Genetic dissection of Pitx2 in craniofacial development uncovers new functions in branchial arch morphogenesis, late aspects of tooth morphogenesis and cell migration

Pitx2 is a paired-related homeobox gene that was shown to be the gene mutated in Rieger Syndrome type I (RGS I)(Semina et al., 1996), an autosomal dominant, haploinsufficient disorder that includes tooth abnormalities as one of its primary features(Flomen et al., 1998). The craniofacial defects in individuals with RGS I, that have one half dose of Pitx2, include dental hypoplasia, anodontia vera, abnormally shaped teeth and a flattened midface (Amendt et al., 2000). Individuals with RGS I also have ocular anterior chamber disorders, which often result in glaucoma and umbilical abnormalities(Semina et al., 1996). Pitx2 plays a central role in left right asymmetry(Capdevila et al., 2000; Harvey, 1998) and is a component of Wnt-β-catenin signaling in pituitary and cardiac outflow tract development (Kioussi et al.,2002). Experimental evidence supports the idea that the dominant genetics of RGS I results from haploinsufficiency; however, there is evidence for a dominant negative mechanism in a subset of patients(Saadi et al., 2003; Saadi et al., 2001).

Investigation of Pitx2 function using loss-of-function approaches in mice has shown that Pitx2 plays an important role in early stages of tooth development (Gage et al.,1999; Kitamura et al.,1999; Lin et al.,1999; Lu et al.,1999). Pitx2-null mutant embryos had arrested tooth development at placode or bud stage. Consistent with a haploinsufficient mechanism, tooth phenotypes were observed in Pitx2 null ^+/– mice (Gage et al.,1999). Early epithelial-mesenchymal signaling was intact in Pitx2-null embryos as suggested by the presence of a condensed dental mesenchyme (Lin et al., 1999; Lu et al., 1999). Expression of markers such as Shh and mesenchymal Bmp4 and Msx1 also supported the idea that tooth initiation and specification occurred but tooth germ expansion failed in Pitx2-null embryos(Lin et al., 1999; Lu et al., 1999). In situ also showed that Bmp4 expression was expanded, while Fgf8 failed to be expressed or was downregulated in oral epithelium of Pitx2-null embryos (Lin et al., 1999; Lu et al., 1999). Taken together, these data suggest that the initial events in tooth development occurred in the absence of Pitx2, subsequent signaling events were deranged resulting in a premature extinction of Fgf8 expression and failure of demarcation of Bmp4 expression to dental epithelium. These experiments uncovered an early function for Pitx2 in tooth morphogenesis but failed to address any later role for Pitx2 in craniofacial development.

The Pitx2 gene encodes three isoforms, Pitx2a, Pitx2b and Pitx2c in mice and a fourth Pitx2 isoform, Pitx2d,has been identified in humans (Cox et al.,2002). The different isoforms are generated by both alternative splicing and alternative promoter usage(Shiratori et al., 2001)(Fig. 1A,B) and have both overlapping and distinct expression patterns. All Pitx2 isoforms have a common C terminus and distinct N termini(Fig. 1A). Pitx2c is the asymmetrically expressed isoform while Pitx2a, Pitx2b and Pitx2c isoforms are co-expressed in head mesoderm, oral ectoderm,eye, body wall and central nervous system(Kitamura et al., 1999; Liu et al., 2001; Schweickert et al., 2000; Smidt et al., 2000). Pitx2c, but not Pitx2a or Pitx2b, is expressed in hematopoietic stem cells (Degar et al.,2001). Co-expression of Pitx2 isoforms is found in the three developmental fields that are most frequently affected in individuals with RGS I: eyes, teeth and anterior abdominal wall.

The observation that Pitx2 regulated two fundamentally important signaling pathways in craniofacial morphogenesis raised the possibility that haploinsufficiency observed in humans and mice was a consequence of differential sensitivity of these important target pathways to total Pitx2 dose. An alternative idea, suggested by multiple Pitx2isoforms with overlapping expression in developing teeth, was that Pitx2 function in craniofacial development was a consequence of distinct isoform function. For example, it is conceivable that one Pitx2 isoform functions to repress Bmp4 while a separate isoform maintains Fgf8 expression. In addition, Pitx2isoforms have been shown to form heterodimers in vitro suggesting that Pitx2 isoform heterodimers may have distinct target genes(Cox et al., 2002). Overexpression of a Pitx2 engrailed repressor(en^r) fusion protein in left lateral plate of chick embryos revealed that Pitx2c en^r but not a Pitx2a en^r fusion could interfere with endogenous Pitx2cfunction (Yu et al., 2001),consistent with the idea that Pitx2 isoforms have distinct target genes. Experiments performed in Xenopus and zebrafish, as well as tissue culture studies, support the idea that Pitx2 isoforms have distinct targets (Cox et al.,2002; Essner et al.,2000; Faucourt et al.,2001; Suh et al.,2002).

We investigated Pitx2 isoform function in craniofacial morphogenesis by analyzing craniofacial phenotypes of isoform-specific deletions. We used Pitx2 alleles that encode differing levels of Pitx2 to investigate the requirements for total Pitx2 dose in craniofacial morphogenesis (Liu et al.,2001). Our results show that Pitx2 isoforms have interchangeable function in craniofacial development and that signaling pathways that are regulated by Pitx2 respond differently to changes in total Pitx2 dose. The Fgf8 maintenance pathway uses low Pitx2 doses, while Bmp4 repression requires high Pitx2 doses. Our findings uncovered downstream functions for Pitx2 in tooth development and fate mapping experiments with a Pitx2 cre recombinase knock-in allele revealed that Pitx2daughter cells are migratory. Movement of Pitx2 daughters was aberrant in Pitx2 mutants, suggesting that Pitx2 regulates cell movement in craniofacial primordia.

Whole mount and section in situ hybridization performed as described(Lu et al., 1999) with modifications for the use of digoxigenin labeled probes. Bmp4, Barx1,Pax9, Fgf8, Pitx2c and myogenin probes were described(Lu et al., 1999; Mitsiadis et al., 1998; Peters et al., 1998; Trumpp et al., 1999; Winnier et al., 1995; Liu et al., 2002).

Mouse embryos were fixed in Bouin's, dehydrated and embedded in paraffin wax. Sections were cut (7-10 μm) and stained with Hematoylin and Eosin. lacZ staining was as previously described(Lu et al., 1999).

The Pitx2 δabc^null,δ ab^hypoc, δab and δcalleles have been described previously(Liu et al., 2001; Liu et al., 2002; Lu et al., 1999). For Pitx2 δabc^creneo allele, a targeting vector was constructed that introduced cre recombinase neofrt into PvuII and Nru1 sites in Pitx2 fifth exon. Crosses to a rosa26 eFlp deletor strain resulted in neomycin removal(Farley et al., 2000). Crosses to Pitx2 δabc^null allele confirmed that Pitx2 δabc^creneo was a null allele and in situ hybridization experiments showed cre expression recapitulated endogenous Pitx2.

Total mRNA was extracted using SV total RNA isolation system (Promega) and cDNA produced with M-MLV reverse transcriptase (Invitrogen). Four Pitx2 primers detected Pitx2 isoform expression: exon 2(5′-attgtcgcaaactagtgtcgg-3′), exon 3(5′-ccgtgaactcgacctttttga-3′), exon 4(5′-tcctgggactcctccaaacat-3′) and exon 5(5′-gtttctctggaaagtggctcc-3′). A 104 bp Pitx2b fragment was amplified with exon 2 and exon 3 primers, 159 bp Pitx2a fragment with exon 2 and exon 5 primers and a 207 bp Pitx2c fragment with exon 4 and exon 5 primers.

The Pitx2 δabc^null allele, a homeobox deletion, removes function of all isoforms, while theδ ab^hypoc and δab alleles delete the Pitx2a and Pitx2b specific exons and leave Pitx2cintact (Fig. 1A,C). Theδ ab^hypoc allele, which retains PGKneomycin, encodes less Pitx2c function than the δab allele in which PGKneomycin was removed (Liu et al.,2001). We generated a deletion of the Pitx2c isoform(Liu et al., 2002), theδ c allele, that was a replacement of the Pitx2c-specific exon 4 with a LoxP flanked PGKneomycin. In the finalδ c allele, PGKneomycin has been removed by crossing to the CMVcre deletor strain (Liu et al., 2002) (Fig. 1C). To study the developmental progression of Pitx2daughter cells (see below), we generated Pitx2δ abc^creneo, a Pitx2 cre recombinase knockin allele (Fig. 1B; see Materials and methods). We introduced cre into Pitx2 exon 5 that resulted in a Pitx2 null allele and expressed cre in the same spatiotemporal pattern as endogenous Pitx2 (see below). Excision of the PGKneomycin cassette by crossing to the rosa26 eFlp deletor strain resulted in the Pitx2 δabc^cre allele.

We studied Pitx2a and Pitx2b isoform expression using theδ ab^hypoc and δab alleles that contain a lacZ knock-in into Pitx2 exon 2 and deletes Pitx2exon 3 (Fig. 1C-F). As lacZ was introduced into exon 2, this analysis provides information about Pitx2a and Pitx2b specific expression but does not distinguish between these two isoforms because Pitx2a uses exon 2 and Pitx2b uses both exon 2 and exon3(Fig. 1A). We used RT-PCR to distinguish between Pitx2a and Pitx2b expression (see below). We also performed in situ analysis using a Pitx2c probe. At 10.5 dpc, lacZ was expressed uniformly throughout the oral ectoderm,while at 14.5 dpc, lacZ expression was found in dental epithelium and primary enamel knot of cap stage tooth(Fig. 1D-F). Using a Pitx2c probe for in situ, we detected Pitx2c expression throughout the 10.5 dpc oral ectoderm (Fig. 1G). At 14.5 dpc, Pitx2c was expressed in dental epithelium similarly to Pitx2a and Pitx2b(Fig. 1H,I). To distinguish between Pitx2a and Pitx2b isoform expression in oral ectoderm, we performed RT-PCR with a primer set that distinguished between Pitx2a, Pitx2b and Pitx2c. We identified all three isoforms in the mandibular arch epithelium at 10.5 and 12.5 dpc(Fig. 1J). These data suggest that the Pitx2a, Pitx2b and the Pitx2c isoforms are coexpressed in oral ectoderm and, at later stages, within tooth epithelial structures.

Co-expression of Pitx2 isoforms suggests a number of possibilities for the regulation of target pathways by Pitx2. It is possible that Pitx2 isoforms would regulate distinct target genes in tooth formation or Pitx2 isoforms may have redundant functions. Isoform co-expression also supports the idea that some Pitx2 target genes have a requirement for Pitx2 heterodimers(Cox et al., 2002). To address these ideas, we analyzed forming teeth ofδ ab^–/– andδ c^–/– embryos.

As a control, we analyzed teeth of δab;δcmutant embryos. We reasoned that this allelic combination should encode near normal levels of all Pitx2 isoforms, albeit from different chromosomes, and should be functionally similar toδ abc^null heterozygous embryos. Analysis of coronal and sagittal sections through the teeth of δab;δ c embryos at 14.5 and16.5 dpc revealed that tooth development was normal (Fig. 2A-D). From this, we conclude that the δab and δc alleles encode adequate levels of Pitx2 isoforms to support normal tooth development.

To test the idea that Pitx2 isoforms had distinct target genes and thus distinct functions in tooth development, we analyzed the teeth ofδ ab^–/– embryos at two timepoints, 16.5 dpc and 18.5 dpc. We found that teeth of δab homozygous mutant embryos that lack Pitx2a and Pitx2b are normal suggesting that there is redundant function between the Pitx2a, Pitx2b and Pitx2c isoforms in tooth development or that Pitx2c has the major role in tooth development (Liu et al., 2001) and Fig. 2G,H,J,K). Sections through Pitx2c mutant teeth at 16.5 and 18.5 dpc revealed normal molar tooth morphology suggesting that Pitx2 a, Pitx2b and Pitx2c isoforms have redundant function in tooth morphogenesis (Fig. 2G,I,J,L). These data argue against an absolute requirement for either Pitx2isoform-specific target genes or Pitx2 isoform heterodimers in branchial arch morphogenesis and tooth development. These results suggest that common Pitx2 target genes are differentially regulated by total Pitx2 dose (Table 1).

Previous data suggested that Fgf8 expression was absent in Pitx2 δabc^null homozygous mutants(Lu et al., 1999) but was diminished only in embryos homozygous mutant for an independently generated Pitx2-null allele (Lin et al.,1999). One idea to explain this discrepancy is that Fgf8expression was induced but not maintained in Pitx2-null mutant embryos. To determine if Pitx2 was required for the maintenance of Fgf8 expression, we examined Fgf8 expression in Pitx2-null mutants at earlier timepoints than previously reported. In 9.5 dpc δabc^null homozygous mutant embryos, low levels of Fgf8 mRNA was expressed in the oral ectoderm(Fig. 3A,B). Sectioning revealed that the Fgf8 expression domain was restricted to a small region of oral ectoderm at the proximal aspect of the mandibular process in Pitx2 δabc^null homozygous mutants when compared with wild-type embryos (Fig. 3C,D). In the absence of Pitx2, the majority of the oral ectoderm loses the competency to express Fgf8, suggesting that Pitx2 has a role in the demarcation of the Fgf8 expression domain to the proximal aspect of the mandibular and maxillary processes. At later timepoints, Fgf8 expression is lost in Pitx2-null mutants (Lu et al., 1999) (see below).

We examined expression of genes that are proposed Fgf8 targets in mandibular mesenchyme. Lhx6 expression was shown to be dependent on Fgf8 function as Lhx6 failed to be induced in mutants with an oral ectoderm specific inactivation of Fgf8(Trumpp et al., 1999). In Pitx2 δabc^null mutants, Lhx6expression was reduced (Fig. 3E,F). The residual Lhx6 expression in the Pitx2δ abc^null embryos was in the proximal mandible near the region where Fgf8 was expressed in the Pitx2δ abc^null mutant embryos(Fig. 3D). Expression of Pitx1, normally expressed in the oral ectoderm and proximal mandibular mesenchyme, has been shown to be induced by implantation of an Fgf8 bead (St Amand et al.,2000). Pitx1 expression was reduced in the proximal aspect of the Pitx2 δabc^null mutant mandibular arch mesenchyme at 10.5 dpc(Fig. 3G,H). Expression of Dlx2 in mandibular mesenchyme has also been shown to be upregulated by Fgf8 bead implantation (Thomas et al., 2000). We found that the mesenchymal expression of Dlx2 was reduced in Pitx2 δabc^nullmutants (Fig. 3I,J). As previous data suggested that induction of Pitx1 and Dlx2expression was independent of Fgf8, our results suggest that Fgf8 functions to maintain pitx1 and dlx2expression in the mandibular mesenchyme(Trumpp et al., 1999). Expression of endothelin 1 (Edn1), also dependent on Fgf8function, was downregulated in the mandibular arch ectoderm of Pitx2δ abc^null mutants(Fig. 3K,L). It is notable that expression of Lhx6, Pitx1, and Dlx2 in the maxillary primordial of Pitx2 δabc^null mutants was also reduced; however, further experiments are necessary to rule out the possibility that this was secondary to reduction in the outgrowth of the forming maxilla (Fig. 3E-J).

We noted that Dlx2 was still expressed in the caudal aspect of the Pitx2 mutant mandibular mesenchyme(Fig. 3I,J). As Pitx2expression is restricted to the rostral mandibular arch ectoderm, continued expression of Dlx2 in caudal mandibular mesenchyme suggested that Fgf8 signaling from the caudal aspect of the mandibular ectoderm was intact in the Pitx2 δabc^null mutant embryos and that patterning of the mandibular process was disrupted in the Pitx2 δabc^null mutants. Goosecoid(Gsc), an Fgf8 responsive homeobox gene, is normally expressed in the caudal mandibular arch mesenchyme. Caudal Gscexpression is normally maintained via a Fgf8 repressive pathway that inhibits Gsc expression in the rostral mandibular process(Tucker et al., 1999). We reasoned that if maintenance of Fgf8 signaling was disrupted in Pitx2 δabc^null mutants, then Gscexpression should be expanded rostrally. We found that Gsc expression was weakly expanded in a subset of Pitx2δ abc^null mutants embryos(Fig. 3M,N), while in the remainder of mutant embryos Gsc expression was caudally restricted(data not shown). The incomplete penetrance of expanded Gscexpression suggests that in the subpopulation of Pitx2 mutant embryos with correct Gsc expression, the early Fgf8 expression was sufficient to specify the correct Gsc expression domain.

Correct patterning of the mandibular mesenchyme is necessary for formation of Meckel's cartilage (Tucker et al.,1999). Based on the weak expansion of Gsc expression, we expected that Pitx2-null mutants would have a weak Meckel's cartilage phenotype. To assess this, we performed whole-mount cartilage staining on Pitx2 δabc^null mutants and control wild-type littermate embryos. The Pitx2δ abc^null mutants had a variable deficiency of Meckel's cartilage supporting the notion that rostral caudal polarity of the mandibular process was weakly affected by loss of Pitx2 function(Fig. 3O,P). Taken together,these data suggest that in the absence of Pitx2, Fgf8 expression in oral ectoderm fails to be maintained. In the absence adequate Fgf8signaling, Fgf8-dependent signaling to underlying mesenchyme is reduced leading to defective mandibular arch rostral caudal polarity.

To address the idea that Pitx2 target pathways have distinct requirements for total Pitx2 dose, we examined Fgf8 and Bmp-signaling pathways in Pitx2 allelic combinations that encode differing levels of Pitx2 activity(Liu et al., 2001). We used the δabc^null allele, in conjunction with theδ ab and δab^hypoc alleles that encode reduced levels of Pitx2c in the absence of Pitx2a and Pitx2b to generate Pitx2 allelic combinations with intermediate levels of Pitx2 activity. Previously, we showed that theδ abc^null+/– embryos expressed ∼58% of homozygous wild-type Pitx2c mRNA levels while theδ abc^null;δab andδ abc^null; δab^hypoc allelic combinations expressed ∼50% and 38% of wild-type Pitx2c mRNA levels respectively (Liu et al.,2001).

At 10.5 dpc, Fgf8 expression was not detectable in the Pitx2 δabc^null homozygous mutant oral ectoderm, supporting the idea that Pitx2 was required for maintenance of Fgf8 expression in the oral ectoderm(Fig. 3Q,R)(Lin et al., 1999; Lu et al., 1999). In the rostral mandibular process of Pitx2δ abc^null mutant embryos, Barx1 and Pax9, mesenchymal targets of Fgf8 signaling pathways(Neubuser et al., 1997; Tucker et al., 1998), were not expressed or had greatly diminished expression(Fig. 3T,U,W,X). Caudal mandibular arch expression of Barx1 was maintained in Pitx2δ abc^null mutant embryos as this expression is probably dependent on Fgf8 and Edn1 signaling from the caudal aspect of the mandibular process that does not express Pitx2(Fig. 3R,S). By contrast, theδ abc^null;δab andδ abc^null;δab^hypoc allelic combinations, that encode reduced levels of Pitx2c mRNA and lack Pitx2a and Pitx2b (Liu et al., 2001) (Fig. 1C), expressed Fgf8 in the oral ectoderm of 10.5 dpc embryos (Fig. 3Q-S and data not sown). Barx1 and Pax9 were expressed in theδ abc^null;δab^hypoc embryos that encode low levels of Pitx2(Fig. 3T,W).

We investigated whether repression of Bmp signaling by Pitx2 was also rescued in theδ abc^null;δab^hypoc allelic combination that encodes low levels of Pitx2 function. To assess expansion of Bmp signaling, we examined Bmp4 expression in oral ectoderm of 10.5 dpc Pitx2 mutant embryos. In contrast to the Fgf8 signaling pathway, Bmp repression required high levels of Pitx2 function. In Pitx2δ abc^null–/– embryos Bmp4expression was expanded laterally in mandibular process ectoderm(Fig. 4A,B)(Lu et al., 1999). In wild-type embryos, Bmp4 expression is found in the medial mandibular process and the distal aspect of the ectoderm of the maxillary process at 10.5 dpc (Fig. 4B,E). In Pitx2δ abc^null;δab^hypoc andδ abc^null;δab allelic combinations, Bmp4 expression in the mandibular process was weakly expanded. Moreover, in the maxillary process ectoderm of Pitx2δ abc^null;δab^hypoc andδ abc^null;δab mutants, Bmp4expression failed to be distally restricted and was detected all the way to the junction with the mandibular process(Fig. 4C-G).

We examined expression of Msx1 and Msx2 that are mesenchymal targets of Bmp signaling(Barlow and Francis-West, 1997; Vainio et al., 1993). In Pitx2 δabc^null–/– embryos andδ abc^null;δab^hypoc andδ abc^null;δab allelic combinations,expression of Msx2 (Fig. 4H-K) and Msx1 (Fig. 4L-O) was expanded proximally in the mandibular and maxillary processes. These data also revealed that expression of Msx1 and Msx2 was more obviously expanded than the Bmp4 ligand,particularly in the mandibular process in Pitx2 mutant allelic combinations. We noted that expression of Msx1 and Msx2 was expanded in the branchial arch mesenchyme that probably contributes to the developing heart in some Pitx2 mutant embryos(Fig. 4K,O). Taken together,these results suggest that maintenance of Fgf8 expression and repression of Bmp-signaling pathways have distinct requirements for total Pitx2 dose in the branchial arches (summarized in Table 1).

We investigated the tooth morphology of theδ abc^null;δab^hypoc andδ abc^null;δab allelic combinations using histological analysis. Sections through 18.5 dpc wild-type, andδ abc^null;δab mutant embryos revealed well-formed molars. We found that in the δabc^null;δ ab embryos, the orientation of the molar tooth was abnormal(Fig. 5A,C,E,G). Inδ abc^null;δab^hypoc 18.5 dpc mutant embryos, analysis of serial sections revealed that molar teeth were absent (Fig. 5B,F). As lacZ marks cells fated to express Pitx2a and Pitx2b, serving as a marker of dental epithelium, we performed lacZ staining on serial cryosections from heads of 14.5 dpc Pitx2 allelic combinations. In δab^+/–and δabc^null; δab embryos,well-formed cap stage molar teeth were clearly evident with lacZstaining (Fig. 5I,J). Inδ abc^null; δab^hypoc mutant embryos, the dental lamina invaginated but failed to form the dental cap(Fig. 5K). In Pitx2δ abc^null homozygous mutant embryos, tooth development arrested at the placode or bud stage. The molar phenotype inδ abc^null; δab^hypocembryos, with a more developed dental lamina, suggests that tooth development progressed further than in δabc^null mutant embryos. These data show that as the dose of Pitx2 decreases there is evidence of increasingly severe defects in tooth morphogenesis.

From these results, we conclude that Pitx2 has a late function in molar orientation and in morphogenesis of the cap stage tooth. The intermediate tooth phenotypes observed in theδ abc^null; δab^hypoc andδ abc^null; δab mutants most probably reflects a direct role for Pitx2 in morphogenesis of dental epithelium. Although it is possible that expression of Fgf8 in the Pitx2 δabc^null;δ ab^hypoc and δabc^null;δ ab oral ectoderm is inadequate to completely rescue molar tooth development, the expression of Pax9 and Barx1 in dental mesenchyme of these allelic combinations suggests that Fgf8signaling to mesenchyme is intact in these mutant embryos and argues that Pitx2 directly regulates epithelial morphogenesis.

Our previous data revealed that Pitx2 functioned to regulate local cell movement in heart development (Liu et al., 2002). To determine if a similar mechanism was at work in craniofacial development, we used the δabc^cre knock in allele and the Gtrosa 26 reporter mouse to follow the movement of Pitx2 daughter cells within the first branchial arch. At 9.5-11.0 dpc, cre expression was detected in the oral ectoderm in bothδ abc^cre+/– andδ abc^cre; δabc^null embryos,although by 11.0 dpc cre expression was diminished in theδ abc^cre; δabc^null embryos(Fig. 6A,B and not shown). Cre expression was restricted to oral ectoderm and was not found in facial ectoderm or epithelium lining the oral cavity(Fig. 6C-E). Fate mapping with the GtRosa26 reporter showed that Pitx2 daughters were detected in the oral ectoderm, periocular mesenchyme, guts, heart and body wall (Fig. 6F,G).

In the craniofacial region, Pitx2 daughters moved outwards from the oral ectoderm to the facial ectoderm in both wildtype and mutant embryos(Fig. 6H-K). As cremRNA expression was restricted to oral ectoderm, these data reveal that lacZ-positive migrating cells were Pitx2 daughters that had extinguished Pitx2 expression. There were differences in the pattern of daughter migration in Pitx2-null mutant compared with wild-type embryos. In wild-type embryos, Pitx2 daughters moved a short distance to cover the outer aspect of the mandibular and maxillary process. Some Pitx2 daughters also contributed to the nasal process of wild-type embryos (Fig. 6H,J). In Pitx2 mutant embryos, daughter cells moved aberrantly in a dorsal direction just inferior to the eye and failed to contribute to the mutant nasal process (Fig. 6I,K).

Pitx2 daughters extensively populated the floor and roof inside the forming mouth (Fig. 6L-O). In Pitx2 mutants, fewer daughter cells populated the oral cavity roof as compared with wild type (Fig. 6N-Q). Pitx2 daughters contributed to Rathke's pouch and dental epithelium, of both the wild type and mutant although in the Pitx2 mutant tooth morphogenesis was arrested(Fig. 6N-S and not shown). These data reveal that Pitx2 daughter cells exit the oral ectoderm and contribute to both facial ectoderm and the ectoderm lining the oral cavity and Pitx2 function is necessary for correct deployment and expansion of daughter cells.

In craniofacial development, the mechanisms that organize growth and morphogenesis of the branchial arches remain poorly understood. We investigated Pitx2 isoform function in craniofacial morphogenesis using Pitx2 exon-specific deletions. Analysis of Pitx2allelic combinations encoding different levels of Pitx2 also uncovered the influence of variations in total Pitx2 dose on Fgf8 and Bmp4 signaling(Table 1). Our data indicate that Pitx2 isoforms have interchangeable function in craniofacial development and that Pitx2 target pathways have distinct requirements for total Pitx2 dose. Reduced Pitx2 levels resulted in unbalanced interplay between Fgf8 and Bmp4 signaling pathways in craniofacial morphogenesis. We found that Pitx2 daughter cells are migratory, eventually populating intraoral and facial ectoderm, and that Pitx2 function is required for this movement. We provide evidence that Pitx2 connects overall growth and morphogenesis of the first branchial arch through a mechanism involving differential sensitivity of target pathways to total Pitx2 dose.

Deletion of Fgf8 in oral ectoderm revealed a role for Fgf8 in survival and outgrowth of mandibular mesenchyme(Trumpp et al., 1999), while pharmacological suppression of Fgf signaling in explants suggested that Fgf functioned primarily by signaling to the underlying mesenchyme(Mandler and Neubuser, 2001). Bead implantation also suggested an early role for Fgf8 in establishing the maxillo-mandibular region of the chick embryo(Shigetani et al., 2000). Importantly, antagonistic interactions between Fgf and Bmp signaling has been implicated in proximodistal mandibular arch patterning, placement of tooth organ formation and determination of the maxillo-mandibular region of the early embryo (Neubuser et al.,1997; Shigetani et al.,2000; Tucker et al.,1998).

Our data reveal that Pitx2 maintains Fgf8 expression in branchial arch ectoderm. Expression of prospective Fgf8 target genes,such as Barx1 and Pitx1, was severely reduced in Pitx2 δabc^null homozygous mutant embryos. Consistent with a role of Fgf8 signaling in mandibular rostral caudal polarity, expression of Gsc was expanded rostrally in the mandibular process of Pitx2 δabc^null mutants. In addition, as Pitx2 is normally expressed in rostral mandibular arch ectoderm that contributes to oral ectoderm, Pitx2δ abc^null homozygous mutants lose Barx1expression in the rostral but not caudal mandibular arch. The Pitx2δ abc^null; δab^hypoc andδ abc^null; δab mutant embryos express Fgf8 and Fgf8 target genes, suggesting that maintenance of this pathway requires only low doses of Pitx2.

In contrast to Fgf8, high doses of Pitx2 are required for repression of Bmp signaling. In theδ abc^null; δab^hypoc andδ abc^null; δab mutants, expression of Bmp4 was expanded in maxillary ectoderm while Msx1 and Msx2 expression was expanded in mesenchyme of both maxillary and mandibular processes. Thus, expression of the Bmp target genes was more significantly expanded than expression of Bmp4 ligand. This may reflect the induction of a signal relay cascade in the mandibular process. It is also interesting to note that Dpp has been shown to act as a classical morphogen in the wing imaginal disc of Drosophila(Entchev et al., 2000; Teleman and Cohen, 2000).

We found that in δabc^null;δ ab^hypoc and δabc^null;δ ab Pitx2 mutants components of Bmp4 and Fgf8signaling pathways, such as Msx1 and Barx1, are co-expressed in mandibular mesenchyme. Previous work suggested an antagonistic interaction between these two signaling pathways(Neubuser et al., 1997; Tucker et al., 1998). It is likely that in the Pitx2 mutant allelic combinations, Bmp signaling is only weakly expanded and this is insufficient to antagonize expression of Barx1 in mandibular mesenchyme.

These data provide insight into the normal function of Pitx2 in regulating gene expression. The Fgf8 pathway and the Bmpsuppression pathway have different requirements for total Pitx2 dose. As Pitx2, Fgf8 and Bmp4 are co-expressed in many cells of the oral ectoderm, one can envision a mechanism where Pitx2 would directly regulate Fgf8 and Bmp4 expression. In this model,one idea to explain the different requirements for Pitx2 dose in regulating Bmp4 and Fgf8 would be that the regulatory regions of Bmp4 and Fgf8 contain different numbers of high-affinity Pitx2-binding sites, a mechanism suggested to underlie the haploinsufficiency of individuals with Holt-Oram syndrome that are heterozygous for tbx5 (Bruneau et al., 2001). Thus, Pitx2 target genes with more Pitx2-binding sites would require higher doses of Pitx2 for correct levels of gene expression. However, this model is complicated by in vitro observations showing that Pitx2 can cooperatively bind DNA(Dave et al., 2000; Wilson et al., 1993),suggesting that low levels of Pitx2 can form higher order complexes on DNA. It is likely that there are other mechanisms, such as interaction with co-factors, to constrain or augment the ability of Pitx2 to activate target genes. Further experiments are necessary to rule out the possibility that Pitx2 indirectly regulates the Fgf8 and Bmp4pathways.

Pitx2-null embryos have arrest of tooth development at the placode or bud stage (Gage et al.,1999; Lin et al.,1999; Lu et al.,1999). In the Pitx2 δabc^null;δ ab^hypoc and δabc^null;δ ab embryos, molar tooth morphogenesis was partially rescued in that an invaginated dental lamina formed without a cap or the orientation of the dental cap was abnormal. Our in situ studies showed that Fgf8 was expressed in the oral ectoderm of δabc^null;δ ab^hypoc and δabc^null;δ ab embryos. Moreover, expression of Pax9 was also detected in the prospective dental mesenchyme and Barx1 was expressed in proximal mandibular mesenchyme of these embryos revealing that Fgf signaling to mandibular mesenchyme is intact in the Pitx2 hypomorphic embryos. Although expanded Bmp signaling could account for tooth defects in the δabc^null;δ ab^hypoc and δabc^null;δ ab embryos, the abnormal tooth morphology was not suppressed by reducing Bmp4 dose using a Bmp4-null allele (W.L. and J.F.M., unpublished). Based on these data, we favor the notion that Pitx2 regulates tooth morphogenesis through a pathway that is distinct from Fgf8 and Bmp4 signaling, although further experiments are required to investigate these ideas.

Our fate-mapping studies show that Pitx2 daughter cells move from oral ectoderm to populate facial and inner oral cavity ectoderm. Pitx2-expressing cells make a decision to extinguish Pitx2and become motile. It may be that Pitx2 expression promotes cell compaction or inhibits cell motility. It is notable that one of the phenotypes of the Pitx2-null embryos was failure of compaction and differentiation of the periocular mesenchyme(Lu et al., 1999). Fgf8 signaling was implicated in cell movement as Fgf8-null embryos had defects in cell migration through the primitive streak. Analysis of Xenopus sprouty2, an inhibitor of Fgf signaling, revealed that Fgf signaling in Xenopus regulated both mesoderm induction and convergent extension movements (Nutt et al.,2001). Thus, it is plausible that Pitx2 regulates cell movement in the craniofacial primordia through an Fgf8-mediated pathway.

A direct connection of Pitx2 to cytoskeleton and morphogenetic movement has been made by the observation that Pitx2 controls Rho GTPase activity by regulating expression of the guanine nucleotide exchange factor, Trio (Wei and Adelstein, 2002). It has recently been proposed that Pitx2 is a target of canonical Wnt β-catenin signaling pathway in pituitary and cardiac development(Kioussi et al., 2002). This work uncovered a genetic interaction between Pitx2 and dishevelled 2,a Wnt pathway branchpoint, in the heart. Other studies showed that Rho family GTPases are downstream components of non-canonical planar cell polarity (PCP) pathway (Habas et al.,2003; Strutt et al.,1997; Winter et al.,2001). Although further experiments are required, our data showing that Pitx2 daughters are migratory supports the idea that Pitx2 may be a component of a non-canonical Wnt pathway in craniofacial development.

The phenotypes in individuals with Rieger syndrome with PITX2mutations are heterogeneous. Our data reveal that slight changes in Pitx2 dose can have a large influence on resulting phenotypes. This is illustrated most clearly by comparing theδ abc^null; δab^hypoc andδ abc^null; δab mutants that have only slight changes in Pitx2 activity but dramatic differences in tooth morphogenesis (Liu et al.,2001). Many organ systems, such as heart and lungs, cannot distinguish between these small differences in Pitx2 activity(Liu et al., 2001).

The isoform deletions of Pitx2 reveal functional redundancy between isoforms in tooth development. These data are consistent with the observation that all Pitx2 mutations detected in individuals with Rieger syndrome are in regions common to all isoforms(Alward, 2000; Kozlowski and Walter, 2000; Priston et al., 2001; Saadi et al., 2001). Our data suggest that the Pitx2 N terminus does not have a significant function in tooth morphogenesis because this region is not conserved between Pitx2a, Pitx2b and Pitx2c. This differs from pituitary and skeletal muscle where the N terminus has an influence on Pitx2function (Kioussi et al.,2002; Suh et al.,2002). It is also clear that Pitx1 functions cooperatively with Pitx2 in pituitary organogenesis and limb development (Marcil et al.,2003). As Pitx1 is co-expressed with Pitx2 in developing teeth, it will be interesting to investigate potential cooperative functions of Pitx1 and Pitx2 in oral and dental epithelium.

We thank A. Bradley, P. Soriano, R. Behringer for reagents; A. McMahon, G. Martin, R. Balling, T. Mitsiadis, B. Hogan and W. Klein for in situ probes. Supported by grants NIDCR (2R01DE/HD12324-06 and R01DE013509) and by Grant Number 5-FY97-698 from March of Dimes (J.F.M.).