Tbx5 is dispensable for forelimb outgrowth

The forelimb (FL) and hindlimb (HL) develop from outgrowths of the lateral plate mesoderm (LPM) at precise positions along the body axis. Although the precise hierarchical relationships are not completely understood, many of the genes that participate in the initiation and subsequent patterning of the limb bud have been identified and their activities analysed. Two T-box transcription factors, Tbx5 and Tbx4, which are expressed in the FL and HL, respectively, have recently been shown to play crucial roles in the formation of each limb type.

Experiments in a range of vertebrate model systems have shown that Tbx5 is required for FL and heart development (e.g. Agarwal et al., 2003; Ahn et al., 2002; Rallis et al., 2003). Using a conditional knockout allele of Tbx5, it has been demonstrated that when the gene is deleted prior to, or during, limb initiation, no FL initiates(Agarwal et al., 2003; Rallis et al., 2003). Similarly, deletion of Tbx4 in the mouse leads to failure of HL development, although the phenotype is not as profound as that seen in the FL in the absence of Tbx5 (Naiche and Papaioannou, 2003).

Although the precise hierarchy remains unclear, and may vary between species, Tbx5 and Tbx4 play pivotal roles during FL and HL initiation,respectively, by activating Fgf10 in the limb mesenchyme(Agarwal et al., 2003; Logan, 2003; Naiche and Papaioannou, 2003; Rallis et al., 2003). Fgf10,in turn, activates Fgf8 in the apical ectodermal ridge (AER), a distinct strip at the distal extreme of the overlying ectoderm. Fgf8 signaling from the AER to the underlying mesoderm is subsequently required for the maintenance of mesenchymal Fgf10 expression, thereby creating a positive-feedback loop of FGF signaling between the two tissue layers. Tbx5 and Tbx4 appear to play analogous roles in the FL and HL, respectively. Using a gene deletion and replacement strategy, the FL deficiency defect of the Tbx5 conditional knockout can be rescued by Tbx4, demonstrating that Tbx4 has the ability to carry out the functions of Tbx5 in FL initiation(Minguillon et al., 2005).

Interestingly, Fgf10 mutant mice initiate a limb bud(Sekine et al., 1999),suggesting that Tbx5 may also regulate other targets. In addition, Tbx5 continues to be expressed throughout later stages of limb outgrowth. Misexpression of a dominant-negative form of Tbx5 in the chick wing bud leads to the downregulation of Fgf4, Fgf8, Fgf10, Bmp2 and Wnt3a, and truncated FLs result(Rallis et al., 2003; Rodriguez-Esteban et al.,1999). These observations have led to the proposal that, in addition to its activity during initiation, Tbx5 also regulates later FL outgrowth and may be required for the orchestration of the distinct patterning events taking place in the growing limb(Logan et al., 1998; Rallis et al., 2003; Rodriguez-Esteban et al.,1999).

In humans, mutations in TBX5 are associated with the dominant disorder Holt-Oram syndrome (HOS; OMIM 142900)(Basson et al., 1997; Li et al., 1997), which leads to upper(fore)limb and heart deformities. Although haploinsufficiency of TBX5 is fully penetrant, the severity of the limb phenotypes can be variable,ranging from severe deletion deformities of many of the skeletal elements of the limb (aplasia), to more subtle phenotypes such as extended phalangeal elements of the thumbs (Newbury-Ecob et al., 1996). Similarly, mutations in TBX4 are associated with Small Patella syndrome (SPS; OMIM 147891), which is characterised by abnormalities of the lower(hind)limb. The limb skeletal defects in SPS are also characterised by deletion deformities with variable penetrance; however, they are generally less severe than those observed in HOS. Although the results from gene deletion experiments in the mouse are consistent with the phenotypes observed in HOS and SPS, they do not provide any information on the temporal requirement for Tbx5 and Tbx4 to form individual elements of the limb at later stages in development.

Here, we test whether Tbx5 plays a role beyond the initiation stages of FL formation. Using a tamoxifen (TM)-inducible Cre transgenic line(Prx1CreERT2), we have deleted Tbx5 in the nascent FL buds in embryos ranging from 21 somites [early embryonic day (E) 9.5] to E10.5. Surprisingly, although Cre-catalysed deletion of the Tbx5 conditional allele was apparent in the FL, limb outgrowth was not impaired. These results demonstrate that Tbx5 is not required for outgrowth of the limb bud and/or patterning of the FL skeleton. Furthermore, these findings have important implications for our understanding of the aetiology of the limb skeletal defects present in HOS.

Mouse embryos were staged according to Kaufman(Kaufman, 2001). Noon on the day a vaginal plug was observed, was taken to be E0.5. The mouse lines carrying a conditional allele of Tbx5(Bruneau et al., 2001), Rosa26RlacZ (Soriano,1999) and a Prx1Cre transgene(Logan et al., 2002) have been described previously. Construction of the Prx1CreERT2 construct was carried out using standard ligation procedures. The Prx1Cre backbone was digested with EcoRI to excise the Cre ORF, which was then replaced by an EcoRI CreERT2 fragment.

TM preparation and induction were performed as described on the Joyner laboratory webpage(http://saturn.med.nyu.edu/research/dg/joynerlab/protocols.html). Briefly, mice were gavaged with 6.5 mg (from a 20 mg/ml stock) of TM at the indicated time points.

To follow the rate of Tbx5 exon 3 recombination, we used primers flanking this exon and monitored the loss of PCR product resulting from its recombination. Briefly, E10.5 FL mesenchymal cells were dissected and the genomic DNA isolated for use in the quantitative PCR analysis. Quantification of the results was carried out using the ABI Prism 7700 Sequence Detection System, User Bulletin 2, Comparative C_T Method. Primers used were:Tbx5 exon3 Fwd, 5′-GGCATGGAAGGAATCAAGGT-3′; Tbx5 int 3-4 Rev,5′-ATTCCCTCCAATGACTGTCC-3′. The amount of template was normalised using primers that amplify a fragment of the cardiac actin promoter: mCA Fwd,5′-CCCCCTGGCTGATCCTCTAC-3′; mCA Rev,5′-TGGTCGCCTTAGCACCAT-3′. All reactions were carried out in triplicate.

Whole-mount and section in situ hybridisation were carried out essentially as described previously (Riddle et al.,1993; Schaeren-Wiemers and Gerfin-Moser, 1993). A minimum of three mutant embryos were analysed with each probe at each stage described. Most of the probes used have been described previously: Shh(Echelard et al., 1993), Fgf10 (Bellusci et al.,1997), Fgf8 (Crossley and Martin, 1995), Tbx5(Rallis et al., 2003), Tbx5 ex3 (kindly provided by C. Minguillon, NIMR, UK), Scx(Schweitzer et al., 2001), Sox9 (Morais da Silva et al.,1996), Col2a (Col2a1 - Mouse Genome Informatics)(Metsaranta et al., 1991), Myod (Davis et al.,1987), Pax3 (Goulding et al., 1991), Sall4 (a kind gift of S. Harvey, NIMR,UK), Tbx3 (kindly provided by C. Goding, Marie Curie Research Institute, UK) and Tbx15 (Singh et al., 2005).

The cartilage and bone elements of newborn mouse pups were stained with Alcian Blue and Alizarin Red, respectively, essentially as described previously (McLeod, 1980).

Detection of proliferating or apoptotic cells was carried out as previously described (Rallis et al.,2003). Briefly, mitotic cells were identified using a rabbit anti-phosphorylated histone H3 primary antibody (Upstate Biotechnology) and an HRP-conjugated goat anti-rabbit IgG secondary antibody (Jackson Laboratory). Apoptotic cell death was assayed by TdT-mediated dUTP nick end labelling(TUNEL) according to the manufacturer's protocol (Q-Biogene).

The extent of the anteroposterior axis was determined by measuring the distance between the anterior-most and posterior-most extremes of wild-type(n=15) and mutant (n=12) limbs at their proximal base.

Tbx5 has a pivotal role during initiation of the FL where it activates Fgf10 expression in the mesenchymal cells of the emerging limb bud(Agarwal et al., 2003; Rallis et al., 2003). Deletion of Tbx5 leads to the disruption of FL bud formation and all elements of the limb fail to form (Rallis et al.,2003). Such early and profound defects have prevented the study of potential roles of the gene later in limb development. It therefore remains unclear whether Tbx5 plays any crucial role in regulating outgrowth and coordinating patterning of the bud after its initiation. Expression of Tbx5 is not restricted to the initiation stages of limb bud development, but rather it is retained throughout the limb outgrowth stages(Fig. 1 and data not shown). Serial sections of the FL at different embryonic stages (E11.5, E12.5 and E15.5) were examined for the expression of Tbx5 along with markers of the prospective cartilaginous precursors (Sox9 or Col2a),tendons (Scx) and muscles (Pax3, Myod or muscle myosin). At E11.5, Tbx5 was expressed in all mesenchymal cells of the limb and was co-expressed in domains overlapping with bone, tendon and muscle precursors (Fig. 1A-D). At E12.5, Tbx5 expression was retained in all mesenchymal cells of the bud, but not as uniformly as that observed at E11.5(Fig. 1E). Tbx5transcripts were detected, but levels appeared reduced along the dorsal and ventral sides of the limb overlapping with the Myod and Scxexpression domains. However, Tbx5 expression remained elevated in domains overlapping with Col2a-expressing cartilage precursors in the centre of the limb (Fig. 1E-H). At E15.5, Tbx5 expression levels declined and were maintained in only a few cells at the tip of the digits (not shown).

The pan-mesenchymal expression of Tbx5 in the early stages of limb outgrowth is consistent with the gene playing a role in the maintenance and coordination of limb outgrowth. To confirm this hypothesis, we used a conditional Tbx5 allele (Bruneau et al., 2001) in combination with transgenic mouse lines expressing TM-inducible Cre recombinase (CreERT2)(Feil et al., 1997) under the influence of the Prx1 promoter, leading to expression of CreERT2 in the mouse limbs (Prx1CreERT2; Fig. 2). Administration of TM at specific time points and the consequent activation of Cre activity leads to the deletion of Tbx5at distinct stages of limb development.

To test the efficacy of the inducible Cre lines that we generated, we compared their activity with that of the previously described Prx1Cretransgenic line by crossing Prx1CreERT2 or Prx1Cre mice to the Rosa26RLacZ reporter line(Soriano, 1999)(Fig. 2 and data not shown). To induce Cre activity from the onset of CreERT2 expression driven by the Prx1 promoter, we tested oral gavage regimes of a single TM dose at E8.5 and a double dose at E7.5 and E8.5. The single dose regime of TM produced comparable levels of recombination as the double dose (data not shown). Gavage regimes at E7.5 or E6.5 resulted in little or no Cre activity,respectively. We therefore used a single oral gavage at E8.5 to detect the earliest Cre activity in the Prx1CreERT2 lines (see Materials and methods). This comparison reveals that the Prx1Cre transgene has more robust Cre activity at early stages than does the Prx1CreERT2 line. Prx1Cre was active in cells of the LPM at the level of the forming bud in the 14-somite stage embryo and, once a bud was visible, Cre activity was evident throughout the mesenchymal cells(Fig. 2A-C). By contrast,during limb initiation stages, Prx1CreERT2 induced recombination in a relatively small number of sparsely spaced cells(Fig. 2D,E,G,H), and only once the embryo had reached the 21- to 23-somite stage, after initiation of the bud had occurred, was recombination seen throughout the bud(Fig. 2F). From this time point onwards, the Prx1CreERT2 transgene activity in the reporter was indistinguishable from the Prx1Cre transgene(Fig. 2I-M and data not shown). Therefore, the difference between the activities of the Prx1CreERT2and Prx1Cre deleter transgenes in the FL was the patchy recombination in the FL prospective region in the 14- to 21-somite stage embryos that express the TM-inducible Cre.

To test whether Tbx5 is required beyond the initiation stages(when the Prx1Cre deleter is active), and is necessary for later outgrowth of the FL (when the Prx1CreERT2 is fully active), we set up crosses to produce Tbx5^lox/lox; Prx1CreERT2embryos. Cre activity was induced by gavaging the mother at E8.5 and harvesting embryos at E10.5. At E10.5, the limb buds are easily detectable and signaling centres such as the AER and zone of polarising activity (ZPA) that control proximodistal and anteroposterior (AP) growth of the limb,respectively, are well established. Unexpectedly, mutant limb buds that are devoid of a functional Tbx5 allele do form(Fig. 2L,M, Fig. 3). To demonstrate that Cre has been active in these limbs, we carried out the same cross but with a Rosa26RlacZ reporter allele also present in the background. Cre was active in all cells of the Tbx5 mutant limb bud to the same extent as in the wild-type limbs (Fig. 2I-M). To directly follow the amount of recombination of the Tbx5 conditional allele, we performed quantitative PCR using genomic DNA extracted from FL mesenchymal cells of E10.5 Tbx5^lox/lox; Prx1CreERT2 embryos gavaged at E8.5. This analysis confirmed that 87%-97% of Tbx5 exon3 (Tbx5 ex3) (the sequence deleted following the Cre-catalysed recombination to generate the Tbx5 null allele), had been deleted following TM-dependent Cre induction (Fig. 2N). To further confirm the extent of deletion of the Tbx5 allele, we performed in situ hybridisation with a probe that recognises Tbx5 ex3 and thereby identifies cells in which recombination of Tbx5 has not occurred. In wild-type or heterozygous limbs, this probe mimicked the endogenous Tbx5 expression pattern and stained both the FL and the heart (Fig. 3A,B and data not shown). In mutant embryos, because the Prx1 promoter is active in the limbs but not in the heart, expression of the unrecombined, functional conditional allele was detected in the heart(Fig. 3B, arrow). In the limbs,however, no signal was detected, demonstrating that Tbx5 ex3 had been deleted and that no functional transcripts of Tbx5 were detectable,reinforcing our observations that Cre is active throughout the mesenchyme of the mutant limb (Fig. 2I,L-N).

These results suggest that the deletion of Tbx5 in the prospective FL of embryos at the 21-somite stage does not impair further outgrowth of the FL bud. During limb initiation, Tbx5 is required for the upregulation of Fgf10 in mesenchymal cells of the bud(Agarwal et al., 2003; Rallis et al., 2003), and in turn Fgf10 induces Fgf8 in the overlying ectoderm. However, despite the deletion of Tbx5, no differences in Fgf10 expression were detectable (Fig. 3C,D). Consistent with normal Fgf10 signaling, Fgf8 was expressed in an equivalent domain, and at comparable levels throughout the AER, to that of wild-type or heterozygous littermates (Fig. 3E,F). An orthologue of Drosophila spalt major, Sall4,has recently been shown in zebrafish and mice to be a target of Tbx5 and to participate in the FGF signaling feedback loop(Harvey and Logan, 2006; Koshiba-Takeuchi et al.,2006). To test whether Sall4 is also sensitive to Tbx5 deletion after limb initiation, we performed in situ hybridisation with Sall4. Tbx5^lox/lox-deleted limbs had reduced Sall4 levels when compared with their wild-type counterparts,a phenotype most evident in the anterior domain of the limb(Fig. 3G,H, arrow). To determine whether the AP patterning of the mutant limbs was normal, we analysed the expression patterns of a range of markers expressed in nested domains across the AP axis. Tbx3 was expressed in anterior and posterior domains, and this pattern was not perturbed in the mutant limb(Fig. 3I,J). Tbx15 was expressed in an apparently reciprocal pattern to Tbx3, comprising a broad medial stripe, and this was unchanged in the Tbx5^lox/lox mutant limb(Fig. 3M,N). The Shhexpression that marks the ZPA at the posterior of the limb was not disrupted in Tbx5^lox/lox-deleted limbs(Fig. 3K,L). Therefore, the initial expression of Tbx5 in embryos up to the 21-somite stage is sufficient for two key signaling centres in the limb to be established: the AER and the ZPA.

Although the expression of molecular markers of two key signaling centres, Shh (ZPA) and Fgf8 (AER), were established normally, the appearance of the Tbx5^lox/lox-deleted limbs was not entirely normal. The mutant limbs were narrower in their anterior to posterior extent (Fig. 2J-M). We compared the number of somites that the limbs spanned by analysing the expression of Fgf8, which marks the AER, and Myod, which marks the somites. Whereas the normal limb spanned approximately five somites(Fig. 3O), the Tbx5^lox/lox limbs spanned only four somites(Fig. 3P). Significantly,although both mutant and wild-type limbs were positioned at the equivalent somite level at their posterior extreme, the anterior extent of the mutant limb was reduced. This anterior bias of the limb phenotype in the Tbx5 mutant limbs resembles the HOS phenotype, in which the more anterior structures are most severely affected (see Discussion). Measurement of the mean AP extent of mutant and control limbs at E10.5 indicated that whereas wild-type limbs spanned on average 850 μm, the mutant limbs spanned on average only 700 μm (18% narrower). This reduction in limb size was due to a small, yet detectable, reduction in cell proliferation, and not to an apparent increase in the rate of cell apoptosis in the limb mesenchyme. The number of cells staining positive for phospho-histone H3, a marker for cells undergoing mitosis, was greater in wild-type than in mutant limbs. However, no differences between wild-type and mutant limbs were seen for cells staining positive following TUNEL, a marker of cells undergoing apoptosis (see Fig. S1 in the supplementary material).

Our molecular analysis demonstrated that limb buds lacking a functional allele of Tbx5 from approximately the 21-somite stage onwards,although slightly smaller, do continue the outgrowth programme, indicating that the gene does not play an essential role during limb outgrowth. To further analyse whether Tbx5 plays additional roles in later stages of FL outgrowth, we analysed limbs from mice gavaged once at E8.5, E9.5 or E10.5. Induction of Cre activity at these later stages results in recombination in all, or almost all, cells of the FL when embryos were harvested 2 days after gavage (Fig. 4A-C). By E16.5, all cells of the limb were lacZ-positive(data not shown), even when embryos were gavaged as late as E10.5. Because females gavaged at E8.5 consistently miscarried at late stages of gestation,we harvested embryos at E14.5. Skeletal elements were analysed by staining E14.5 or E16.5 embryos with Alizarin Red and Alcian Blue. Surprisingly, mutant limbs resulting from TM administration at E8.5, E9.5 and E10.5 contained all skeletal elements (Fig. 4D-I). Furthermore, the overall skeletal patterning was largely normal. Interestingly, only in mutant limbs derived from TM administration at E8.5 were the left limbs more severely affected than the right (data not shown). In the limbs that developed following gavage at E8.5 or E9.5 (n=12),holes were present in the scapula blade and there were obvious notches at the proximal border (Fig. 4E″,G″ and data not shown). The deltoid tuberosity of the humerus was also missing (Fig. 4E,E″,G,G″, arrows). Significantly, the phalangeal elements of digit 1, which are equivalent to the human thumb, were elongated and resembled the tripalangeal defect observed in HOS(Fig. 4E′,G′,arrowheads). In one example, digit 1 was elongated and duplicated in a manner similar to that observed in individuals with HOS(Fig. 4E′, arrowhead)(Newbury-Ecob et al., 1996). The remaining four digits appeared completely normal. These phenotypes were not observed in embryos resulting from TM administration at E10.5, where all the skeletal elements were normal (Fig. 4I). These results are reinforced by the absence of abnormal limb phenotypes when a Col2a-Cre deleter line(Ovchinnikov et al., 2000) is used to delete Tbx5 function in the cartilage progenitors of the limb skeleton from E11.5 onwards (P.H. and M.P.O.L., unpublished). These results confirm that Tbx5 is not required for outgrowth of the FL and demonstrate that Tbx5 is not required for formation or patterning of the limb skeleton.

Mutations in TBX5 are associated with the human dominant disorder HOS that is characterised by multiple upper-limb and heart defects. In vertebrates, Tbx5 plays a pivotal role in the initiation of the FL by activating Fgf10 (Agarwal et al.,2003; Ahn et al.,2002; Ng et al.,2002; Rallis et al.,2003), and Tbx5 knockout mice do not initiate a FL(Agarwal et al., 2003; Rallis et al., 2003). These and other experiments have lead to the placement of Tbx5 (and its paralogue Tbx4 in the HL) at the top of the genetic cascade required for limb formation. However, Tbx5 (and Tbx4), continue to be expressed at later stages of limb development(Fig. 1), and misexpression of dominant-negative forms of Tbx5 in the chick leads to limb truncations and the downregulation of genes required for limb outgrowth(Rallis et al., 2003; Rodriguez-Esteban et al.,1999). It has therefore been proposed that Tbx5 might continue to play a crucial role in limb outgrowth and in later skeletal patterning events in the growing limb(Rallis et al., 2003; Rodriguez-Esteban et al.,1999). Our results, and those of others(Naiche and Papaioannou,2007), suggest that this is not the case, and that Tbx5and Tbx4 do not have a major role in limb outgrowth and formation of the limb skeleton.

The deletion of Tbx5 using Prx1Cre results in the complete failure of FL formation, whereas deletion with Prx1CreERT2does not lead to the same extreme phenotype and early limb markers are expressed relatively normally (Fig. 3). As the resulting mutant limbs do not express an active Tbx5 gene (Fig. 2N, Fig. 3B), the difference in phenotypes (i.e. limbless versus limb, respectively) is a consequence of the timing of the gene deletion. Our analysis shows that Prx1Cre is active in the LPM of 14-somite stage embryos at the level of the forming bud. By contrast, Prx1CreERT2, although initially active at the same stage(starting at 14 somites), leads to recombination in a small and sparse number of cells in the bud. Only in 21- to 23-somite stage embryos is activity seen throughout the FL mesenchyme in a manner comparable with that of Prx1Cre. The temporal disparity in Cre activity between the two Cre deleter lines, and the distinct phenotypes produced, demonstrate the critical time window within which Tbx5 is required for FL initiation. This allows us to identify two distinct phases in early limb development: a limb initiation phase that is Tbx5 dependent(Fig. 5, top), followed by a limb outgrowth phase that is Tbx5 independent(Fig. 5, bottom). During the initiation phase, Tbx5 activates Fgf10 expression, and this phase is complete by the 21-somite stage. At this time, key signaling centres in the limb, the AER and the ZPA, are established. Deletion of Tbx5 after the 21-somite stage (TM administration at stages prior to E10.5) does not impair limb outgrowth, although the limb is not entirely normal. This is consistent with recent observations showing that, in addition to regulating Fgf10, Tbx5 also activates Sall4. Sall4 is required for the expression of mesenchymal FGFs, possibly by facilitating establishment of the FGF signaling feedback between the mesenchyme and overlying ectoderm(Harvey and Logan, 2006; Koshiba-Takeuchi et al.,2006), thereby enabling maintenance of Fgf8 expression in the AER. In humans, mutations in SALL4 are associated with Okihiro syndrome (OS; OMIM 607323) (Kohlhase et al., 2002), and the limb abnormalities characteristic of OS share many similarities with those found in HOS. Strikingly, the skeletal abnormalities resulting from TM administration at E8.5 and E9.5 are reminiscent of both HOS and OS. It is thus plausible that the phenotypes observed in mice gavaged at E8.5 and E9.5 are due to the reduction in Sall4 activation by Tbx5 (e.g. Fig. 3H)(Harvey and Logan, 2006; Koshiba-Takeuchi et al.,2006). Crucially, deletion of Tbx5 function at later stages does not produce HOS-like skeletal phenotypes. This defines the outgrowth phase when the feedback loop between mesenchymal Fgf10 and ectodermal Fgf8 has been established, can operate and is maintained independently of Tbx5 (Fig. 5, bottom).

Our results demonstrate that deletion of Tbx5 during distinct phases of limb development leads to phenotypes resembling the clinical manifestations of HOS. When Tbx5 is deleted during early outgrowth(gavage at E8.5), FL buds are narrower and span the extent of four, rather than five, somites along the main body axis(Fig. 3); this appears to be at least in part due to a reduced rate of cell proliferation, rather than to any increase in cell death (see Fig. S1 in the supplementary material). This phenotype is reminiscent of one of the characteristics of HOS: the loss of anterior structures. These phenotypes are also observed in OS(Kohlhase et al., 2002), in which a reduction in anterior expression is also observed.

Skeletal analysis of Tbx5 mutant limbs following TM administration up to E9.5 revealed additional phenotypes characteristic of HOS. In individuals with HOS, the left limb is commonly more severely affected than the right (Newbury-Ecob et al.,1996). In embryos generated following a gavage regime at E8.5, but importantly not at later stages of TM administration, the left limb was consistently more severely affected than the right limb. Furthermore,following gavage at E8.5 and E9.5, the FLs of the resulting embryos had anterior abnormalities, in particular an extended digit 1(Fig. 4), similar to the thumb abnormalities characteristic of HOS (R. Newbury-Ecob, personal communication)(Newbury-Ecob et al., 1996). Significantly, however, when Tbx5 was deleted later (gavage at E10.5), no skeletal malformations were observed, demonstrating that HOS skeletal phenotypes are caused by loss of Tbx5 activity at earlier stages. Furthermore, we did not observe several prominent features of HOS,such as aplasia of skeletal elements. In combination, these results imply that different HOS manifestations can be attributed to distinct stages of Tbx5 activity.

Recently, suggested roles for Tbx4 and Tbx5 in the specification of limb-type identity have been ruled out(Minguillon et al., 2005). Our data and those of others (Naiche and Papaioannou, 2007) suggest that these genes likewise play no role in regulating limb outgrowth, despite their expression being maintained during the outgrowth stages of limb development. It is possible that the activities of Tbx5 and Tbx4 during these later stages of limb development are controlled at a post-transcriptional and/or post-translational level. Evidence for post-translational regulation has come from the description of a PDZ-LIM protein, Lmp4, which is expressed in the limb. When co-expressed in COS-1 cells, Lmp4 is able to bind to Tbx5 and Tbx4, leading to cytoplasmic localisation of the complex via association with the actin cytoskeleton(Krause et al., 2004). It remains to be determined, however, whether Tbx5 and Tbx4 are regulated in this manner during limb development. Similarly, in Xenopus, the canonical T-box protein Xbra has been shown to bind Smad1, a binding that modulates its activities (Messenger et al.,2005). Because Smad1 activity is dynamically regulated(Massague et al., 2005), it is possible that a similar situation exists for Tbx5 and Tbx4 during limb outgrowth, and that their activities are controlled via the regulation of co-factors. Our results for Tbx5, and those of others for Tbx4, demonstrate that the activities of these proteins are regulated during limb development. Future work will determine if Tbx5 and Tbx4 are required for other processes during limb development, and whether their activities are regulated at a post-transcriptional or post-translational level.

We thank C. Minguillon and S. Harvey for providing reagents; H. Reardon and M. Caulfield of the Biological Services section (NIMR) for animal husbandry;and Catherine Shang for advice on quantitative PCR. We thank Ruth Newbury-Ecob for personal communications and Pierre Chambon for providing the CreERT2 clone. We are grateful to Naiche Adler and Virginia Papaioannou for communicating results prior to publication; T. Heanue for critical reading of the manuscript; and A. DeLaurier for fruitful discussions. P.H. is funded by an EMBO Long-term Fellowship; 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.