A genetic link between Tbx1 and fibroblast growth factor signaling

Tbx1, a T-box putative transcription factor, is required for the segmentation of the pharyngeal apparatus, and homozygous mutation causes multiple developmental defects, including cardiovascular, craniofacial, ear, thymic and parathyroid defects (Jerome and Papaioannou, 2001; Lindsay et al., 2001; Vitelli et al., 2002). Tbx1 expression has been extensively analyzed by RNA in situ hybridization (Chapman et al., 1996; Garg et al., 2001; Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001) and by a lacZ reporter knocked into the endogenous locus (Vitelli et al., 2002). The expression pattern is dynamic during development and is mainly localized in the head mesenchyme, pharyngeal endoderm, core mesenchyme of the cranial pharyngeal arches, outflow tract of the heart, mesenchyme surrounding the dorsal aortae, otocyst and sclerotome. Tbx1 haploinsufficiency causes aortic arch patterning defects in mice (Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001) of the same type as those observed in a mouse model, named Df1/+ of the chromosomal deletion occurring in individuals with DiGeorge syndrome (Lindsay et al., 1999). The mouse deletion Df1 and the human deletion del22q11 both include Tbx1, making this gene the most likely candidate for a pathogenetic role in this syndrome.

Tbx1 haploinsufficiency causes early growth and remodeling defects of the fourth pharyngeal arch arteries (PAAs), first evident at embryonic day (E) 10.5. The caudal PAAs (third, fourth and sixth) form sequentially as symmetric vessels connecting the aortic sac with the dorsal aortae. From ∼E11.5, the PAAs undergo a major, asymmetric remodeling that leads to the mature aortic arch and great vessel patterning (Srivastava and Olson, 2000). In particular, the left fourth PAA contributes to the section of the mature aortic arch between the origins of the left common carotid artery and the left subclavian artery. Developmental failure of the left fourth PAA causes interruption of the aortic arch type B (IAA-B). The right fourth PAA provides the connection of the right subclavian artery with the innominate artery. Developmental failure of the right fourth PAA causes aberrant origin of the right subclavian artery, most commonly from the descending aorta, via a retroesophageal vessel.

lacZ-knock-in experiments have shown that Tbx1 is not expressed in the structural components of the fourth PAA but in the surrounding pharyngeal endoderm, suggesting that the role of Tbx1 in the growth of this artery is cell non-autonomous (Vitelli et al., 2002). The fibroblast growth factor (FGF) signaling has been shown to interact with the function of other T-box genes, and the presence of regulatory loops between T-box transcription factors and Fgf genes has been suggested (Casey et al., 1998; Ohuchi et al., 1998; Rodriguez-Esteban et al., 1999; Takeuchi et al., 1999). Therefore, we tested whether FGF signaling could mediate Tbx1 functions especially as it relates to the pathogenesis of cardiovascular defects, the major cause of morbidity and mortality in individuals with DiGeorge syndrome. Our results identify Fgf8 as the first known genetic interactor of Tbx1 and suggest that other Fgf family members may mediate the role of Tbx1 in the development of other derivatives of the pharyngeal apparatus.

The generation of Tbx1 mice carrying the allele Tbx1^tm1Bld (referred to as Tbx1^–) has already been described (Lindsay et al., 2001). Mutants were maintained on a C57BL/6×129SvEvBrd(129S5) mixed genetic background. Fgf8^+/– animals (carrying the null allele Fgf8^Δ2,3 (Meyers et al., 1998) originally maintained in a ICR outbred background, were back-crossed twice in a C57BL/6×129SvEvBrd(129S5) and bred with Tbx1^+/– mutants. Embryos were collected at embryonic day (E) 18.5 or E10.5, considering E0.5 as the day on which the vaginal plug was observed. Mice or embryos were genotyped by PCR of DNA extracted from tail biopsies or yolk sacs, using previously published PCR primer pairs (Lindsay et al., 2001; Meyers et al., 1998). Phenotype scoring was performed before genotypic analysis.

Radioactive or non-radioactive in situ hybridization experiments were performed on sectioned or whole-mount embryos, respectively, using a published protocol (Albrecht et al., 1997). Sense and antisense riboprobes were prepared by reverse transcription of DNA probes and labeled by incorporation of digoxigenin-conjugated UTP (Roche) or ³⁵S-UTP (ICN).

Embryos were fixed in paraformaldehyde and then processed for X-gal staining, according to standard procedures. Embryos were photographed as wholemounts and then embedded in paraffin wax and cut in 10 μm histological sections. Sections were counterstained with Nuclear Fast Red.

India ink was injected intracardially in E10.5 embryos using pulled glass needles. To avoid scoring growth delayed embryos, we only considered embryos in which the sixth PAAs could be clearly visualized by ink injection. A total of 8 litters was analyzed, in which 45 embryos were scorable. Three fourth PAA phenotypes were scored: normal (similar or larger size than the third PAA), small (considerably smaller than the third PAA) and non-patent to ink, according to previously published criteria (Lindsay and Baldini, 2001). Embryos showing fourth PAA non-patent to ink were embedded in paraffin wax for histological examination.

Cell death was detected on whole mount embryos using Lysotracker (Molecular Probes) essentially as described (Zucker et al., 1999). Embryos were then embedded in paraffin and 10 μm histological sections were examined under a fluorescence microscope.

The Tbx1 haploinsufficiency phenotype is characterized by hypoplasia of the fourth PAAs. The abnormalities become apparent from ∼E10, soon after the formation of the arteries (Lindsay et al., 2001). Tbx1 is mainly expressed in the pharyngeal endoderm surrounding the arteries but not in their structural components, suggesting a cell non-autonomous role in the early growth and remodeling of the vessels (Vitelli et al., 2002). We hypothesized that Tbx1 triggers a molecular signal in the pharyngeal endoderm directed to the underlying mesenchyme to support vessel growth. FGF signaling has been shown to interact with other T-box genes and therefore it is possible that FGF molecules may mediate the role of Tbx1 in fourth PAA growth. The Fgf8 gene expression pattern overlaps with that of Tbx1 in the pharyngeal endoderm around the time when the fourth PAA hypoplasia becomes apparent (Fig. 1A,B). To determine whether the expression of Fgf8 may be altered in the endoderm of Tbx1 mutants, we performed in situ hybridization of Tbx1^+/– embryos at E10 (Fig. 1E-F) and E11.5 (not shown), but no alteration could be detected. However, in situ hybridization is not quantitative and subtle variations in gene expression may not be apparent. Therefore, we tested Fgf8 expression in Tbx1^–/– embryos. Results show that in these embryos, the Fgf8 expression in the pharyngeal endoderm is lost, while the other expression domains are maintained (Fig. 1C,D,G,H). Loss of Fgf8 in the pharyngeal endoderm may be due to gene downregulation or lack of development/survival of endodermal cells expressing Fgf8. Indeed, Tbx1^–/– mutants have severely hypoplastic pharynx and lack pharyngeal pouches (Jerome and Papaioannou, 2001; Lindsay et al., 2001; Vitelli et al., 2002). To test whether the residual pharyngeal endoderm expresses endodermal markers, we hybridized in situ Shh, Nkx2-5 and Pax9 on tissue sections. All these markers are robustly expressed in the endoderm of homozygous mutants (not shown). In addition, using our Tbx1-lacZ-knock-in allele, we could demonstrate β-gal activity in the endoderm of Tbx1^–/– embryos, indicating that Tbx1 function is not required for the contribution of Tbx1-expressing cells to the endoderm (Vitelli et al., 2002). Overall, these results suggest that the endoderm of Tbx1^–/– embryos is properly specified.

In addition to pharyngeal endoderm, Tbx1 is expressed in the core mesenchyme of pharyngeal arches 1, 2 and 3 (Fig. 2A,C), which are considered to be of paraxial mesoderm derivation. Interestingly, Fgf10 is also expressed in the core mesenchyme of the arches with a similar pattern (Fig. 2B,D), raising the question of whether Tbx1 may be required for Fgf10 expression in this tissue. Indeed, Fgf10 expression is abolished in the core mesenchyme of arches 1 and 2 of Tbx1^–/– mutants and, in the first arch, it has a low-level, diffuse expression in the arch mesenchyme (Fig. 2E,F). Pharyngeal arch 3 does not form in these mutants, while pharyngeal arch 2 is hypoplastic and arch 1 is of apparently normal size but is slightly mis-shapen. However, arches 1 and 2 express normal levels (though with size reduction in the second arch) of mesenchymal markers such as Dlx2 (Vitelli et al., 2002), Foxd1 and Bmp4 (E. A. L. and A. B., unpublished). Core mesenchyme cells that normally express Tbx1 are also present and normally localized in the arches of Tbx1^–/– embryos, as shown by β-Gal staining (Vitelli et al., 2002). These results are consistent with at least two interpretations: one is that Fgf10 expression is downregulated in the core mesenchyme, the other is that Fgf10-expressing cells are misplaced and diffused in the arch mesenchyme. Fgf10 expression is normal in other tissues of Tbx1^–/– embryos, with one notable exception. Fgf10 is expressed in a region of splanchnic mesoderm located dorsocaudally to the junction of the outflow tract with the pharyngeal apparatus, in the dorsal wall of the pericardial cavity, Tbx1-expressing cells are also present in this region (Fig. 3A-C). Fgf10-expressing cells from this region are thought to migrate and contribute to the muscle wall of the outflow tract, as they are part of the anterior or secondary heart field (Kelly et al., 2001; Mjaatvedt et al., 2001; Waldo et al., 2001). This Fgf10 expression domain could not be detected in Tbx1^–/– embryos (Fig. 3D), while expression of Nkx2-5, also present in the secondary heart field (Waldo et al., 2001), was reduced but conserved (not shown). These results suggest that the Tbx1 function in outflow tract development, revealed by severe outflow tract defects in Tbx1^–/– mutants, may be mediated by the FGF signaling in the secondary heart field.

The gene expression data shown above suggest that Fgf8 may interact with Tbx1 in the pharyngeal endoderm and be a mediator of the cell non-autonomous role of Tbx1 in fourth pharyngeal arch artery development. If this is the case, dosage reduction of Fgf8 should enhance the Tbx1 haploinsufficiency phenotype. To test this hypothesis, we bred Tbx1^+/– mice with Fgf8^+/– mice (Meyers et al., 1998) and scored the phenotype of the progeny at E18.5 in order to understand whether combined heterozygosity affects the mature aortic arch anatomy. A total of 137 embryos were analyzed by dissection of the heart and great vessels and all genotypes were recovered according to mendelian ratio. Tbx1^+/+;Fgf8^+/+ embryos (n=30) and Tbx1^+/+;Fgf8^+/– embryos (n=30) were normal. Of the 35 Tbx1^+/–;Fgf8^+/+ embryos analyzed, 11 (27%) presented with aortic arch patterning defects of the types previously described in this mutant and with a similar penetrance previously reported for Df1/+ animals (Lindsay and Baldini, 2001), which carry a multi-gene heterozygous deletion that includes Tbx1. Tbx1^+/–;Fgf8^+/– animals presented with high penetrance of arch defects (50%, n=36), significantly higher than that found in Tbx1^+/–;Fgf8^+/+ mutants (P=0.036). Interestingly, the type of defects observed was not affected by Fgf8 mutation (Table 1), all defects were attributable to a failure of the fourth PAA development, only one Tbx1^+/–;Fgf8^+/+ and one Tbx1^+/–;Fgf8^+/– embryo presented with a ventricular septal defect associated with IAA-B. These were the only intracardiac defects observed in any genotype. The most common abnormality observed in compound heterozygous mutants is an interruption of the aortic arch (type B as it occurs between the left common carotid artery and the left subclavian artery), associated with a retroesophageal connection between the ascending aorta (on the right side of the body axis) and the descending aorta on the left side of the body axis (Fig. 4). Because of this connection, the defect can be described as a right aortic arch (RAA). We interpret the connecting vessel as a compensatory persistence of the embryonic right dorsal aorta, which normally regresses.

To study the early fourth PAA phenotype, we injected ink into the heart of E10.5 embryos obtained from the same breeding scheme indicated above, a total of 56 embryos were scored (Table 2). As previously reported for Df1/+ embryos (Lindsay and Baldini, 2001), the penetrance of fourth PAA abnormalities caused by Tbx1 haploinsufficiency at E10.5 is much higher than the penetrance of aortic arch defects at E18.5. Out of 14 Tbx1^+/–;Fgf8^+/+ embryos examined, 13 were abnormal; out of 14 Tbx1^+/–;Fgf8^+/– embryos, 13 were abnormal. However, the latter were more severely affected because all but one had bilateral defects, and most had ‘absent’ fourth PAA phenotype, i.e. were non-patent to ink (Table 2). Indeed, the number of arteries scored as non-patent were significantly higher in Tbx1^+/–;Fgf8^+/– than in Tbx1^+/–;Fgf8^+/+ embryos (21 out of 28 and 7 out of 28, respectively, P=0.0002). Histological examination of embryos with fourth PAAs non-patent to ink showed that the arteries were small but present; hence, excluding a defect of formation (not shown). The third and sixth PAAs were normal in all the genotypes. These data indicate that Tbx1 and Fgf8 interact genetically during the early phases of fourth PAA development, consistent with the time of Fgf8 and Tbx1 expression overlap in the pharyngeal endoderm. We also examined the progeny of crosses between Tbx1^+/–;Fgf8^+/– and Tbx1^+/–;Fgf8^+/+ animals. The cardiovascular phenotype of Tbx1^–/–;Fgf8^+/+ and Tbx1^–/–;Fgf8^+/– embryos was essentially identical (n=7 and 5, respectively) (Table 3). Although the number of embryos analyzed is small, these results are predicted by the hypothesis that the phenotypically significant interaction between the two genes occurs during the development of the fourth PAAs. Because the fourth PAAs do not form in Tbx1^–/– animals, the phenotypic enhancing effect of Fgf8 mutation is pre-empted.

In addition to aortic arch defects, we noticed that at E18.5, 14 out of 30 double heterozygous animals presented with thymic hypoplasia and/or lobe asymmetry (not shown). This phenotype was also seen in six out of 37 Tbx1^+/–;Fgf8^+/+ animals, in three out of 28 Tbx1^+/+;Fgf8^+/– animals, and in one out of 21 Tbx1^+/+;Fgf8^+/+ animals. Although this phenotype may be in the range of normal variability of the thymus in this genetic background, its penetrance in double heterozygous mutants is significantly higher than that in the other genotypes (P=0.0068). Because Fgf8^neo/– embryos (where Fgf8^neo is a hypomorphic allele) also present with thymic defects, including hypoplasia (Abu-Issa et al., 2002), and because the thymus primordia develop from the third pharyngeal pouch where both Tbx1 and Fgf8 are expressed (Fig. 1A,B), it is likely that the thymic phenotype in double heterozygous mutants is also a sign of Tbx1-Fgf8 interaction in the pharyngeal endoderm.

Complete loss of function of Tbx1 causes migration defects of neural crest cells (Vitelli et al., 2002), while severe reduction of Fgf8 dose in Fgf8^neo/– animals causes increased neural crest cell death (Abu-Issa et al., 2002). To test whether compound heterozygosity may be associated with neural crest migration defects, or increased apoptosis of neural crest cells, we have examined double heterozygous E9.5-E10.5 embryos and compared them with wild type and single mutants. Neural crest migration was tested by in situ hybridization with Crabp1 and cell death by staining with Lysotracker. These tests produced very similar results in all the genotypes tested, suggesting that either potential differences are below the sensitivity of the methods used, or that the Fgf8 modifier effect is independent from neural crest cell migration or survival.

The haploinsufficiency phenotype of the multigene deletion mutant Df1/+ recapitulates several aspects of the DiGeorge/del22q11 syndrome, including cardiovascular defects (Lindsay et al., 1999), thymic and parathyroid abnormalities (Taddei et al., 2001), and behavioral abnormalities (Paylor et al., 2001). The haploinsufficient gene responsible for the cardiovascular defects of Df1/+ mice is Tbx1 (Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001), while the genes responsible for the other aspects of the Df1-associated phenotype are still to be identified. An important aspect of the human syndrome is phenotypic variability, which contrasts with a remarkable homogeneity of the genetic defect (Lindsay, 2001; McDermid and Morrow, 2002). The heterogeneity of the clinical picture is due mainly to reduced penetrance of individual clinical findings and, to a lesser extent, to variable expressivity (Matsuoka et al., 1998). We have shown that the reduced penetrance of aortic arch defects in Df1/+ mice is related to ‘recovery’ from hypoplasia of the fourth PAA, mostly occurring between E10.5 and E11.5 (Lindsay and Baldini, 2001). At E10.5, all the mutant embryos have at least one of the two fourth PAAs affected. The same phenomenon occurs in Tbx1^+/– animals (I. T. and E. A. L., unpublished). This recovery process is, at least in part, under genetic control, as it is affected by the genetic background of animals (Taddei et al., 2001). Our data indicate that Fgf8 mutation enhances the primary defect of Tbx1^+/–, i.e. fourth PAA hypoplasia. Hence, this is a different mechanism from that underlying penetrance differences caused by genetic background.

Severe reduction of Fgf8 activity in Fgf8^neo/– embryos causes abnormalities of the third, fourth and sixth PAAs, thymic defects, as well as other abnormalities (Abu-Issa et al., 2002). By contrast, the phenotypic enhancing effects of Fgf8 heterozygous mutation on Tbx1 haploinsufficiency is restricted to the fourth PAA and to the thymus, consistent with expression overlap of the two genes in the fourth and third pouches. A phenotypic effect on the third and sixth PAA, which are also close to the third and fourth pouches, may require further reduction of Fgf8 and/or Tbx1 proteins.

The nature of Tbx1-Fgf8 interactions remains to be clarified, the Fgf8 promoter/enhancer elements have not been fully characterized, nor has the exact DNA binding site and transcription activity of the Tbx1 protein. A direct induction of Fgf8 by Tbx1 is possible, but it would be restricted to the pharyngeal endoderm because the expression of the two genes does not overlap in any other tissue. While this hypothesis is simple and consistent with phenotypic observations, it is also possible that the Tbx1-Fgf8 interaction is non-specific. For example, Fgf8 dose reduction may have a generic negative impact on arch mesenchymal cells, for example, aggravating the development of an already compromised fourth PAA. We think that this scenario is unlikely because (1) Tbx1 is required for the expression of at least two Fgf genes in shared expression domains, suggesting specific interactions with FGF signaling, and (2) Fgf8^neo/– embryos have fourth PAA defects (Abu-Issa et al., 2002), indicating that Fgf8 contributes to the development of these arteries.

We propose that FGF signaling mediates the role of Tbx1 in the development of various derivatives of the pharyngeal arches and pouches, and perhaps different Fgf genes may be involved in the development of different derivatives. We have shown here that Fgf8 interacts with Tbx1 in aortic arch and in thymic development in vivo. Tbx1^–/– mice do not have thymus (Jerome and Papaioannou, 2001; Lindsay et al., 2001) and Df1/+ mice, in certain genetic backgrounds, have thymic hypoplasia (Taddei et al., 2001). We have also shown that Fgf10 expression is affected in Tbx1^–/– mice in the core mesenchyme and in the secondary heart field. Fgf10 is required for lung, limb (Min et al., 1998; Sekine et al., 1999) and thymic development (Revest et al., 2001), but the cardiovascular phenotype in Fgf10^–/– has not been described. It would be of interest to cross-breed Tbx1 and Fgf10 mutants, when available, to investigate whether the two genes interact during outflow tract and/or thymic development.

The significance of our findings for DiGeorge/del22q11 syndrome remains to be addressed. Our data should prompt the analysis of FGF loci to establish whether allelic variants are associated with increased risk of cardiovascular defects, thymic defects, or other pharyngeal pouch/arch developmental abnormalities in individuals with del22q11 syndrome.

We thank Tuong Huynh and Hedda Sobotka for valuable technical help. This research has been supported by grants RO1-HL51524, RO1-HL64832 and PO1-HL67155 from the National Heart Lung and Blood Institute, NIH (to A. B.), and by Grant 0060099Y from the American Heart Association, Texas Affiliate (to E. A. L.). F. V. is recipient of a fellowship from the Italian Telethon.