Tbx5 is required for forelimb bud formation and continued outgrowth

The forelimb and hindlimb buds are derived from territories of the lateral plate mesoderm located at defined rostrocaudal levels along the main body axis(Capdevila and Izpisua Belmonte,2001). Cells of the lateral plate mesoderm within these limb`fields' respond to axial cues that initiate limb bud outgrowth and subsequently produce a morphologically distinct limb bud by stage 16 in the chick (Hamburger and Hamilton,1951). Classical embryological experiments in the chick have demonstrated that limb-type specification occurs prior to the initiation of overt limb bud outgrowth. Grafts of lateral plate tissue from limb forming regions from as early as stage 8 (four-somite stage) will develop into limbs of the appropriate type – wing or leg – when grafted to the flank or coelomic cavity (Chaube,1959; Rudnick,1945; Wolff,1934).

Several genes have been identified that are expressed exclusively in either the developing forelimb or hindlimb. The T-box transcription factor Tbx5 is first detected in the prospective forelimb mesenchyme prior to overt limb bud outgrowth and this limb-type restricted expression pattern is maintained throughout limb development stages(Gibson-Brown et al., 1996; Isaac et al., 1998; Logan et al., 1998; Ohuchi et al., 1998). A closely related T-box gene, Tbx4, and a paired-like homeodomain factor, Pitx1, are both expressed in a reciprocal pattern in the developing hindlimb mesenchyme(Gibson-Brown et al., 1996; Isaac et al., 1998; Lanctot et al., 1997; Logan et al., 1998; Ohuchi et al., 1998). Misexpression experiments in the chick have demonstrated that ectopic expression of Tbx5 in the leg bud is capable of transforming the hindlimb to a more forelimb character(Takeuchi et al., 1999). Conversely, misexpression of Tbx4 or Pitx1 in the developing wing bud is capable of transforming the forelimb to a more hindlimb character(Logan and Tabin, 1999; Rodriguez-Esteban et al.,1999; Takeuchi et al.,1999). Direct evidence for a role of Tbx5 in forelimb development has been provided by the discovery that mutations in human TBX5 cause Holt-Oram Syndrome (HOS, OMIM 142900), a dominant disorder characterised predominantly by upper(fore) limb defects and heart abnormalities (Basson et al.,1997; Li et al.,1997). Targeted deletion of Tbx5 in the mouse has demonstrated that this gene is essential for normal heart development(Bruneau et al., 2001). Tbx5-null embryos die at or around embryonic day (E) 10 because of the severity of the heart defects. Diminished TBX5 function in human,however, does not obviously affect limb-type identity but instead produces deletion deformities (Basson et al.,1994). Therefore, Tbx5 may have roles related to growth and differentiation of the embryonic limbs that are distinct from, or intrinsically linked to, its role in defining limb-type identity.

To examine the role of Tbx5 in forelimb development we have undertaken two strategies to disrupt its function in the developing limb bud. We have used a conditional knockout strategy to delete Tbx5 function in the developing limbs while leaving the gene intact in other areas of the developing embryo. This approach avoids the complication of phenotypes arising from Tbx5 loss-of-function in regions of the embryo other than the limb, in particular the heart. The second approach involves using avian retroviruses to misexpress dominant-negative Tbx5 constructs to knock down Tbx5 function in the developing wing bud. As a complementary strategy, we also misexpressed full-length and dominant-active forms of the gene.

Mouse embryos were staged according to Kaufman(Kaufman, 1992). Noon on the day a vaginal plug was observed was taken to be E0.5 days of development. The mouse lines carrying a conditional allele of Tbx5(Bruneau et al., 2001) and a Prx1Cre transgene (Logan et al.,2002) have been described previously. Fertilised chicken eggs(Needle's Farms, Winter's Farms) were incubated at 37°C and staged according to Hamburger Hamilton (HH)(Hamburger and Hamilton,1951).

PCR analysis to genotype pup tail and embryonic material (E10, 30 somites)was carried out in a single reaction using three primers that identify the endogenous Tbx5 allele, and both the conditional (floxed) and deleted(floxed-out) Tbx5 allele (Bruneau et al., 2001).

Cloning of retroviral constructs and production of concentrated retroviral supernatants were carried out as described previously(Logan and Tabin, 1998). The Tbx5Δ construct contains amino acids 1-274 of the full-length chick Tbx5 clone (Accession Number AF069396). The Tbx5^en construct contains amino acids 1-274 of the full-length Tbx5 clone fused to amino acids 2-298 of the Drosophila Engrailed protein(Jaynes and O'Farrell, 1991). The Tbx5^vp16 contains amino acids 1-274 of Tbx5fused to a duplex of the lambda hinge region and VP16(Ohashi et al., 1994). Cells of the prospective forelimb were infected between HH stages 8-10 with concentrated viral supernatants as previously described(Logan and Tabin, 1998).

Whole-mount in situ hybridisation was carried out essentially as previously described (Riddle et al.,1993). A minimum of two mutant embryos were analysed at each stage described with each probe. Most probes have been described previously: chick Shh (Riddle et al.,1993), mouse Shh(Echelard et al., 1993), chick Msx (Ros et al.,1992), chick Hoxc4(Nelson et al., 1996), mouse Fgf10 (Bellusci et al.,1997), mouse Pea3(Chotteau-Lelievre et al.,2001), mouse Fgf8(Crossley and Martin, 1995),mouse Tbx4 (Bruneau et al.,2001), mouse Pitx1(Logan and Tabin, 1999) and chick Fgf8 (Vogel et al.,1996). A fragment of the chick Lhx9 sequence was isolated from a chick plasmid library (Logan et al., 1998) and its identity confirmed by sequencing and comparison with published sequences. A chick Groucho homologue Grg4 was generously provided by Johan Ericson (Muhr et al., 2001). Section in situ hybridisation was performed on 20μm paraffin wax-embedded sections of stage 21 chick embryos. Additional chick Groucho genes were cloned by degenerate PCR from chick limb cDNA prepared from HH stages 20-23. The degenerate PCR primers lie in the highly conserved Q domain and WD40 domains of Drosophila, mouse and human Groucho genes: 5′-AA-RACIGARATGCARMGICAY-3′,5′-IGCYTCICCICCIACDATIAR-3′. PCR using a 45°C annealing temperature yielded multiple products, including a 1.1 kb fragment that was cloned into the pGEM-T vector (Promega). Sequencing of this clone revealed similarity to human TLE3.

The cartilage and bone elements of newborn mouse pups were stained with Alcian Blue and Alizarin Red, respectively, essentially as described previously (Hogan et al.,1994). Apoptotic cell death was assayed with TdT-mediated dUTP nick end labelling (TUNEL). Mouse embryos were fixed overnight in 4%paraformaldehyde and then processed in whole mount using TUNEL reagents(Q-BIOgene) following the manufacturer's protocol. Chick embryos were fixed overnight in 4% paraformaldehyde, washed in PBT and embedded in OCT (BDH,Merck). Transverse sections (12 μm) were assayed by TUNEL according to the manufacturer's protocol. To detect cells in mitosis, a rabbit anti-phosphorylated histone H3 primary antibody (Upstate Biotechnology) and Cy3-conjugated goat anti-rabbit IgG secondary antibody (Jackson Laboratory)were used following the protocol described previously(Yamada et al., 1993).

To examine the function of Tbx5 in forelimb development, we genetically inactivated the gene in the limb buds by generating mice carrying the Tbx5 conditional allele, Tbx5^lox/lox(Bruneau et al., 2001) and a transgene that expresses Cre recombinase in the limb bud, Prx1Cre (Logan et al.,2002). Previous work using the Z/AP reporter indicated that the Prx1Cre transgene is capable of inducing recombination of floxed alleles in the forelimb by E9.5. Cre-catalysed recombination of the Tbx5 conditional allele in Tbx5^lox/lox;Prx1Creembryos was confirmed by PCR (Fig. 1D). The deleted allele was present in DNA samples prepared from hindlimb buds at E10 (30 somites). Mice in which Tbx5 has been deleted in the developing limb die perinatally but it was not possible to identify the exact cause of mortality. Analysis of the gross anatomy indicated there were no obvious abnormalities of any of the internal organs. In all examples, however, the ribcage failed to fuse and the sternum was absent. This corresponds to regions of the embryo in which both Tbx5 and the Prx1-Cre transgene are expressed at later embryonic stages (data not shown). It is likely that, as a result of these defects, the mutant pups cannot breathe independently and that this is the most probable cause of death. Survival of Tbx5^lox/lox;Prx1Cre pups throughout the entire gestation period extends the developmental time window for studying the loss-of function phenotype in the limb. The forelimbs of newborn (P0) pups are completely absent although the hindlimbs are unaffected (Fig. 1A,B). In some examples a small, rudimentary flap of skin is present on one side of the embryo where the forelimb would have formed(Fig. 1B). In the majority of cases, however, the skin where the forelimb would normally form is uniform and indistinguishable from the rest of the inter-limb flank. All skeletal elements of the forelimb, including the elements of the pectoral girdle, the scapula and clavicle, are clearly identifiable in control littermates(Fig. 1E-G) but are absent in Tbx5^lox/lox;Prx1Cre P0 pups(Fig. 1H-J).

Scanning electron micrographs of control and Tbx5^lox/lox;Prx1Cre embryos indicate that the limb defect is manifest by embryonic day 10.5 (E10.5) (Fig. 2B,C). In wild-type E10.5 embryos, the limb bud is a prominent outgrowth from the flank of the embryo and the apical ectodermal ridge (AER) is clearly visible(Fig. 2A). In Tbx5^lox/lox;Prx1Cre embryos, no limb outgrowth is visible and no morphologically distinguishable AER is present(Fig. 2B). In most cases, the region where the forelimb would normally form is indistinguishable from other regions of the embryo flank. In a minority of examples, a small mass of tissue is present in the forelimb region (Fig. 2C). This vestige of the forelimb bud most probably forms in examples where the Cre recombinase has failed to delete Tbx5 function completely from every cell of the prospective forelimb mesenchyme. This small mass of tissue may ultimately give rise to the small flap of tissue occasionally observed in Tbx5^lox/lox;Prx1Cre P0 pups(Fig. 1B).

Although analysis of E10.5 embryos clearly demonstrated a loss of a morphological forelimb in Tbx5^lox/lox;Prx1Cre embryos, we were curious to learn if molecular markers of the limb were expressed in the forelimb region. Limb bud formation involves the establishment of key signalling centres within the nascent bud. Classical embryological experiments in the chick have demonstrated the requirement of the AER for limb bud outgrowth. Members of the fibroblast growth factor (Fgf) family expressed in the AER have been shown to mediate the effects of this signalling centre(Fallon et al., 1994; Niswander et al., 1993). At E9.5, Fgf8 is expressed in the nascent AER in both the forelimb and hindlimb and provides a sensitive molecular marker of this structure(Fig. 2D). In the same stage Tbx5^lox/lox;Prx1Cre embryo, Fgf8 expression is not detected in the forelimb region(Fig. 2E). Expression of Fgf8 is detected in the AER of the hindlimb bud, consistent with the normal development of the hindlimb in Tbx5^lox/lox;Prx1Crepups (Fig. 2E). The secreted signalling molecule Shh expressed by cells of the zone of polarising activity (ZPA) is essential for normal limb patterning(Echelard et al., 1993; Riddle et al., 1993). At E10.5, Shh is expressed in the distal posterior of the limb buds, in the cells of the ZPA (Fig. 2F). In Tbx5^lox/lox;Prx1Cre embryos, however, Shh is not expressed in the forelimb region, although normal expression is detected in the hindlimb bud (Fig. 2G).

Previous studies in the chick have suggested that Tbx5 expressed in the prospective forelimb region may repress expression of the closely related gene Tbx4 which is normally restricted to the hindlimb(Takeuchi et al., 1999). We were therefore interested to determine whether, in the absence of Tbx5, hindlimb markers would be ectopically expressed in the forelimb region. In Tbx5^lox/lox;Prx1Cre embryos, Tbx4expression is detectable in the hindlimb but is not detected in the forelimb region at E9.5 or E10.5 (Fig. 2I; data not shown). Similarly, Pitx1 expression is restricted to the hindlimb in the mouse(Fig. 2J). After deletion of Tbx5 in the forelimb, Pitx1 expression remains restricted to the hindlimb and is not expressed in the forelimb region at E9.5 or E10.5(Fig. 2K; data not shown). Taken together, these data suggests that Tbx5 does not normally function to repress the expression of hindlimb markers, Tbx4 or Pitx1, in the forelimb region as, in the absence of Tbx5,the expression patterns of Tbx4 and Pitx1 are unaffected. These results are in agreement with similar studies in which Tbx5function has been deleted in all cells of the developing embryo(Agarwal et al., 2003).

The absence of any morphological forelimb bud led us to suspect a defect at earlier stages of limb development. Fgf10 is expressed in the lateral plate mesoderm (LPM) of the prospective limb field prior to the expression of Fgf8 in the prospective AER(Ohuchi et al., 1997). Fgf10 and other members of the Fgf family, are capable of inducing ectopic limb bud formation when applied to cells in the interlimb flank(Cohn et al., 1995; Martin, 1998). The functional importance of Fgf10 in limb formation was further demonstrated by the observation that mice carrying a null mutation in Fgf10 fail to form an AER and pups are born with severely truncated limbs(Min et al., 1998; Sekine et al., 1999). In the absence of Tbx5 function, Fgf10 is not expressed in the prospective forelimb bud mesenchyme by E9.5 (21 somites) (Fig. 3B,D). Pea3 is an Ets-related transcription factor that is expressed in the prospective limb mesenchyme(Chotteau-Lelievre et al.,2001) in a complementary pattern to Fgf10(Fig. 3E). It has been proposed to mediate the nuclear response to Fgf signalling directly(Raible and Brand, 2001) and it therefore provides a molecular read-out of Fgf signalling. Pea3 is not expressed in the forelimb region of Tbx5^lox/lox;Prx1Cre embryos by E9.5 (21 somites)(Fig. 3F) consistent with a failure of Fgf signalling. Pea3 is also expressed in the intermediate mesoderm lateral and caudal to the forelimb(Fig. 3E). This expression domain is not affected in Tbx5^lox/lox;Prx1Cre embryos(Fig. 3F), indicating that the effect on Pea3 is limited to the cells of the prospective forelimb and is not affected at other sites in the developing embryo. TdT-mediated dUTP nick end labelling (TUNEL) analysis demonstrated that by E9.5 (24 somites)cells in the prospective forelimb region of Tbx5^lox/lox;Prx1Cre embryos were first detected to be undergoing an increased extent of apoptotic cell death(Fig. 3H) when compared with control embryos at a similar stage (Fig. 3G). By E10.5 the domain of cell death in Tbx5^lox/lox;Prx1Cre embryos was extensive throughout the forelimb forming region (Fig. 3J) compared with control embryos(Fig. 3I). These results demonstrate that as a result of Tbx5 inactivation, Fgf10 is not expressed and cells of the prospective forelimb subsequently undergo extensive apoptosis.

During normal development, Tbx5 is expressed throughout the limb mesenchyme from the earliest stages of forelimb development(Gibson-Brown et al., 1998; Logan et al., 1998) consistent with a role for Tbx5 in forelimb bud initiation. This limb-type restricted expression pattern is also maintained during later limb bud development stages, suggesting Tbx5 may have additional roles during subsequent stages of limb outgrowth and patterning. To examine these possible later roles of Tbx5, we injected avian retroviruses into chick embryos to misexpress dominant-negative and dominant-active forms of Tbx5 in the developing forelimb. Misexpression of dominant-negative Tbx5 using this method should lead to a disruption of Tbx5function at later limb bud stages than the genetic deletion approach we used in the mouse. Tbx5 is a transcription factor that, by analogy to other members of the Tbx family, is thought to mediate its effects by binding to target sites on DNA via the conserved N-terminal T domain. The transcriptional effector domain resides in the C terminus of the protein(Fig. 4G). Mutations that lead to truncations of the C terminus of TBX5 have been identified in individuals with HOS, indicating that this region is essential for TBX5 protein function. We have generated two putative dominant-negative constructs of Tbx5:a truncated construct that contains residues encompassing the N terminus and T domain only (Tbx5Δ); and a construct that contains the N terminus and T domain fused directly to the transcriptional repressor domain of the Drosophila Engrailed protein(Jaynes and O'Farrell, 1991)(Tbx5^en) (Fig. 4G). Such fusion proteins have been demonstrated to function as dominant-negative constructs in a range of tissues and organisms(Markel et al., 2002; Yu et al., 2001). It should be noted however that the engrailed `active' repressor domain has been shown to be Groucho dependent (Jimenez et al., 1997). We therefore examined the expression of Groucho homologues during chick and mouse limb development. We find that several,including Groucho 1, Groucho3 and Groucho4, are widely expressed in the developing limb (Fig. 4H; data not shown) giving confidence that the engrailed`active' repressor domain is likely to function in the limb. Both dominant-negative constructs would be expected to compete with endogenous Tbx5 protein for DNA-binding sites upstream of Tbx5 target genes. The truncated construct would fail to activate expression of target genes, while the Engrailed-fusion protein would directly repress gene expression. In our experiments, misexpression of both dominant-negative constructs produced indistinguishable results (n=144), suggesting that both forms had similar effects of blocking endogenous Tbx5 gene function. Scanning electron micrographs of wing buds injected with dominant-negative forms of Tbx5 indicate that the injected limb is severely truncated when compared with an uninjected control wing bud (Fig. 4A,B; n=2/2, 100%). In contrast to the phenotype observed after deletion of Tbx5 in the mouse, a limb bud does form in the chick. We presume that this difference is due to incomplete knockdown of gene function using this retroviral misexpression technique. We examined the relative rates of cell proliferation in control and dominant-negative Tbx5 injected limb buds by assaying for phosphorylated histone H3 (PH3), a marker of mitotic cells. No differences were found in the frequency of mitotic cells of injected limb buds (Fig. 4D; n=3/4, 75%) compared with the contralateral uninjected control limb bud (Fig. 4C) when regions containing the same total cell number were compared. The extent of programmed cell death was examined using TUNEL staining. An increased frequency in cell death was observed in injected limb buds(Fig. 4F; n=3/4, 75%)compared with the contralateral uninjected control limb bud(Fig. 4E). Taken together,these results indicate that injected limb buds are smaller because of increased cell death and not because of a decrease in the rate of cell proliferation. These data are consistent with our observations in Tbx5^lox/lox;Prx1Cre mouse embryos and Fgf4/Fgf8 double knockout embryos that have smaller limbs as a consequence of increased cell death, rather than as a result of a reduction in cell proliferation (Sun et al.,2002). A common feature in the generation of these similar phenotypes is the disruption of Fgf signalling.

To understand how normal patterning is affected in the truncated forelimbs after dominant-negative Tbx5 misexpression, we analysed the expression patterns of two of the signalling centres of the limb. We observe that Fgf8 expression is downregulated in the anterior AER and Shh is absent in the injected wing buds(Fig. 5B; n=2/9, 22%)compared with control embryos (Fig. 5A). If injection of the retrovirus is performed at later stages,lower-level infection of the wing bud results (Rallis and Logan, unpublished)(Logan and Tabin, 1998) and we observe more subtle effects on Fgf8 and Shh expression (Fig. 5C,D). In these examples, Fgf8-expressing cells are still present in the distal, anterior of the limb but in a broader, flattened domain of the ectoderm that has not formed a ridge (Fig. 4D,F; n=4/18, 22%). This broad pattern of Fgf8 expression is reminiscent of cells of the immature, nascent AER present in earlier stage limb buds, suggesting a delay of AER formation or a failure to maintain the AER. We conclude that the effect on Fgf8 expression in the ectoderm must be occurring as a secondary effect to the knock down of Tbx5function in the limb mesenchyme because Tbx5 is expressed in the forelimb mesenchyme but not the ectoderm. Furthermore, the injection protocol leads to dominant-negative Tbx5 misexpression in the limb mesenchyme that does not spread to the overlying ectoderm (data not shown)(Logan and Tabin, 1998). Indirect action of misexpressed, mesenchyme-restricted Tbx5 on the ectoderm is consistent with the normal mesenchymal-restricted expression of Tbx5 (Gibson-Brown et al.,1998; Logan et al.,1998). Strikingly, the effect on Fgf8 expression in the AER was observed predominantly in the anterior of the limb bud (Fig. 5B,D).

To investigate the effect of the knock-down of Tbx5 activity in the limb in more depth, we analysed the expression of other markers of the limb mesenchyme. Msx1 is a homeobox gene expressed in the AER and in the distal-most mesenchyme cells that lie just under the AER at early limb bud stages (Fig. 6A)(Ros et al., 1992). Msx1 has been considered a sensitive marker for the influence of the AER on the distal mesenchyme cells as Msx expression is rapidly lost after AER removal (Ros et al.,1992). This occurs more rapidly than the reported wave of apoptosis in the distal mesenchyme that follows AER removal(Dudley et al., 2002; Rowe et al., 1982), although some of the observed downregulation of Msx1 may be attributable to cell death. In the dominant-negative Tbx5-injected wing bud, Msx1 expression is downregulated in the anterior, whereas the expression pattern was not significantly affected in the posterior of the limb(Fig. 6B; n=2/5, 40%). To investigate the apparent anterior bias of the limb truncations produced by the dominant-negative Tbx5 constructs, we analysed the effect of expression of markers of the anterior limb mesenchyme. Lhx9, a homeodomain factor, is normally expressed in the anterior of the limb bud(Fig. 5C)(Retaux et al., 1999). A functional role of Lhx9 in limb development has not been established(Birk et al., 2000) but for our purposes it serves as a marker of anterior limb mesenchyme. After misexpression of dominant-negative Tbx5, the expression domain of Lhx9 is reduced in the anterior mesenchyme(Fig. 6D; n=5/9, 56%). The homeobox gene Hoxc4 is also expressed in the anterior of the forelimb bud (Fig. 6E)(Nelson et al., 1996);however, this expression domain is lost after expression of dominant-negative Tbx5 (Fig. 6F; n=4/6, 67%). By contrast, expression of snail is not affected (4/4, 100%; data not shown) after expression of dominant-negative Tbx5. These results confirm our earlier observations that knock down of Tbx5 function after misexpression of Tbx5dominant-negative constructs predominantly affected the anterior mesenchyme of the limb.

As a complementary strategy to misexpressing dominant-negative forms of Tbx5, we also generated dominant-active constructs by fusing the N-terminal region of Tbx5, including the DNA-binding T domain, to the VP16 transcriptional activation domain (Fig. 4G) (Ohashi et al.,1994). This fusion construct would be expected to bind to the endogenous DNA-binding sites upstream of Tbx5 target genes and to activate their expression. After misexpression of dominant-active Tbx5, the wing bud is enlarged and expanded anteriorly (data not shown, Fig. 6H,J). Analysis of Fgf8 expression indicates that the domain of Fgf8-expressing cells in the AER extends more anteriorly than in the uninfected wing bud, but is not expanded in the posterior of the bud (Fig. 6G,H; n=5/7, 71%). The domain of Lhx9 expression is expanded anteriorly but not posteriorly in the wing bud (Fig. 6I,J; n=4/5, 80%). A similar anterior expansion of the expression domains of Fgf10 (n=2/2; 100%) and Msx (n=3/3;100%) was also observed (data not shown). Similar anterior expansions of the limb mesenchyme were also obtained after misexpression of full-length Tbx5 constructs (n=17, data not shown). This finding and phenotypes observed with dominant-negative forms of Tbx5 support the conclusion that Tbx5 is acting as a transcriptional activator in the developing limb bud.

Tbx5 is expressed in the prospective forelimb mesenchyme of the lateral plate prior to overt limb bud outgrowth and coincident with the time that these cells are specified to their limb-type fate(Gibson-Brown et al., 1998; Isaac et al., 1998; Logan et al., 1998; Ohuchi et al., 1998). Tbx5 expression precedes that of Fgf10(Agarwal et al., 2003) and Tbx5 transcripts are present in the limb-forming region of Fgf10 knockout mice (Min et al.,1998; Sekine et al.,1999). By contrast, when Tbx5 is inactivated in all cells of the embryo prior to limb bud stages, Fgf10 is never expressed demonstrating that Tbx5 is required to initiate Fgf10expression in cells of the prospective forelimb(Agarwal et al., 2003). Moreover, inactivation of Tbx5 in cells of the prospective forelimb at early limb induction stages, E9-E9.5, using the Prx1Cre transgenic leads to loss of Fgf10 expression and the failure to form a morphological limb bud. Together, these data establish a genetic hierarchy required for forelimb initiation. Tbx5 is genetically upstream of the Fgf10 and is essential for the induction and maintenance of Fgf10 that, in turn, is required for the initiation of forelimb bud formation. TUNEL staining of mutant embryos indicated that failure to establish the positive-feedback loop of Fgf signalling between the limb mesenchyme and overlying ectoderm cells leads to programmed cell death in the prospective forelimb-forming region. Tbx5 is therefore the earliest known marker of the prospective forelimb mesenchyme and is essential for forelimb bud formation.

A striking additional observation of conditional knockout of Tbx5in limb mesenchyme is the complete absence of all the elements of the appendicular skeleton. Tbx5 is required for the formation of the clavicle and scapula of the pectoral girdle in addition to the skeletal elements of the limb proper. This phenotype is more severe than that observed in the forelimb of Fgf10-null mice. This indicates that Tbx5has a broader influence on forelimb development than Fgf10 and is absolutely required for the formation of all elements derived from the forelimb lateral plate mesoderm. Although all skeletal elements of the limb proper and most of the pectoral girdle are formed from lateral plate tissue of the prospective forelimb region, the proximal portion of the scapula blade is derived from the somites that lie medial and adjacent to the forelimb(Burke, 2000; Huang et al., 2000). After limb ablation in the chick, the somite-derived hypaxial myoblasts that form the limb musculature, but do not express Tbx5, are never recruited to the limb field (Gumpel-Pinot et al.,1984). By extension, the most likely explanation for the loss of the entire scapula in Tbx5^lox/lox;Prx1Cre embryos is that,after deletion of Tbx5 in the limb mesenchyme and the failure of early limb bud formation, the somite-derived (and at that stage non-Tbx5-expressing) scapula precursors are not recruited into the limb field.

Recently, the requirement of tbx5 for the formation of the pectoral fin in zebrafish has been demonstrated using morpholino antisense oligonucleotides to knock down tbx5 function(Ahn et al., 2002; Garrity et al., 2002). This observation is consistent with work in other species by ourselves and others(Agarwal et al., 2003; Ng et al., 2002). In one report (Ahn et al., 2002), the authors conclude that the function of tbx5 in the development of the zebrafish forelimb involves the directed migration of lateral plate mesodermal cells to the future limb-bud-forming region. However, the results of our experiments in the chick and mouse do not support a model in which Tbx5 is involved in directing migration of cells of the prospective forelimb. In higher vertebrates, cells of the prospective forelimb do not undergo a similar migration (Saunders et al., 1957; Saunders et al.,1959; Searls,1967; Searls and Janners,1969) and Tbx5-expressing cells detected in the prospective forelimb region are not migratory(Gibson-Brown et al., 1998; Isaac et al., 1998; Logan et al., 1998; Ohuchi et al., 1998). Somite-derived cells that contribute to the limb do not express Tbx5(data not shown). Furthermore, in Tbx5-null embryos, patterning of the lateral plate mesoderm at limb levels is intact and only limb bud outgrowth is affected by constitutive loss of Tbx5(Agarwal et al., 2003). Although our results and those of Ahn et al. are phenotypically similar, our results demonstrate a different mechanism for Tbx5 action. Although in lower vertebrates tbx5 may have a role in migration of limb precursor cells, in higher vertebrates Tbx5 is required to regulate inductive signalling interactions essential for limb bud initiation and continued limb outgrowth.

Our misexpression experiments in the chick demonstrate that Tbx5is not only required for limb initiation but is also required at later stages of limb development for continued outgrowth. Knock down of Tbx5function by misexpression of dominant-negative forms of the protein leads to the disruption of the induction and maintenance of the AER. This role for Tbx5 can be placed within our current models of limb initiation and outgrowth(Fig. 6K). Fgf10expressed in the lateral plate mesoderm is initially required for the induction of Fgf8 in cells of the nascent AER(Fig. 6K)(Ohuchi et al., 1997; Yonei-Tamura et al., 1999). Fgf8 expressed by cells of the AER is required, in turn, for the maintenance of Fgf10 in distal mesenchyme underlying the AER. After AER formation, a positive feedback loop is established between Fgf8-expressing cells in the AER and cells of the distal mesenchyme expressing Fgf10 (Fig. 6K) (Ohuchi et al.,1997). Our results are consistent with a requirement for Tbx5 in the AER-mediated maintenance of Fgf10 expression in the distal mesenchyme at later stages of development. Disruption of this epithelial-mesenchymal induction loop would lead to severe truncation of limb outgrowth and could explain the reduction deformities observed in HOS which,in the most severe cases, results in an almost complete absence of all elements of the limb (phocomelia). The more commonly observed characteristic features of HOS phenotypes are deformities of anterior elements of the limb,such as the thumb, thenar elements or radius. After knock down of Tbx5 by misexpression of dominant-negative constructs we observed defects in AER formation and maintenance primarily in the anterior of the infected wing bud. Loss of Fgf8 expression is observed in the anterior AER and a concomitant failure of Fgf-mediated signalling to the underlying mesenchyme is indicated by the loss of expression of various markers in the anterodistal mesenchyme. Our chick misexpression protocol generated a phenotype consistent with characteristic abnormalities presented in individuals with HOS. The downregulation of these anterior markers therefore provide a molecular context to understand the genesis of HOS deformities. Unfortunately, we are unable to analyse the skeletal deformities that would have resulted from the disruption of gene expression at early limb bud stages, because continued spread of the retrovirus produces heart defects that lead to embryonic lethality (data not shown). A direct, positive regulatory relationship between a T-box gene and an Fgf gene has previously been demonstrated between brachyury and eFgf during mesoderm induction in Xenopus(Schulte-Merker and Smith,1995). Our results suggest that this regulatory relationship has been conserved and reused in the context of the limb. This is consistent with results demonstrating that Tbx5 binding sites are present in the promoter of Fgf10 (Agarwal et al.,2003; Ng et al.,2002). Tbx4, a closely related T-box family member(Agulnik et al., 1996), may play an analogous role to Tbx5 in the hindlimb.

Misexpression of both types of dominant-negative Tbx5 constructs produced essentially identical results and are consistent with defects observed because of haploinsufficiency of TBX5 in HOS(Basson et al., 1997; Li et al., 1997) and defects in the Tbx5 heterozygous knockout mouse(Bruneau et al., 2001). In the converse experiment, misexpression of full-length Tbx5 or constructs containing the T-domain of Tbx5 fused to the VP16 transcriptional activation produced identical results. Moreover, the phenotypes observed complement those observed with dominant-negative constructs. Together, these observations support the conclusion that Tbx5 is acting as a transcriptional activator in the forelimb. These results are consistent with reports demonstrating that Tbx5 can transactivate expression from constructs containing a region of the Fgf10 promoter(Agarwal et al., 2003; Ng et al., 2002).

The results of loss-of-function studies in mouse, chick and zebrafish,combined with previous misexpression studies performed in the chick reveal dual roles for Tbx5 in limb development. Tbx5 is required to induce and maintain Fgf10 expression that in turn is necessary for forelimb initiation and continued outgrowth(Fig. 6K). Tbx5 also acts to specify forelimb identity by influencing the response of cells of the forelimb to common patterning cues(Takeuchi et al., 1999). As Tbx5 is required for limb initiation and outgrowth, loss-of-function approaches have not been informative as to the role of Tbx5 in specifying limb-type identity. Further studies will be required in which the requirement for Tbx5 in limb outgrowth can be uncoupled from its role in specifying limb-type identity.

We thank L. Hurst at the Electron Microscopy core (NIMR) for producing the scanning electron micrographs, P. Mealyer and A. Hewett of the Biological Services section (NIMR) for animal husbandry, and K. Mathers and O. Rufian of the Procedural Services section (NIMR) for assistance re-deriving mouse lines. We thank Hans-Georg Simon for collaboration in the original cloning of chick Tbx5. We thank T. Ogura and J.-C. Izpisua-Belmonte for communicating results prior to publication, and T. Heanue for critical reading of the manuscript. B.G.B. is funded by the Canadian Institutes of Health Research and the Heart and Stroke Foundation of Ontario. C.J.T. and C.E.S. are funded, in part, by a joint grant from NIH (P50 HL61036). C.R., J.D-B. and M.P.O.L. are funded by the Medical Research Council (UK). M.P.O.L. is an EMBO Young Investigator.