Tbx1 controls cardiac neural crest cell migration during arch artery development by regulating Gbx2 expression in the pharyngeal ectoderm

The pharyngeal apparatus is a transient embryonic structure present during vertebrate development. It forms gradually as the pharynx segments into a series of arches and pouches in a cranial-to-caudal order. Each pharyngeal arch (PA) possesses an outer pharyngeal surface ectoderm (PSE) layer and an inner pharyngeal endodermal (P.endo) layer. In the centre of each PA, the pharyngeal arch artery (PAA) is encircled by a mesodermal core structure surrounded by a layer of neural crest cells (NCCs). The bi-lateral PAA system develops sequentially as the pharynx segments and allows blood to flow from the heart tube to the dorsal aortae(DeRuiter et al., 1993). At E10.5, five PAAs have formed, and whereas PAAs 1 and 2 have regressed into capillary beds, PAAs 3, 4 and 6 persist in their respective PAs. Starting from E11.5, these caudal PAAs undergo an extensive asymmetrical remodelling and persist later as parts of the great vessels arising from the aorta(Srivastava and Olson, 2000). Defective development of the fourth PAAs is responsible for severe vascular anomalies, such as interrupted aortic arch type B, which is postnatally lethal(Vitelli et al., 2002a). Perturbation of normal pharyngeal development is a feature of congenital diseases, such as DiGeorge syndrome, in which the main cause of death is congenital heart defects (CHDs) (Lindsay,2001).

Haploinsufficiency of the T-box transcription factor TBX1 is responsible for many features of DiGeorge syndrome(Paylor et al., 2006; Stoller and Epstein, 2005; Yagi et al., 2003). Tbx1 is expressed dynamically in the developing pharyngeal apparatus in the mouse embryo and Tbx1 homozygous mutants display severe cardiovascular, craniofacial, thymic, parathyroid and ear defects(Jerome and Papaioannou, 2001; Lindsay et al., 1999; Merscher et al., 2001). In particular, CHDs found in Tbx1 mutant embryos consist of great vessel disruptions arising from defective fourth PAA development(Jerome and Papaioannou, 2001; Lindsay et al., 2001), and/or septation and alignment defects of the outflow tract (OFT) of the heart(Jerome and Papaioannou, 2001; Merscher et al., 2001). The essential role of Tbx1 in pharynx organogenesis relies on its ability to interact with crucial signalling pathways during development, such as the fibroblast growth factor (Fgf) (Hu et al.,2004; Vitelli et al.,2002b; Xu et al.,2004), hedgehog (Yamagishi et al., 2003) and retinoic acid(Guris et al., 2006; Roberts et al., 2006)signalling pathways. Furthermore, Tbx1 is required at different times(Xu et al., 2005), in different tissues (Arnold et al.,2006; Xu et al.,2005; Zhang et al.,2005; Zhang et al.,2006) and at different levels (Liao et al., 2004; Zhang and Baldini, 2008) to pattern all pharyngeal-derived structures. Thus, the analysis of Tbx1 effectors during embryonic development should advance our understanding of how the pharyngeal apparatus develops and, furthermore, aid in the identification of essential genes implicated in CHDs.

The special predisposition of the fourth PAA system toward interruption and/or hypoplasia is linked to the unique vascular morphology of these arteries (Bergwerff et al.,1999). Lineage-tracing in the mouse combined with morphological analysis of specific arterial segments has shown that muscular and non-muscular components of the fourth PAA system originate from cNCCs(Bergwerff et al., 1999; Jiang et al., 2000). cNCCs are pluripotent embryonic cells derived from the neuroepithelium(Le Douarin and Kalcheim,1999). They originate from the neural folds in a region between the middle of the otic vesicle and the caudal edge of somite 3(Kirby et al., 1985). They migrate through PA 3, 4 and 6(Miyagawa-Tomita et al., 1991)and form the vascular smooth muscle layer of these arteries(Le Lievre and Le Douarin,1975). cNCCs also contribute to the connective tissue of the thymus, parathyroid and thyroid glands(Bockman and Kirby, 1984; Le Lievre and Le Douarin,1975) and are essential for the formation of the aorticopulmonary septum (Kirby et al., 1983). cNCCs migrate long distances and, like NCCs in general, are extremely sensitive to external signalling from adjacent tissues, enabling them to interact with, and respond to, their changing milieu(Trainor and Krumlauf, 2000). Deletion of genes involved in cNCC migration(Feiner et al., 2001),self-renewal (Teng et al.,2008), survival (Macatee et al., 2003) and patterning(High et al., 2007), or ablation of premigratory cNCCs in the chick(Nishibatake et al., 1987),results in defective organogenesis of the PAA system. Therefore, the development of certain DiGeorge-like syndromic features, including great-artery defects, result, at least in part, from defective development of the cNCC lineage.

Notably, whereas Tbx1 is expressed as early as E7.5 in all three germ layers (Chapman et al.,1996), it is not expressed in cNCCs(Garg et al., 2001); thus, the effects of Tbx1 on cNCCs in fourth PAA development are non-cell-autonomous. Endothelial cells lining the fourth PAAs derive from mesodermal cells (Kirby and Waldo,1995) and Tbx1-fated cells are found in the endothelium of the fourth PAA (Vitelli et al.,2002a; Zhang et al.,2005). Nevertheless, conditional deletion of Tbx1 in endothelial cells does not generate fourth PAA defects(Zhang et al., 2005). Conditional deletion of Tbx1 from all mesodermal expression domains has a severe impact on pharynx segmentation, with loss of PAs 3 to 6 and consequent developmental failure of the PAA system(Zhang et al., 2006). However,mesoderm-specific restoration of Tbx1 expression in a mutant background rescues most of these defects, but not fourth PAA formation(Zhang et al., 2006). Subsequently, the PSE and P.endo, both of which express Tbx1, have been shown to be required for fourth PAA development(Arnold et al., 2006; Zhang et al., 2005). Also, the PSE has been shown to be a crucial source of Fgf signals for fourth PAA formation and remodelling (Macatee et al.,2003).

Although the diverse spatiotemporal functions of Tbx1 have been extensively studied and characterised, less is known about the genetic networks regulated by this protein during specific aspects of pharynx and heart development. Our laboratory and others have started to investigate potential downstream Tbx1 targets by analysing gene expression arrays obtained using available mouse models (Ivins et al., 2005; Liao et al., 2008). In this report, we provide new insights into how Tbx1 regulates important aspects of fourth PAA development by controlling the expression of the homeobox gene Gbx2. We employed a genetic approach to abrogate Gbx2activity downstream of Tbx1 and analysed resulting morphological and molecular defects. These studies revealed abnormalities in cNCC migration that lead to defective development of the PAAs and we propose a new mechanism by which Tbx1/Gbx2 might regulate cNCC migration via the Slit/Robo signalling pathway.

The mouse lines used in this study have been described previously: Tbx1^+/- (Lindsay et al., 2001), Tbx1^neo2/+(Zhang et al., 2006), Tbx1^flox/flox (Arnold et al., 2006; Xu et al.,2004); Tbx1^Cre(Brown et al., 2004), Gbx2^+/- (Wassarman et al., 1997), Gbx2^flox/flox(Li et al., 2002), AP2α^IresCre (referred to as AP2α^Cre in the text)(Macatee et al., 2003), Slit2^+/- (Plump et al., 2002) and R26R (Soriano,1999). Tbx1^+/-,AP2α^Cre and Tbx1^neo2/+lines were maintained on a C57Bl/6 background. Tbx1^Cre,Gbx2^+/- and Gbx2^flox/flox founders used in the crosses described below came originally from three different unknown,mixed backgrounds.

Tbx1^Cre/+;Gbx2^flox/- were generated by crossing Tbx1^Cre/+;Gbx2^+/- animals to Gbx2^flox/flox. Likewise, AP2α^Cre/+;Gbx2^+/- animals were crossed to Gbx2^flox/flox to generate Gbx2-PSEconditional mutant embryos: AP2α^Cre/+;Gbx2^flox/-. Finally,two types of Tbx1-PSE conditional mutants were generated: AP2α^Cre/+;Tbx1^+/- animals were crossed to Tbx1^flox/flox to generate AP2α^Cre/+;Tbx1^flox/- and AP2α^Cre/+;Tbx1^+/flox animals were crossed to Tbx1^flox/flox to generate AP2α^Cre/+;Tbx1^flox/flox. Fgf8^neo/δ2,3 embryos were obtained by crossing Fgf8^Δ2,3/+ mice with Fgf8^neo/+ mice and were genotyped as described(Meyers et al., 1998).

lacZ-expressing cells from E9.5 Df1^+/-;Tbx1^+/-, effectively Tbx1^-/- (Lindsay et al., 1999), and Tbx1^+/- embryos were labelled using the CMFDG Labelling Kit (Molecular Probes) and collected by FACS. RNA was isolated using Trizol and amplified by PCR using the Microarray Target Amplification Kit (Roche). Quantitative real-time PCR was carried out on an ABI7000 and expression values were obtained using DART-PCR Analysis Excel software (Peirson et al.,2003). Slit2 relative expression corresponds to n=3 technical replicates. Each sample represents cells collected from at least 10 E9.5 embryos.

Whole-mount RNA in situ hybridisation was performed based on methods described previously (Wilkinson,1992) using digoxygenin-labelled probes for Gbx2, Slit1,Slit2, Slit3, Robo1 and Sox10. Ink injection was performed on E10.5 embryos fixed in 4% paraformaldehyde overnight at 4°C by targeting the OFT with a microinjection needle filled with Indian ink (Pélican). Mixed-stage embryos were stained with X-Gal following standard procedures(Nagy et al., 2003). Lysotracker Red (Invitrogen) was used at 5 μM for 20-30 minutes at 37°C. Anti-PECAM antibody (Mec13:3, Pharmingen) was used at 1:50,anti-Tbx1 antibody (Zymed) at 1:50, anti-GFP antibody (Sigma G6539) at 1:100 and anti-AP2α (Tcfap2α) antibody at 1:25 (3B5, Developmental Studies Hybridoma Bank). Whole-mount immunolabelling was carried out as described (Schwarz et al.,2008) using anti-p75 or anti-HNK-1 IgM antibody (Zymed MHCD 5701)at 1:100. Anti-Robo1 antibody was used at 1:2000(Tamada et al., 2008) and signal was detected using biotinylated secondary antibody coupled with Streptavidin Alexa Fluor 488 conjugate. Alexa Fluor-conjugated secondary antibodies (Molecular Probes) were used at 1:200.

Loss of Gbx2 is associated with abnormal development of the PAAs and hindbrain (Byrd and Meyers,2005; Wassarman et al.,1997). In order to investigate whether Gbx2 could mediate some of the functions of Tbx1 during fourth PAA development, as previously suggested(Ivins et al., 2005), we compared Gbx2 and Tbx1 expression in wild-type (wt) and Tbx1^-/- mouse embryos at E8.5, the specific time at which Tbx1 is required to pattern the PAAs (Xu et al., 2005). As shown in Fig. 1, both Tbx1 and Gbx2 are strongly expressed in the pharyngeal region of E8.5 wt embryos(Fig. 1A,D). Section analysis showed that Gbx2 is expressed predominantly in the PSE, although some staining could also be detected in the P.endo(Fig. 1B). Tbx1 and Gbx2 expression overlap in these two epithelial domains[Fig. 1B,E, PSE (red arrow),P.endo (yellow arrow)], which are both potentially involved in PAA development(Zhang et al., 2005). Tbx1 was additionally detected in the head mesoderm and in the secondary heart field (SHF) mesoderm, neither of which expressed Gbx2(Fig. 1B,E). From these observations, we conclude that Gbx2 and Tbx1 show overlapping expression domains at the time of fourth PAA specification. Later in development, Tbx1 and Gbx2 expression is detected in the region where PAAs 3 to 6 are going to form (the bracketed domain in Fig. 1C,F) and is ultimately restricted to the P.endo (Fig. 1I,J) (Zhang et al.,2005).

We next examined Gbx2 expression in Tbx1^-/-embryos at E8.5 and found that it was lost from both the PSE and the P.endo(Fig. 1G,H), but was maintained at the midbrain/hindbrain junction and in the tail bud region(Fig. 1G). By contrast, when we examined Gbx2 expression in Tbx1 PSE conditional mutants(Tbx1^flox/flox;AP2α^Cre/+,referred to here as Tbx1-PSE) (AP2α is also known as Tcfap2a - Mouse Genome Informatics), we found that it was lost in the PSE, whereas it was maintained in the P.endo (compare Fig. 1M with 1N and Fig. 1B with 1O). Therefore, decreased Gbx2 expression is more likely to be due to gene downregulation than to a lack of development/survival of Gbx2-expressing cells in the PSE. At later stages, Tbx1^-/- mutants displayed severe hypoplasia of the pharynx, lacked the caudal PAs and lost Gbx2-expressing tissues (Fig. 1K,L).

Overall, our results showed that at the time of PAA specification, Gbx2 expression domains, which consist of the PSE and P.endo, are dependent on Tbx1 activity.

The gene expression data presented above suggest that Gbx2 might interact with Tbx1 in the PSE and/or P.endo. To test this hypothesis,we crossed Tbx1^+/- with Gbx2^+/- mice and analysed the PAA phenotype of the progeny at E10.5 and at E15.5 by ink injection. At E10.5, we scored a total of 42 embryos and found a variety of defects, as described in Fig. 2A-H and Table 1. In our mixed genetic background (see Materials and methods), we found that 5 out of 11 Tbx1^+/-;Gbx2^+/+ embryos (45.5%) were abnormal (Table 1). Under these conditions, dosage reduction to one copy of the Gbx2 allele strongly increased the penetrance and severity of the fourth PAA phenotype: 12 out of 14 Tbx1^+/-;Gbx2^+/- embryos (85.7%) were abnormal, which is significantly higher than that found for Tbx1^+/-;Gbx2^+/+ embryos (P=0.04). Furthermore, whereas the type of defects found in Tbx1^+/-;Gbx2^+/+ embryos mostly affected the fourth PAA on one side only (3 out of 5 presented unilateral defects, Table 1), those found in Tbx1^+/-;Gbx2^+/- embryos were much more severe,with 10 out of 12 affected on both sides, and 7 out of 10 harbouring no fourth PAA (NP/NP) (Fig. 2A-H; Table 1). The third and sixth PAAs were normal in all genotypes.

Similar to our findings in E10.5 embryos, the penetrance of arch defects was significantly higher in Tbx1^+/-;Gbx2^+/-mutants (59%, n=22) than in Tbx1^+/-;Gbx2^+/+ embryos at E15.5 (26%, n=19) (see Table 1; Fig. 2I-L). Defects included an aberrant origin of the right subclavian artery (A-RSA), either alone or combined with interruption of the aortic arch type B (IAA-B)(Fig. 2J-L). Overall, our results showed that Tbx1 and Gbx2 interact genetically during fourth PAA development.

So far, our results suggest that Gbx2 might mediate a number of important Tbx1 functions during PAA morphogenesis. However, tissues expressing Gbx2 and Tbx1 interact closely during pharyngeal development, raising the issue of whether Gbx2 might be required in one or more domains, possibly extending beyond Tbx1-expressing tissues, to contribute to the development of the PAAs. We used lineage-specific and tissue-specific gene ablation experiments to address these points.

In order to test whether cells expressing both Tbx1 and Gbx2 are responsible for the fourth PAA phenotype, we performed lineage-specific gene ablation using a Tbx1-Cre transgenic mouse line combined with Gbx2^+/- and Gbx2^flox/+alleles (see Materials and methods). The Tbx1-Cre transgenic mouse has been used successfully to investigate Fgf8 functions downstream of Tbx1 (Brown et al.,2004).

Tbx1^Cre/+;Gbx2^flox/- embryos were collected at E10.5 and fourth PAA formation was evaluated by ink injection(Table 2). Twenty-seven percent of conditional mutant embryos displayed an abnormal fourth PAA on one or both sides of the embryo; no other defects were found for any other genotypes produced in this cross. The low penetrance of the fourth PAA phenotype found in Gbx2 conditional mutants did not recapitulate the percentage of defects found in Gbx2^-/- embryos (50%, Table 2). This discrepancy is likely to be due to the low efficiency of Tbx1-Cre-driven transgene recombination in epithelial tissues at the time of fourth PAA specification(see Fig. S1 in the supplementary material), as previously reported(Macatee et al., 2003). Nevertheless, these results clearly demonstrate that deletion of Gbx2in Tbx1-expressing cells is sufficient to generate a fourth PAA phenotype, in agreement with a model in which Tbx1 regulates Gbx2 in a cell-autonomous manner.

We also noted that the genetic background affected the incidence of defects found in Gbx2^-/- embryos, which reached 50% on a mixed background (Table 2) as compared with 100% on a C57Bl/6 background(Byrd and Meyers, 2005). This phenotypic variability was also found in different genetic backgrounds for the same Tbx1 deletion (Taddei et al., 2001). For this reason, and also because the fourth PAA defect recovers partially over time(Lindsay and Baldini, 2001),we decided to focus on PAA formation scored at E10.5 and not to assay PAA remodelling at later stages.

In order to abrogate Gbx2 activity specifically in the PSE during embryonic development, we used an AP2α-Cre deleter mouse line, which has been successfully used to define Fgf8 functions in that tissue (Macatee et al.,2003). Lineage-tracing analyses of the AP2α-Cre driver line detect Cre activity in the PSE as early as E8.5 (Macatee et al.,2003) (see Fig. S1 in the supplementary material), and also indicate labelling of pharyngeal NCCs(Macatee et al., 2003), a cell population that expresses neither Gbx2, Tbx1 nor Fgf8(Fig. 1J, asterisk)(Bouillet et al., 1995; Chapman et al., 1996; Frank et al., 2002; Garg et al., 2001). Therefore,using the AP2α-Cre driver in our model corresponds to ablating Gbx2 expression in the PSE at the onset of PAA formation.

We first analysed fourth PAA formation at E10.5 by performing ink injection on Gbx2-PSE conditional mutants. We found that the incidence of defects attributable to abnormal fourth PAA development was the same in Gbx2^-/- and Gbx2-PSE embryos, with 47% of Gbx2-PSE embryos displaying abnormal fourth PAA as compared with 50%for Gbx2^-/- (Fig. 3A-E; Table 2).

We next investigated PAA formation defects found in Gbx2-PSEmutants at the cellular level. A PECAM (Pecam1 - Mouse Genome Informatics)study of a thin-patent (Th-P) Gbx2-PSE conditional mutant at E10.5 showed that endothelial cells were present in the developing fourth PAA, but many failed to organise into tubes, as previously reported for Gbx2^-/- mutants (Fig. 3G,G′) (Byrd and Meyers,2005). In order to compare the Gbx2-type with the Tbx1-type PAA defects, we stained a non-patent (NP) Tbx1^+/- embryo and a NP Tbx1-PSE(AP2α^Cre/+;Tbx1^flox/flox)conditional mutant embryo for PECAM and observed the same endothelial cell organisation defects at E10.5 (Fig. 3H,I). We therefore conclude that ablation of Gbx2 in the PSE is sufficient to recapitulate the fourth PAA defects found in Gbx2^-/- embryos.

In order to investigate the origin of the fourth PAA abnormalities, we first tested whether Gbx2-PSE mutants recapitulate the cNCC patterning defects found in Gbx2^-/- embryos(Meyers et al., 1998). Analysis of Sox10 expression at E10.5, which marks migrating NCCs(Kuhlbrodt et al., 1998),revealed migration defects in 50% of embryos (n=8)(Fig. 4B). Whereas two discrete streams of cells were observed in wt embryos, these streams were partially or completely fused in Gbx2-PSE embryos [compare yellow (normal) with red (abnormal) arrows in Fig. 4A,B]. This aberrant migration pattern was accompanied by a loss of cNCCs entering the fourth PA specifically [compare yellow (normal) with red(abnormal) asterisks in Fig. 4A,B]. In addition, we found that two types of Tbx1-PSEconditional mutants(AP2α^Cre/+;Tbx1^flox/flox and AP2α^Cre/+;Tbx1^flox/-) suffered from similar cNCC migration defects (Fig. 4C,D). Furthermore, cNCC migration defects were more severe in Tbx1^+/-;Gbx2^+/- embryos as compared with Tbx1 or Gbx2 heterozygous embryos(Fig. 4E-H), in agreement with a genetic interaction between Gbx2 and Tbx1 during fourth PAA development (Fig. 2; Table 1).

In addition to the cNCC migration phenotype observed in all Gbx2and Tbx1 mutants, the decreased number of cNCCs entering the fourth PA could result from a lack of survival of these cells within this specific PA, as previously reported (Arnold et al.,2006; Macatee et al.,2003). In order to test this, we performed dual fluorescence staining at E10.5 using p75 (Ngfr - Mouse Genome Informatics) antibody and Lysotracker reagent, which mark NCCs (Rao and Anderson, 1997) and areas containing apoptotic cells,respectively (Barbosky et al.,2006). In wt or Gbx2^+/- embryos, cNCC patterning occurred normally and cNCCs migrated into the caudal fourth PA,which was filled with green, p75-positive cells(Fig. 5Aa,Ad). In Gbx2-PSE embryos, cNCC migration was abnormal (yellow arrows in Fig. 5Ba) and the number of p75-positive cells within the fourth PA was substantially decreased (yellow asterisks in Fig. 5Ba,Bd). This aberrant migration was not accompanied by an increase in the number of apoptotic cells in the region caudal to the otocyst along the migratory path(compare the Lysotracker reagent pattern in Fig. 5Ab and Bb), none of these cells being p75 positive (note the absence of double-labelled cells in Fig. 5Ac and 5Bc). Similar results were observed at E9.5 (data not shown). Furthermore, we did not detect any increase in apoptosis in the PSE, P.endo or mesodermal tissues within the fourth PA(Fig. 5Ad,Bd). Similar results were obtained with Tbx1^+/- and Gbx2^+/-;Tbx1^+/- embryos(Fig. 5C,D). In all mutants,cNCC deficiency was accompanied by a missing fourth PAA(Fig. 5Bd,Cd,Dd). We therefore conclude that cNCC deficiency at the level of the fourth PA does not originate from a lack of survival signals triggered by the PSE, but rather from an incorrect migration signal that misguides these cells from their intended destination.

Surprisingly, we found that 100% of Tbx1^+/- embryos(n=6) displayed cNCC migration abnormalities(Fig. 4F,G; Fig. 5C) in a c57Bl/6 background, a defect that has not been documented previously and is contrary to studies suggesting that cNCC migration occurs normally in Tbx1^+/- embryos(Kochilas et al., 2002; Lindsay and Baldini, 2001). Our analysis shows that although the number of cNCCs migrating into the caudal PAs is not affected (Fig. 4F,G; Fig. 5C)(Kochilas et al., 2002), their migration paths are disrupted, which in turn leads to a cNCC deficiency at the level of the fourth PA. To elaborate on these findings, we investigated how this cNCC response varied with Tbx1 dosage. We examined cNCC migration defects in a Tbx1 hypomorphic series (see Materials and methods) and compared this with the incidence of fourth PAA defects(Zhang and Baldini, 2008). We found that progressive diminution of Tbx1 mRNA levels in vivo, which is associated with increased penetrance of the fourth PAA phenotype(Zhang and Baldini, 2008),correlated with an increased incidence of cNCC migration defects(Fig. 4I-N). Overall, our data present for the first time a non-cell-autonomous function for Tbx1 in cNCC migration and correlate fourth PAA deficiency with cNCC migration defects.

In summary, our data suggest that loss of Tbx1 and Gbx2in the PSE results in fourth PAA abnormalities associated with cNCC deficiency in the fourth PA. Our data also demonstrate that this striking reduction of cNCCs in the fourth PAA is linked to a severe diversion of cNCC routes, rather than to any increased apoptosis of these cells. Finally, we propose a central function for cNCCs in arterial development and endothelial cell organisation at E10.5, prior to their differentiation into smooth muscle cells at E11/E11.5(Bergwerff et al., 1999; Byrd and Meyers, 2005).

Fgf8 is expressed in the PSE(Crossley and Martin, 1995),interacts genetically with Tbx1 in the development of the fourth PAA(Vitelli et al., 2002b) and specific deletion of Fgf8 in the PSE causes fourth PAA defects(Macatee et al., 2003). It is therefore possible that Gbx2 performs some of the developmental functions of Fgf8 in this tissue, as previously suggested(Byrd and Meyers, 2005; Liu et al., 1999). In order to test this, we performed whole-mount in situ hybridisation for Gbx2 on Fgf8^neo/Δ2,3 hypomorphic embryos at E8.5 and E9.5. No change in Gbx2 expression in the PSE was detected (see Fig. S2 in the supplementary material). As a result, we conclude that Gbx2 acts downstream of Tbx1, but independently of Fgf8, in the development of the fourth PAA.

The Slit/Robo signalling pathway is involved in trunk neural crest migration (Jia et al., 2005)and is also implicated in Drosophila cNCC-like motility and assembly into the OFT (Zmojdzian et al.,2008). The gene encoding the secreted protein Slit2 was identified as a potential downstream target of Tbx1 in a new series of gene expression arrays, in which Tbx1-expressing cells, isolated directly from embryos, were submitted to transcriptome analysis (K.L.V.B. and P.J.S.,unpublished data; see Materials and methods). Slit2 was found to be downregulated in Tbx1^+/-;Df1^+/- cells[effectively Tbx1^-/-(Lindsay et al., 1999)] as compared with Tbx1^+/- cells [to 0.529±0.125(mean±s.d.), P<0.01]. Therefore, we investigated elements of this pathway in wt and mutant embryos at the time of fourth PAA development by in situ hybridisation (Figs 6 and 7).

At E8.75, Slit2 was detected in two major ectodermal regions covering both sides of the otic vesicle(Fig. 6A). Slit2 was strongly expressed in the PSE, whereas it was barely detectable in the P.endo(Fig. 6B), a pattern which resembles that of Gbx2 (Fig. 1B). In the PSE, Slit2 and Tbx1 colocalise at the time of PAA formation (Fig. 6E-G). Additionally, Slit2 was detected in the floorplate of the neural tube and in the notochord (Fig. 6B). At E9.25, Slit2 expression is robustly detected in the pharyngeal pouches (Fig. 6C)(Yuan et al., 1999) and was increased in the P.endo by E9.5 (Fig. 6D). Conversely, Slit2 expression in the PSE was not longer detected at that stage (Fig. 6D). We also noted a potential Slit ligand redundancy as both Slit1 and Slit3 were expressed in the PSE at the time of PAA formation and in overlapping domains with Slit2 (see Fig. S3 in the supplementary material).

Slit proteins mediate their effects through the Robo family of receptors,which consists of four members (Robo1-4). Slit2 can bind to all four Robo receptors (Brose et al., 1999; Camurri et al., 2005; Park et al., 2003), with Robo4 expressed exclusively in the vascular endothelium(Park et al., 2003). When examining Robo1-3 expression at the time of cNCC migration, we found that Robo1 had the most relevant pattern of expression(Fig. 6H-K), although Robo2 and Robo3 were expressed in overlapping regions with Robo1 (data not shown). Robo1 was first detected in the neural tube at E8.5 (Fig. 6H)and labelled streams of migrating NCCs in the vagal and trunk regions of E9.5 embryos (Fig. 6I,J). Robo1-positive cNCC streams were observed as early as E9.0 (data not shown) and migrated through PAAs 4-6, which exhibited strong Robo1labelling at E11 (Fig. 6K). When comparing Robo1 expression with two other NCC markers at E9.5,we found that the Robo1-positive stream of cells corresponded to that detected by Sox10 or Hnk1 (B3gat1 - Mouse Genome Informatics)labelling (Fig. 6L-N,L′-N′). As Robo1 is also expressed in the dermomyotome (Sundaresan et al.,2004), we performed a double immunofluorescence on E9.5 wt embryos, which confirmed Robo1 colocalisation with AP2α-positive migrating NCCs (Mitchell et al.,1991) (Fig. 6O-Q′).

We next investigated the expression of Slit/Robo pathway components in Gbx2 and Tbx1 mutants as a possible mechanism contributing to aberrant cNCC migration (Fig. 7).

At E8.75, the ectodermal expression of Slit2 was strongly downregulated in the post-otic region in Gbx2^-/- embryos,whereas the pre-otic region of Slit2 expression appeared unaffected(Fig. 7A,B). The same results were obtained in Gbx2-PSE and Tbx1-PSE(AP2α^Cre/+;Tbx1^flox/-) embryos and confirmed downregulation of Slit2 specifically in the PSE(Fig. 7D,I). Slit2expression was also downregulated in Tbx1^-/- embryos in both regions spanning either side of the otic vesicle(Fig. 7E,F) and in the post-otic region of Tbx1^+/-;Gbx2^+/- embryos(Fig. 7H). Slit2downregulation was more severe in Tbx1-PSE and Tbx1^-/- mutants than in Gbx2-PSE and Gbx2^-/- mutants (Fig. 7B,D,F,I), suggesting that Tbx1 might also regulate Slit2 expression independently of Gbx2.

When examining Robo1 expression in E9.5 Gbx2^-/- embryos, we found that the number of Robo1-positive cNCCs migrating towards caudal PAs was reduced(compare yellow arrow in Fig. 7J with black arrow in Fig. 6M). Similarly, Robo1-positive cells appeared less abundant in E10.5 Gbx2-PSE embryos(Fig. 7L) than in wt embryos(black bracketed region in Fig. 7K). We also observed fewer Robo1-positive cells at the level of the caudal arches, and the overall Robo1 expression domain appeared disorganised (yellow arrows in Fig. 7L). A similar decrease in Robo1-expressing domains was observed in Tbx1-PSEconditional mutants(AP2α^Cre/+;Tbx1^flox/-) and Tbx1^-/- mutants (Fig. 7M,N). These findings suggest that in response to a lower dosage of Slit2 in the PSE, Robo1-expressing domains are affected and Robo1 expression appears downregulated in the cNCC lineage of Gbx2 and Tbx1 mutants. These results are consistent with examples of the tight regulation of Robo expression in other systems(Andrews et al., 2008; Plachez et al., 2008). In conclusion, our data provide the first evidence for a Slit/Robo-mediated signalling pathway implicated in cNCC migration during PAA formation.

In this study, we have described a new function for the T-box transcription factor Tbx1 in controlling cell migration during PAA formation. We found that within the PSE, Tbx1 regulates the expression of the homeobox gene Gbx2, which in turn directs accurate cNCC navigation to caudal PAs. We investigated two potential mechanisms of action and found that this control is Fgf8-independent but imparts the activation of the Slit/Robo signalling pathway.

A number of genes, including Fgf8, Crk1 (Mapk14) and Vegf¹⁶⁴ (Vegfa), interact genetically with Tbx1 during fourth PAA development(Guris et al., 2006; Stalmans et al., 2003; Vitelli et al., 2002b). However, no potential downstream effectors of Tbx1 during PAA development have been reported so far.

Our cross-breeding experiments showed that Gbx2 interacts genetically with Tbx1, causing significant enhancement of the fourth PAA Tbx1 haploinsufficiency phenotype at both E10.5 and E15.5. Recovery from fourth PAA defects has previously been observed in Df1/+ mutants (Lindsay and Baldini, 2001) and did not appear to be affected by mutation of Gbx2, as a similar proportion of embryos overcame the early defect in Tbx1^+/- mutants and Tbx1^+/-;Gbx2^+/- double mutants by E15.5. These results suggested that the interaction with Tbx1 contributes to the primary defect, i.e. early in fourth PAA morphogenesis, rather than at the subsequent remodelling stage.

We generated a series of conditional mutants to test a putative role for Gbx2 downstream of Tbx1 in fourth PAA development. We employed a Tbx1-Cre line to show that Gbx2 is required in Tbx1-expressing cells and validated a model in which Gbx2 is a downstream effector of Tbx1 during fourth PAA development. Although three consensus T-box-binding sequences have been described within the first 5 kb of the Gbx2 promoter(Ivins et al., 2005), we have so far failed to identify any transactivation of reporters coupled to promoter fragments containing these elements (data not shown). The molecular mechanism underlying differential expression of Gbx2 in Tbx1 mutants remains to be determined.

Tissue-specific expression of Tbx1 in the SHF mesoderm is required for proper OFT development (Zhang et al.,2006) and some OFT-related defects have been reported to occur at a low frequency in Gbx2^-/- mutants(Byrd and Meyers, 2005). Although it is unlikely that Gbx2 mediates the effects of Tbx1 in OFT development, as Gbx2 expression was not detected in the SHF, it is still possible that Tbx1 and Gbx2 could regulate adjacent pathways that control different aspects of OFT morphogenesis. Nevertheless, no cardiac defects were observed, neither in Tbx1^+/-;Gbx2^+/- embryos (apart from a single ventricular septal defect, data not shown), nor in Gbx2-PSE(n=4) or Gbx2;Tbx1-Cre (n=23) mutant embryos. In addition, no thymic defects were noted. Thus, the phenotype-enhancing effect of Gbx2 appeared to be restricted to fourth PAA defects.

Our work confirmed the crucial role played by the PSE during PAA development. Since the same incidence of defects was found in Gbx2-PSE embryos as in Gbx2^-/- embryos, we concluded that Gbx2 is required in the PSE to pattern the fourth PAAs. These findings are in agreement with Gbx2 being a downstream effector of Tbx1 in the ectoderm, as defective fourth PAA development has also been reported in Tbx1-PSE conditional mutants (E.A.I., unpublished data). As Gbx2 expression was also downregulated in the pharyngeal endoderm of Tbx1^-/- embryos, it remains possible that this tissue plays a synergistic role in conjunction with the PSE in patterning fourth PAAs. Nevertheless, this role is likely to be subtle, as Gbx2-PSEmutants recapitulate cNCC migration anomalies as well as the endothelial cell organisation defects and fourth PAA abnormalities found in Gbx2^-/- embryos.

We found that two genes, Gbx2 and Slit2, which are both implicated in the Tbx1-related fourth PAA phenotype (the role of Slit2 is discussed below) are expressed in overlapping ectodermal domains with Tbx1. Notably, at ∼E8.5, which corresponds to the time point at which Tbx1 is required for fourth PAA patterning(Xu et al., 2005), Gbx2 and Slit2 displayed very strong expression in the PSE compared with the P.endo, and this expression was not maintained later in development. A similar transient ectodermal expression pattern has previously been described for Tbx1 (Zhang et al., 2005). We speculate that this tight spatiotemporal control of gene expression is required for correct patterning of cNCCs. Indeed, cNCCs emigrate from the neural tube at the 7-9 somites stage in the mouse embryo(Chan et al., 2004; Trainor, 2005). Therefore, the timing of Tbx1, Gbx2 and Slit2 expression, accompanied by the position of the PSE, immediately juxtaposed to cNCCs as they start to migrate, is likely to be instructive. It would not be surprising to find that other genes expressed in the PSE also regulate cNCC function in relation to cardiovascular development, as is already known to be the case for Fgf8 (Crossley and Martin,1995; Macatee et al.,2003).

The best-characterised functions of Tbx1 during pharyngeal and heart development are its cell-autonomous effects on cell proliferation, cell fate decision and progenitor cell expansion (Xu et al., 2004; Xu et al.,2007; Zhang et al.,2005; Zhang et al.,2006). Little is known about the non-cell-autonomous role of Tbx1 in cNCC development.

Our study describes a new function for Tbx1 in regulating cNCC migration during PAA development and constitutes the second set of evidence that Tbx1 plays a role in cNCC development. Previously, histological analysis of the fourth PAAs in Df1/+ and Lgdel/+ embryos, two chromosome deletion mouse models for DiGeorge syndrome reducing Tbx1 to hemizygosity (Lindsay et al.,1999; Merscher et al.,2001), revealed a role for Tbx1 in cNCC differentiation into vascular smooth muscle (Kochilas et al.,2002; Lindsay and Baldini,2001) as a possible explanation of the Tbx1haploinsufficiency phenotype. Nevertheless, additional defective signalling pathways are likely to be involved in the fourth PAA phenotype at E10.5 as cNCC differentiation into vascular smooth muscle is not stably detected at the level of the fourth PAA until E11.5(Bergwerff et al., 1999; Byrd and Meyers, 2005). Our work demonstrated that the fourth PAA phenotype originates from a cNCC deficiency at the level of the fourth PA in both Gbx2 and Tbx1 mutants, highlighting a central function for the cNCCs in arterial development prior to their differentiation into smooth muscle cells.

Evidence for a role for Tbx1 in regulating cNCC migration has been obtained previously. In Lgdel/+ mice, the number of cNCCs reaching the cardiac cushions of the OFT was affected(Kochilas et al., 2002), and in Tbx1^-/- embryos, post-otic cranial nerves IX and X,which derive in part from cNCCs and follow similar migratory streams as they develop, were abnormally fused in their distal part(Vitelli et al., 2002a). We demonstrated that Tbx1 acts on cNCC migration by controlling the expression of Gbx2, a gene implicated exclusively in the migratory aspect of cNCCs(Byrd and Meyers, 2005). We provided evidence that migrating streams of cNCCs are disorganised in Gbx2-PSE, Tbx1^+/- and Tbx1-PSE mutant embryos by analysing three markers of migrating cNCCs: Sox10, p75 and Robo1. This dysregulation of migratory path-finding results in a reduced number of these cells reaching the fourth PAs and is associated with abnormal organisation of adjacent endothelial cells into tubes and,subsequently, with fourth PAA abnormalities.

Finally, our data have shown that the Tbx1/Gbx2 pathway acts independently of Fgf8 in the PSE during PAA development, which is in agreement with separate roles for Tbx1/Gbx2 and Fgf8 in controlling cNCC migration and survival,respectively (Macatee et al.,2003). Indeed, in Tbx1 and Gbx2 mutants, cNCCs do not apoptose within the fourth PA, but instead fail to reach this location because their migration paths have been diverted.

cNCCs represent a transitional population between the cranial and the trunk crest region and share properties common to both cells(Suzuki and Kirby, 1997). It is therefore likely that signalling molecules implicated in cranial and trunk crest cell migration, such as Semaphorins, Ephrins and Slits, are also used by cNCCs to reach their targets. So far, very little is known about how cNCC migration is regulated, although one series of reports has implicated the Semaphorin family of secreted ligands(Brown et al., 2001; Feiner et al., 2001; Toyofuku et al., 2008).

We showed that Slit2 is downregulated in the PSE of Gbx2^-/-, Gbx2-PSE, Tbx1-PSE and Tbx1^-/- embryos at the time of fourth PAA development. We also showed that at least one member of the Roundabout family, Robo1,is expressed in the NCC lineage. These findings support a new role for Tbx1 in regulating cNCC migration and indicate a possible pathway utilised by cNCCs to reach their targets during development. NCCs migrate in segmentally restricted streams, the boundaries of which are maintained by signalling molecules such as those of the Slit/Robo family(De Bellard et al., 2003). In Gbx2^-/-, Gbx2-PSE, Tbx1-PSE and Tbx1^-/- embryos, fusion of cNCC streams suggests that the molecular basis of distinct path formation has been disrupted. Whether Slit/Robo signalling mediates repulsive or attractive signals to maintain separated streams during cNCC migration is currently under investigation. Interestingly, a recent study demonstrated that both repulsive and attractive signals provided by the Semaphorin/Plexin family are required for precise navigation of cNCCs (Toyofuku et al.,2008).

Robo4 is a specific marker of endothelial cells and activation of Robo4 by Slit2 stabilises the vascular network by inhibiting endothelial tube formation and permeability as well as vascular leak(Jones et al., 2008). In addition, Robo1 and Robo4 are involved in filopodia formation and endothelial cell motility (Sheldon et al.,2009). Notably, Gbx2^-/-;Gbx2-PSE,Tbx1^+/- and Tbx1-PSE embryos displayed abnormally organised endothelial cells (Byrd and Meyers, 2005). Therefore, a non-cell-autonomous action of Tbx1 on fourth PAA development could act in two synergistic ways by controlling both cNCC migration/differentiation and endothelial cell organisation. Likewise,investigating the role of the Slit/Robo pathway in fourth PAA development will require careful analysis of multiple potential players.