Ablation of specific expression domains reveals discrete functions of ectoderm- and endoderm-derived FGF8 during cardiovascular and pharyngeal development

The pharyngeal arches (PAs) consist of mesenchyme encased in ectoderm and endoderm. PA mesenchyme is derived from paraxial mesoderm and neural crest(NC, ectomesenchyme). Mesoderm-derived mesenchyme contributes muscular and endothelial precursors to the PAs. Some of the ectomesenchyme gives rise to skeletal structures, and to pericytes and smooth muscle cells of the pharyngeal arch arteries (PAAs), while a different population traverses the PAs en route to forming parts of the outflow tract (OFT) and heart(Jiang et al., 2000; Li et al., 2000). A PAA forms within each arch, connecting the heart to the bilateral dorsal aortae; this initially symmetric vascular array is extensively remodeled during development. This morphogenetic process is coordinately regulated by a number of signaling pathways that, if perturbed, can result in aberrant formation and/or remodeling of the PAAs and lethal congenital cardiovascular malformations.

Fgf8 encodes a crucial member of the FGF family(MacArthur et al., 1995). Secreted FGF8 protein provides survival, mitogenic, anti/pro-differentiation and patterning signals to adjacent tissues, and may also have autocrine activity. Complete ablation of Fgf8 function in mice results in early embryonic lethality at approximately embryonic day (E) 8.5(Meyers et al., 1998; Moon and Capecchi, 2000; Sun et al., 1999). We and others have thus employed hypomorphic and conditional alleles to study its role in limb, face, brain, cardiovascular and pharyngeal development(Abu-Issa et al., 2002; Frank et al., 2002; Garel et al., 2003; Meyers et al., 1998; Meyers and Martin, 1999; Moon et al., 2000; Moon and Capecchi, 2000; Storm et al., 2003; Sun et al., 2000; Trumpp et al., 1999).

Fgf8 mutations in several species demonstrate its role(s) in early cardiovascular and PA development. Zebrafish acerebellar Fgf8 mutants have abnormal cardiogenesis, ventricular hypoplasia and circulatory failure(Reifers et al., 2000). Removal of Fgf8-expressing endoderm adjacent to precardiac mesoderm in chicks alters expression of cardiac markers such as Nkx2.5(Alsan and Schultheiss,2002).

In the mouse, Fgf8 is expressed in several temporospatial domains that are potentially relevant to cardiovascular and pharyngeal development. These include the precardiac mesoderm(Crossley and Martin, 1995),the early foregut endoderm and later, in restricted regions of PA endoderm and ectoderm. Murine Fgf8 hypomorphic mutants (mice with globally decreased FGF8 signaling throughout embryogenesis) have severe cardiovascular and pharyngeal defects, including altered cardiac outflow tract (OFT)alignment and septation, disrupted pharyngeal vascular development, and abnormal formation of the thymus and parathyroids(Abu-Issa et al., 2002; Frank et al., 2002). These Fgf8 hypomorphs phenocopy human syndromes associated with deletion of chromosome 22q11 (del22q11) such as DiGeorge syndrome(Epstein, 2001; Frank et al., 2002; Lindsay, 2001; Scambler, 2000). Furthermore, Fgf8 genetically interacts with Tbx1(Vitelli et al., 2002), a gene located in the human del22q11 region known to play a crucial role in generating human del22q11 phenotypes(Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001). Thus,delineating the function of specific FGF8 signaling domains and downstream pathways will provide insight into how their dysfunction results in the spectrum of birth defects seen in human del22q11syndromes.

Although the Fgf8 hypomorphic model provides enormous insight into the importance of FGF8 signaling during cardiovascular and pharyngeal development, it does not allow us to test the role of local (pharyngeally produced) FGF8 in development of the pharynx or cardiovascular system, or to dissect the respective role(s) of different Fgf8 expression domains that are relevant to these morphogenetic processes. Furthermore, analyses of the molecular and cellular pathways that are disrupted by loss of a specific Fgf8 expression domain cannot be assessed with this system.

We have postulated that the Fgf8 hypomorph cardiovascular and pharyngeal phenotypes result from disrupted local FGF8 signaling from the epithelia of PAs 3-6 to mesenchymal cells populating and migrating through these arches, including cardiac neural crest en route to the OFT(Frank et al., 2002). To test this hypothesis, and to determine whether FGF8 signals emanating from the pharyngeal ectoderm and endoderm perform discrete functions, we generated a unique series of Fgf8 conditional alleles and Cre recombinase-expressing drivers designed to ablate FGF8 in different pharyngeal epithelial domains.

The strategy for homologous recombination in ES cells used to target Fgf8, and the other loci noted below, was as previously described(Moon et al., 2000).

The conditional alleles employed in this study are shown schematically in Fig. 1. Insertion of the GFP reporter gene into the 3′ untranslated region of Fgf8 was performed as previously described for the Fgf8^APNhypomorphic, conditional reporter allele(Frank et al., 2002; Moon and Capecchi, 2000). The Fgf8^null allele, resulting from removal of exon 5, has also been reported (Moon and Capecchi,2000). Mutant embryos are in a 75% C57Bl6, 25% SV129 background.

An ectodermal domain-specific `Cre driver' was generated by targeting an IRESCre cassette into the 3′ untranslated region of the AP2α locus (Fig. 2A). This cassette contains an Internal Ribosomal Entry Site, IRES(Jackson et al., 1990; Jang and Wimmer, 1990),upstream of the Cre recombinase gene(Sauer and Henderson, 1988)and an frt-flanked neomycin phosphotransferase gene(Fig. 2A). Placing the IRESCre cassette between the stop codon and the endogenous poly-adenylation signal allows regulation of Cre expression by the AP2α locus with intact AP2α gene function.

To ablate FGF8 in both the endoderm and ectoderm of PAs 3-6, we targeted the IRESCre cassette into the hoxa3 locus in a genomic ApaI site located 3′ of the stop codon(Fig. 3A).

Alkaline phosphatase staining was performed as previously described(Moon et al., 2000).

Somite/stage-matched embryos were fixed and whole mount terminal UTP nick end labeled (TUNEL) assays performed as described previously(Frank et al., 2002; Stadler et al., 2001).

Whole-mount green fluorescent protein (GFP) detection was performed on whole embryos using a rabbit anti-GFP and FITC-conjugated anti-rabbit IgG secondary antibodies (1:1000 and 1:500, respectively, both from Molecular Probes).

For cryosectioned specimens, embryos were protected in sucrose and gelatin-embedded. Cryosections (12 μm) were cut transversely, parallel to the third PAA. The Ap2α transcription factor, which is expressed in neural crest and ectoderm, was detected with a mouse monoclonal anti-AP2α-antibody (1:25, 3B5, Developmental Studies Hybridoma Bank) and a FITC-conjugated anti-mouse IgG secondary (1:500, Molecular Probes). Simultaneous TUNEL was performed by adding the TMR Red in situ cell death detection reagents (Roche) to secondary antibody incubation. Sections were preserved in Fluoromount-G (Southern Biotechnology Associates) and analyzed by confocal microscopy. FITC and Texas red fluorescence were recorded using a BioRad MRC 1024 laser-scanning confocal imaging system fitted to a Leitz Aristoplan microscope. A digital Kalman averaging filter was used to reduce background fluorescence.

Since Fgf8 null homozygotes die at approximately E8.5 due to defects in gastrulation (Meyers et al.,1998; Moon and Capecchi,2000; Sun et al.,1999), conditional mutagenesis is required to study the different roles of this protein during murine development.

We previously described a conditional reporter allele, Fgf8^APN, that is hypomorphic due to the presence of a neo^r gene in the 3′ untranslated region of the Fgf8locus (Frank et al., 2002; Moon and Capecchi, 2000)(Fig. 1A). Mice bearing this allele and a null allele of Fgf8 are hypomorphs (genotype Fgf8^APN/null) and die at birth with the aforementioned complex phenotype (Frank et al.,2002).

For the current study, we modified the Fgf8^APN allele by removing the frt-flanked neo^r gene with flp-mediated recombination in the germline of founder animals(Dymecki, 1996)(Fig. 1B). The resulting Fgf8^AP allele is not hypomorphic. Fgf8^AP/null compound heterozygotes survive and are phenotypically normal (see Fig. 4A). The Fgf8^AP allele was also designed as a conditional reporter allele: exon 5 is flanked with loxP sites and coding sequences for human alkaline phosphatase (AP) are positioned in the 3′untranslated region of Fgf8. Cre-mediated recombination of this allele removes exon 5 and allows expression of the AP reporter gene under control of the Fgf8 promoter.

We also generated a second nonhypomorphic conditional reporter allele, Fgf8^GFP, using the strategy described for the Fgf8^AP allele. Fgf8^GFP functions identically to Fgf8^AP except that green fluorescent protein (GFP) is produced upon Cre-mediated recombination of the allele.

Expression of either reporter gene depends on: (1) Cre-mediated recombination of Fgf8^AP or Fgf8^GFP to generate the Fgf8^APR and Fgf8^GFPR null reporter alleles (Fig. 1C) and(2) activity of the Fgf8 locus. These reporters permit precise determination of the temporospatial inactivation of Fgf8 in cells in which it is expressed (Moon et al.,2000; Moon and Capecchi,2000).

Deletion of exon 5 from the of Fgf8^AP or Fgf8^GFP conditional alleles depends on expression and activity of Cre-recombinase. We obtained spatial and temporal control over Cre by inserting an IRES followed by Cre-encoding sequences into the 3′untranslated region of two genes that are expressed in pharyngeal domains of interest, the Ap2α and hoxa3 loci (see Materials and methods; Fig. 2A, Fig. 3A). As described in detail below, to determine whether expression of these targeted Cre drivers recapitulated the pattern of the endogenous loci, we examined the expression of the global Cre reporter gene, Rosa26^lacZ(Soriano, 1999). Cre-mediated recombination of the Rosa26^lacZ allele results inβ-galactosidase production from the recombined, constituitively expressed Rosa26 locus. Furthermore, we determined the domains of functionally relevant recombination of Fgf8 obtained with these Cre drivers by characterizing expression of recombined Fgf8 conditional reporter alleles in the developing PAs.

We ablated Fgf8 gene function in its PA ectodermal expression domains from the time of PA formation with the targeted AP2α-IRESCre driver(Fig. 2A,B). The AP2α gene is expressed in many regions of the mouse embryo(Mitchell et al., 1991),including pharyngeal NC and ectoderm(Brewer et al., 2002). Lineage analyses of the AP2α-IRESCre driver in the PAs were performed by crossing this allele into mice bearing the Rosa26^lacZ allele (genotype AP2α^IRESCre/+; Rosa26^lacZ/+, Fig. 2C). β-Galactosidase activity is detectable in developing PAs1 and 2, indicating that onset of Cre activity occurred at approximately the 10 somite stage (ss, indicated in lower right corner of each panel). Caudal ectoderm over the region that will form PAs3-6 is also stained prior to definitive arch formation(Fig. 2C, large red arrowheads).

Cells that express both AP2α-IRESCre and Fgf8 were identified by staining for alkaline phosphatase (AP)activity in an Fgf8^AP/+; AP2α^IRES-Cre/+, 21 ss embryo. Analysis of whole mount and coronal sections demonstrates that AP is expressed specifically in the ectoderm of PAs 1-3(Fig. 2D). This expression pattern reflects cells of the AP2α lineage that express Fgf8^APR (generated by AP2α-IRESCreactivity), and defines the functionally relevant domains of Fgf8inactivation.

We confirmed that AP2α-IRESCre ablates Fgf8 throughout its ectodermal expression domains by comparing expression of GFP in Fgf8^GFP/+;AP2α^IRES-Cre/+ embryos versus Fgf8^GFP/+;deleterCre embryos(Fig. 2E, left and right panels, respectively). The deleterCre transgene is active in germ cells so Fgf8^GFP is recombined in all cells of the embryo from the earliest developmental stages(Schwenk et al., 1995). Therefore, GFP expression in Fgf8^GFP/+;deleterCre embryos depends only on the activity of the Fgf8 locus. Importantly, Fgf8^GFPR expression in the PA ectoderm is the same whether it results from the action of AP2α-IRESCre or deleterCre (Fig. 2E). Although Fgf8 is also expressed in the PA endoderm at this stage(discussed below), the endodermal domain of GFP expression in the Fgf8^GFP/+;deleterCre embryo(Fig. 2E, right panel) is obscured by overlying ectodermal signal. Coronal sections of a 20 ss Fgf8^GFP/+; AP2α^IRES-Cre/+ embryo(Fig. 2F,G) reveal that AP2α-IRESCre is not active in the endoderm, and that the ectoderm of PAs 1, 2 and developing PA3 express GFP.

All of these data provide confirm that in Fgf8;AP2α-IRESCre conditional mutants(Fgf8^GFP/null; AP2α^IRES-Cre/+),FGF8 is ablated specifically in the PA ectoderm.

To simultaneously ablate Fgf8 gene function in both the endoderm and ectoderm of PAs 3-6, we used the hoxa3-IRESCre driver(Fig. 3A,B). Lineage analysis of hoxa3-IRESCre in hoxa3^IRESCre/+;Rosa26^lacZ/+ embryos shows that hoxa3-IRESCrerecapitulates the caudal to rostral progression of the endogenous hoxa3 locus (Fig. 3C,10-28 ss whole-mount panels). Previous reports described hoxa3expression in endoderm and mesenchyme of PAs3 and 4(Manley and Capecchi, 1995),but we also found lacZ staining in the ectoderm of PAs 3-6 (see Fig. 3C, sectioned 20 ss and 23 ss embryos). Ectodermal lacZ staining is lighter than in mesenchyme(which may explain inability to detect this domain of hoxa3 mRNA by in situ hybridization). Ectodermal expression was consistent in all embryos and was also confirmed using the Fgf8^GFP reporter allele.

To evaluate functionally relevant activity of hoxa3-IRESCre in Fgf8-expressing cells, we analyzed Fgf8^GFPR expression in Fgf8^GFP/+;hoxa3^IRESCre/+ coronally sectioned embryos(Fig. 3D) in comparison with Fgf8^GFP/+;deleterCre embryos(Fig. 3E). The relative planes of ventral and dorsal coronal sections are demonstrated by the black lines labeled v and d in the 20 ss whole mount in Fig. 3C. Fgf8^GFPR is clearly expressed in the endoderm and ectoderm of developing PAs 3-6, including pharyngeal pouches and clefts. At the 20 ss,ventral and dorsal sections (Fig. 3D, 20v and 20d, respectively) show Fgf8^GFPRexpression throughout the PA3 endoderm, including the developing third pouch(3p), and ectoderm. By the 24 ss, Fgf8^GFPR expression is decreasing in the rostral endoderm of PA3, but persists in caudal endoderm and pouch (Fig. 3D, 24v and 24d,yellow arrowheads). At the 27 ss, GFP is detected throughout endoderm and ectoderm of PA4 as it forms. Remarkably, the expression pattern of GFP was the same in Fgf8^GFP/+;deleterCre and Fgf8^GFP/+;hoxa3^IRESCre/+ embryos. Note that Fgf8 is not expressed in PA mesenchyme (Fig. 2D,F,G; Fig. 3D,E).

In concert, the data in Fig. 3 clearly show that in Fgf8^GFP/null;Hoxa3^IRES-Cre/+ mutants (Fgf8;hoxa3-IRESCre mutants),FGF8 is ablated throughout its expression domains in endoderm and ectoderm of developing PAs 3-6. hoxa3-IRESCre activity is reproducibly present in the endoderm from the 18-19 ss, and from the 16 ss in the ectoderm (data not shown). Note that thymic and parathyroid epithelia arise from the third endodermal pouch, and that `cardiac' neural crest migrates from rhombomeres 6-8 through PAs 3-6 into the OFT (Kirby et al., 1997; Kirby and Waldo,1990; Kirby and Waldo,1995). These are the relevant domains of hoxa3-IRESCreactivity.

The consequence of specifically ablating FGF8 in the PA ectoderm was determined by comparing Fgf8;AP2α-IRESCreectodermal domain mutants (genotype Fgf8^GFP/null or ^AP/null; AP2α^IRES-Cre/+) with controls (genotypes Fgf8^+/+, Fgf8^AP/+,or Fgf8^+/null) and Fgf8 hypomorphs (genotype Fgf8^APN/null) at multiple developmental stages using anatomic and molecular assays (Figs 4, 5, 6, 7).

Fgf8 hypomorphs survive to birth, but 100% die in the neonatal period with a complex phenotype that includes growth delay, cyanosis,craniofacial, cardiovascular and pharyngeal gland defects(Fig. 4B). Hypomorphs have severe cardiac OFT septation and alignment defects. Their vascular abnormalities result from abnormal development of the fourth pharyngeal arch artery (PAA) (Abu-Issa et al.,2002; Frank et al.,2002).

By contrast, only 75% of Fgf8;AP2α-IRESCremutants survive to birth and the rest die postnatally due to lethal vascular defects in 30% and severe craniofacial malformation in 100%. The craniofacial defect results from complete ablation of FGF8 in PA1(Fig. 4C). The pharyngeal phenotype is detected as early as E9.5 as severe PA1 hypoplasia and hypoplasia and fusion of more caudal PAs (data not shown).

The fourth PAAs form between E9.5-10.0 (25-29 ss); the right fourth PAA forms the proximal right subclavian artery, while the left fourth PAA becomes the aortic arch between the left common carotid and left subclavian arteries(the segment of aorta missing in IAAB). Ninety-five percent of Fgf8;AP2α-IRESCre mutants have vascular defects at birth resulting from abnormal formation of the fourth PAAs(Fig. 4G,O-S; Fig. 5; Tables 1 and 2). Thirty percent of Fgf8;AP2α-IRESCre E18.5 conditional mutants have the postnatally lethal vascular malformation, interrupted aortic arch type B (IAAB, Fig. 4G,P,Q). In addition to IAAB and subclavian artery anomalies, we observed circumflex right aortic arch (RAA, Fig. 4O) and RAA with right ductus arteriosus (Fig. 4E,R,S) in these mutants; defects also attributable to failed left PAA4 formation.

To evaluate fourth PAA formation in these mutants, we performed ink injections in E10.5 embryos. Ninety-five percent of Fgf8;AP2α-IRESCre embryos have abnormal fourth PAA formation: 33% display bilateral aplasia of the fourth and sixth PAAs, a phenotype we did not observe in Fgf8 hypomorphs(Frank et al., 2002). The incidence of bilateral aplasia of PAAs 4 and 6 at E10.5 (33%), compared with survival of mutants to birth (75%), reveals that some embryos with this lesion at E10.5 survive. Thirty-three percent of these mutants display bilateral PAA4 aplasia (Fig. 5, Table 2). Recovery of the fourth and/or sixth PAAs by vascular remodeling was noted in some E11.5 mutants (data not shown).

Overall, the incidence of defects attributable to abnormal fourth PAA formation, such as IAAB, RAA or aberrant subclavian artery, is the same in Fgf8 hypomorphs and Fgf8;AP2α-IRESCre mutants. However, the severity of the defects at E10.5 resulting from complete ablation of FGF8 in the PA ectoderm is much greater in Fgf8;AP2α-IRESCre mutants when compared with the globally deficient hypomorphs. This indicates that the pharyngeal FGF8 ectodermal domain has required function(s) during PAA4 vasculogenesis.

We have previously shown that PAA4 vasculogenesis is specifically disrupted in Fgf8 hypomorphs: endothelial cells (ECs) are specified and differentiated in the fourth PA as they express the VEGF receptor, Flk1 and the cell adhesion molecule PECAM (Cleaver and Krieg, 1999). However, the ECs fail to organize into primitive vascular tubes (Frank et al.,2002). To evaluate vasculogenesis in Fgf8;AP2α-IRESCre mutants, we examined whether migration and specification of ECs in the fourth PA proceeds normally. Flk1/PECAM-expressing cells were detected in hypoplastic fourth PAs of Fgf8;AP2α-IRESCre mutants at the 25-27 ss(during PAA4 formation) in clusters that are indistinguishable from controls, Fgf8 hypomorphs, or hoxa3-IRESCre mutants(discussed below, and data not shown). However, in all classes of Fgf8 mutants at the 35-37 ss, ECs in PAA4 remain disorganized and fail to form primitive vascular tubes, long after this vessel is patent and pericyte recruitment is under way in controls (data not shown)(Frank et al., 2002). Thus,migration and early differentiation of PAA4 ECs occurs normally in the absence of FGF8, but subsequent vascular organization fails.

Remarkably, and unlike Fgf8 hypomorphs, Fgf8;AP2α-IRESCre mutants have normal OFT development: that is, alignment, septation and rotation of the aortic and pulmonary arteries are normal. In the mutant shown in Fig. 4G, the aorta and ductus arteriosus/pulmonary artery are normally aligned and septated, but the mutant has the IAAB vascular defect. Fgf8;AP2α-IRESCre mutants also have coronary artery anomalies (Fig. 4M,N),in isolation or associated with other vascular defects(Table 1).

In further contrast to Fgf8 hypomorphs, Fgf8;AP2α-IRESCre mutants have normal thymic and parathyroid glands (Fig. 4H, Table 1, Table 3).

The presence of vascular phenotypes and the absence of OFT or glandular defects in Fgf8;AP2α-IRESCre mutants reveal a distinct functional and anatomic FGF8 signaling domain that is essential for PAA formation and separable from the FGF8 expression domains required for OFT and pharyngeal gland development.

We used the hoxa3-IRESCre driver to ablate FGF8 from both the endoderm and ectoderm of PAs 3-6 as they develop. In marked contrast to both Fgf8 hypomorphs and Fgf8;AP2α-IRESCre mutants, all Fgf8;hoxa3-IRESCre mutants survive to birth and most survive beyond the neonatal period. These mutants have normal craniofacial morphology because Fgf8 is intact anterior to PA3(Fig. 4D). Thirty percent of Fgf8;hoxa3-IRESCre mutants die as neonates because of the same lethal cardiovascular malformations described in Fgf8;AP2α-IRESCre(Table 1 and below).

Notably, Fgf8;hoxa3-IRESCre mutants display thymic ectopy and hypoplasia and parathyroid ectopy, hypoplasia and aplasia(Table 3, Fig. 6A-G). Parathyroid and thymic epithelia are derived from the anterior and posterior third pouch endoderm, respectively. The incidence of abnormal thymic and parathyroid development in Fgf8;hoxa3-IRESCre mutants(Table 3) was comparable with that of Fgf8 hypomorphs (Frank et al., 2002). However, the severity of these defects in hoxa3-IRESCre mutants was less than in globally deficient hypomorphs because no hoxa3-IRESCre mutants had bilateral thymic or parathyroid aplasia, which occurred frequently in Fgf8 hypomorphs. Fourth pouch-derived thyroid C-cells were detected normally in Fgf8;hoxa3-IRESCre mutants, Fgf8 hypomorphs and controls, as assessed by anti-calcitonin immunohistochemistry (data not shown). Notably, glandular defects are not seen in Fgf8;AP2α-IRESCre mutants(Fig. 4G, versus Fig. 6A-G, Table 3).

As FGF8 is also ablated in the ectoderm of PAS 3-6 in Fgf8;hoxa3-IRESCre mutants, it was not surprising to find the same PAA and coronary vascular defects in these mutants seen in Fgf8;AP2α-IRESCre mutants (see above and Table 1). However, it is quite remarkable that ablation of FGF8 in both the endoderm and ectoderm in Fgf8;hoxa3-IRESCre mutants does not increase either incidence or severity of PAA or coronary vascular defects at any developmental stage (Tables 1, 2, 4). The same defect in vascular tube formation in the fourth PA described in Fgf8 hypomorphs and Fgf8;AP2α-IRESCre mutants was also detected in Fgf8;hoxa3-IRESCre mutants. These findings confirm that the ectodermal domain of FGF8 is specifically required for normal PAA and coronary vascular development.

Twenty-three percent of Fgf8;hoxa3-IRESCre mutants(Fig. 6I-K, Table 1) had bicuspid aortic valves (BAV). By contrast, BAV was not seen in Fgf8;AP2α-IRESCre mutants and was found in Fgf8 hypomorphs in association with severe OFT lesions. Given the high frequency of BAV in Fgf8;hoxa3-IRESCre mutants, we were surprised to find that only 1/33 of these mutants had severe perturbation of OFT septation and alignment (Tetralogy of Fallot and BAV, Fig. 6J,M,O).

The glandular and valvar phenotypes of Fgf8;hoxa3-IRESCremutants indicate that the FGF8 endodermal domain ablated by hoxa3-IRESCre has distinct functional roles in pharyngeal and aortic valve development.

We previously reported that neural crest (NC) cells in developing PAs 3-6 undergo abnormal apoptosis at the 25-29 ss in Fgf8 hypomorphs(Frank et al., 2002). As the cardiovascular features of hypomorphs recapitulate those of NC-ablated chicks(Kirby et al., 1985), we hypothesized that the high incidence of severe OFT defects in hypomorphs(85%), resulted from abnormal NC survival in PAs 3-6 prior to entering the OFT(Frank et al., 2002). As we have now demonstrated that differential ablation of FGF8 in PA ectoderm and endoderm separates vascular from glandular and OFT defects, we questioned whether domain-specific mutants would display unique patterns of NC apoptosis. Therefore, we compared apoptosis in all three classes of Fgf8 mutants by whole-mount TUNEL. We detected the same abnormal areas of apoptosis in PA3 and developing PAs 4 and 6 in Fgf8 hypomorphs and domain-specific mutants (Fig. 7A-H, compare white circled regions and white arrowheads in controls with yellow arrowheads in mutants). Note that similar domains of normal (transient, stage specific)apoptosis are present in the otocysts of all embryos, indicating appropriate stage matching.

These observations were confirmed by assaying for apoptosis and expression of the transcription factor Ap2α (labels NC) using double fluorescent immunohistochemistry on serial cryosections of 25 ss embryos(Fig. 7I-L). Minimal NC apoptosis was detected in controls (Fig. 7, row I), whereas all three mutant classes had large abnormal domains of apoptosis in NC migrating from rhombomeres 6-8 into the lateral regions of PA3, and developing PAs 4 and 6 (yellow arrowheads). The hypomorph shown had hypoplastic third PAAs (Fig. 7, row J) and large domains of abnormal apoptosis. Note the abnormally large third PAA in Fgf8;AP2α-IRESCre and Fgf8;hoxa3-IRESCre mutants(Fig. 7, rows K and L,respectively; white asterisks mark third PAAs); consistent with abnormal persistence and enlargement of the third PAA in mutants when the fourth PAAs do not form (see Fig. 5C).

Mice with globally decreased FGF8 signaling throughout embryogenesis(Fgf8 hypomorphs) display severe cardiovascular and pharyngeal defects, and phenocopy the cardiovascular, craniofacial and glandular abnormalities seen in humans with del22q11 syndromes(Abu-Issa et al., 2002; Frank et al., 2002). We previously postulated that Fgf8 may play a role in the developmental dysfunction that results in human del22q11 phenotypes(Frank et al., 2002). This hypothesis has subsequently been supported by a description of a genetic interaction between Fgf8 and Tbx1, a gene located within the del22q11 region (Vitelli et al.,2002).

To determine if local FGF8 signals in the arches contribute to normal pharyngeal and cardiovascular development, and to evaluate which of the many potentially relevant Fgf8 expression domains might be responsible for different aspects of Fgf8 hypomorphic pharyngeal and cardiovascular phenotypes, we generated a system of nonhypomorphic Fgf8 conditional reporter alleles and domain-specific Crerecombinase drivers. Appropriate combinations of these alleles in murine embryos allowed us to differentially ablate FGF8 in distinct temporospatial expression domains in the developing PAs.

AP2α-IRESCre unequivocally ablates FGF8 throughout its expression domains in PA ectoderm from the time of PA formation, at least by the 10-somite stage. By contrast, hoxa3-IRESCre is active in all three germ layers of developing PAs 3-6 from the 16 ss in the ectoderm and the 18-19 ss in the endoderm.

These detailed expression analyses allowed us to determine the precise location and timing of overlap between Fgf8 expression and Cre expressed by the different temporospatially restricted drivers. This is crucial to discerning how ablation of FGF8 in a given domain relates to the distinct phenotypes obtained by domain-specific conditional mutagenesis and in comparison with the complex Fgf8 hypomorphic phenotype.

Fgf8 domain-specific mutants display vascular, coronary artery,aortic valve and pharyngeal gland phenotypes due to loss of specific, local FGF8 signals from separate expression domains in the PA ectoderm and endoderm.

Our Fgf8 allelic series and conditional mutagenesis system reveals a remarkable dosage sensitivity of regional vascular development to FGF8 levels in the caudal PA ectoderm. This domain of expression is required for normal formation of the fourth (and frequently sixth) PAA (Tables 1, 2, Figs 4, 5). Although the overall incidence of PAA4-related vascular defects (IAAB, RAA, subclavian artery anomalies) was the same in Fgf8;AP2α-IRESCre mutants and FGF8-deficient hypomorphs, the severity of PAA defects at E10.5 was much greater in Fgf8;AP2α-IRESCre mutants because of complete absence of FGF8 in the PA ectoderm from the earliest stages of PA formation. hoxa3-IRESCre mutants had (statistically) the same incidence and severity of PAA and coronary vascular defects as Fgf8;AP2α-IRESCre mutants, in spite of the fact that FGF8 was ablated in both the PA endoderm and ectoderm (Tables 1, 2, 4). Ablation of FGF8 with a Tbx1Cre transgene that is active in early endoderm and precardiac mesoderm (but not in arch ectoderm) results in OFT and glandular defects, but no fourth PAA-related vascular defects (J. Epstein, personal communication). These observations indicate that FGF8 signaling from PA ectoderm is critical for normal PAA4 development and that the vascular defects seen in all classes of Fgf8 mutants are attributable to FGF8 deficiency in the PA ectoderm.

By contrast, FGF8 ablation from the endoderm of PAs 3-6 in Fgf8; hoxa3-IRESCre mutants results in thymic, parathyroid and BAV defects. Thus the FGF8 endodermal domain makes important and distinct contributions to development of these structures. Thyroid, parathyroid and thymic epithelial cells are derived from pharyngeal pouch endoderm; these glands initially have a NC component, and interactions between NC and endoderm play an important role in early thymic and parathyroid development(Auerbach, 1960; Bockman and Kirby, 1984; Graham, 2001; Graham and Smith, 2001; LeDouarin and Jotereau, 1975; LeLievre and LeDouarin, 1975). Because FGF8 is ablated from both epithelial layers of the third pouch/cleft of Fgf8;hoxa3-IRESCre mutants, it is possible that glandular hypoplasia and ectopy in these mutants represents a combined effect of absence of FGF8 in both domains. This is unlikely given that the glandular hypoplasia and ectopy resulting from complete ablation of FGF8 in hoxa3-IRESCre mutants is less severe when compared with the frequent glandular aplasia seen in Fgf8 hypomorphs. It is likely that an earlier endodermal FGF8 domain has a crucial influence on the earliest precursors of these glands. Indeed, the aforementioned Fgf8;Tbx1Cre mutants have thymic aplasia (J. Epstein,personal communication).

The same regions of abnormal NC apoptosis are observed in the PAs of all classes of Fgf8 mutants. Importantly, the only abnormal phenotypes common to these different mutants are PAA4-derived defects and coronary vascular anomalies. We hypothesize that signaling between NC-derived ectomesenchyme and endothelial cells (ECs) is required for normal PA vascular development prior to differentiation of NC into the vascular smooth muscle cells (VSMC) and pericytes supporting the PAAs. This signaling may be perturbed by abnormal death of NC migrating from rhombomeres 6-8 into and through PAs 3-6 in Fgf8 mutants. Disruption of EC/ectomesenchymal interactions may contribute to both the PAA and coronary vascular defects seen in all classes of Fgf8 mutants in our series.

Approximately 50% of Fgf8;AP2α-IRESCre and Fgf8;hoxa3-IRESCre mutants have coronary artery (CA)defects. Most commonly, we see a single CA arising from the right cusp of a normal aortic valve. Coronary vessels are derived from cells in the proepicardial organ (PEO, a hepatic-derived structure) that invade the tubular heart and become ECs, VSMCs and pericytes of the coronary vasculature and main CAs (Mikawa and Gourdie,1996). Although NC does not contribute structurally to the CAs(Jiang et al., 2000; Li et al., 2002), CA defects have been described in NC-ablated chicks(Hood and Rosenquist, 1992; Waldo et al., 1994) and in murine mutants of Connexin43 (Li et al.,2002). Cx43 is expressed in both NC and PEO cells; however, the phenotype of Cx43-null mice suggests that the CA defects are due to abnormal migration and/or survival of PEO-derived VSMCs in combination with perturbed interactions between NC and PEO-derived cells in the region of the developing CAs and aortic valve (Li et al.,2002). FGF8 is produced in PA ectoderm adjacent to the PEO;deficiency or ablation of FGF8 in this domain may have similar effects. Further investigation of the entire cardiac vasculature and expression of Cx43 and other intercellular adhesion molecules that participate in CA development,such as α4-integrin or VCAM1 (Kwee et al., 1995; Yang et al.,1996) in Fgf8 mutants is warranted, in addition to examination of survival, proliferation and differentiation of PEO cells. We have not yet determined if Fgf8 or our Cre-drivers are expressed in the proepicardial organ; if so, FGF8 autocrine or paracrine actions could play a role in coronary vascular development.

Bicuspid aortic valve (BAV) is a mild form of OFT defect and is the most common human congenital cardiac malformation. Semilunar (aortic and pulmonary)valve formation requires interactions between endocardial and NC-derived mesenchymal cells and myocardium (Ya et al., 1998). NC progeny are present during formation of the aortic valve leaflets at E13.5 and postnatally in the leaflets and tissues adjacent to the CA orifices (Jiang et al.,2000). NC ablation or dysfunction results in severe OFT defects and mild semilunar valve defects, such as BAV. However, existing mouse models of abnormal semilunar valve development result from disruption of endocardially expressed genes (de la Pompa et al., 1998; Lee et al.,2000; Ranger et al.,1998; Ya et al.,1998). BAV in Fgf8;hoxa3-IRESCre mutants is attributable to FGF8 ablation in the endoderm since Fgf8;AP2α-IRESCre mutants have normal aortic valves. However, the cellular and molecular etiologies of this defect require further investigation in the context of altered FGF8 signaling.

The OFT septum is derived largely from NC, and ablation or dysfunction of NC results in severe OFT defects (Bockman et al., 1989; Conway et al.,1997a; Conway et al.,1997b; Conway et al.,1997c; Epstein et al.,2000; Franz, 1989; Goulding et al., 1993; Jiang et al., 2000; Kirby and Waldo, 1990). Whole-mount TUNEL and analyses of NC apoptosis in the different classes of Fgf8 mutants presented herein indicate that abnormal NC apoptosis at the 25-27 ss does not cause the severe OFT defects observed in Fgf8hypomorphs. Most NC destined for the OFT may have already traversed PAs 3-6 by this somite stage, although OFT septation has not begun. In fact, separation of severe OFT defects (which probably involve some manner of NC dysfunction)from glandular and vascular defects in different classes of Fgf8mutants, implies that FGF8 influences on NC are location and time dependent. This raises the exciting possibility that subsets of NC, with distinct structural fates, migrate through the PAs at different times and are supported by distinct domains of FGF8 expression.

Our findings indicate that the frequent, severe defects in cardiac OFT alignment, septation and growth seen in Fgf8 hypomorphs result from altered FGF8 signaling at an earlier stage in the endoderm, or from abnormalities resulting from decreased FGF8 in the primary and/or putative secondary heart field. Because only one out of 33 Fgf8;hoxa3-IRESCremutants had a severe OFT defect and BAV, we hypothesize that onset of hoxa3-IRESCre activity in the endoderm (at approximately the 18 ss)may be at a relatively late stage of endoderm function during OFT development,and that BAV is a manifestation of late endodermal FGF8 deficiency (similar to our hypothesis regarding the glandular defects in these mutants). Expected`wobble' in this biological system could result in occasional earlier Cre activity in the endoderm and the rare occurrence of more severe OFT defects.

Ablation of FGF8 in early endoderm and cardiac mesoderm (but not PA ectoderm) with a Tbx1Cre transgene results in severe OFT defects, but normal development of fourth PAA-derived vessels (J. Epstein, personal communication). Although this experiment confirms the crucial roles of distinct Fgf8 expression domains in different aspects of pharyngeal and cardiovascular morphogenesis, the question remains whether the crucial source of FGF8 for OFT development is derived from mesoderm or endoderm. Other groups have reported that Tbx1 is not expressed in precardiac mesoderm (Yamagishi et al.,2003), which would suggest that early endoderm is the requisite source. We are developing early endoderm-specific and precardiac mesoderm Cre drivers to definitively address this question.

We thank Ethan Reichert, Tyler Gasser and Jennetta Hammond for technical assistance; Dr Mario Capecchi and the Capecchi laboratory for helpful discussions and support; Dr Gary Schoenwolf, Dr Kirk Thomas, Dr Lisa Urness,Dr Guy Zimmerman and members of the Li laboratory for critical reading of the manuscript; Karl Lustig, Carol Lenz, Gail Peterson, Sheila Barnett and Julie Tomlin and all vivarium staff for their expertise. B.H. and D.F. are supported by the University of Utah Children's Health Research Center. A.M.M. is supported by the Program in Human Molecular Biology and Genetics, and NIH R01HD044157-01.