|Home | About | Journals | Submit | Contact Us | Français|
Osr1 and Osr2 are the only mammalian homologs of the Drosophila odd-skipped family developmental regulators. The Osr1 protein contains three zinc-finger motifs whereas Osr2 exists in two isoforms, containing three and five zinc-finger motifs respectively, due to alternative splicing of the transcripts. Targeted null mutations in these genes in mice resulted in distinct phenotypes, with heart and urogenital developmental defects in Osr1−/− mice and with cleft palate and open eyelids at birth in Osr2−/− mice. To investigate whether these contrasting mutant phenotypes are due to differences in their protein structure or to differential expression patterns, we generated mice in which the endogenous Osr2 coding region was replaced by either Osr1 cDNA or Osr2A cDNA encoding the five-finger isoform. The knockin alleles recapitulated endogenous Osr2 mRNA expression patterns in most tissues and completely rescued cleft palate and cranial skeletal developmental defects of Osr2−/− mice. Mice hemizygous or homozygous for either knockin allele exhibited open-eyelids at birth, which correlated with differences in expression patterns between the knockin allele and the endogenous Osr2 gene during eyelid development. Molecular marker analyses in Osr2−/− and Osr2Osr1ki/Osr1ki mice revealed that Osr2 controls eyelid development through regulation of the Fgf10-Fgfr2 signaling pathway and that Osr1 rescued Osr2 function in maintaining Fgf10 expression during eyelid development in Osr2Osr1ki/Osr1ki mice. These results indicate that the distinct functions of Osr1 and Osr2 during mouse development result from evolutionary divergence of their cis regulatory sequences rather than distinct biochemical activities of their protein products.
The odd-skipped (odd) gene was first identified in a large mutagenesis screen for developmental control genes in Drosophila (Nusslein-Volhard and Wieschaus, 1980). In the trunk of odd mutant embryos, cuticular defects appear limited to alternate, odd-numbered segments, hence the gene name and its designation as a pair-rule gene (Nusslein-Volhard and Wieschaus, 1980). Molecular cloning of the odd locus showed that it encodes a protein containing four contiguous C2H2-type zinc finger repeats (Coulter et al., 1990). By screening a Drosophila embryonic cDNA library using an odd DNA probe, Hart et al. (1996) identified two odd-cognate genes, brother of odd with entrails limited (bowl) and sister of odd and bowl (sob), which each encodes a protein with five highly conserved C2H2-type zinc finger tandem repeats. The fourth member of the odd-skipped gene family in Drosophila, drumstick (drm), was identified during characterization of genes regulating embryonic hindgut elongation (Green et al., 2002). In contrast to odd, bowel, and sob, drm encodes a small protein containing only two zinc finger motifs, of which only the first finger conforms to the canonical C2H2 sequence and shows high sequence identity to the first zinc finger in the other three Odd-skipped family proteins (Green et al., 2002).
The odd, sob, and drm genes are clustered together on the second chromosome and exhibit near identical patterns, but different levels, of expression during Drosophila embryogenesis, suggesting that they may function partially redundantly (Hart et al., 1996; Green et al., 2002; Johansen et al., 2003). Indeed, extensive mutagenesis screens have failed to recover a mutation in sob with any embryonic developmental defect (Johansen et al., 2003). Although ODD and DRM proteins have limited amino acid sequence identity, ectopic expression of odd and drm in the same tissues often resulted in similar developmental defects (Green et al., 2002; Hao et al., 2003; de Celis Ibeas and Bray, 2003; Hatini et al., 2005; Bras-Pereira et al., 2006). However, odd is required for embryonic segmentation whereas mutations in drm disrupted hindgut and proventriculus morphogenesis without affecting segmentation (Nusslein-Volhard and Wieschaus, 1980; Coulter and Wieschaus, 1988; Green et al., 2002).
Bowl, on the other hand, is expressed in a largely distinct pattern and exhibits clearly distinct functions from the other odd-skipped family genes during embryogenesis and tissue morphogenesis (Hart et al., 1996; Wang and Coulter, 1996; Hao et al., 2003; Johansen et al., 2003). Although bowl and odd have partially overlapping expression domains during embryogenesis and bowl mutants occasionally exhibited a mild segmentation defect, their mutant phenotypes are distinct from each other and odd/bowl double mutants did not show any synergism (Wang and Coulter, 1996). Moreover, whereas bowl and drm mutants exhibit similar defects in proventriculus and hindgut morphogenesis, bowl drm odd and sob compound mutants had no additional defects in gut development even though both odd and sob are co-expressed with drm and the domains of their expression also partially overlaps with that of bowl during gut development (Johansen et al., 2003). Furthermore, although there is significantly more extensive amino acid sequence identity between ODD and BOWL than that between ODD and DRM proteins, ectopic expression of bowl almost always resulted in different phenotypes from that of ectopic expression of odd or drm under the same conditions (Hao et al., 2003; Bras-Pereira et a., 2006). For example, ectopic expression of drm throughout the developing hindgut caused dramatic expansion of the small intestine whereas ectopic expression of bowl using the same driver had little effect on the morphology and patterning of the gut (Johansen et al., 2003). Together, these data indicate that the protein products of odd and its cognate genes have largely distinct functions during Drosophila development.
In the mammalian genome, only two odd-skipped related genes, Osr1 and Osr2, exist. Osr1 encodes a protein with three zinc fingers and Osr2 encodes both a three-finger (Osr2B) and a five-finger (Osr2A) protein due to alternative splicing of the pre-mRNA (So and Danielian, 1999; Lan et al., 2001; Katoh, 2002; Kawai et al., 2005). While their structural homology to the Drosophila ODD family proteins is limited to the zinc finger motifs, Osr1 and Osr2B show 65% amino acid sequence identity overall and 98% amino acid sequence identity in their zinc finger domains (Lan et al., 2001). During mouse embryonic development and organogenesis, Osr1 and Osr2 exhibit distinct as well as partially overlapping expression patterns (So and Danielian, 1999; Lan et al., 2001; Lan et al., 2004; Stricker et al., 2006). Targeted null mutations in Osr1 and Osr2 in mice resulted in distinct phenotypes, with heart and urogenital developmental defects in Osr1−/− mice and with cleft palate and open-eyelids at birth in Osr2−/− mice (Lan et al., 2004; Wang et al., 2005; James et al., 2006). Whereas the early embryonic lethality of Osr1−/− mutant mice precluded direct analysis of the roles of Osr1 in many developmental processes in those mutants, the correlation of developmental defects in the Osr2−/− mutants to specific tissues that normally do not express Osr1 and the lack of phenotypes in Osr2−/− mutants in many tissues where Osr1 and Osr2 are normally co-expressed suggest that Osr1 and Osr2 function partly redundantly during mouse embryonic development (Lan et al., 2004; Wang et al., 2005). However, Kawai et al. (2005) reported that the two Osr2 isoforms, containing three (Osr2B) and five (Osr2A) zinc finger repeats respectively, when fused to the Gal4 DNA binding domain, exhibited opposite transcriptional activity in a cell culture based reporter assay. Since the Osr1 gene product only contains three zinc-finger repeats, it is possible that Osr1 and Osr2B may have similar biochemical functions due to their overall amino acid sequence homology whereas Osr2A may have evolved distinct functions. To directly address whether the mammalian Osr1 and the two Osr2 isoforms have evolved distinct biochemical functions in vivo, we have generated mice with the Osr2 gene coding region replaced by either an Osr1 cDNA or an Osr2A cDNA through targeted gene replacement in mice. Here we show that these two gene substitution strategies similarly rescued the developmental defects of the Osr2−/− mutant mice.
A129/SvEv strain mouse BAC clone containing the entire Osr2 genomic region, as reported previously (Lan et al., 2004), was used for gene targeting vector construction. To replace the Osr2 coding region with either Osr1 cDNA or Osr2A cDNA, a 2 Kb genomic fragment immediately 5’ to the translation start codon (ATG) of the Osr2 gene was amplified by PCR, subcloned, sequenced, and used as the 5’ homology arm of the targeting vectors. A hexameric Myc epitope-tag coding region from the pCS+MT vector (Rupp et al., 1994) was ligated in-frame to the 5’ end of PCR-amplified cDNA fragments containing either Osr1 or Osr2A coding sequences, each ending 3’ to their respective translation stop codon. The targeting vectors contain the 2 Kb 5’ homology arm described above, either the Myc-Osr1 or the Myc-Osr2A cDNA cassette, followed by a loxP-flanked PGK-neo expression cassette for positive selection, a 3.3 kb 3’ homology arm containing the 3’ untranslated region of the Osr2 gene, and a PGK-DTA expression cassette for negative selection (Fig. 1). Correct targeting of the Osr2 locus results in deletion and replacement of the 2.6 Kb region from the translation start codon in Exon 2 through the middle of the 3’ untranslated region in Exon 4 with the Myc-Osr1 or Myc-Osr2A cDNA, and the neo expression cassette (Fig. 1).
The targeting vectors were linearized and electroporated into CJ7 ES cells. ES cell culture and Southern screening of ES clones were carried out as previously described (Swiatek and Gridley, 1993; Lan et al., 2004). Two independent correctly targeted ES cell clones for each of the targeting vectors were injected into blastocysts from C57BL/6J mice and the chimeric male mice were bred with C57BL/6J females to test for germline transmission. Tail DNA of F1 mice were used for genotyping by PCR and Southern hybridization. Mice and embryos from subsequent generations were genotyped by PCR. PCR with primer 1 (5’- GAT ACG GGT AAG ACA GAA ACT G-3’) and primer 2 (5’- CTA CAA GGA TCT AGC ACA TGC TG-3’) amplifies a 490 bp band from the wild-type Osr2 allele. PCR with primer 2 and primer 3 (5’-CTT CTT GAC GAG TTC TTC TGA GG-3’) amplifies a 460bp allele-specific product from either the Osr2Osr1ki or the Osr2Osr2Aki allele. Heterozygous F1 mice were either crossed with the Osr2+/− mice (heterozygous for the Osr2tm1Jian allele, Lan et al., 2004) to generate Osr2Osr1ki/− and Osr2Osr2Aki/− hemizygous mice or intercrossed to generate Osr2Osr1ki/Osr1ki and Osr2Osr2Aki /Osr2Aki homozygous knockin mice, respectively.
Skeletal preparations of newborn mice were carried out following the protocol described by Martin et al. (Martin et al., 1995). For histology and in situ hybridization, embryos were dissected at desired developmental stages. Tail DNA or yolk sac DNA was extracted and genotypes of the embryos were determined by allele-specific PCR. For histology analysis, embryos were fixed in Bouin’s fixative, then dehydrated through graded ethyl alcohols, embedded in paraffin, sectioned at 7 µm thickness and stained with hematoxylin and eosin. For in situ hybridization, embryos were fixed in 4% paraformaldehyde in PBS overnight at 4°C. Sectioned in situ hybridization was carried out as previously described (Lan et al., 2001) with digoxygenin-labeled antisense RNA probes.
For detection of cell proliferation in the developing eyelids, timed pregnant female mice were injected intraperitoneally on gestational day 13.5 with BrdU (Roche) labeling reagent (45 µg/g body weight). One hour after injection, embryos were dissected, fixed in Carnoy’s fixative, dehydrated through graded alcohols, embedded in paraffin and sectioned in the coronal plane at 5 mm thickness. Immunodetection of BrdU was performed using the BrdU labeling and detection kit (Roche) according to manufacturer’s instructions and the sections were counterstained with nuclear fast red to visualize all cellular nuclei. The total number of cell nuclei as well as the number of BrdU-labeled nuclei on sections through the middle of the upper and lower eyelids were counted and recorded separately for the epithelium and mesenchyme from ten continuous sections. The cell proliferation index was calculated as percentage of the cell nuclei with BrdU labeling. Students’ t–test was used to analyze the significance of difference and a P value less than 0.05 was considered statistically significant.
To directly analyze whether Osr1 and Osr2 have equivalent or distinct functions in vivo, we used the gene targeting technology to replace the Osr2 coding region with a Myc-Osr1 cDNA together with a loxP-flanked neo expression cassette (Fig. 1A). Chimeric mice were generated from two correctly targeted ES cell clones and germ line transmission of the targeted allele was confirmed with Southern hybridization of F1 mouse progeny tail DNA samples (Fig. 1B). F1 mice heterozygous for the Osr2Osr1ki allele were indistinguishable phenotypically from wild-type mice and were fertile. Osr2Osr1ki/+ heterozygotes were either crossed with Osr2+/− heterozygous mice to generate Osr2Osr1ki/− hemizygotes or intercrossed to generate Osr2Osr1ki/Osr1ki mice.
Since a similarly targeted Osr2tm1Jian allele, in which the Osr2 gene was deleted from the middle of Exon 2 to the 3’ untranslated region in Exon 4 and replaced by an in-frame fusion of the lacZ reporter gene, resulted in LacZ expression recapitulating endogenous Osr2 expression patterns during early mouse embryogenesis (Lan et al., 2004), we expected that the Osr2Osr1ki allele will result in expression of Myc-Osr1 in the same pattern as endogenous Osr2 mRNA expression. To test this, we compared Osr1 and Osr2 mRNA expression at selected stages of palate development by in situ hybridization analysis. As reported previously (Lan et al., 2001; Lan et al., 2004), Osr2 mRNA is expressed in a lateral-medial gradient, with lower levels in the medial side, in the developing palatal shelves at E13.5 in wild-type embryos (Fig. 2A). In contrast, Osr1 mRNA is weakly expressed only in the lateral half, but not in the medial half, of the developing palatal mesenchyme at this stage (Fig. 2D). Both Osr1 and Osr2 mRNAs are expressed in the mandibular mesenchyme lateral to the developing molar tooth buds (Fig. 2A, D). In addition, whereas Osr2 mRNA is strongly expressed in the maxillary and mandibular mesenchyme on the lingual side of the molar tooth buds (Fig. 2A), Osr1 is only weakly expressed in a subset of mesenchymal cells lingual to the mandibular tooth bud but not in the maxillary molar tooth mesenchyme (Fig. 2D). Consistent with the differential expression of Osr1 and Osr2 in the developing tooth mesenchyme, we recently discovered that Osr2 antagonizes the Msx1-Bmp4 molecular pathway to pattern the buccolingual axis of the molar tooth developmental field and that Osr2−/− mutant mice exhibit supernumerary tooth formation lingual to the normal molar teeth (Zhang et al., submitted, and Fig. 3B). No Osr2 mRNA expression was detected in all tissues and developmental stages examined in the Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous embryos (Fig. 2B,C, and data not shown), consistent with deletion of the coding region of the endogenous Osr2 gene in these embryos. Osr1 mRNA expression, in contrast, is clearly detected in the Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous embryos in the tissues that normally express Osr2 but not Osr1 mRNA, including the medial side of the developing palatal shelves and in the lingual side of the developing molar tooth mesenchyme (Fig. 2E, F). Thus, the Osr2Osr1ki allele drives Osr1 mRNA expression under the control of the Osr2 cis regulatory sequences and makes it possible to directly test whether Osr1 and Osr2 are functionally equivalent in vivo.
Osr2Osr1ki/+ mice were crossed with Osr2+/− mice or intercrossed to generate Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki mice, respectively. The Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous mice were born at expected Mendelian ratios and both genotypes survived postnatally, exhibited similar life span to wild-type and heterozygous mice and were fertile. Litter size from either intercrosses of these genotypes or outcrosses to wild-type mice was similar to wild-type mice in the same genetic background. Detailed histological analysis of Osr2Osr1ki/− and Osr2Osr1ki/Osr1ki fetal mice showed that palate development occurred normally in these mice, indicating that the Myc-Osr1 expressed from the Osr2 locus rescued the cleft palate defect of Osr2−/− mutant mice (Fig. 3A–D). Similarly, tooth development appeared normal in the Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous mice (Fig. 3D, and data not shown). In addition, examination of skeletal preparations of newborn mice did not detect any skeletal abnormalities in comparison with wild-type and heterozygous littermates. Specifically, while the Osr2−/− mutant pups had defects in palatine bones, the Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous mice all had similar palatine bones with the wild-type littermates (Fig. 3E–H). In addition, while the tympanic rings of newborn Osr2−/− mutant mice are thickened and significantly reduced in size, in comparison with wild-type littermates (Lan et al., 2004, Fig. 3I,J), the tympanic rings of newborn Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous mice are normal in size although it is still slightly thickened in the Osr2Osr1ki/− hemizygous mice (Fig. 3K, L). These data indicate that Osr1, when expressed under the control of the regulatory sequences of Osr2 locus, can carry out all the essential functions of Osr2 during palate, tooth, and tympanic ring development.
As reported previously (Lan et al., 2004), Osr2−/− mutant mice are born with open eyelids (Fig. 4B). We found that Osr2Osr1ki/− hemizygous mice are born with an open eyelid defect similar to that of the Osr2−/− mutants (Fig. 4C). The severity of the open eyelid defect is greatly reduced in Osr2Osr1ki/Osr1ki mice, however, it is still identifiable as a small opening in the middle part of the eyelids at birth (Fig. 4D). Since the Osr2Osr1ki allele contains a loxP-flanked PGK-neo expression cassette, which has been shown to interfere with gene expression in some gene-targeted mouse strains (Olson et al., 1996; Seidl et al., 1999; Holzenberger et al. 2000), it is possible that the open eyelid defect in the Osr2Osr1ki/− and Osr2Osr1ki/Osr1ki mice might result from interference of Myc-Osr1 expression by the PGK-neo cassette. To test this possibility, Osr2Osr1ki/+ heterozygous mice were crossed to E║a-Cre transgenic mice (Lakso et al., 1996) to delete the PGK-neo expression cassette from the targeted Osr2 locus. Osr2Osr1ki/+E║a-Cre double heterozygous mice were crossed to Osr2+/− mice or intercrossed to generate Osr2Osr1ki/−E║a-Cre mice and Osr2Osr1ki/Osr1kiE║a-Cre mice, respectively. Southern hybridization and PCR analyses of tail genomic DNA samples confirmed that the PGK-neo cassette had been deleted from the targeted Osr2 locus in these mice (data not shown). However, they were still born with an open eyelid phenotype with severity similar to the Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous mice, respectively (data not shown).
One possible reason why the Osr2Osr1ki/− and Osr2Osr1ki/Osr1ki mice are born with open eyelids is that the Osr2Osr1ki allele may have failed to completely recapitulate the endogenous Osr2 gene expression pattern during eyelid development. Thus, we compared the expression patterns of Osr1 mRNA in the developing eyelid regions in wild-type and Osr2Osr1ki/Osr1ki embryos with that of Osr2 mRNA in wild-type embryos. In wild-type embryos at E13.5, Osr2 mRNA was strongly expressed in both the epithelium and mesenchyme of developing eyelids (Fig. 5A). In contrast, endogenous Osr1 mRNA expression was only expressed in a small subset of mesenchymal cells at the base of the eyelids at this stage (Fig. 5B). In the Osr2Osr1ki/Osr1ki homozygous embryos at E13.5, Osr1 mRNA was moderately expressed in the developing eyelid mesenchyme but expression in the eyelid epithelium was much weaker (Fig. 5C), in contrast to the strong Osr2 mRNA expression in the same areas of the developing eyelids in the wild-type embryos (Fig. 5A). To investigate whether the differences in the expression patterns between Osr2Osr1ki allele and the endogenous Osr2 gene was due to deletion of Osr2 gene sequences in the knockin allele, we analyzed lacZ expression from the previously targeted Osr2tm1Jian allele, which carries deletion of the same genomic region containing the second and third introns, in Osr2+/− mice. As shown in Fig. 5D, lacZ expression in the developing eyelids in Osr2+/− mice is also much weaker in the epithelium than in the mesenchyme. These data suggest that the open-eyelids phenotype in the Osr2Osr1ki/− hemizygous and Osr2Osr1ki/Osr1ki homozygous mice resulted from deletion of cis regulatory sequences in the Osr2 gene necessary for complete recapitulation of the endogenous Osr2 gene expression pattern by the Osr2Osr1ki allele.
The Osr2 gene encodes two mRNA isoforms, Osr2A and Osr2B, differing by the presence or absence of an 82-nucleotide sequence corresponding to the beginning of the fourth exon, due to alternative splicing (Kawai et al., 2005). Whereas Osr2A encodes a protein containing five contiguous zinc finger repeats, homologous to the Drosophila BOWL and SOB proteins, the 82-nucleotide deletion in Osr2B results in a frame shift and an in-frame translation stop codon shortly after the third zinc finger-coding region. Kawai et al. (2005) reported that Osr2A and Osr2B exhibited opposite transcriptional activity when fused with the Gal4 DNA-binding domain and co-transfected into COS-7 cells with a luciferase reporter under the control of five tandem Gal4 binding sites. Since the mouse Osr1 and Osr2B proteins both contain only three zinc finger motifs and they share 65% identity in overall amino acid sequences and 98% identity in the three zinc finger motifs (Lan et al., 2001), it is possible that the Myc-Osr1 expressed from the Osr2Osr1ki allele could only replace the function of the Osr2B isoform but not that of the five-finger Osr2A isoform, which may be part of the reason why expression of Myc-Osr1 from the Osr2Osr1ki allele failed to completely rescue the eyelid developmental defects of Osr2−/− mutant mice. To test this possibility, we replaced the endogenous Osr2 coding region with the Osr2A cDNA and a loxP-flanked PGK-neo expression cassette using a gene targeting vector essentially the same as the Osr2Osr1ki targeting vector but using Myc-Osr2A cDNA in place of the Myc-Osr1 cDNA (Fig. 1A). We obtained five correctly targeted ES clones and used two independent clones to generate chimeric mice. Chimeric males were crossed to C57BL/6J female mice and tail genomic DNA were analyzed by Southern hybridization and allele-specific PCR to confirm germline transmission of the targeted allele (data not shown). Mice heterozygous for the Osr2Osr2Aki allele are normal and fertile. Osr2Osr2Aki/+ heterozygotes were crossed with Osr2+/− mice or intercrossed to generate Osr2Osr2Aki/− and Osr2Osr2Aki/Osr2Aki mice, respectively. The Osr2Osr2Aki/− and Osr2Osr2Aki/Osr2Aki mice were found at expected Mendelian ratios at weaning and exhibit comparable lifespan with wild-type littermates. Similar to the Osr2Osr1ki mice, Osr2Osr2Aki/Osr2Aki mice showed normal palate, tooth, and tympanic ring development (data not shown). Also similar to the Osr2Osr1ki mice, both Osr2Osr2Aki/− and Osr2Osr2Aki/Osr2Aki mice were born with open eyelids (Fig. 6), with the Osr2Osr2Aki/Osr2Aki homozygous mice displaying less severe open-eyelid defects than the Osr2Osr2Aki/− hemizygous mice. Thus, the Osr2Osr2Aki/− and Osr2Osr2Aki/Osr2Aki mice essentially phenocopy the Osr2Osr1ki/− and Osr2Osr1ki/Osr1ki mice, respectively. Furthermore, we crossed the Osr2Osr1ki/Osr1ki mice with Osr2Osr2Aki/Osr2Aki mice and generated Osr2Osr1ki/Osr2Aki transheterozygous mice, which were also born with a mild open-eyelid phenotype similar to either parent (data not shown). These data indicate that Osr1 and Osr2A display equivalent biochemical activities in vivo when expressed in the same cells, in spite of their differences in the number of zinc finger repeats.
Development of the eyelid requires coordinated growth, epithelial migration and fusion. In mice, eyelid outgrowth is initiated at E11.5, forming deep grooves above and below the developing optic vesicle by E13.5 (Tao et el., 2005). Compared to wild-type littermates, Osr2−/− mutant mice exhibited obvious retardation in eyelid outgrowth by E13.5 (Fig. 7A,B). In contrast, Osr2Osr1ki/Osr1ki embryos showed only slightly reduced eyelid primordial size compared with the wild-type embryos at this stage (Fig. 7C). By E15.5, both upper and lower eyelid epithelium formed leading edges that migrated toward each other over the surface of the developing cornea in wild-type embryos (Fig. 7D, G). The developing eyelids in Osr2−/− littermates at this stage appeared round and did not form the periderm leading edges seen in the wild-type littermates (Fig. 7E, H). The leading edges of the developing eyelids in Osr2Osr1ki/Osr1ki embryos at E15.5 formed rudimentary protruding periderm clumps but did not extend over the cornea (Fig. 7F, I). The wild-type eyelids had completed fusion by E16.5 (Fig. 7J), whereas the Osr2−/− mutant upper and lower eyelids were still separated wide apart at E18.5, although their leading edges started to form rudimentary epithelial ridges (Fig. 7K). By E18.5, the leading edge periderm cells of Osr2Osr1ki/Osr1ki upper and lower eyelids migrated over the developing cornea, but their eyelid bodies were too far apart to complete fusion over the central region of the cornea (Fig. 7L).
We performed BrdU incorporation assay to investigate the eyelid growth retardation phenotypes in Osr2−/− and Osr2Osr1ki/Osr1ki embryos. At E13.5, Osr2−/− mutant embryos showed about 50% reduction in cell proliferation in both upper and lower eyelid epithelium, compared to wild-type littermates (Fig. 8A–C). In addition, mesenchymal cell proliferation was also significantly reduced in both upper and lower eyelids in Osr2−/− mutant embryos (Fig. 8A–C). In contrast, no significant differences in cell proliferation in lower eyelid epithelium and mesenchyme as well as in upper eyelid mesenchyme between Osr2Osr1ki/Osr1ki embryos and their wild-type littermates were detected (Fig. 8D–F). However, cell proliferation in the upper eyelid epithelium of E13.5 Osr2Osr1ki/Osr1ki embryos was still reduced by approximately 20%, compared with the wild-type littermates (Fig. 8D–F). These data indicate that Osr2 plays an essential role in regulating eyelid outgrowth.
The eyelid developmental defects in Osr2−/− mutant mice, including reduction in cell proliferation and defects in leading edge epithelial morphology, are most similar to that in mice lacking Fgf10 (Tao et al., 2005). Thus, we examined Fgf10 expression at different stages during eyelid development in wild-type, Osr2−/−, and Osr2Osr1ki/Osr1ki mice. At E13.5, Fgf10 mRNA was highly expressed in the central domain of the developing eyelid mesenchyme in wild-type embryos (Fig. 9A). Fgf10 expression was dramatically reduced in the eyelid mesenchyme of Osr2−/− mutant littermates (Fig. 9B). In contrast, Fgf10 mRNA expression in the eyelid mesenchyme in E13.5 Osr2Osr1ki/Osr1ki embryos (Fig. 9C) was similar to that in the wild-type embryos (Fig. 9A). At E14.5, the Osr2−/− embryos continued to show significantly lower levels of Fgf10 mRNA expression in the developing eyelid mesenchyme than that in wild-type littermates (Fig. 9D, E), while the Osr2Osr1ki/Osr1ki embryos exhibited similar levels of Fgf10 expression in the developing eyelid mesenchyme to that in wild-type embryos (Fig. 9D, F). These data indicate that Osr2 regulates Fgf10 mRNA expression during eyelid development.
Fgf10 most likely signals through the Fgfr2 receptor to regulate eyelid development, since Fgfr2 is also required for eyelid outgrowth (De Moerlooze et al., 2000; Li et al., 2001). Interestingly, although Fgfr2 expression was similar in the developing eyelid tissues in Osr2−/− and wild-type littermates at E13.5 (data not shown), decreased Fgfr2 expression in both the epithelium and mesenchyme of the developing eyelid was consistently detected in Osr2−/− mutant embryos at E14.5 (Fig. 9G, H). Fgfr2 expression was also reduced in both the eyelid epithelium and mesenchyme in Osr2Osr1ki/Osr1ki embryos at E14.5, compared to the wild-type embryos (Fig. 9G, I). At E15.5, Osr2−/− mice still exhibited significantly reduced Fgfr2 expression in the developing eyelids (Fig. 9K), compared with that in wild-type littermates (Fig. 9J). However, Fgfr2 expression is restored to wild-type levels in both the epithelium and mesenchyme of the developing eyelids in Osr2Osr1ki/Osr1ki mice at E15.5 (Fig. 9L), which correlated with the delayed leading edge epithelial morphogenesis in these mice. These data indicate that Osr2 controls eyelid development, at least in part, through regulation of Fgf10 and Fgfr2 expression.
We previously showed that targeted deletion of the Osr2 gene resulted in cleft palate and open eyelids at birth in Osr2−/− homozygous mutant mice (Lan et al., 2004). Interestingly, whereas Osr2 is expressed in a graded pattern along the medial-lateral axis of the vertically growing palatal shelves, with lower levels of expression in the medial side than in the lateral side of the palatal mesenchyme, the Osr2−/− mutant mice exhibited specific reduction in cell proliferation in the medial but not lateral half of the developing palatal shelves (Lan et al., 2001; 2004). Osr1 mRNA is expressed at relatively low abundance in the early developing palatal mesenchyme and becomes restricted to only the lateral half of the palatal mesenchyme. Thus, the palatal mesenchyme proliferation defect in the Osr2−/− mutant embryos was restricted to the domain of the palatal mesenchyme that normally only expressed Osr2 but not Osr1 mRNA (Lan et al., 2004). Similarly, the eyelid defects in the Osr2−/− mutant mice also correlate with distinct expression patterns between Osr1 and Osr2 during eyelid development (this report). No apparent defects were detected in the Osr2−/− mutant mice in tissues that normally express both Osr1 and Osr2, including the developing kidney, limb and the proximal mandible (Lan et al., 2004). These data suggest that Osr1 and Osr2 may function redundantly during mouse embryonic development. However, the early embryonic lethality of the Osr1−/− mutant mice (Wang et al., 2005) precludes detailed analysis of possible functional redundancy between Osr1 and Osr2 in the double null mutant mice.
To directly analyze whether Osr1 can carry out all developmental functions of Osr2 when expressed under the regulatory sequences of the Osr2 locus, we generated mice in which the Osr2 coding sequence is replaced by a Myc-Osr1 cDNA cassette. We found that expression of Myc-Osr1 from the Osr2 locus completely rescued the cleft palate, supernumerary tooth, and tympanic ring defects of the Osr2−/− mutant mice. The only defect of the Osr2−/− mutant mice not completely rescued by the Osr2Osr1ki allele is the open eyelids at birth phenotype. We found that Myc-Osr1 expression in the Osr2Osr1ki mice did not completely recapitulate the endogenous Osr2 mRNA expression patterns during eyelid development, most likely due to deletion of eyelid-specific regulatory sequences in the intronic regions of the Osr2 locus. Moreover, we show that the Osr2Osr2Aki mice, which differ from the Osr2Osr1ki mice only in the replacement cDNA cassette, phenocopy the Osr2Osr1ki mice. The open eyelid phenotypes in the single homozygous and transheterozygous knockin mice further indicate that failure of the respective knockin alleles to completely rescue the eyelid developmental defects of Osr2−/− mutant mice is not due to the structural differences between Osr1 and Osr2A proteins. To the contrary, these data indicate that Osr1 and Osr2A have equivalent functions, when expressed under the control of the regulatory sequences of the Osr2 gene, during mouse development.
Our finding that Osr1 and Osr2A function equivalently during mouse development is in contrast to previous reports of largely distinct functions of the Drosophila Odd-skipped family members (Nusslein-Volhard and Wieschaus, 1980; Coulter et al., 1990; Wang and Coulter, 1996; Green et al., 2002; de Celis Ibeas and Bray, 2003; Johansen et al., 2003; Hatini et al., 2005). During Drosophila embryogenesis, odd and its cognate genes sob and drm are expressed in almost identical patterns (Hart et al, 1996; Johansen et al., 2003), but odd and drm mutants display completely distinct phenotypes whereas no sob mutation has been isolated and no synergism was detected in the compound triple mutants (Nusslein-Volhard and Wieschaus, 1980; Coulter et al., 1990; Wang and Coulter, 1996; Green et al., 2002; Johansen et al., 2003). The odd, sob, and drm genes are clustered together within a 200 Kb genomic region and their near identical expression patterns suggest that they share regulatory enhancer sequences. However, their gene structures and protein products have diverged substantially such that the only structural similarity in the family is limited to the zinc finger motifs. The C2H2 zinc fingers in the Odd-skipped family proteins are typically involved in sequence-specific DNA binding, consistent with the proposed roles of these factors as transcriptional regulators. Indeed, both Drosophila ODD and mouse Osr2 have recently been shown to bind to similar specific DNA sequences (Meng et al., 2005; Kawai et al., 2007). In addition, both BOWL and DRM have been shown to directly bind to the LINES protein through their first zinc finger motif (Green et al., 2002; Hatini et al., 2005). Thus, the Odd-skipped family proteins share some functional similarities at the molecular level because of their conserved zinc finger motifs, which explains why ectopic overexpression of odd, sob, and drm during Drosophila tissue morphogenesis sometimes resulted in similar developmental effects (Green et al., 2002; Hao et al., 2003; de Celis Ibeas and Bray, 2003; Hatini et al., 2005; Bras-Pereira et al., 2006), whereas the distinct mutant phenotypes likely reflect their functional divergence from the distinct protein structures outside of the conserved zinc finger motifs.
Phylogenetic analysis indicated that two odd-skipped family genes were present in an ancestral metazoan, which further duplicated in the arthropod clade and gave rise to the four cognate genes in Drosophila (Buckley et al., 2004). In contrast, mammalian Osr1 and Osr2 share extensive amino acid sequence similarity throughout their entire length of the protein products (Lan et al., 2001), suggesting that these two genes arose by a more recent duplication from one of the ancestral odd-skipped genes, with apparent loss of the other, during metazoan evolution (Buckley et al., 2004). Many other paralogous transcription factors exist in mammals as pairs or groups that apparently arose by gene duplication early during vertebrate evolution (e.g., Noll, 1993; Hanks et al., 1995; Greer et al., 2000; Singh and Hannenhalli, 2008). These transcription factor paralogs often share significant structural similarities and partially overlapping expression patterns. Several examples exist where similar gene replacement strategies have been used to evaluate possible functional equivalence of closely related transcription factors in mice (e.g., Hanks et al., 1995; Bouchard et al., 2000; Coppola et al., 2005; Kellerer et al., 2006). Whereas some transcription factor paralogs have been found functionally interchangeable, such as between En1 and En2 as well as between Pax2 and Pax5 (Hanks et al., 1995; Bouchard et al., 2000), others showed only partial functional overlap, such as between Phox2a and Phox2b as well as between Sox8 and Sox10 (Coppola et al., 2005; Kellerer et al., 2006). By studying a large collection of human paralogous transcription factor pairs, Singh and Hannenhalli (2008) concluded that paralogous transcription factors diversify their functions mainly through divergence in either their DNA binding site motifs or in expression. Our finding that expression of Osr1 from the Osr2 locus rescued the developmental defects of Osr2−/− mutant mice indicate that the mammalian Osr1 and Osr2 genes evolved distinct developmental roles mainly through divergence of the cis regulatory sequences controlling their spatiotemporal expression rather than changes in their coding sequences. Because Osr1 and Osr2 are co-expressed during many developmental processes, including craniofacial, kidney, limb, and synovial joint development (Lan et al., 2001; 2004; Stricker et al., 2006), further studies using conditional gene inactivation approaches are warranted for the elucidation of the roles of Osr1 and Osr2 in these developmental processes.
Our findings that either Osr1 or Osr2A, when expressed from the endogenous Osr2 locus, rescued almost all developmental defects of Osr2−/− mice seem at odds with previously reported opposite transcriptional activity for Osr2A and Osr2B in vitro (Kawai et al., 2005). However, although both Osr1 and Osr2B contain only three zinc finger repeats while Osr2A contains five, our study does not directly compare the functions of Osr2A and Osr2B in vivo. It is possible that the Osr2Osr2Aki/Osr2Aki mice may have subtle cellular or physiological deficits that could be attributed to lack of the Osr2B isoform. It is also possible that the different transcriptional activity found in the cell culture assays represents non-physiological activity. In this regard, it is noteworthy, while genetic analyses in Drosophila indicated a primarily transcriptional repressor function for Odd-skipped (Mullen and DiNardo, 1995; DiNardo and O’Farrell, 1987; Kuhn et al., 2000; Bouchard et al., 2000), its overexpression in embryos showed a role in activation (Saulier-Le Drean et al., 1998).
Our characterization of the open eyelid defects in Osr2−/− and Osr2Osr1ki/Osr1ki mice revealed an essential role for Osr2 in regulating eyelid outgrowth. Previous studies indicated that eyelid outgrowth is controlled by Fgf10-Fgfr2 signaling. Fgfr2 exists in two major isoforms, with Fgfr2b expressed primarily in the epithelium and Fgfr2c primarily in the mesenchyme, due to alternative splicing (Ornitz et al., 1996). Fgfr2b function in the epithelium is required for eyelid initiation (De Moerlooze et al., 2000; Li et al., 2001). A major ligand for the epithelial Fgfr2b receptor during eyelid development is Fgf10, which is expressed in the mesenchyme (Tao et al., 2005). Mice lacking Fgf10 exhibited morphological and cell proliferation defects during eyelid development that are highly similar to that in the Osr2−/− mice (Tao et al., 2005), including reduction in eyelid epithelial cell proliferation and leading edge morphogenesis. We found that Fgf10 expression in the developing eyelid mesenchyme was dramatically reduced in the Osr2−/− mutant embryos by E13.5, compared with that in their wild-type littermates, indicating that Osr2 is involved in the regulation of Fgf10 expression during eyelid outgrowth. That decreased Fgf10 expression is at least part of the mechanism underlying eyelid growth defects in Osr2−/− mice is supported by the correlation of restoration of Fgf10 expression with the partial rescue of eyelid growth in the Osr2Osr1ki/Osr1ki mice.
Interestingly, Fgfr2 expression was affected in the Osr2−/− mutant mice after E13.5. It is possible that the reduction in Fgfr2 expression is a secondary effect of Fgf10 down-regulation during eyelid development in the Osr2−/− mutant mice. However, Fgfr2 expression in the Osr2Osr1ki/Osr1ki embryos was also reduced, which correlated with incomplete recapitulation of endogenous Osr2 gene expression, in particular in the inner eyelid epithelium, suggesting that Osr2 may also play a primary role in maintaining Fgfr2 expression during eyelid development. In the Osr2Osr1ki/Osr1ki mice, although the leading edge epithelium of the developing eyelids formed periderm ridges that migrated over the developing corneal surface, the eyelids failed to fuse in the central regions due to the delayed eyelid development prior to E15.5. Further elucidation of the roles of the Osr1 and Osr2 transcription factors, in particular identification of downstream target genes, will be necessary to understand how these transcription factors interact with the Fgf10-Fgfr2 and other signaling pathways during mouse development.
We thank Tom Gridley for the CJ7 mouse ES cells, Qingru Wang for technical assistance, and the University of Rochester Transgenic Mouse Facility for assistance in generation of chimeric mice from targeted ES cells. This work was supported by NIH/NIDCR grant R01DE013681 to RJ.