A role for the mesenchymal T-box gene Brachyury in AER formation during limb development

The limb bud is induced in the lateral plate mesoderm in response to signals thought to be relayed in several steps from the embryo midline to the periphery (reviewed by Martin,1998). Growth and patterning of the limb bud is organized by three signaling centers that arise before or soon after initial budding, and whose activities polarize pattern formation along the three axes of the limb:anteroposterior (AP), dorsoventral (DV), proximodistal (PD). Multiple cross-regulatory interactions between these signaling centers further coordinate development.

Limb outgrowth along the PD axis is regulated by FGF signals from a specialized columnar ectoderm that normally forms along the DV boundary at the limb apex, the apical ectodermal ridge (AER) (reviewed byMartin, 1998;Lewandoski et al., 2000;Moon and Capecchi, 2000;Sun et al., 2002). Interacting FGF and WNT cascades are critical in early limb induction, AER formation and reciprocal epithelialmesenchymal interactions necessary for AER maintenance(reviewed by Martin, 1998;Tickle and Munsterberg, 2001). Several WNTs and FGFs, acting directly or indirectly, are able to induce ectopic limbs in the flank (e.g. Cohn et al., 1995; Ohuchi et al.,1997; Kawakami et al.,2001). Ultimately, FGF10 in the prospective limb lateral plate induces ectodermal Wnt (Wnt3a in chick) and Fgf8expression in the forming AER (Ohuchi et al., 1997; Kengaku et al.,1998; Galceran et al.,1999). FGF8 from the AER ectoderm also maintains high levelFgf10 expression in limb mesoderm. Thus, reciprocal FGF8 and FGF10 signals form a positive regulatory loop by which AER and subridge mesoderm functionally maintain each other (Ohuchi et al., 1997; Revest et al.,2001). Sonic hedgehog (Shh), expressed in posterior limb bud mesoderm, regulates AP polarity and number of skeletal elements. Activation of Shh also depends on ridge signals and subsequently positive feedback between FGF4 from posterior ridge and SHH maintains both signaling centers (reviewed byCapdevila and Johnson,2000).

AER induction, as well as later maturation, is normally closely linked to DV boundary formation although they can be uncoupled (reviewed byZeller and Duboule, 1997;Tickle and Munsterberg, 2001). Early upstream BMP signaling regulates both processes in parallel(Ahn et al., 2001;Pizette et al., 2001);contributing to AER induction by upregulating Msx genes and to establishing DV polarity by activating En1 expression in ventral ectoderm. EN1 represses Wnt7a expression in ventral ectoderm(Loomis et al., 1996;Logan et al., 1997), while dorsal ectodermal WNT7a signals activate Lmx1b in the underlying mesoderm to regulate dorsal fates (Parr and McMahon, 1995; Riddle et al., 1995; Vogel et al.,1995; Cygan et al.,1997; Chen et al.,1998).

The murine AER forms from a broad zone of Fgf8-expressing ventral ectoderm in the prospective limb region, both by movement of pre-AER cells to the DV border and loss of AER-gene expression in remaining ventral cells, to form a sharply demarcated ridge during maturation(Kimmel et al., 2000)(reviewed by Tickle and Munsterberg,2001). This process is also regulated by En1 in the ventral ectoderm, via repressing Wnt7a and setting a ventral border for mature AER (Cygan et al.,1997; Loomis et al.,1998; Kimmel et al.,2000). Although the chick differs somewhat from mouse in thatFgf8 expression is more restricted from its onset(Crossley et al., 1996), a pseudostratified, columnar AER also forms from a broader zone of flat ectodermal precursors that compact into a sharp ridge along the DV apex(Altabef et al., 1997;Michaud et al., 1997). Regulation of maturation is likely to be conserved in chick; for example,En1 also affects AER formation(Laufer et al., 1997;Logan et al., 1997;Rodriguez-Estaban et al., 1997). The role of mesodermal signals in later morphogenesis of a mature AER is unclear. Disruption of mesodermal FGF10 signaling interferes with early stages of AER induction, obscuring later function (Min et al., 1998;Xu et al., 1998;Arman et al., 1999;Sekine et al., 1999;Revest et al., 2001).

We previously noted expression of Brachyury (T) in early limb buds on northern blots (Knezevic et al., 1997a). T is the founding member of the T-box family of transcription factors which share a conserved DNA binding domain and play multiple roles during development (reviewed byPapaioannou and Silver, 1998),several of which are expressed in developing limb(Gibson-Brown et al., 1996;Gibson-Brown et al., 1998;Isaac et al., 1998;Logan et al., 1998;Ohuchi et al., 1998).T plays an essential role in primary mesoderm formation (reviewed byHerrmann, 1995) and has been implicated in both WNT and FGF signaling pathways during gastrulation(Smith et al., 1997;Yamaguchi et al., 1999a;Arnold et al., 2000;Tada and Smith, 2000;Galceran et al., 2001), but has not been reported to be expressed elsewhere. We found that T is expressed at low levels in lateral plate at the onset of limb initiation and persists in the subridge mesoderm of the limb bud, as well as several other sites associated with WNT and FGF signaling. Retroviral misexpression ofT in chick causes anterior AER extension along the limb apex and skeletal phenotypes consistent with extended AER function, whereas loss ofT in mutant mouse embryos disrupts normal AER morphogenesis. These results suggest T functions in the mesoderm to direct AER maturation along the DV limb border. Taken together, altered subridge Fgf10expression levels in response to in vivo changes in T expression, and the ability of ridge-specific WNT and FGF signals to induce T in vitro suggest that T may also regulate maintenance of a mature AER as a component of the reciprocal epithelial-mesenchymal signaling mediated by FGF and WNT pathways.

White leghorn chicken eggs (Truslow Farms) were incubated at 38.5°C and staged according to Hamburger and Hamilton(Hamburger and Hamilton,1951). The mouse T^+/- mutant lineT^2J (in C57BL6), containing a ∼80 kb genomic deletion including the entire T gene(Herrmann et al., 1990), was obtained from the Jackson Laboratory. Extended survival ofT^-/- embryos (to ∼E10.5) was achieved by breedingT^2J/C57BL6 line to wild-type FVB/N.T^-/- embryos were easily identified phenotypically. Noon on the day of the appearance of the vaginal plug was considered E0.5.

Conditions for probe preparation and in situ hybridization were as previously described (Knezevic et al.,1997a; Knezevic et al.,1997b). For chick T probe, the color reaction contained 10% polyvinyl alcohol (Barth and Ivarie,1994) to enhance detection sensitivity, and was developed for 12-18 hours. Other chick probes used included Fgf4 (L. Niswander),Fgf8 (J.-C. Izpisua-Belmonte), Fgf10 (S. Noji), Shh(R. Riddle), and Wnt5a (T. Nohno). Mouse probes used includedBmp4 (B. Hogan), Fgf8 (G. Martin), Fgf10 (D. Ornitz), En1 (A. Joyner), Lmx1b (R. Johnson), Msx1,Msx2 (R. Maas) and Shh, Wnt7a (A. McMahon). Whole-mount embryos were paraffin embedded or frozen in OCT and sectioned at 5-10 μm, or embedded in 3% agarose to cut 50 μm vibratome sections. Immunostaining of cryosections or paraffin sections (after antigen retrieval by steaming) was detected with peroxidase-linked secondary antibodies and Vectastain kit. Affinity purified DLX-antibody was a gift from G. Panganiban (for details, see(Panganiban et al., 1997).

Mitosis-specific anti-phospho-histone H3 (Upstate Biotechnology) was used to immunostain multiple sections. Mitotic cells in equal-sized areas were counted and averaged for comparisons. Apoptosis was assessed on sections by TUNEL assay (Apoptag kit, Intergen) using fluoro-dNTP incorporation and direct fluorescence microscopy with propidium iodide counterstaining, or by whole-mount staining of fresh embryos with 1/50,000 Nile Blue sulfate in PBS.

The chick T coding region(Knezevic et al., 1997a) was cloned as an NcoI-EarI fragment into theNcoI-SmaI sites of Cla12Nco vector, and transferred into RCASBP as a ClaI fragment. Virus production and infection were as described by Morgan and Fekete (Morgan and Fekete, 1996), except that a Picospritzer II (Parker instruments)was used for injection. Parallel control infections with alkaline phosphatase(AP)-expressing virus (a gift from B. Morgan) were done to assess phenotype specificity, and adequacy of hindlimb infection by AP staining.T-virus infected embryos were also hybridized with a T probe as a control to evaluate adequacy of infection. Overall, 55-75% ofT-infected embryos at early or late times post-infection had abnormalities. The extents of viral T expression in hindlimb suggested that the lack of visible effect of infection in the remaining embryos was due to inherent variability in infection spread (data not shown). Variability in amplification of T-virus also occurred in cell culture. For skeletal analysis, infected day 9-10 embryos were fixed in 5%TCA, stained with 0.1% Alcian Green 2GX in acid-alcohol, washed and dehydrated in absolute alcohol, and cleared in methyl salicylate.

T/GST fusion protein from plasmid expressing N-terminal (amino acids 1-123)T/GST fusion protein (a gift from B. Herrmann) was used to immunize rabbits using standard techniques. Antibody was affinity purified as described(Kispert and Herrmann, 1993). Protein extracts were prepared from dissected embryonic tissues by lysis in PBS containing protease inhibitor cocktail, 10 μg/ml AEBSF, 1% SDS, and 0.5 mM DTT. Proteins, electrophoresed on 4-12% NuPAGE gradient gels (Novex), were blotted and probed with antibody using standard techniques. Monoclonal anti-α-tubulin (Sigma) served as an internal protein loading control. For immunohistochemistry, affinity-purified anti-T was pre-incubated with blotted 100 kDa protein (24 hours) to remove cross-reacting epitopes.

Primary cell cultures from stage 19-20 chick limb buds, as described(Knezevic et al., 1997b), were incubated with recombinant FGF4 (a gift from Genetics Institute, Inc.) or FGF8b (R&D systems) at 750 ng/ml and heparin sulphate (100 ng/ml). Cultures infected with Wnt3a-expressing retrovirus (a gift from Cliff Tabin) at a multiplicity of 3-5 IU/cell, were compared to RCASBP control. Northern analysis with chick T 3′UTR and control probes was as described previously (Knezevic et al.,1997a).

We previously detected chick T transcripts in limb buds by northern analysis (stages 20-24) (Knezevic et al., 1997a), and therefore evaluated the expression pattern by in situ hybridization. Weak expression of T in the prospective forelimb lateral plate and adjacent somites was first detected at about stage 15 (Fig. 1A,D). Subsequently,expression in forelimb and hindlimb bud increased(Fig. 1B) and localized to a thin strip of distal mesenchyme just beneath the AER(Fig. 1E,F). Weak expression was also seen in somitic dermomyotome and in Wolffian/nephric ducts(Fig. 1D,E). Expression wanted by stage 22/23 in limb buds and adjacent somites(Fig. 1C,G) and became undetectable by stage 26-27 (not shown). Surprisingly, hybridization with a sense probe revealed antisense T transcripts at several sites including somites and limb buds (data not shown).

T protein expression in limb was verified with immunoblots of protein from different chick and mouse tissues, probed with affinity-purified anti-T. T protein was detected in early stage chick and mouse limb buds, albeit at much lower levels than tailbud (Fig. 2A). A highly abundant, cross-reacting, ubiquitous 100 kDa protein(see Fig. 2A) hampered in situ immunostaining even after affinity purification, necessitating extensive pre-adsorption against blotted 100 kDa proteins to deplete cross-reacting epitopes. Depleted antibody revealed mouse T protein in subridge limb mesoderm and somites, as in chick (Fig. 2B). The intriguing proximity of T expression to the forming AER was investigated using retroviral misexpression to perturbT in developing limb.

The prospective hindlimb lateral plate of stage 9-11 chick embryos was infected with RCASBP retrovirus expressing full-length T protein. Changes in limb bud morphology were discernible by stage 19-20 (∼48 hours post-infection). The AER, visualized by Fgf8 expression as a functional marker, extended farther anteriorly along the DV edge of infected limb buds, accompanied by variable broadening of the anterior limb bud (50%,7/15, Fig. 3A,B). In some embryos, the `extended' AER was widened and irregular (e.g. seeFig. 3B), but truly ectopic AERs away from the DV limb margin were not seen, except for rare, small isolated spots of Fgf8 expression (data not shown). Older infected embryos (stage 24, ∼60 hours post-infection) also had anteriorly extended AERs (40%, 12/29, Fig. 3C) and some embryos without extension still showed stronger Fgf8 expression in the AER on the infected side (20%, 6/29 embryos, not shown), which may correlate with mild soft tissue phenotypes seen at day 10 (see below), perhaps reflecting later infection onset or spread. Fgf4 expression, normally restricted to posterior AER, was also extended anteriorly in Tinfected limb buds (52%, 16/31, Fig. 3D), consistent with expanded AER extent and function.

At 10 days, T-infected embryos showed skeletal and soft tissue phenotypes predicted by earlier AER changes (75%, 21/28; no abnormalities in control-virus infections, 0/15). Skeletal abnormalities (54%, 15/28) included anterior digit duplications (Fig. 4B-D), in some cases with posterior transformation of anterior digits (Fig. 4B). Sometimes the anterior-most metatarsal was also thickened(Fig. 4C) or duplicated(Fig. 4D). The proportion of infected embryos with anterior digit changes correlated well with the incidence of anterior AER extension at earlier stages. SHH plays a central role regulating AP digit patterning and is often a factor in the production of anterior polydactyly (discussed inKnezevic et al., 1997b). Moreover, changes in AER extent might also be associated with ectopicShh expression, driven by increased FGFs (reviewed byMartin, 1998). However, no ectopic anterior Shh was detected in a set of infected limb buds showing clear anterior broadening and AER extension (0/11,Fig. 3C). This may not be surprising since augmented AER function is not invariably linked to ectopicShh induction (e.g. Pizette and Niswander, 1999; Zhang et al.,2000) and digit I specification does not require SHH (seeLewis et al., 2001).

T misexpression caused milder interdigital soft tissue changes in some embryos (21%, 6/28), including soft tissue broadening(Fig. 4E,F), delayed loss of webbing (Fig. 4E) and ectopic cartilage condensations (Fig. 4D,E). Inhibition of AER regression is associated with interdigital soft tissue overgrowth, apparently due to prolonged Fgf8expression in the AER at late stages(Pizette and Niswander, 1999). Stronger late Fgf8 expression seen in some T-infected embryos may have a similar effect, suggesting T may delay AER regression.

T is a transcription factor so we expected direct targets to be expressed in the mesoderm, where T is normally expressed, and to include secreted signals that regulate the AER. Since Wnts andFgfs are T targets during gastrulation(Smith et al., 1997;Tada and Smith, 2000), we checked Fgf10 and Wnt5a, both expressed in early distal limb mesenchyme. Fgf10, which regulates limb initiation and AER formation(Ohuchi et al., 1997;Min et al., 1998;Sekine et al., 1999), was increased in the distal subridge mesoderm in T-infected limb buds(7/14 total, stage 19-24) and was already upregulated by stage 19-20 (4/7 embryos, Fig. 3E), suggesting that Fgf10 may be an early target of T in the subridge zone.Wnt5a, which plays a role in limb outgrowth(Yamaguchi et al., 1999b), was unchanged in stage 19-21 infected embryos (0/8, data not shown) and later was expanded slightly anteriorly (4/8, Fig. 3F), most likely as a very indirect effect.

Prominent T expression in apical subridge mesoderm suggested regulation by AER signals. FGFs and WNTs also function upstream of Tduring gastrulation (Smith et al.,1997; Yamaguchi et al.,1999a; Arnold et al.,2000; Galceran et al.,2001), so AER-specific WNTs and FGFs were tested. Since limbT expression was very weak, we evaluated quantitative effects onT expression in primary cultures of stage 19/20 limb mesenchyme.T was induced within 18 hours after infection with aWnt3a-expressing virus (Fig. 5). Recombinant FGF8b induced T within 8 hours(Fig. 5), while recombinant FGF4 protein had no effect early or later (data not shown). Hence, both FGF8b and WNT3a ridge signals may participate in activating or maintaining subridgeT expression. Fgf4, expressed in posterior AER, functionally overlaps with Fgf8 (Sun et al.,2002) but shows some differences in receptor binding(Ornitz et al., 1996) and ability to regulate mesodermal genes (e.g.Haramis et al., 1995;Mahmood et al., 1995;Kimura et al., 2000), perhaps explaining its failure to induce T.

To assess whether T plays a role in normal AER formation, we analyzed mouse T^-/- embryos. The T^-/-null mutant is an early embryonic lethal (∼E10.5), precluding skeletal analysis, but allowing evaluation of forelimb AER formation. No hindlimb bud forms owing to disrupted posterior mesoderm formation and ensuing posterior truncations. T^-/- forelimb buds were smaller than wild type and often more rounded at mature AER stages, lacking a sharp DV edge (see Figs 6,7).

From its onset at E9, Fgf8 expression in pre-AER ectoderm was mottled and weaker in T^-/- embryos(Fig. 6A). Later, Fgf8continued to be irregular in the T^-/- pre-AER, revealing its failure to compact normally toward the DV apex(Fig. 6B-E). At more mature stages (E9.75-10) the T^-/- AER was wavy, zigzagging into both dorsal and ventral ectoderm, with variably weaker Fgf8 levels(Fig. 6D,E). AER morphology was assessed with anti-DLX, since Dlx gene expression marks AER progenitors similarly to Fgf8(Panganiban et al., 1997;Loomis et al., 1998). DLX expression during AER morphogenesis (Fig. 7A-H′) was irregular and sometimes less clearly restricted to ventral ectoderm in T^-/- embryos (e.g.Fig. 7C,D), and revealed that progression to a highly compact, pseudostratified AER was delayed and erratic. Even in late stage T^-/- limb buds with the mildest phenotype, the extent of AER maturation varied dramatically in neighboring sections (e.g. Fig. 7H vs H′).

Bmp4 is highly expressed in AER progenitors and BMP signalling has been implicated in AER induction and DV patterning(Ahn et al., 2001;Pizette et al., 2001).Bmp4 expression in T^-/- embryos was altered in parallel with Fgf8, starting weak and patchy at E9(Fig. 6F), and remaining broad and irregular at later stages (Fig. 6G-J), consistent with abnormal AER maturation. SinceBmp4 is expressed in ectoderm and in mesenchyme, the distribution was examined on sections, revealing abnormalities in AER morphogenesis similar to those seen with DLX antibody (Fig. 7I-N). Mesodermal Bmp4 expression, though comparatively quite weak between E9-10, was detectable early and reduced inT^-/- limb buds (Fig. 7I,J). BMP downstream target genes Msx1 and Msx2are expressed throughout ventral ectoderm as well as forming AER, and underlying limb mesoderm (Ahn et al.,2001; Pizette et al.,2001). Despite reduced expression of Bmp4, Msx1 andMsx2 expression was quantitatively preserved (data not shown), as was BMP target En1 (below), indicating that the BMP pathway was largely intact.

Pre-AER gene expression and morphology indicated disrupted maturation.En1 in ventral ectoderm regulates compaction and positioning of mature ventral AER borders by repressing Wnt7a(Cygan et al., 1997;Loomis et al., 1998;Kimmel et al., 2000). InT^-/- embryos, Wnt7a and En1 were expressed in dorsal and ventral limb bud ectoderm, respectively. However, the distal ectodermal expression borders for each of these genes was not sharp as it is in wild type (Fig. 6K-S),and remained jagged and irregular even at late stages.

Fgf10 induces AER formation and was upregulated by Tmisexpression. Fgf10 expression declined progressively over time inT^-/- embryos compared to wild type(Fig. 6T-X), but did not appear very different until E9.5, after changes in Fgf8 and Bmp4expression were discernible.

Proliferation rates, assessed with the mitosis-specific antiphospho-histone H3 (Gurley et al., 1978), were similar in T^-/- and wild-type limb buds at E9.5 and E9.75(Fig. 8A, and not shown). Normal mitotic rates are not inconsistent with reduced FGF8 inT^-/- limb buds; proliferation is unchanged even inFgf8/Fgf4 nulls despite reduced limb bud size(Sun et al., 2002). Levels of apoptosis, analyzed in sections and whole mount, were similarly very low in wild-type and T^-/- limb buds from E9.5-E10(Fig. 8B, and not shown), even at E10 when extensive apoptosis was present in T^-/- axial tissues. Loss of ridge FGF survival signals causes mesenchymal apoptosis, but does not become appreciable until about E10-10.5(Moon and Capecchi, 2000;Revest et al., 2001;Sun et al., 2002). Thus,abnormalities in T^-/- limb buds are not due to general growth arrest and lost viability. They are also unlikely to be due to a general developmental delay. Lmx1b, initially expressed uniformly and later restricted to dorsal mesoderm after the AER forms(Loomis et al., 1998), showed normal dorsal restriction in T^-/- embryos at E10(Fig. 8C). However, some mesodermal gene expression was lost. Shh, which depends on ridge FGF signals (reviewed by Capdevila and Johnson,2000), was absent in T^-/- limb buds(Fig. 8D). InFgfr2-IIIb null mutants, which fail to form a mature AER, loss ofShh induction is evident before other mesodermal changes such as decreased Msx1, or increased apoptosis(Revest et al., 2001). The results suggest that T plays a role in AER morphogenesis and that altered AER function in T^-/- embryos causes some changes in mesodermal gene induction. However, the possibility that AER changes in theT^-/- embryo are an indirect consequence of lost early midline signals cannot be excluded at present (seeDealy, 1997).

T is a well-characterized transcription factor, highly expressed in axial mesoderm and primitive streak, with key roles in mesoderm formation,migration and notochord function (reviewed byHerrmann, 1995). Texpression has not until now been documented at other sites, presumably because of its very low levels. The timing and pattern of Texpression in lateral plate and early limb bud suggest a potential role in AER formation. Misexpression in chick prospective limb produced phenotypes consistent with anteriorly extended AER formation and function, while formation of a mature AER compacted towards the DV limb edge was disrupted inT^-/- mouse embryos, confirming a role for T in AER regulation during normal development.

Normal T expression in limb bud is restricted to the immediate subridge mesoderm. This may be due to a very stringent dependence ofT expression on signals from the ridge. In addition to maintenance by ridge signals, lateral inhibitory signals may also serve to limit the zone ofT expression and/or function. In the ectoderm, Cux1 is induced along the edges of the forming AER and seems to function to restrict AER formation from spreading laterally (Taveres et al., 2000). Perhaps one of the Sprouty genes, which are induced as feedback inhibitors during FGF signaling, could serve a similar role in the limb mesoderm (seeMinowada et al., 1999). Even when T is uniformly misexpressed throughout the early lateral plate and limb mesoderm, its functional effects are nevertheless restricted to the DV boundary along which AER formation normally occurs, resulting in AER extension along the limb apex. In contrast, certain other factors can induce ectopic AERs in any orientation in limb ectoderm (e.g.Laufer et al., 1997;Rodriguez-Estaban et al., 1997; Kengaku et al., 1998; Pizette et al.,2001). This observed functional restriction could be due to lateral inhibition of T function by other limb mesodermal factors. Alternatively, T function may require positive mesodermal co-factors that are also restricted to the DV boundary, and so are unavailable to misexpressed T elsewhere in the limb. In either case, such restrictions of T activity may provide a mesodermal mechanism to ensure that normal AER formation occurs only along the DV boundary in the ectoderm.

The consequences of loss of T are also consistent with such a function, and suggest that mesodermal factors may play a role, along with ectodermal factors such as En1(Kimmel et al., 2000), in regulating ridge maturation and positioning. However, unlike En1, Tis unlikely to function in AER formation by regulating DV boundary positions. At present, it remains uncertain what mesodermal signal T may act through to contribute to AER maturation. At very early stages, BMPs regulate both DV polarity and AER formation in parallel, by inducing En1 in ventral ectoderm and by activating Msx expression, respectively(Ahn et al., 2001;Pizette et al., 2001). Although Bmp4 expression was reduced early in theT^-/- forelimb region, DV polarity was preserved in both ectoderm and mesoderm. Likewise, expression of other BMP targets, Msxgenes, was unchanged. Consequently, we feel that the BMP signaling pathway is not primarily affected by loss of T function; reduced Bmp4and irregular En1 and Wnt7a distal expression borders are more likely secondary to some other causative defect in AER morphogenesis. FGF10 plays some role in formation of a mature AER, as well as in limb initiation. In the complete absence of FGF10 no bud or pre-AER forms(Min et al., 1998;Sekine et al., 1999), but loss of the ectodermal FGFR2-IIIb isoform, which FGF10 binds, disrupts AER formation shortly after induction and the expression of Fgf8 and of early ridge-dependent mesodermal genes is initiated but not maintained(Revest et al., 2001). InT^-/- limb buds, Fgf10 expression is not substantially reduced until E9.5, but earlier quantitative changes in distal/apical Fgf10 expression that are difficult to appreciate could still contribute to a hypomorphic effect. However, whether decreasedFgf10 is directly due to a loss of T, or is an indirect effect of reduced AER function cannot be distinguished at present.

Restricted T expression along a narrow strip of apical mesoderm could strongly promote AER formation and morphogenesis along the DV boundary by ensuring high focal Fgf10 expression along the limb apex, or via another mesodermal signal. In the T^-/- limb bud,Fgf8 expression and pre-AER compaction are not properly reinforced along this distal DV border and remain broad and irregular. Alternatively,T could play some other role in the mesoderm, such as changing properties of apical mesodermal cells to facilitate compaction of pre-AER ectoderm towards the DV edge.

During limb induction, WNT signals maintain high Fgf10 expression in prospective limb and FGF10 activates ectodermal Wnt3a andFgf8 expression, initiating AER formation(Kawakami et al., 2001). AER signals subsequently also maintain mesodermal Fgf10 expression(Ohuchi et al., 1997).T transcripts are first clearly detected at stage 15, at the onset ofWnt3a and Fgf8 activation in the ectoderm(Kengaku et al., 1998). Both the ability of WNT3a and FGF8 to induce T expression, and the ability of T to increase subridge expression of Fgf10 early after misexpression suggest that T may be a component of the mesodermal response to developing AER signals that maintains high Fgf10 apically and thereby also maintains the forming AER, establishing a regulatory loop between ectoderm and mesoderm.

Several T-box genes are expressed in developing limb(Gibson-Brown et al., 1996;Gibson-Brown et al., 1998;Isaac et al., 1998;Logan et al., 1998;Ohuchi et al., 1998), raising the question of specificity of T function. Tbx2 andTbx3 are expressed along the anterior and posterior borders of limb mesoderm and in AER as well, but their expression and regulation suggest other than a primary role in AER formation. Both Tbx2 and Tbx3 are also expressed in the flank between limb buds and Tbx3 is down-regulated upon ectopic limb bud induction by FGF(Isaac et al., 1998). Tight regulation of Tbx3 by SHH (Tumpel et al., 2002) and loss of posterior forelimb elements in humans with TBX3 mutations [Ulnar-Mammary Syndrome(Bamshad et al., 1997)] both suggest a more direct role for Tbx3 in AP patterning. Tbx5and Tbx4 are expressed selectively in forelimb or hindlimb mesoderm,respectively, and regulate limb identity(Rodriguez-Esteban et al.,1999; Takeuchi et al.,1999; Bruneau et al.,2001). Loss of Tbx5 in zebrafish interferes with pectoral fin bud induction and Tbx5 misexpression in very early chick lateral plate induces ectopic limbs (Ahn et al.,2002; Ng et al.,2002), showing that Tbx5 also regulates limb outgrowth. However, misexpression of Tbx5 or Tbx4 in prospective limb field cause transformations in limb identity and limb truncations(Rodriguez-Esteban et al.,1999; Takeuchi et al.,1999) quite different from T misexpression phenotypes. The identities of T-box family targets coordinated throughout limb formation remain to be explored. Because of their different expression patterns and distinct misexpression phenotypes, it is unlikely that T and other T-box factors regulate targets redundantly. One interesting possibility is that T and Tbx5 may cooperate in AER regulation, given the dimerization potential of T-box proteins(Muller and Herrmann, 1997). Such dimerization might regulate novel targets, since binding site analyses indicate that different T-box proteins, including T and TBX5, recognize distinct targets in vivo and in vitro(Conlon et al., 2001;Ghosh et al., 2001).

Intriguingly, T is expressed in association with several sites proposed to `relay' signals for limb induction from midline to periphery(reviewed by Martin, 1998)including node/early notochord, somites and nephric duct epithelium adjacent to nephrogenic mesoderm. Kawakami et al.(Kawakami et al., 2001)propose that such relay sites may represent points of cross talk for FGF and WNT signals, raising the possibility that T plays a role in signal relay at these sites. Such questions are best addressed with genetic approaches, such as a conditionally targeted T allele, currently being developed.

We are grateful to Bernard Hermann, Brigid Hogan, Steve Hughes, Alex Joyner, Randy Johnson, Andy McMahon, Richard Maas, Gail Martin, Bruce Morgan,Lee Niswander, Tsutomu Nohno, Sumihare Noji, Dave Ornitz, Bob Riddle and Cliff Tabin for providing plasmids, Grace Panbanigan for the generous gift of affinity purified DLX-antibody, the Genetics Institute for recombinant FGF4 protein, and to Chuxia Deng and Dave Levens for discussions and comments on the manuscript.