Required, tissue-specific roles for Fgf8 in outflow tract formation and remodeling

Cardiogenesis requires specification, differentiation, proliferation and migration of cell populations from diverse sites. Inductive interactions between these populations and regionally-specific transcription factor hierarchies regulate distinct morphogenetic events during heart development;their disruption by mutation or environmental insult usually results in malformations that affect discrete cardiac substructures, leaving the rest of the organ intact (de la Cruz et al.,2001; Harvey,2002). Given the duration and complexity of heart development, it is not surprising that nearly one percent of humans are born with congenital cardiovascular defects (Creazzo et al.,1998; Harvey,2002; Olson and Schneider,2003).

Myocardial precursors leave the primitive streak and reside briefly in a`cardiac crescent' of bilateral mesoderm where their fate is influenced by adjacent endoderm (Lough and Sugi,2000; Schultheiss et al.,1995; Yutzey and Kirby,2002). Clonal analyses revealed that two distinct lineages contribute to the heart (Zaffran et al.,2004), and studies in mouse and chick indicate that crescent mesoderm contains two precursor populations: a `primary heart field' (PHF)contains the earliest cells to undergo myocardial differentiation, which contribute to the primitive tubular heart and ultimately to the left ventricle(LV), whereas the `anterior heart field' (AHF) contributes right ventricular(RV) and outflow tract (OFT) myocardium(Buckingham et al., 2005; Zaffran et al., 2004; Cai et al., 2003; de la Cruz et al., 1977; Kelly et al., 2001; Mjaatvedt et al., 2001; Waldo et al., 2001). Although Isl1 is expressed by undifferentiated cells in the AHF, it is also expressed by many LV precursors (Cai et al.,2003) (this manuscript). To date, no proteins have been identified whose expression specifically identifies either field and it is unknown if cells are developmentally restricted within each field. Rather than indicating a distinct molecular identity, the concept of the AHF is currently a teleologic one, reflecting the relative position of a cell in the crescent,the time it contributes to the heart, and its ultimate location within the heart. For clarity, in this manuscript we employ AHF to refer to mesodermal cells residing in the dorsomedial crescent and, subsequently, in anterior splanchnic mesoderm (SM) ventral to the foregut, which are required to form the RV and the OFT.

Transcription factor networks regulate cardiomyocyte specification (Gata4 and Nkx2.5), differentiation (Mef2c), and regional identity within the heart(Brand, 2003; Bruneau, 2002; Cripps and Olson, 2002; Firulli and Thattaliyath,2002; Fishman and Chien,1997; Srivastava and Olson,2000). AHF expression of Mef2c is regulated by at least two intronic elements, one of which confers responsiveness to Isl1 and Gata(Dodou et al., 2004) and the other to a Tgfβ/Foxh1/Nkx2.5 pathway(von Both et al., 2004). Although expression of Mef2c, Isl1 and Foxh1 is not restricted to the AHF, their inactivation primarily disrupts the formation of AHF-derived structures, suggesting that this network regulates the proliferation and survival of undifferentiated AHF cells(Cai et al., 2003; Lin et al., 1997; von Both et al., 2004).

Much less is known about the signaling molecules that trigger cardiomyocyte differentiation, or guide the subsequent formation of cardiac substructures. Endodermal Wnt, Bmp and Fgf signals influence cardiomyocyte differentiation(Brand, 2003), and members of these families have later roles in OFT and valve morphogenesis(Abu-Issa et al., 2002; Armstrong and Bischoff, 2004; Delot et al., 2003; Frank et al., 2002; Liu et al., 2004; Sanford et al., 1997).

Fgf8 and Fgf10 are secreted signaling proteins expressed in the cardiac crescent and in SM (Crossley and Martin,1995; Kelly et al.,2001). Although Fgf10^-/- mice survive embryogenesis without cardiac defects (Min et al., 1998; Sekine et al.,1999), germline ablation of Fgf8 is embryonic lethal(Sun et al., 1999). The complex cardiac and OFT phenotypes seen in Fgf8-deficient hypomorphic mice indicate that this protein has required functions at different times and locations during cardiogenesis and the subsequent OFT remodeling. Severe hypomorphs have looping defects, reflecting a role for Fgf8 in embryonic and/or thoracic left-right axis determination(Meyers and Martin, 1999). Milder hypomorphs have hypoplastic OFTs and defects reflecting disrupted ventriculoarterial alignment (double outlet right ventricle, transposition of the great arteries) or failed OFT septation (persistent Truncus Arteriosus)(Abu-Issa et al., 2002; Frank et al., 2002).

The source(s) of Fgf8 required for outflow tract formation and subsequent remodeling are unknown. OFT development is normal after Fgf8loss-of-function in pharyngeal ectoderm(Macatee et al., 2003). Early widespread Fgf8 ablation recapitulates the entire spectrum of OFT defects observed in hypomorphs (Brown et al., 2004), whereas later ablation throughout the caudal pharynx causes only a low incidence of an alignment-type OFT defect(Macatee et al., 2003)(A.M.M., unpublished).

In the wake of these studies, several important questions remain. Does the Fgf8 expression domain in the cardiac crescent include both PHF and AHF cells, and what are the cellular targets of Fgf8 signaling in the crescent? What are the source(s) and timing of Fgf8 signals required for OFT formation and for Fgf8-dependent morphogenetic events during subsequent OFT remodeling? Is the source of Fgf8 required for OFT septation separable from that supporting rotation/alignment?

In order to address these questions, we determined the onset and breadth of Fgf8 expression in the cardiac crescent, heart tube and pharynx. By precisely characterizing the activity of a series of Cre drivers relative to these Fgf8 expression domains, we were able to examine and interpret the effects of Fgf8 loss-of-function at specific times, and in particular cell populations, during the earliest stages of cardiogenesis and subsequent OFT remodeling. We found that autocrine Fgf8 signaling is required at the crescent stage to support formation of the heart tube and the RV/OFT,and for looping. Unlike the variable, complex phenotypes of previously reported Fgf8 mutants, our approach generated discrete OFT defects in the face of normal embryonic LR axis specification and revealed specific,required roles for endoderm and mesoderm-derived Fgf8 in different aspects of OFT remodeling.

Fgf8^-, Fgf8^GFP, MesP1Cre,Mef2c-AHFCre and Tbx1^- alleles have been described previously (Jerome and Papaioannou,2001; Macatee et al.,2003; Moon and Capecchi,2000; Saga et al.,1999; Verzi et al.,2005). A novel Fgf8^C conditional allele was generated by inserting a loxP site 5′ to exon 4, and frt-flanked neomycin^r and a loxP site into the 3′ untranslated region. Construction of Isl1^Cre will be detailed elsewhere(C.-L.C. and S.E., unpublished); Cre was inserted in frame into exon 1 of Isl1, creating a Cre knock-in, Isl1knockout.

Primary and secondary antibodies, and immunohistochemical methods have been previously described (Macatee et al.,2003). Anti-Isl1 antibody was described previously(Pfaff et al., 1996). Experiments were repeated a minimum of three times per somite stage.

Embryos were stained for β-galactosidase activity resulting from Cre recombination of the Rosa26lacZ reporter(Soriano, 1999) using standard protocols; counterstaining was with nuclear Fast Red.

Embryos were harvested, fixed and hybridized using a standard protocol(Moon et al., 2006). Mutants and controls were processed in the same vial throughout. Experiments were repeated a minimum of three times; representative results are shown.

To precisely characterize the timing and location of Fgf8expression within crescent mesoderm and pharynx, we used the Fgf8^GFP conditional reporter allele(Macatee et al., 2003) and a universally expressed Cre recombinase, HPRT^Cre(Tang et al., 2002). Production of Fgf8GFP in Fgf8-expressing cells requires Cre-mediated recombination of Fgf8^GFP and Fgf8GFP faithfully recapitulates the expression of the endogenous Fgf8 locus throughout embryogenesis (Boulet et al.,2004; Ladher et al.,2005; Macatee et al.,2003).

In Fgf8^GFP/+;HPRT^Cre/+ embryos,myocardial precursors migrating anterolaterally from the primitive streak initiate Fgf8 expression as they accumulate in the cardiac crescent at late headfold (LHF) to 1-somite stage (ss)(Fig. 1A,B). It has been proposed that Fgf8 expression is limited to the AHF(Buckingham et al., 2005; Kelly and Buckingham, 2002);however, many LV precursors in the ventral heart tube express Fgf8GFP(Fig. 1C-E). At 10ss, Fgf8GFP persists in splanchnic mesoderm (SM) and nascent RV/OFT, but not in the LV or inflow (Fig. 1F,G). Rapid loss of GFP in the LV and inflow (already waning by 6ss, not shown) indicates that the half-life of the Fgf8GFP fusion protein is <10 hours in these cells. In the pharyngeal epithelia, Fgf8GFP is detectable in pouch endoderm and surface ectoderm by 5ss (Fig. 1D,E).

Given the broad expression of Fgf8 in the crescent and early pharynx, and the complex phenotypes of previously reported Fgf8mutants, we assembled a panel of Cre-expressing `drivers' to systematically generate Fgf8 loss-of-function in subsets of myocardial precursor mesoderm and early endoderm. A detailed comparison of the activity of these drivers relative to Fgf8 expression is crucial to interpreting and understanding the consequences of differential ablation of Fgf8; as will become apparent, subtle differences in the timing and location of Fgf8 function profoundly influence ultimate phenotype.

We used MesP1^Cre(Saga et al., 1999) to ablate Fgf8 in all cardiac precursor mesoderm, as evidenced by lacZexpression from the Rosa26-Cre reporter throughout the cardiac crescent of presomite embryos, in the unlooped HT, and in all myocardium of embryonic day(E) 10.5 hearts (Fig. 1H-K)(Saga et al., 2000; Saga et al., 1999). In Fgf8^GFP/+;MesP1^Cre/+ embryos,Fgf8GFP is broadly detected in the crescent at LHF (1ss; Fig. 1A1,B1), in dorsal and ventral cells of the 5ss heart tube (Fig. 1C1,D1), and in AHF-derived RV/OFT(Fig. 1F1,G1). Mesodermal Fgf8GFP expression resulting from HPRTCre and MesP1Cre activity is indistinguishable (compare Fig. 1A-G with 1A1-1G1), indicating that MesP1Cre completely ablates Fgf8 in the precardiac mesoderm.

Fgf8 is expressed in the pharyngeal endodermal pouches, but its role in this tissue is unknown. Because Isl1 is expressed broadly in pharyngeal endoderm and throughout the AHF(Cai et al., 2003), we generated a new, highly efficient `Isl1Cre'-expressing line to determine whether the loss of endodermal Fgf8 contributes to the cardiac or OFT defects observed in previously described Fgf8 mutants. Isl1Cre-mediated recombination of the Rosa26-lacZ reporter is first evident in pharyngeal endoderm and crescent mesoderm at the 1-2ss(Fig. 1L). Its activity expands rapidly such that, by 5ss, the first pouch and adjacent endoderm, and many cells in the ventral heart tube, are lacZ-positive(Fig. 1M-O2).

Fgf8GFP is not reproducibly detected in the cardiac crescent of Fgf8^GFP/+;Isl1^Cre/+ embryos until 2-3ss (Fig. 1A2,B2). Notably,fewer GFP-positive cells are present in the ventral 5ss heart tube of Fgf8^GFP/+;Isl1^Cre/+(Fig. 1C2,D2) than in either Fgf8^GFP/+;MesP1^Cre/+ or Fgf8^GFP/+;HPRT^Cre/+ embryos, suggesting that the Fgf8 expression domain is broader than that of Isl1 (and the AHF)at the crescent stage. At 5ss, Fgf8GFP is already present in pouch endoderm and in scattered ectodermal cells. Endodermal Fgf8GFP expression resulting from Isl1Cre and HPRTCre activity is comparable from 5ss forward, indicating that Isl1Cre ablates Fgf8 throughout the Fgf8 endodermal expression domains (compare Fig. 1D-G with 1D2-G2).

To assess whether Fgf8 is required independently in AHF mesoderm,we employed Mef2c-AHFCre. Consistent with published activity of this driver(Verzi et al., 2005), Fgf8GFP is present in few AHF cells of Fgf8^GFP/+;Mef2c-AHFCreembryos at 2-3ss, but throughout the AHF by 5ss(Fig. 1A3-E3). In contrast to MesP1Cre and Isl1Cre, no Fgf8GFP-positive cells were present in LV precursors in the ventral heart tube of 5ss Fgf8^GFP/+;Mef2c-AHFCreembryos (Fig. 1D3,R).

The later onset and restricted domains of recombination in cardiac mesodermal precursors of Mef2c-AHFCre and Isl1Cre relative to MesP1Cre are summarized in Fig. 1P-S:Fgf8GFP is present in few crescent cells of Fgf8^GFP/+;Isl1^Cre/+ or Fgf8^GFP/+;Mef2c-AHFCre embryos prior to 2-3ss, which represents at least six hours of ongoing mesodermal Fgf8 signaling in these mutants relative to MesP1Cre (Fig. 1A1,B1 compare with A2,B2 and A3,B3). By 5ss, all three drivers are active throughout the SM/AHF, but only Isl1Cre has endodermal activity(Fig. 1R,S).

To generate Fgf8 conditional mutants in the absence of confounding hypomorphic effects, we used a new loxP-containing allele with minimal exogenous sequence (abbreviated Fgf8^C, see Materials and methods). Fgf8^C/C and Fgf8^C/- animals are phenotypically normal and are indistinguishable from Fgf8^C/+ littermates.

Because MesP1^Cre and Isl1^Cre are loss-of-function alleles, we looked for evidence of genetic interaction with Fgf8. Fgf8^C/+;MesP1^Cre/+ embryos were morphologically and molecularly indistinguishable from Fgf8^C/+ or Fgf8^C/- littermates and were present at the expected Mendelian ratios (E8-E11.5: n=79, 105%of predicted; E18.5/newborn: n=30,107% of predicted; Fig. 2A-C,L). Nor did we detect evidence of a genetic interaction between Fgf8 and Isl1 in Fgf8^C/+;Isl1^Cre/+ embryos(Fig. 2F-H,N; E8-E11.5: n=68, 100% of predicted; E18.5/newborn: n=26, 100% of predicted).

Early, complete ablation of Fgf8 function in both heart fields in Fgf8^C/-;MesP1^Cre/+ conditional mutants(Fgf8;MesP1Cre) disrupted HT formation, looping, and accretion of OFT/RV myocardium. These severely affected embryos (66/103, 65%) had hypoplastic heart tubes (Fig. 2E), short or absent OFTs, and a single dilated ventricle and atrium (Fig. 2E′), and died by E10.0 with anasarca and pericardial effusion(Fig. 2M′). This phenotype is also seen after ablation of Fgf8 with Nkx2.5Cre (E. Meyers, unpublished). The lower incidence of this severe phenotype in Fgf8^C/-;Isl1^Cre/+ mutants (26/73, 35%; Fig. 2J,J′,O′)correlates with a later onset and a restricted domain of Isl1Cre activity in crescent mesoderm. By contrast, AHF-limited Fgf8 ablation driven by Mef2c-AHFCre permitted all Fgf8;Mef2c-AHFCre mutants to survive to birth (n=16, 105% of expected) after normal initial formation of the heart tube, OFT and RV (Fig. 2K,K′,P′).

Differences in Fgf8 signaling obtained after MesP1Cre, Isl1Cre and Mef2c-AHFCre ablation were not only manifest in the frequency of cardiac hypoplasia and embryonic death, but also in the OFT remodeling phenotypes of surviving mutants. Thirty-five percent of Fgf8;MesP1Cre mutants survived but had short, narrow OFTs and small RVs at midgestation(Fig. 2D,D′,M). Based on their embryonic phenotypes, we expected these mild mutants to have hypoplastic RVs at birth, but they did not; rather, 30% had D-transposition of the great arteries (TGA, n=15; Fig. 2R). This indicates that OFT rotation/alignment was perturbed by the loss of mesodermal Fgf8 and the resulting deficiency of RV/OFT myocardium at the time of OFT remodeling. Intrathoracic and intrabdominal LR axes appeared to be normal. Forty percent of Fgf8;MesP1Cre newborns also had a bicuspid aortic or pulmonary valve (the LV-associated valve; Fig. 2R′). No OFT or valve defects were observed in Fgf8^C/- or Fgf8^C/+;MesP1^Cre/+ controls(n=28). The importance of post-crescent stage mesodermal Fgf8 for subsequent OFT alignment was confirmed by the occurrence of TGA and double outlet right ventricle (DORV) in Fgf8^C/-;Mef2c-AHFCremutants (25%, n=16; Fig. 2S,S′). These findings allow us to attribute the OFT rotation/alignment and valve defects previously described in Fgf8hypomorphs, Fgf8;Hoxa3IresCre mutants and Fgf8;Tbx1Cremutants (Abu-Issa et al., 2002; Frank et al., 2002; Macatee et al., 2003; Brown et al., 2004)specifically to disrupted autocrine Fgf8 function in AHF mesoderm.

Rather than alignment-type OFT defects, some Fgf8 hypomorphs have persistent Truncus Arteriosus (PTA), in which failed OFT septation leaves a single vessel arising from both ventricles(Abu-Issa et al., 2002; Frank et al., 2002). Septated OFTs in all surviving Fgf8;MesP1Cre and Fgf8;Mef2c-AHFCremesodermal mutants, and after ablation of Fgf8 from pharyngeal ectoderm(Macatee et al., 2003),indicates that septation is independent of these sources and is likely to depend on endodermal Fgf8.

Most Fgf8^C/-;Isl1^Cre/+ mutants survived embryogenesis (E18.5/newborn: n=18, 65% of predicted). At E8.75-E10.5 they had small pharyngeal arches and severe RV hypoplasia, with a markedly altered relationship between the LV and OFT(Fig. 2I,I′,O). Fgf8^C/-;Isl1^Cre/+ conotruncal cushions were hypocellular relative to controls; fusion of the conotruncal cushions to form the AP septum was already underway in E10.5 Fgf8^C/+;Isl1^Cre/+ controls, but not in Fgf8^C/-;Isl1^Cre/+ mutants(Fig. 2N,O). Remarkably, 100%of E18.5/newborn Fgf8;Isl1Cre mutants had PTA. The truncal vessel was abnormally positioned over the RV (Fig. 2T-T″), which is surprising given the severity of RV hypoplasia observed at midgestation.

In order to understand the etiology of the early shared phenotype of MesP1Cre and Isl1Cre mutants, namely the disrupted formation of primary heart tube and OFT/RV, we sought to identify Fgf8-sensitive tissues at crescent and early somite stages and thus examined the expression of Erm. Erm is regulated by Fgf8 and encodes an Ets-domain transcription factor that is an effector of Fgf/Fgf receptor signaling(Brent and Tabin, 2004; Firnberg and Neubuser, 2002; Raible and Brand, 2001). In severe Fgf8;MesP1Cre and Fgf8;Isl1Cre mutants, decreased size and lower intensity of the Erm expression domain are apparent relative to controls and to mildly affected Fgf8;MesP1Cre mutants(Fig. 3A-F). Erm was absent in proximal OFT myocardium, and decreased in SM of both classes of mutants (Fig. 3A′-F′). Endodermal Erm expression was more severely decreased in Fgf8;Isl1Cre mutants, suggesting autocrine Fgf8 activity in the endoderm (Fig. 3F′).

Isl1 is a LIM-homeodomain transcription factor required for the survival and proliferation of AHF mesoderm (Cai et al., 2003). We examined Isl1 expression by in situ hybridization and Isl1 protein production with an anti-Isl1 antibody(Pfaff et al., 1996). Early disruption of Isl1 expression in Fgf8;MesP1Cre and Fgf8;Isl1Cre mutants mirrored those observed in Erm(Fig. 3G-L). This is not attributable to the Isl1 loss-of-function allele in Fgf8^C/-;Isl1^Cre/+ mutants because Isl1 mRNA and protein production are decreased in MesP1Cre mutants(Fig. 3I), and are more severe in Fgf8^C/-;Isl1^Cre/+ mutants than in Fgf8^C/+;Isl1^Cre/+ controls(Fig. 3L,K). Analysis of Isl1 protein revealed both fewer Isl1-producing cells in SM, and less intense Isl1 immunostaining in cells in SM, proximal OFT and endoderm of conditional mutants relative to controls (Fig. 3M-P″). Levels of Isl1 mRNA correlated with the severity of OFT defects observed in E9.5 MesP1Cre or Isl1Cre mutants: embryos dying with severe OFT/RV hypoplasia and incomplete looping had a markedly decreased intensity of Isl1 expression in SM and endoderm(Fig. 3Q-V′).

We investigated whether the survival or expansion of cardiac precursors was decreased by a loss of Fgf8 in the crescent mesoderm and reasoned that PHF/AHF dysfunction must occur early in cardiogenesis in mutants, because abnormal heart tube and OFT morphology, and decreased Isl1 immunostaining were evident at 7-9ss. No reproducible increase in apoptosis was observed in mutant mesoderm at 0-4ss (Fig. 4A-F;data not shown), although apoptosis in neuroectoderm and ventral endoderm of mutants was increased relative to controls after 0ss(Fig. 4C-F). At 7-9ss, we detected excess apoptotic cells in ventral endoderm and adjacent midline SM accumulating to the OFT in both MesP1Cre and Isl1Cre mutants(Fig. 3M-P′, Fig. 4E-J), relative to double heterozygote controls (P=0.004, paired t-test; controls mean=9 cells/high-power field, n=6; mutants mean=22 cells/high-power field, n=4 Fgf8;MesP1Cre and n=5 Fgf8;Isl1Cre mutants).

Anti-PHH3 labels cells in all phases of mitosis(Li et al., 2005). At 0ss,high levels of proliferation were detected in both mutants and controls;notably the mutants had more PHH3-positive cells in the neuroectoderm, but fewer in the mesoderm and endoderm than controls(Fig. 4A,B). 4ss mutants had an average of 46% fewer proliferating cells in crescent mesoderm(Fig. 4E,F; P=0.005,paired t-test; controls mean=66%, n=4; Fgf8;MesP1Cre mutants mean=20%, n=4) and this difference persisted at 9ss. Fewer PHH3-positive cells were also noted in the proximal OFT and pharyngeal epithelia of 9ss conditional mutants(Fig. 4G-J).

To characterize the molecular events associated with disrupted growth of the OFT/RV in Fgf8;MesP1Cre and Fgf8;Isl1Cre conditional mutants, we examined gene expression in the AHF/nascent OFT. Expression of Mef2c is regulated by Isl1 and Foxh1 via AHF-specific enhancers(Dodou et al., 2004; von Both et al., 2004). When RV/OFT-fated myocardium first accrues to the heart (7-8ss in this system),expression of Mef2c in the AHF localizes to the right SM and nascent RV/OFT myocardium. This localization occurred in conditional mutants, but the level of Mef2c mRNA in OFT and SM was markedly decreased, whereas expression in the first pharyngeal arch and inflow was intact(Fig. 5A,A′). We attribute this alteration in Mef2c expression to the demonstrated decrease in Isl1 activity, because Foxh1 expression was intact in Fgf8;MesP1Cre and Fgf8;Isl1Cre mutants (not shown).

Wnt11 influences cell adhesion, migration and polarity(McEwen and Peifer, 2000); it is expressed in truncal myocardium (Cai et al., 2003) and is required for OFT alignment and septation in mice(W. Zhou, L. Lin, A. Majumdar, X. Li, X. Zhang, W. Liu, L. Etheridge, Y. Shi,J. Martin, W. Van de Ven, V. Kaartinen, A. Wynshaw-Boris, A. McMahon, M. G. Rosenfeld and S.E., unpublished). Wnt11 expression correlated with the severity of RV/OFT dysplasia: it was undetectable in severe Fgf8;MesP1Cre and Fgf8;Isl1Cre mutants, decreased in mild variants, and preserved in Fgf8;Mef2c-AHFCre mutants(Fig. 5B,B′).

Disrupted heart looping is a feature of embryonic left-right (LR) axis disruption and it has been proposed that TGA and PTA may be manifestations of globally disrupted LR asymmetry (Bamforth et al., 2004). Fgf8 functions upstream of nodal and Pitx2c in embryonic LR axis determination (Boettger et al., 1999; Fischer et al.,2002; Kioussi et al.,2002; Liu et al.,2002; Meyers and Martin,1999; Schneider et al.,1999; Welsh and O'Brien,2000). We examined Pitx2 expression to assess whether the looping and OFT defects observed in MesP1Cre and Isl1Cre Fgf8conditional mutants reflect the altered specification/maintenance of the embryonic LR axis and found that Pitx2 localized appropriately to the left SM, OFT and atrial myocardium in mild and severely affected mutants at the 10-16ss (Fig. 5C-C″;data not shown). Normal expression of Ventricular myosin light chain 2 (Franco et al., 1999)indicated that myocardial differentiation occurred in the hypoplastic right ventricles of conditional mutants (not shown).

Tbx1 is required for OFT septation, as well as for cell proliferation and Fgf10 expression in the AHF(Arnold et al., 2006; Xu et al., 2004). As Tbx1 and Fgf10 expression were intact in Fgf8;MesP1Cre and Fgf8;Isl1Cre mutants(Fig. 5D,E), disruption of these factors does not cause the observed OFT defects.

Based on the shared phenotypes of Fgf receptor 1 (Fgfr1) and Fgf8 mutants, on recent data indicating that Fgfr1 is activated by Fgf8 in vivo (Moon et al.,2006), and on our unpublished observation that OFT septation is sensitive to Fgfr1 function, Fgfr1 probably mediates Fgf8 signals in target cells required for OFT septation. Moreover, decreased Erm expression in the endoderm of Fgf8;Isl1Cre mutants, which develop PTA, suggests that Fgf8 has an autocrine role in the endoderm required for OFT septation. It has previously been demonstrated that Fgf8 expression is markedly decreased in the endoderm of Tbx1^-/- mutants, in which PTA is attributable to loss of Tbx1 in the endoderm(Xu et al., 2004); thus, we questioned whether loss of Tbx1 also disrupts Fgfr1expression and found a dosage-sensitive decrease in the expression of Fgfr1 in the endoderm and mesenchyme, specifically in the caudal pharyngeal arches of Tbx1 heterozygotes and null mutants(Fig. 6).

Our studies indicate that Fgf8 signaling to mesodermal myocardial precursors is required at presomite stages for the normal expression of Isl1 and Mef2c, and for the proliferation of myocardial precursors during formation of the primary heart tube, RV and OFT. Our results also demonstrate that the sources and targets of Fgf8 required for OFT septation and rotation/alignment are not only temporally distinct from those for heart tube and OFT formation, but are spatially separate from one another(Fig. 7).

Although no markers exist for the PHF, greater numbers of Fgf8GFP-positive LV precursors in the ventral heart tubes of Fgf8^GFP/+;MesP1^Cre/+ and Fgf8^GFP/+;HPRT^Cre/+ embryos than Fgf8^GFP/+;Isl1^Cre/+ embryos strongly suggest that Fgf8 is expressed in both the PHF and the AHF. MesP1Cre reliably ablates early Fgf8 function throughout the cardiac crescent,generating a high percentage of mutants with severe heart tube and RV/OFT hypoplasia (Fig. 7B). Although a small subset of Fgf8;Isl1Cre mutants display this severe phenotype,AHF-restricted mesodermal mutants (Fgf8;Mef2c-AHFCre) do not. The differences in timing and domain of these Cre drivers thus reveal a previously unknown role for autocrine Fgf8 signaling in the crescent mesoderm to support normal formation of the heart tube and AHF-derived tissues. Although we attribute the death of severe mutants to cardiac insufficiency from a decreased pump size and OFT obstruction, abnormal Fgf8 signaling may also disrupt myocardial calcium handling and cardiac function(Farrell et al., 2001).

Loss of Fgf8 in the AHF/SM resulted in decreased Ermexpression within the SM itself, providing additional evidence for autocrine Fgf8 activity in the AHF. Furthermore, we detected altered apoptosis/proliferation and decreased numbers of Isl1- and Mef2c-expressing cells in the AHF of mutants, suggesting that Fgf8 maintains the Isl1/Mef2c transcriptional cascade required for normal survival and accrual of AHF-derived cells to the heart. As Wnt11 expression is undetectable in the OFT myocardium of severe MesP1Cre and Isl1Cre mutants (and in Fgf8;Nkx2.5Cre conditional mutants)(Ilagan et al., 2006), but is preserved in Fgf8;Mef2cAHFCre mutants, we conclude that Wnt11 expression in AHF-derived myocardium depends on autocrine Fgf8 function in the AHF prior to the 2-3 ss.

It is unlikely that the survival of some Fgf8;MesP1Cre mutants is due to incomplete MesP1Cre recombination because, in our experience, and in work reported previously (Saga et al.,1999), this driver is active in all anterior primitive streak-derived mesoderm. Thus, we interpret the survival of some Fgf8;MesP1Cre mutants to indicate that other factor(s) can support heart tube and RV/OFT formation. Fgf10 is a candidate for functional redundancy, as its expression was intact in the mesoderm of these mutants. Experiments testing Fgf8/Fgf10 functional redundancy are underway.

The distinct OFT phenotypes of MesP1Cre, Isl1Cre and Mef2c-AHFCre mutants also reveal domain-specific functions of Fgf8 during OFT remodeling. TGA and DORV, observed in Fgf8;MesP1Cre and Fgf8;Mef2c-AHFCre mesoderm-only mutants,(Fig. 7D), exist along a spectrum of OFT alignment/rotation defects in which the Truncus is septated,but the great vessels are not aligned correctly with the ventricles. Conotruncal septal spiraling, conal involution and positioning, and OFT rotation (Lamers and Moorman,2002) may be affected directly by loss of Fgf8 signaling, or indirectly by an insufficient quantity of OFT myocardium(Yelbuz et al., 2002).

In contrast to TGA/DORV, PTA (observed in 100% of Fgf8;Isl1Cremutants, Fig. 7E) reflects a complete failure of OFT septation. Because PTA was observed in Fgf8hypomorphs and Fgf8;Tbx1Cre mutants (both of which have intact Isl1 loci), it is unlikely that the failed septation in Fgf8;Isl1Cre mutants is due to the additive effects of Fgf8loss-of-function and Isl1 heterozygosity. Furthermore, Isl1^Cre/+ and Fgf8^C/+;Isl1^Cre/+ OFTs are normal. However,formal assessment of the contribution of Isl1 haploinsufficiency to OFT defects observed in Fgf8;Isl1Cre mutants requires a Cre-expressor that ablates Fgf8 in the same temporospatial domains as Isl1Cre.

Notably, in most human and mouse examples of PTA, the Truncus is normally aligned, straddling both ventricles(Bartelings and Gittenberger-de Groot,1989; Xu et al.,2004; Yelbuz et al.,2002). However in Fgf8;Isl1Cre mutants (and Tbx1^-/-) (Xu et al.,2004; Zhang et al.,2005) the Truncus is abnormally aligned, 100% committed to the RV,indicating that the mechanism(s) that disrupt the alignment/rotation after loss of mesodermal Fgf8 in Fgf8;MesP1Cre mutants also operate in Fgf8;Isl1Cre endodermal/mesodermal mutants.

Our conclusion that the endoderm mediates crucial Fgf8 signals to regulate OFT septation is consistent with studies indicating that PTA in Tbx1^-/- mutants results from a loss of endodermal Tbx1,accompanied by a loss of Fgf8 in this tissue(Arnold et al., 2006; Xu et al., 2004). Our finding of decreased Fgfr1 expression in the endoderm of Tbx1^-/- mutants indicates that Tbx1 function influences Fgf signaling at multiple levels, and provides new evidence that such crosstalk regulates OFT septation.

The OFT septum is formed by contributions from the endothelium (by epithelial to mesenchymal transformation) and the cardiac neural crest (CNC),so these are potential paracrine targets of endodermal Fgf8(Fig. 7D,E). However, ablation of Fgfr1 and/or Fgfr2 in premigratory CNC does not cause OFT septation defects, so CNC is not a direct target of endodermal Fgf8 (L. Francis and A.M.M., unpublished) (Ilagan et al., 2006). By contrast, the ablation of Fgfr1 and one copy of Fgfr2 with Isl1Cre generates PTA at high penetrance (L. Francis, S.E. and A.M.M., unpublished). Because the Isl1Cre expression domain includes the mesodermal precursors of OFT endothelium(Fig. 1O1 and data not shown)and endoderm, further tissue-restricted receptor ablation analyses are required to determine whether PTA in Fgf8 and Fgfr mutants results from disrupting paracrine and/or autocrine endodermal Fgf8 signaling.

When considered in the context of earlier studies, these findings provide new insight into the requisite timing of Fgf8 signals in different expression domains. Ablation of Fgf8 with Nkx2.5Cre(Ilagan et al., 2006), or in precardiac mesoderm at E7.75-E8.0 (LHF-1ss), disrupts PHF and AHF function,causing severe phenotypes in Fgf8;MesP1Cre and some Fgf8;Isl1Cre mutants. Later ablation of Fgf8 with a Tbx1Cre transgene (expressed in multiple domains from E8.5) generated both septation and alignment-type OFT defects(Brown et al., 2004). By contrast, only a low incidence of alignment-type defects occurred with Fgf8 ablation after E9.0 using Hoxa3IresCre(Macatee et al., 2003). Thus,Fgf8 signals at E7.75-E8.0 in precardiac mesoderm are required for normal PHF and AHF function, whereas signaling from pharyngeal endoderm and splanchnic mesoderm, required for OFT septation and alignment/rotation, respectively,occurs prior to E9.0. This is an important finding, as invasion of the AP septum and OFT cushion fusion begin at least one day later(Lamers and Moorman,2002).

Pitx2 mutants revealed that local, organ-specific LR axes can be determined independently of the embryonic LR axis and at different morphogenetic stages (Kioussi et al.,2002; Kitamura et al.,1999; Liu et al.,2001; Liu et al.,2002). It has been unclear whether looping defects and pulmonary isomerism of severe Fgf8 hypomorphs(Meyers and Martin, 1999)reflect embryonic LR-axis disruption due to Fgf8 deficiency in the primitive streak or node, or later dysfunction in the SM that perturbs the intrathoracic LR axis. Preserved Pitx2 lateralization in early stage Fgf8;MesP1Cre and Fgf8;Isl1Cre mutants (including those with altered looping) supports the former hypothesis. It remains to be determined whether disruption of a later, cardiac-specific LR pathway contributes to OFT remodeling defects in these conditional mutants.

The survival/proliferation and molecular analyses presented here focused on heart tube formation and the accumulation of RV/OFT myocardium in the absence of mesodermal Fgf8. Further studies are underway to examine these processes during OFT remodeling and to determine the bases of the distinct OFT defects observed in mesodermal versus endodermal Fgf8 mutants. The primitive cardiovascular system must support the entire organism while crucial aspects of its development are still underway and vulnerable to a variety of genetic or environmental insults. Genetic systems that reliably create specific cardiovascular defects are an important step towards ultimately dissecting how structural and hemodynamic inputs interact with, and modify, the genetic and molecular programs that control cardiac morphogenesis in normal and pathological conditions.

We thank Yumiko Saga and Virginia Papaioannou for kindly providing MesP1Cre and Tbx1^- alleles, respectively, Charles Murtaugh for the generous gift of Isl1 antibody, and Jim Martin for the Pitx2 riboprobe. We thank Erik Meyers for discussing unpublished results. We are grateful to Kirk Thomas, Chris Lehman, James Martin and Yukio Saijoh for critically reading the manuscript and suggestions. Tim Macatee and Jaclyn Tygesen provided technical assistance. This work was supported by grants from NIH/CHHD (R01HD044157, A.M.M.) and the AHA (A.M.M.), and from the University of Utah Children's Health Research Center (D.U.F.).