 4A-C). The number of proliferating cells in the upper eyelid epithelium was significantly reduced as well (P<0.0005; Fig. 4C). To check whether the reduced protrusion of the Fgf10-null eyelid was exacerbated by an increase in apoptosis, we also analyzed sections using an in situ TUNEL assay. No prominent changes in the apoptotic cell numbers were detected in Fgf10-null eyelids compared with their wild-type littermates (data not shown). These observations indicate that the loss of Fgf10interferes with eyelid formation by E11.5 due to impaired cell proliferation.

By E16, the normal eyelids were fused closed(Fig. 3O), whereas the mutant eyelids remained wide apart, above and below the eye, leaving the cornea exposed (Fig. 3Q). Distinct accumulation of periderm cells was seen in the normal eyelid(Fig. 3P). In the mutant eyelid, on the other hand, a protruding ridge was formed at the leading edge of the lower eyelid primordium (Fig. 3S) but did not continue to grow later, and a round malformed eyelid leading edge was often observed in the mutant(Fig. 3R).

Scanning electron microscopy (SEM) observation of E15 normal eyes showed that rounded periderm cells were present in clumps all around the eyelid margin (Fig. 5A; B for Fgf10^–/– mutant). In Fgf10-null eyelids, however, typical peridermal clumps were merely seen at the inner canthus (Fig. 5D; C for comparison). The periderm cells were rather scattered on the upper and lower eyelid margins of the mutant (Fig. 5F;E for comparison), and they were hardly seen at the outer canthus(Fig. 5H; G for comparison). Transmission electron microscopy further revealed the formation of filopodia in the leading edge periderm cells (Fig. 5I). It is known that filopodia are pivotal for epithelial fusion:they scan the opposing leading edge, playing an integral role in finally knitting the epithelial hole closed. In the Fgf10^–/– mutant leading edge, epidermal cells still developed filopodia, although these were distinctly fewer and shorter(Fig. 5J) (n=3). Thus,FGF10 is not required for the formation of filopodia per se during eyelid fusion, but seems necessary for their growth and maturity. At E16, SEM observation verified that the normal eyelids were fused with the epidermis that was streaming towards the point of fusion, where periderm cells accumulated along the junctional region(Fig. 5K). In contrast,wide-open, Fgf10-null eyelid rudiments had an epithelial ridge on their margin, whose subsequent growth collapsed(Fig. 5L).

A decrease in proliferating cells of the epithelium was not detected in the Fgf10-null epithelium at E15 (Fig. 4I-M), although it was detected at E13.5 (P=0.03)(Fig. 4D-H). The TUNEL assay did not show any differences in the numbers of apoptotic cells between the mutant and normal eyelid territories at either E13.5 or E15 (data not shown). Thus, Fgf10-null eyelids fail to maintain peridermal clumps on the lid margin at around E15, which may results from cellular events independent of cell proliferation and apoptosis.

It has been reported that cell migration is a crucial process during eyelid epithelial fusion, in which EGF/TGFα and/or activin-Jun cascades are involved (Li et al., 2003; Zenz et al., 2003; Zhang et al., 2003). Therefore, it is possible that FGF10 controls the epithelial cell migration required for eyelid closure. To assess the role of FGF10 in cell migration during eyelid closure, we isolated primary keratinocytes as major constituents of the eyelid epidermis from wild-type and Fgf10-null fetuses, and cultured cells with EGF as the only growth factor supplement. To determine whether FGF10 signaling might be involved in cell migration, we employed an in vitro scratch assay, in which a `wound' was introduced in cultured monolayers of keratinocytes (Fig. 6A);these cells were cultured in the presence of mitomycin C and were thus mitotically inactive. In the wild-type cultures, the cells responded to the scratch by migrating into the gap and effected closure within 20 hours. The Fgf10-null keratinocytes also migrated into the wound in considerable numbers during the same period (Fig. 6A).

We then characterized the migration response to the scratch by observing the cellular levels and distribution of F-actin, which is essential for cell movement. Staining of F-actin with phalloidin (within 5 hours) revealed that in wild-type leading edge cells, the actin distribution was polarized and formed into stress fibers, which accumulated in the anterior lamellipodia oriented toward the wound (Fig. 6B). In Fgf10-null cells, the formation of lamellipodia was observed, but their orientation appeared rather irregular; they did not face the gap (Fig. 6C). The formation of filopodia scanning the opposing leading edge was discernible in the mutant as well as the wild type (Fig. 6D,E). The filopodia in the mutant cells, however, appeared thinner than those in the wild-type cells. Thus, Fgf10-null keratinocytes exhibit normal motility and ability to form filopodia and lamellipodia as revealed by an in vitro scratch assay.

To test whether FGF10 also regulates actin polymerization in the developing eyelid epithelium, we examined the formation of actin filaments in the eyelid tissues of E15 fetuses. In both wild-type and Fgf10^+/– fetuses, the eyelid epithelial cells developed prominent F-actin networks as demonstrated by whole-mount phalloidin staining (Fig. 7A,D-F). By contrast, in the homozygous mutant, only a few cells that were mostly confined to a single cell layer at the eyelid tip, formed actin cables(Fig. 7G). Histological sections revealed that F-actin accumulated in the leading edge cells of the wild type (Fig. 7B,C), but not of the mutant (Fig. 7H,I). These results demonstrate that FGF10 regulates actin stress fiber formation in epithelial leading edge cells of the developing eyelid, which is probably associated with epithelial movement and eyelid closure.

Since cell polarization is used to mediate physical fates, as in orientated cell migration (for review, see Macara,2004), we further examined the polarity of the eyelid epithelial cells in Fgf10-null eyelid tips. We performed immunostaining ofγ-tubulin to reveal a centrosome: a microtubule-organizing center(MTOC), localized apically in the epithelial cell(Rizzolo and Joshi, 1993). The initial stratification of the single-layer ectoderm during embryonic development gives rise to an outer periderm layer and an inner basal layer. In the wild type, γ-tubulin was localized in the apical side of the epidermal cells in the basal layer (Fig. 7J). It is known that migrating sheets of cells recognize the direction of migration and polarize so that protrusive activity is restricted to the front, and that the MTOC re-orients itself in front of the nucleus to face the direction of migration(Etienne-Manneville and Hall,2002). Notably, under the experimental conditions employed here,γ-tubulin expression was not observed in the leading edge cells, the periderm cells, but rather in the inner basal layer of the epidermis(Fig. 7J,J′). This finding might be related to the fact that the eyelid epithelial cells lose their typical apicobasal polarity by degrees in migration, as is found in the epithelial-mesenchymal transition, or that the expression was just obscured by unknown factors. Even in the Fgf10^–/– mutant eyelid, the inner basal layer exhibited γ-tubulin expression apically in the cell (Fig. 7K,K′);however, one row of basal cells appeared wavy, and the number ofγ-tubulin-expressing cells seemed to decrease (n=4). Although we found some variability in the expression pattern of γ-tubulin,depending on the developmental stages of eyelid extension and/or the embryos,these results suggest that polarity of the epidermal cells is initially established but later impaired to some extent without FGF10.

Epithelial to mesenchymal transition (EMT) is a crucial morphogenetic process in animal development, in which epithelial cells become motile and mesenchymal in form, losing both their typical apicobasal polarity and their regular arrayed structure. This transition is observed in many developmental processes, such as gastrulation, formation of neural crest cells and fusion process of two primordia such as heart and palate(Savagner, 2001). The development of the eyelid involves temporary fusions of the epithelial layers. To know whether an EMT-like transition might be observed during the eyelid epithelial fusion, we next examined the expression pattern of vimentin in the leading edge migratory eyelid epithelial cells. The intermediate filament protein vimentin is a marker of mesenchymal cells(Savagner, 2001). In agreement, vimentin was expressed by a small population of epithelial cells at the advancing margin of the eyelid, as well as in the eyelid mesenchyme at E15(Fig. 7L,L′)(n=5). In the mutant eyelid tip, there was positive staining for vimentin, but staining was less extensive than in wild-type embryos at an equivalent stage of migration (Fig. 7M,M′) (n=3). Taken together, these results indicate that the EMT-like transition seems to be involved in temporary eyelid epithelial fusion and is blocked to some extent in eyelid migratory epithelial cells with disrupted FGF10 signaling.

It was reported that Shh, Ptch1 and Ptch2 are expressed in the developing mouse eyelid at E14.5(Motoyama et al., 1998). We re-examined the expression of Shh in the eyelid epithelium at E13.5(Fig. 8A), and histological sections verified that Shh is expressed by the basal epithelium abutting the dermis as reported previously(Motoyama et al., 1998) (data not shown). Interestingly, more abundant expression of Shh was seen on the upper eyelid margin and in the outer canthus region (n=4). By E15.5, expression of Shh was found along the eyelid margin and dotted expression was observed on the upper eyelid margin, corresponding to eyelash anlagen (Fig. 8B,C; n=3). Ptch1 and Ptch2 were also expressed on the margin of the developing eyelid at E13.5, and more intensely in the upper eyelid and in the outer canthus region (data not shown). By E15.5, Ptch1 and Ptch2 were expressed in the putative eyelash primordia (not shown). Thus, the expression patterns of Fgf10 and Shh signaling molecules appears complementary to some extent: Fgf10 expression is expressed in the mesenchyme and is more intense in the lower eyelid (Fig. 2G),whereas Shh, Ptch1 and Ptch2 are expressed in the epithelium(Motoyama et al., 1998), and more intensely in the upper eyelid.

We then examined the expression patterns of Shh, Ptch1 and Ptch2 in Fgf10-null eyelid primordia. As expression of the three genes was similar only Shh is shown. At E13.5, expression of Shh, Ptch1 and Ptch2 was observed on the mutant eyelid margin at a very low level (n=3 for each gene)(Fig. 8E; D, control; data not shown for Ptch1, Ptch2). Around E15, however, these genes became expressed along the eyelid margin in the mutant(Fig. 8G; F, control). The expression levels of all three genes appeared to increase on the mutant eyelid margin as compared with the wild type (Fig. 8I; H, control) (n=3). Furthermore, in the wild-type eyelid the Shh-expressing cells were kept compact on the lower margin(Fig. 8J), whereas the cells were scattered on the mutant eyelid margin(Fig. 8K). In a later stage at E16.5, however, the expression patterns of Shh, Ptch1 and Ptch2 were not dotted, as the eyelid remained wide open and eyelash primordia did not develop (not shown). Thus, in the absence of FGF10, the onset of gene expression for Shh, Ptch1 and Ptch2 is delayed. In addition to impaired growth of the eyelid protrusion, the impaired expression of SHH signaling molecules supports the notion that Fgf10-null eyelid primordia have defects by E13.5.

Since morphological analysis indicated that the integrity of epithelial leading edge cells of the eyelid primordia was disrupted in the absence of FGF10, we examined the expression of such peridermally expressed genes as activin βB and Tgfa. We chose activin βB and Tgfaas the mice deficient in these genes exhibit open eyelids at birth(Vassalli et al., 1994; Matzuk et al., 1995; Luetteke et al., 1993; Mann et al., 1993). In normal embryos at around E15, activin βB expression was detected on the eyelid margin (Fig. 8L). Section in situ hybridization indicated that activin βB was expressed in the leading edge of the eyelid epithelium (Fig. 8P). At around E16, the expression of activin βB was observed in the periderm of the fusion line (Fig. 8M). In the Fgf10-null eyelid at E15, however, activinβB expression was down-regulated (Fig. 8O; N for comparison). Section in situ hybridization indicated that activin βB was diffusely expressed in the eyelid epithelium at a very low level (Fig. 8Q).

We next examined the expression of Tgfa and Egfr; mice lacking these genes have no eyelids (Miettinen et al., 1999; Threadgill et al., 1995). Tgfa was expressed in the leading edge epithelial cells of normal eyelids (Fig. 8R) and later in the fusion line and the adjacent periderm(Fig. 8S)(Berkowitz et al., 1996). Since Egfr was diffusely expressed in the epithelium (data not shown)(Berkowitz et al., 1996), we examined the expression of Tgfa as a peridermal marker gene rather than Egfr. Tgfa expression was down-regulated in the mutant periderm and was not concentrated in the leading edge cells(Fig. 8U,W; T,V, control).

To determine whether the absence of FGF10 protein is directly involved in eyelid defects in Fgf10-null mice, we carried out an explant culture of normal eyelid anlagen and implanted an FGF10-soaked bead in the mesenchyme of the lower eyelid. We examined whether activin βB and Tgfawere up-regulated or ectopically induced after FGF10-bead application in normal eyelid anlagen. We found that within 12 hours, activin βB(Fig. 9B,C; A,E for comparison)and Tgfa (Fig. 9H,I;G,J for comparison) were ectopically induced in the thickened epidermis of the FGF10-bead-implanted eyelid(Fig. 9D; F for comparison). Furthermore, the area showing accumulation of F-actin in the leading edge appeared enlarged after FGF10 application, although no ectopic accumulation was observed near the bead (n=4)(Fig. 9K; L for comparison). By contrast, Shh was not ectopically induced by FGF10 protein in the eyelid explants derived from E15 mice (data not shown).

We next assessed whether FGF10 beads could promote epithelial extension in cultured Fgf10-null eyelids by applying quantitative morphometry. We measured the area of the eyelid epithelium covering the cornea before and after FGF10-soaked or PBS-soaked bead application, as shown in Fig. 9M. Although FGF10 did not notably promote epithelial extension in the wild-type eyelid anlagen (data not shown), the area of epithelial extension after FGF10 application was considerably wider than that for the PBS-soaked bead on the mutant eyelid anlagen (Fig. 9N). These results indicate that FGF10 can up-regulate the expression of activin βB and Tgfa in the normal eyelid epithelium and retrieve the eyelid epithelial extension in the Fgf10^–/– eyelid anlagen.

We have reported the expression pattern of Fgf10 during embryonic eyelid development and the eyelid defects in Fgf10-null mice. Fgf10 is expressed in the eyelid mesenchyme from the initiation phase of eyelid development. The nascent eyelid epithelia of Fgf10-null mice have defects in proliferation and later in the integrity of the eyelid periderm forming on the eyelid margin. Fgf10-null keratinocytes exhibit normal motility, and the formation of filopodia and lamellipodia in a scratch assay. Actin fibers, the gene products of activin βB and Tgfa, however, are not accumulated or up-regulated in the epithelial leading edge cells of Fgf10-deficient eyelids, and the onset of Shh expression is delayed in the mutant eyelid epithelium. In an explant culture, the addition of FGF10 protein can up-regulate the expression of activin βB and Tgfa in normal eyelid epithelia and retrieve the eyelid epithelial extension, to some extent, in the Fgf10^–/– eyelid anlagen. These results show that FGF10 has a dual role in embryonic eyelid development, in that it is required for both proliferation (Fig. 10A) and coordinated migration of epithelial leading edge cells(Fig. 10B).

Here, we compare the phenotype of open eyelids at birth found in representative mutants with that of Fgf10-null mice. First, the failure of eyelid induction in Fgfr2 IgIII-deleted embryos initiates much earlier and more severely than in Fgf10^–/– mutants; no grooves are formed above or below the eye, indicating that the loss of all FGFs-FGFR2 signaling blocks eyelid formation at its earliest stages(Li et al., 2001). By contrast, in Fgf10^–/– mutants, shallow but distinct eyelid grooves are formed and epithelial ridge formation is initially observed, although the eyelid protrusion is smaller and the integrity of peridermal clumps is disrupted. Thus, it is conceivable that multiple ligands of FGFR2 must be involved in eyelid development, among which FGF10 is critical for proliferation and maintenance of peridermal clumps at the leading edge of the developing eyelid margin. The residual signaling by FGFR2 viaFGF10-related molecules such as FGF7 is likely to function in the absence of FGF10 and thus the phenotype of Fgf10-null eyelids must be milder than that in FGFR2b-null eyelids.

The EGF receptor (EGFR) is activated upon binding of a family of polypeptides that includes EGF, TGFα, amphiregulin, heparin-binding EGF,betacellulin and epiregulin. Egfr-null mice have open eyelids at birth (Miettinen et al.,1995), whereas Tgfa null mice display a failure or a delay in prenatal eyelid growth and fusion(Berkowitz et al., 1996). This indicates the importance of TGFα-EGFR signaling in prenatal eyelid development and that other ligands of EGFR could also elaborate eyelid development. The eyelids of Tgfa null embryos exhibit variable extension across the cornea but no contact with the opposing epithelia(Berkowitz et al., 1996). This phenotype appears milder than that found in Fgf10-null mice, as Fgf10-null mice have open eyelids in which epithelial extension is not observed. Taken together with the down-regulation of Tgfa gene expression in the absence of FGF10, the Tgfa gene may be a target of the FGF10 signaling during early eyelid development.

The expression analysis of Tgfa and Egfr revealed that both genes are expressed in the developing eyelid epithelium: TgfamRNA is concentrated in the distal tips of the eyelids, whereas EgfrmRNA is prevalent throughout the epithelia on the eyelids and the cornea(Berkowitz et al., 1996). This implies that the TGFα-EGFR signaling may be involved in eyelid development in an autocrine or juxtacrine mode. By contrast, Fgf10 is expressed in the eyelid mesenchyme, while its receptor gene Fgfr2b is expressed in the epithelium (Li et al.,2001) (Fig. 2J-M),indicating that FGF10-FGFR2b acts in a paracrine manner between the epithelium and the mesenchyme. This suggests that there are two important tissue interactions during eyelid formation: intra-epithelial and epithelial-mesenchymal, mediated by TGFα-EGFR and FGF10-FGFR2b signaling, respectively. Since it has been shown that TGFα stimulates keratinocyte proliferation and migration(Barrandon and Green, 1987), it is possible that some functions of FGF10 in eyelid formation, such as eyelid epithelial proliferation and migration could be mediated by TGFαsignaling.

SHH is a counterpart of Drosophila Hedgehog (Hh), a secreted molecule implicated in the formation of embryonic structures and in tumorigenesis. The Patched (PTCH1) protein, constituting a receptor complex for Hh molecules is thought to oppose Hh signals by repressing the transcription of genes that can be activated by Hh. In vertebrates, two types of PTCH molecules, PTCH1 and PTCH2, have been identified and transcriptional regulation between SHH and PTCH has been reported(Marigo and Tabin, 1996; Goodrich et al., 1996). This study has demonstrated, in the absence of FGF10 signaling, that there is a delay in its interpretation for induction of Shh expression. Distinct Shh expression in the mutant eyelid at E15 is reminiscent of its capability to restrict adhesion and migration, as reported for neural crest cells (Testaz et al., 2001). Since conventional Shh or Ptch1 knockout mice were shown to suffer from severe developmental defects, the precise role of SHH signaling in eyelid development is still unclear. However, Motoyama et al.(Motoyama et al., 1998) showed that Shh, Ptch1 and Ptch2 are expressed in the basal epithelium of developing eyelids, but not in the mobile periderm. Taken together, it is conceivable that activation of SHH signaling might antagonize cell migration, and that it should be down-regulated when the primitive periderm cells start to stream onto the ocular surface under the control of FGF10 signaling.

The activin βB gene encodes the activin/inhibin βB subunit,constituting the dimeric growth factors of activin B, activin AB and inhibin B, which belong to the TGFβ family. Mice deficient in activin βB are viable but have defective eyelid development(Vassalli et al., 1994; Matzuk et al., 1995). This study showed peridermal expression of activin βB and Tgfa, their down-regulation in Fgf10-null eyelid primordia and up-regulation of expression by the FGF10 protein. Thus, it is likely that activin βB is a downstream component of FGF10 signaling during embryonic eyelid development. Given that activin and basic FGF have been shown to control cell migration in the Xenopus gastrula (Wacker et al., 1998), FGF10 might be involved in cell migration by interacting with activin signaling during mouse eyelid development.

Studies of mice with open eyelids at birth have shown that embryonic eyelid closure requires at least two signaling pathways, involving activin-MEKK1-JNK/p38 and TGFα/EGFR-ERK(Zhang et al., 2003; Xia and Kao, 2004). The known end point of the former pathway is actin stress fiber formation and phosphorylation of the nuclear factor Jun, the expression or activity of which might be of importance for the induction of EGFR and the activation of the second pathway. Although expression of activin βB and Tgfa is found in Fgf10-null eyelid epithelia at a low level, their mRNAs are not accumulated in leading edge cells without FGF10. Thus, this study further supports the notion that FGF10 positively regulates these signaling pathways during mouse eyelid closure. It is not known whether the control of mRNA distribution of activin βB and Tgfa and the accumulation of F-actin through FGF10 signaling are correlated or parallel pathways.

Recently, it has been thought that mammalian eyelid fusion is one of the developmental models for epidermal hole/wound closure, re-epithelialization and even wound healing to some extent, as is the case for dorsal closure in Drosophila embryos. There are two modes of epidermal hole/wound closure: actin purse-string mode and lamellipodial crawling mode(Martin and Parkhurst, 2004). In the actin purse-string mode, during the phase of epithelial sweeping, the leading edge cells accumulate actin and myosin just beneath the cell membrane at their apical edge. This F-actin accumulation forms a contractile cable,which pulls the leading edges (LEs) of the epithelial sheets taut(Jacinto et al., 2002) and drives LE cell apical constriction before further elongation and migration of the LE cells (Martin and Parkhurst,2004). Since accumulation of actin fibers in LE cells was not observed in Fgf10-null eyelids, while the scratch assay of cultured mutant keratinocytes showed the formation of lamellipodia, FGF10 may be required for a kind of wound-healing process by the actin purse-string mode rather than by the lamellipodial crawling mode. Furthermore, FGF10 signaling appears to have a role in maturation of filopodia in migrating eyelid epithelial cells.

The epithelial cells of Fgf10^–/– mutant eyelids exhibit a polarity (shown by γ-tubulin expression) and form prospective periderm cells, suggesting that even in the absence of FGF10 signaling the polarization signal is received by the eyelid LE cells, but that there is a collapse in its integrity without FGF10. This suggests that FGF10 makes the eyelid LE cells competent to maintain a pre-existing polarization signal.

The permissive function of FGF10 signaling translates into the correct coordination of different events in eyelid development, i.e. cell shape changes and proliferation in the early phase(Fig. 10A), and cell migration and polarity in the late phase, by regulating the activity of cytoskeleton and gene transcription (Fig. 10B). In the absence of FGF10, the leading edge cells cannot elongate centripetally,and these defects may well be responsible for the failure of Fgf10^–/– eyelid epidermis to spread over the developing cornea. This study also suggests mouse eyelid epithelial fusion as a new paradigm to elucidate the mechanisms of EMT. It has been reported that several members of the Wnt family are expressed in the developing eyelid primordia of the mouse (Liu et al.,2003). Although the phenotype of open eyelids at birth has not so far been reported for any Wnt mutants, a Wnt pathway was shown to be involved in Drosophila dorsal closure(Morel and Arias, 2004). It is therefore tempting to speculate that a Wnt pathway might be related to FGF10 signaling in mouse eyelid development.

Embryonic wound healing is a rapid process (taking place within 1 day in the case of eyelid closure) involving actin cable formation but no apparent hemostatic or inflammatory response(Martin and Lewis, 1992). Therefore, further elucidation of the mechanisms of eyelid closure will be useful in guiding us to better control the cell behaviors of repair in a clinical scenario.

We are grateful to G. Li and R. S. Johnson for information on the mouse keratinocyte culture. We also thank K. Tsuchida and J. Motoyama for providing reagents, and T. Adachi and K. Ohata for technical assistance. The monoclonal antibody used in this study was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242,USA. This study was partly supported by Grants in Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and by the Tanabe Medical Frontier Conference (TMFC).