An Fgf8 mouse mutant phenocopies human 22q11 deletion syndrome

Pharyngeal arches are transient bulges of tissue in the neck of vertebrate embryos consisting of mesenchyme separated by membranous clefts in which the surface ectoderm directly apposes evaginations of the pharyngeal endoderm (the pharyngeal pouches). Pharyngeal arch mesenchyme is derived from neural crest cells and paraxial mesoderm. Neural crest-derived mesenchyme gives rise to the skeletal structures and to pericytes and smooth muscle cells of the arch arteries (Birgbauer et al., 1995; Etchevers et al., 1999). Some neural crest mesenchyme populates the arches only transiently, en route to final destinations in the outflow tract and heart (Epstein et al., 2000; Jiang et al., 2000). The mesoderm-derived mesenchyme contributes muscular and endothelial precursors to each arch. A vessel forms within each arch and circulates blood from the heart to the bilateral dorsal aortae. This pharyngeal arch arterial system is a transient, initially symmetrical array of vessels; the most rostral vessels in the first and second arches largely regress while the caudal pharyngeal arch arteries (PAAs) are extensively remodeled into adult vascular structures, including the subclavian arteries and the arch of the aorta. Aberrant formation and/or remodeling of the pharyngeal arch arterial system results in congenital vascular malformations ranging in severity from clinically insignificant to lethal.

A growing body of evidence indicates that signals from pharyngeal endoderm pattern the arch neural crest and mesoderm (Couly et al., 2002; LeDouarin, 1982; Piotrowski and Nusslein-Volhard, 2000; Thomas et al., 1998; Veitch et al., 1999; Wendling et al., 2000). This tissue expresses signaling molecules such as bone morphogenetic proteins, fibroblast growth factors (FGFs), Sonic hedgehog, and transcription factors including Hox A3, Pax1, Pax9 and Tbx1 (Dietrich and Gruss, 1995; Garg et al., 2001; Manley and Capecchi, 1998; Peters et al., 1998). Pharyngeal pouch endoderm gives rise to the thyroid, parathyroid and thymus glands. Interactions between neural crest and endoderm are important for early thymic and parathyroid development (Bockman and Kirby, 1984; Le Douarin and Jotereau, 1975; LeLievre and LeDouarin, 1975) and later proliferation of the thymic epithelium has been shown to be dependent on FGF signaling (Revest et al., 2001).

Haploinsufficiency of chromosome 22q11(del22q11) is the genetic abnormality shared by the majority of individuals diagnosed with DiGeorge sequence (DiGeorge, 1965), Velocardiofacial syndrome (Shprintzen et al., 1978) and Conotruncal anomaly face syndrome (Burn et al., 1993) (for reviews, see Emanuel et al., 2001; Lindsay, 2001; Scambler, 2000). The craniofacial, cardiovascular and glandular phenotypes that characterize these individuals indicate that all pharyngeal arch-derived structures may be affected. Affected individuals often present in the neonatal period with life-threatening cardiac malformations. Conotruncal defects such as persistent truncus arteriosus and Tetralogy of Fallot are common, as are aortic arch defects including interrupted aortic arch, right aortic arch and vascular ring (Emanuel et al., 2001).

Molecular defects resulting from deletion of genes in del22q11 are hypothesized to adversely affect neural crest function during cardiac and pharyngeal arch development (Emanuel et al., 2001; Robinson, 1975; Van Mierop and Kutsche, 1986). Mechanical ablation of premigratory ‘cardiac’ neural crest from the posterior hindbrain of chick embryos produces many features of del22q11, including interrupted aortic arch type B and truncus arteriosus (Bockman et al., 1989; Kirby and Waldo, 1990). The results of Wnt1 and Pax3 neural crest lineage analyses provide evidence for the existence of a cardiac neural crest population in mammals as well (Epstein et al., 2000; Jiang et al., 2000). Furthermore, mice homozygous for mutations in the neural crest-expressed Pax3 gene [the splotch (Sp2h) mouse] show decreased numbers of migrating cardiac neural crest cells (Conway et al., 1997a; Epstein et al., 2000) and a severe phenotype that includes cardiovascular features of del22q11 (Conway et al., 1997b; Franz, 1989; Goulding et al., 1993).

An enormous effort has been directed towards understanding how haploinsufficiency for the three megabase typically deleted region (TDR, the deletion found in 90% of affected individuals) of chromosome 22 results in the clinically heterogeneous del22q11syndromes. Investigators have attempted to model these syndromes in mice by deleting the region of murine chromosome 16 corresponding to the TDR (Kimber et al., 1999; Lindsay et al., 1999; Puech et al., 2000). Others inactivated murine orthologs of genes within the TDR (Guris et al., 2001; Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001; Saint-Jore et al., 1998). In all cases, heterozygotes were either phenotypically normal or had partial phenotypes with cardiovascular defects that were infrequent and mild, relative to those seen in humans. Twenty five percent of heterozygous mice bearing an engineered deletion on chromosome 16 called Df1 had vascular malformations, predominantly affecting the right subclavian artery (Lindsay et al., 1999). Crko1 heterozygotes (homologous to the CRKL locus in the human TDR) were normal, but homozygotes had vascular, thymic, parathyroid and craniofacial malformations (Guris et al., 2001). Tbx1, another gene within the TDR, encodes a transcription factor in the T-box family expressed in the endoderm and mesoderm of the pharyngeal arches (Garg et al., 2001). Tbx1-null homozygotes display the most severe features of del22q11; 100% of Tbx1-null homozygotes had truncus arteriosus and lacked thymus and parathyroid glands (Jerome and Papaioannou, 2001).

Data from murine models and from humans suggest that multiple loci on chromosome 22 or elsewhere may function in the pathogenesis or act as modifiers of the del22q11 phenotype (Yamagishi et al., 1999; Emanuel et al., 2001; Lindsay, 2001; Scambler, 2000). Sequence and microdeletion analysis of nondeleted syndromic individuals has not revealed causal mutations in Tbx1 or other genes on 22q11 (Chieffo et al., 1997; Wadey et al., 1999; Emanuel et al., 2001). Humans bearing identical deletions have highly variable phenotypes, both with regard to affected structures and presence and severity of cardiovascular disease. Indistinguishable phenotypes have been observed in association with deletions and translocations on chromosome 10p (Gottlieb et al., 1998; Herve et al., 1984) and as the result of teratogenic exposure. Dysfunction of genes that operate in common developmental or genetic pathways may independently generate or modify the ‘del22q11 phenotype’.

FGF8 is a secreted signaling molecule expressed in the developing brain, face and limbs, and in the endoderm and ectoderm of the pharyngeal pouches and grooves (Crossley and Martin, 1995; MacArthur et al., 1995). During embryogenesis, epithelially produced FGF8 provides survival, mitogenic, antidifferentiation and patterning signals to adjacent mesenchyme (Moon and Capecchi, 2000; Lewandoski et al., 2000; Trumpp et al., 1999). Ectodermal FGF8 signaling has been demonstrated to regulate mesenchymal gene expression in the first pharyngeal arch (Tucker et al., 1999a; Tucker et al., 1999b); conditional ablation of FGF8 in first arch ectoderm results in craniofacial defects caused by death of first arch neural crest mesenchyme and perturbed expression of patterning genes (Trumpp et al., 1999). Severe hypomorphism for FGF8 results in gastrulation and left-right asymmetry defects (Meyers et al., 1998; Meyers and Martin, 1999). We employ a hypomorphic allele of Fgf8 to demonstrate that FGF8 performs unique and required roles in development of murine pharyngeal arch and pouch-derived structures including the pharyngeal arch arteries (PAAs), cardiac outflow tract, heart, thymus and parathyroids. The constellation of anomalies displayed by these Fgf8 hypomorphic mutants is a remarkably complete phenocopy of human del22q11 syndrome. These animals provide an excellent model system in which to examine molecular and cellular pathways that result in this common and frequently lethal human syndrome.

Creation of the hypomorphic and null alleles of Fgf8 has previously been described (Moon and Capecchi, 2000). The official names for the Fgf8 null allele (Fgf8^–) is Fgf8^tm1Mrc and the for the Fgf8 hypomorphic allele (Fgf8^H) is Fgf8^tm2Mrc. The alleles were created in an SV129 background and were backcrossed 6 or more generations intoC57Bl6. The Wnt1Cre mice (Jiang et al., 2000) were kindly provided by A. McMahon.

Total RNA was isolated from E10.0 embryos with Trizol reagent (Gibco BRL) according to manufacturer’s instructions. One microgram of total RNA was reverse-transcribed with a first-strand cDNA synthesis kit (Omniscript, Qiagen) and 2 μl of the product was amplified with primers 5′-AAGCTCATTGTGGAGACCGAT-3′ and 5′-AGCTCCCGCTGGATTCCTC-3′ for X=28 and X+1=29 cycles to detect Fgf8 cDNA, or with primers 5′CCAGTATGACTCCACTCACGGCA 3′ and 5′ATACTTGGCAGGTTTCGCCAGGCG 3′ for X=13 and X+1=14 cycles to detect the GAPDH control cDNA. To assure that the results were quantitative, the cycle number for each primer set was optimized for amplification within the linear range.

Pregnant females were anesthetized with a lethal dose of Avertin by intraperitoneal injection and the embryos removed. Thoracotomy was performed and the pericardium opened. Blue Mercox resin and catalyst (Ladd Research Industries #21246 and Benzoyl Peroxide CAS [94-36-0]) were injected into the left ventricle using a pulled glass needle. After allowing the resin to cure for 24 hours, the preparations were placed in Batson’s #17 Maceration solution (Polysciences, Warrington, PA; catalog number, 07359) for 48-72 hours until tissues were completely dissolved. The casts were rinsed repeatedly in distilled water and stored in water.

Embryos were fixed in 4% paraformaldehyde (E9-E11.5) or Bouins fixative (E18.5). Paraffin embedded E9.5 to E11.5 embryos were sectioned at 5 μm and stained with a monoclonal antibody against Pecam-1 (1:50, Pharmingen), to label endothelial cells within the vasculature. A biotinylinated goat anti-rat secondary antibody (1:75, Pharmingen) was used and sections were developed in 3, 3′-diaminobenzidine with nickel chloride chromagen (Vector Laboratories) and counterstained with Eosin. E18.5 embryos were sectioned at 10 μm and stained with Hematoxylin and Eosin.

For cryosections, embryos were fixed in 4% paraformaldehyde and then protected in a sequential series of 10, 20 and 30% sucrose/PBS solutions, oriented in OCT (Tissue Tek) filled molds and rapidly frozen, then sectioned at 10 μm transversely with respect to neural tube and in parallel with the third pharyngeal arch. Sections were washed with PBS, blocked in 2% bovine serum albumin with 0.5% Triton-X100 and incubated overnight with primary antibodies: mouse monoclonal anti-AP2α (1:25, 3B5, Developmental Studies Hybridoma Bank), rabbit polyclonal anti-class III β-tubulin (1:500,Covance, α-Tuj1) or rabbit polyclonal antiphosphohistone H3 (1:100, Upstate Biotechnology, α-PHH3). Sections were washed, blocked and incubated with FITC conjugated anti-mouse IgG antibody (1:500, Molecular Probes), Cy-5 conjugated anti-rabbit IgG antibody (1:500, Jackson Laboratories), and TMR Red in situ cell death detection reagents (Roche). Sections were preserved in Vectashield anti-fading reagent (Vector Laboratories) and observed with confocal analysis at 20× magnification.

In situ hybridization on intact embryos was carried out as previously described (Moon et al., 2000). The plasmids used to generate riboprobes were generously provided by N. Miura (mfh-1), B. Morrow (tbx-1), D. Srivastava (hrt-1) and D. Ornitz (Fgf10).

Thymocytes from embryos at 18.5 days of gestation were dispersed through nylon mesh into PBS, washed, counted on a hemocytometer using Trypan Blue to exclude non-viable cells, stained for cell surface markers (PE-anti-CD25, PerCP-anti-CD8, APC-anti-CD4, FITC-anti-CD44 from PharMingen), washed, resuspended in PBS and analyzed by flow cytometry on a FACSCaliber® (Becton Dickinson).

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

Somite matched E9.5 embryos were fixed and subjected to Terminal UTP nick end labeling (TUNEL) as described elsewhere (Stadler et al., 2001). Confocal analysis was performed as above. Photomicrographs were taken at 8× magnification.

FITC, CY5, and Texas red fluorescence were recorded using a BioRad MRC 1024 laser-scanning confocal imaging system fitted to a Leitz Aristoplan microscope using filters supplied by the manufacturer. A digital Kalman averaging filter was used to reduce background fluorescence.

In order to bypass embryonic lethality resulting from complete ablation of Fgf8 function in embryonic mice, we investigated the role of Fgf8 in cardiovascular and pharyngeal arch development with a hypomorphic allele. The Fgf8^H allele contains a neomycin phosphotransferase (neo^r) and alkaline phosphatase reporter genes in the 3′ untranslated region of an otherwise intact Fgf8 locus (Fig. 1A). This has significant transcriptional consequences. Quantitative RT-PCR amplification (Fig. 1B), reveals that compound heterozygotes bearing hypomorphic (Fgf8^H) and null (Fgf8^–) alleles of Fgf8 (Fgf8^H/–, henceforth called mutants), reproducibly generate less Fgf8 mRNA, when compared with control littermates. The decreased function of the hypomorphic relative to the wild-type allele clear if one compares the compound heterozygotes (in which the only source of Fgf8 mRNA is the Fgf8^H allele) and null heterozygotes (in which only the Fgf8⁺ allele produces Fgf8 mRNA). We have previously shown that removal of the neo^r gene from the Fgf8^H allele restores its function (Moon and Capecchi, 2000).

Decreased production of Fgf8 mRNA has important phenotypic consequences. Fgf8^H/– mutants are small and die as neonates with cyanosis, respiratory distress, edema and multiple congenital anomalies (Fig. 1C). Craniofacial abnormalities are present in all of these animals: 100% have micrognathia and 30% have bony cleft palate (Fig. 1D); other facial malformations such as fusion of the temporomandibular joint and absence of the tympanic ring are more variable within and between mutants. Central nervous system malformations are seen in all mutants including hypoplasia or aplasia of the cerebellum and olfactory bulbs (data not shown). These phenotypes reflect dose sensitivity for FGF8 because Fgf8^+/– null allele heterozygotes (which produce approximately 50% the wild type amount of Fgf8 mRNA) and Fgf8^+/H hypomorph allele heterozygotes (which produce a quantity intermediate to that of wild type and null heterozygotes) are normal.

Multiple, severe developmental defects contribute to the neonatal demise of some mutants; however, most die a cardiovascular death, with progressive cyanosis and respiratory failure. We discovered cardiovascular defects in 95% of 60 term mutants, most of which were lethal malformations (Fig. 2, summarized in Table 1).

Outflow tract septation, rotation and alignment defects were common in these mutants. Persistent truncus arteriosus was the most embryologically primitive lesion observed and results from complete failure of septation of the primitive truncal vessel to form the ascending aorta and main pulmonary artery. Instead, the branch pulmonary arteries, the great vessels of the head and neck, and the descending aorta all arise from a single vessel (Fig. 2B-D). Transposition of the great arteries is a defect in which the aorta arises abnormally from the right ventricle and the pulmonary artery from the left ventricle (inverted ventriculo-arterial alignment, Fig. 2I-L). After birth and upon the closure of the ductus arteriosus, two parallel circulations form such that the systemic circulation is deprived of oxygen and death rapidly ensues. Double outlet right ventricle (Fig. 2F-H) and Tetralogy of Fallot (overriding aorta, ventricular septal defect, pulmonary stenosis and hypertrophy of the right ventricle) are more complex rotation/alignment defects that were observed.

Vascular defects were observed in the majority of mutants (see Table 1) and included interrupted aortic arch type B, in which the arch of the aorta fails to form between the left common carotid artery and left subclavian artery (IAAB, Fig. 2O); right aortic arch (Fig. 2P); and anomalous origin of the subclavian arteries. These defects indicate that formation, stabilization or remodeling of the fourth PAAs is disrupted in the mutants. The right and left fourth PAAs form between E 9.5-E10.0. The right fourth PAA gives rise to the origin of the right subclavian artery. The left fourth PAA becomes the arch of the aorta between the left common carotid artery and left subclavian artery which is the portion of the aorta that is absent in IAAB.

To evaluate fourth PAA formation, we performed intravascular ink injections in E9.5 to E 11.5 mutants and controls. At E9.5, prior to the formation of the fourth PAA, mutant animals have patent PAAs 1-3, as do controls (Fig. 3A,B). The fourth PAA is not patent to ink injection in E10.0 to E11.5 mutants (Fig. 3C-F). All mutants display abnormal fourth PAA formation bilaterally at E10.5 (Table 2). The absence of these arteries can not be attributed to growth delay, as the sixth PAA is present by E10-10.5 (Fig. 3D,F,H,J). A mutant sectioned coronally through the pharyngeal arches and immunohistochemically stained with the endothelial marker PECAM demonstrates that endothelial cells can be detected in the left fourth arch region, although they are disorganized and no vessel lumen is evident (Fig. 3H, left side). On the right-hand side of this animal, the right fourth vessel appears patent in the section displayed, but serial sections revealed that it never connected with the dorsal aorta.

At E10.5, mutants frequently had hypoplastic fourth pharyngeal arch tissue fused to the third arch and no third pouch (Fig. 3H,J). In combination with the vascular defects reported above, this finding suggests that formation of the fourth arch itself and patterning of the third and fourth arches may be disrupted. To determine if decreased FGF8 in the mutants altered expression of transcription factors expressed during fourth arch development, we performed whole mount in situ hybridization in E9.5 and E10.5 embryos. Mfh1 (Foxc2 – Mouse Genome Informatics) encodes a winged helix transcription factor that has been shown to be expressed in the pharyngeal arch vessels and mesenchyme from E9.5; ablation of Mfh1 function results in interrupted aortic arch type B due to aberrant remodeling of the fourth PAA (Iida et al., 1997). Mfh1 expression appeared normal in Fgf8 mutants (Fig. 4A) Expression of Hrt1, which encodes a basic helix-loop-helix transcription factor expressed in the developing arch mesenchyme and endoderm and is a downstream mediator of Notch signaling (Nakagawa et al., 2000; Nakagawa et al., 1999), also appeared normal in Fgf8 mutants (Fig. 4B). Although Fgf8^H/– and Tbx1 mutants have overlapping phenotypes and are both genes are expressed in the pouch endoderm, we did not detect any change in the pattern of Tbx1 expression at these stages (Fig. 4C).

We have previously shown that ectodermal FGF8 maintains mesenchymal Fgf10 expression in the developing limb bud and questioned whether a similar relationship occurred between FGF8 and Fgf10 in the pharyngeal arch epithelia and mesenchyme (Moon and Capecchi, 2000). Examination of Fgf10 expression at E9.5 and E10.5 (Fig. 4D,E) revealed markedly decreased signal in the pharyngeal arch mesenchyme, including the regions of the developing fourth arch and putative secondary heart field (Kelly et al., 2001).

The thymus and parathyroid glands develop from the third and fourth pharyngeal pouch endoderm and migrate caudally to their final location in the mediastinum. We analyzed mutants by gross dissection and in sections, and found that 50% of mutants had no detectable thymus, while in another 50% the thymus was hypoplastic with small single or double lobes or strands of thymic tissue ectopically located in the rostral region of the neck (Fig. 5A-C). Rarely, ectopic thymic tissue could be found penetrating into the oropharynx of E18.5 mutants (data not shown). The parathyroid glands were aplastic or hypoplastic in 50% of mutants (Fig. 5D-F).

One of the cardinal features of individuals with deletion 22q11 and DiGeorge sequence is a variable immunodeficiency associated with decreased number of T cells (Jawad et al., 2001). Though 50% of the Fgf8 mutants had aplasia of the thymus, we were able to evaluate T cells in those with hypoplastic glands. Fig. 5G shows representative FACS (fluorescence activated cell sort) plots comparing T cells obtained from control and mutant thymi immunostained with anti CD4 and anti CD8 antibodies. In Fig. 5H, the percent (left panel) and total (right panel) number of CD4^–CD8^–double negative (DN) and CD4⁺CD8⁺double positive (DP) thymocytes are compared in E18.5 controls and mutants; mutants had 20% of the normal number of T cells. The thymocyte subpopulations were normal and no difference was detected in CD44 or CD25 staining of double negative-gated thymocytes (data not shown).

The Fgf8^H/– mutant phenotypes are striking in their similarity to human del22q11 syndromes and animal models of neural crest dysfunction. Although Fgf8 expression has not been reported in neural crest in the mouse, it is possible that these defects arise from a deficiency of FGF8 in a previously undetected expression domain in neural crest. To determine if Fgf8 is expressed in premigratory or migratory neural crest, we combined the Wnt1-Cre transgene (Jiang et al., 2000) with the alkaline phosphatase (AP) reporter/conditional features of the Fgf8^H allele (Moon and Capecchi, 2000) (Fig. 1A). LoxP sites flank exon 5 and neo^r, and enable conditional expression of the AP reporter upon Cre-mediated recombination of the Fgf8^H allele. In this experiment, cells of the Wnt1 lineage, including premigratory cardiac neural crest (Echelard et al., 1994), contain the recombined Fgf8^H allele; thus, blue AP staining indicates cells that currently or previously express(ed) Wnt1, and that currently express Fgf8 (Fig. 6). We detected overlapping domains of Wnt1 and Fgf8 expression in the isthmus and frontonasal prominences at E9.5 (Crossley and Martin, 1995; Echelard et al., 1994), but none in the domain of premigratory or migratory cardiac neural crest, which originates in rhombomeres six through eight, at either E8.5 or E9.5. These results suggest that the cardiovascular and glandular phenotypes of Fgf8 mutants result from altered signaling between the pharyngeal epithelium, the mesodermal mesenchyme, and neural crest en route to the arches, outflow tract and heart.

We questioned whether neural crest survival and migration were affected by decreased FGF8 in the pharyngeal ectoderm and endoderm of the pharyngeal arches. Using whole-mount TUNEL analysis, we found an increased number of apoptotic cells and abnormal domains of apoptosis in the pharyngeal arches of mutants when compared with controls (Fig. 7A-C). The expression of Fgf8 is also shown to help the reader visualize how the apoptotic domains relate spatially to regions of Fgf8 expression (Fig. 7D-F). Note that there are domains of apoptosis adjacent and relatively distant to locations of normal FGF8 signaling.

To determine specifically if abnormal apoptosis was occurring in neural crest, we performed fluorescent immunohistochemistry using anti-AP-2α antibody to label migrating neural crest in cryosectioned embryos (Mitchell et al., 1991). Domains of apoptosis in the neural tube and developing ganglia were present in mutants and controls at the 25 somite stage (Fig. 8A,B, blue arrowheads), but the domains were much larger in the mutants (Fig. 8, yellow arrowheads). In the mutants, there were additional abnormal domains of apoptosis and many TUNEL/antiAP-2α double-labeled cells (yellow signal, yellow arrowheads) in neural crest migrating from rhombomeres 6 and 7 into the lateral developing fourth and sixth arches. This finding was reproducible and persisted to the 29 somite stage (Fig. 8C,D). No reproducible differences in proliferation were detected at this stage, as assayed with an antiphosphohistone H3 antibody to label chromatin in proliferating cells. Note that in whole-mount TUNEL and cryosectioned specimens, many of the apoptotic cells are distant from the sources of FGF8 in the pharyngeal pouches and clefts shown in Fig. 7 (Crossley and Martin, 1995; Heikinheimo et al., 1994; MacArthur et al., 1995).

We have described an Fgf8 hypomorphic mouse that displays a range of pharyngeal and cardiovascular phenotypes that closely phenocopies human del22q11 syndrome. We postulate that these developmental defects result from an insufficient dosage of FGF8 locally in the pharyngeal arches. The fact that both Fgf8^+/H and Fgf8^+/– heterozygotes are normal indicates that a crucial decrease in the dose of FGF8 occurs when the hypomorphic and null alleles are combined in a single embryo. If the phenotypes were due to an effect of the neo^r gene on other loci, Fgf8^+/H heterozygotes would be affected. Furthermore, ablation of Fgf8 function in the pharyngeal arch ectoderm using a tissue-specific Cre recombinase and a nonhypomorphic conditional allele reproduces many features of the phenotype (Arenkiel and Moon, unpublished data) and reveals that the roles of FGF8 in left-right specification (Meyers and Martin, 1999) and in pharyngeal and cardiovascular development are separable. Meyers and colleagues have previously described aberrant Fgf8 splicing into a neo^r gene targeted to an intron that resulted in a hypomorphic allele (Meyers et al., 1998); this hypomorphic affect appears to be significantly more severe than that resulting from placement of neo^r in the 3′UTR that we describe. They described a range of defects in their compound heterozygotes including early embryonic death, gastrulation failure, severe CNS malformations and left-right asymmetry defects, which indicates earlier and more severe FGF8 deficiency in these animals (Meyers et al., 1998; Meyers and Martin, 1999). Our Fgf8^H/– mice have focal developmental defects from reduced FGF8 dose, indicating that this hypomorphic allele produces sufficient functional Fgf8 message for the embryo to gastrulate and specify the left-right axis correctly. The work of both groups reveals that different tissues have different dosage requirements for FGF8 likely reflecting redundancy other FGFs (Moon et al., 2000; Moon and Capecchi, 2000; Lewandoski et al., 2000; Sun et al., 2000).

Deficiency of FGF8 causes cardiovascular, craniofacial and glandular phenotypes resulting from abnormal pharyngeal arch development, altered mesenchymal expression of Fgf10 and decreased neural crest survival. Similar to our observation in the developing limb (Moon and Capecchi, 2000), we found that epithelial FGF8 is required for expression of Fgf10 in the pharyngeal arch mesenchyme, including the region of the developing fourth pharyngeal arch and the putative secondary heart field (Kelly et al., 2001). Further experiments will be required to determine if aspects of the outflow tract defects seen in Fgf8 hypomorphs are related to this finding. Tissue-specific ablation of FGF8 in the pharyngeal epithelia may clarify its role in this region.

Although cardiac neural crest does not itself express Fgf8, Fgf8 mutants display increased apoptosis in this tissue, indicating that FGF8 signaling from pouch endoderm and ectoderm is important for its survival. FGF receptors are expressed in the ectoderm and throughout the mesenchyme of the pharyngeal arches (MacArthur et al., 1995; Orr-Urtreger et al., 1993; Orr-Urtreger et al., 1991; Peters et al., 1992). Our whole-mount TUNEL and analysis of apoptosis in sections reveals that some of the apoptosis is (not surprisingly) directly adjacent to sites of epithelial Fgf8 expression (Crossley and Martin, 1995; Heikinheimo et al., 1994; MacArthur et al., 1995) (Fig. 7) and probably reflects the failure of FGF8/FGFR1- or FGF8/FGFR2-mediated signaling. However, other large domains of mesenchymal apoptosis are relatively distant from these sources. This finding suggests that altered FGF8 signaling is sensed or relayed distant to its expression domains; such relays may, or may not, be based on FGF/FGFR interactions. We have reported the same phenomenon in proximal limb mesenchyme after conditional ablation of FGF8 in the distal limb ectoderm (Moon and Capecchi, 2000) and reproduced this finding after ablation of both FGF4 and FGF8 in the limb ectoderm (A. M. Boulet, M. R. C. and A. M. M., unpublished). Ablation of FGF8 in first arch ectoderm results in abnormal apoptosis of neural crest-derived mesenchyme (much of which is distant to the epithelial FGF8 source (Trumpp et al., 1999), similar to that described here.

The phenotypes of Fgf8 and Tbx1 mutants confirm that pharyngeal endoderm plays a crucial role in patterning the pharyngeal neural crest and mesoderm-derived mesenchyme (Couly et al., 2002; LeDouarin, 1982; Piotrowski and Nusslein-Volhard, 2000; Thomas et al., 1998; Veitch et al., 1999; Wendling et al., 2000). The importance of cardiac neural crest in the pathogenesis of the human del22q11 syndromes is supported by neural crest ablations in chick (Creazzo et al., 1998) and the splotch mouse (Conway et al., 1997b; Epstein et al., 2000). Several mouse models with neural crest dysfunction exhibit conotruncal and aortic arch defects (Bamforth et al., 2001; Feiner et al., 2001; Mendelsohn et al., 1994; Schorle et al., 1996; Zhang et al., 1996). Our results concur with accumulating evidence that neural crest cells rely on signals from adjacent tissues for patterning and survival en route to their targets (Trainor and Krumlauf, 2000) and indicate an important role for epithelial-mesenchymal interactions in neural crest function during development.

Aortic arch abnormalities arise from defects in formation or remodeling of the PAAs, although models vary considerably in pathogenesis. In neural crest-ablated chicks, PAA 3, PAA 4 and PAA 6 are formed but rapidly regress (Kirby, 1999). This indicates that interactions between neural crest, the primitive vessel and possibly the mesodermal mesenchyme are required for stabilization and subsequent remodeling of the vessel. In Df1 heterozygotes or Tbx1 haploinsufficient mice, the fourth PAA is specifically affected. These mutants initially form an endothelialized tube that is not stabilized and by E10.5, 100% of animals have a nonpatent vessel (Lindsay and Baldini, 2001; Lindsay et al., 1999; Lindsay et al., 2001). Recovery from this defect is variable resulting in a 30% incidence of primarily right-sided cardiovascular defects. Homozygous mutants of Crkol, Mfh1 or genes in the endothelin signaling pathway all demonstrate fourth PAA defects in which arteries form normally, but regress to a variable extent after E11.5 (Guris et al., 2001; Iida et al., 1997; Kurihara et al., 1995; Yanagisawa et al., 1998). Our Fgf8 mutants exhibit early defects in vessel formation and stabilization similar to the Tbx1 and Df1 mutants described by Lindsay and colleagues (Lindsay and Baldini, 2001; Lindsay et al., 1999; Lindsay et al., 2001). Endothelial cells are aggregated in the fourth arch at E10.5, but patent vessel lumens are not evident. However, we see a greater proportion of left sided and bilateral defects at E10.5, and no evidence of recovery by E11.5. This results in a greater incidence of fourth PAA defects in the newborns, particularly the more severe left-sided phenotypes of IAAB and right aortic arch.

The Fgf8 hypomorphic phenotype suggests that Fgf8 may contribute to the variability of human del22q11 phenotypes as a modifier, or may be a causative locus in syndromic individuals that do not have deletions of 22q11. Complete inactivation of one allele of Fgf8 in conjunction with a decrease in function of the remaining allele in mice provides a spectrum of disease closely resembling that seen in humans. FGF8 may participate directly in molecular pathways affected by deletions in the 22q11 region or function in parallel pathways required for normal development of pharyngeal arch and neural crest-derived tissues.

Given the lack of genotype-phenotype correlation in humans with the 22q11 deletion, it is possible that dysfunction of other developmentally active loci is required to produce the most severe del22q11 phenotypes. To test this hypothesis in mice, we are searching for genetic interactions by evaluating mice bearing combinations of mutant loci known to generate aspects of this phenotype. Screening individuals with del22q11 for mutations on the nondeleted chromosome 22 or for mutations in genes known to contribute to pharyngeal arch development (Fgf8, Pax3, for example) will provide another important test of this hypothesis.

We thank the Capecchi laboratory, Gary Schoenwolf and Michael Bamshad for helpful discussions and critical reading of the manuscript; we also appreciate the tissue culture and vivarium staff for their technical expertise and effort. We appreciate Mary Scriven’s assistance with the figures. D. F. is a PCMC Foundation Fellow and is 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 K08HD1373.