Differential regulation of T-box and homeobox transcription factors suggests roles in controlling chick limb-type identity

In recent years, rapid advances have been made in our understanding of the initiation and patterning of the chick limb bud, mostly focusing on genes that are expressed and have common roles in both wing and leg buds. Although homologous structures, however, each type of limb is distinct in its skeletal elements, tendon attachments, muscles, featherbuds and scales. In addition, the wing and leg buds are formed at particular axial levels along the main body axis.

Experiments have indicated the existence of limb-specific territories in the flank. Nerves deflected from the limb into the flank of the newt can induce ectopic limbs to form from this tissue. Whether a forelimb or hindlimb forms is dependent upon the location to which the nerve is moved (Guyenot et al., 1948; Guyenot and Schott, 1926). Similarly, forelimbs and hindlimbs are induced in the anterior or posterior of the flank, respectively, depending on the location of grafted otic vesicles (Balinsky, 1925; Filatow, 1927) and nasal placodes (Balinsky, 1957; Glick, 1931). More recently, similar results have been obtained using FGF-soaked beads applied to chick interlimb flank mesoderm (Cohn et al., 1995, 1997; Crossley et al., 1996; Vogel et al., 1996).

The molecular nature of positional information responsible for the specification of limb-type identity has remained elusive. We reasoned that, if a gene is involved in the specification of limb-type identity, its expression is likely to be restricted to one or other limb type and that this restricted pattern would be maintained throughout the early stages of limb formation. While a number of genes of the HoxC cluster are differentially expressed in the hindlimb or forelimb (Simon and Tabin, 1993; Oliver et al., 1988; Nelson et el., 1996), none of these genes are expressed throughout the limb mesenchyme and their relatively late onset of expression in the limb bud indicates they do not play an early, formative role in specifying limb identity. Recently, three genes have been reported that fit these criteria for controlling limb identity. Two T-box family transcription factors, Tbx5 and Tbx4, are expressed in the forelimb bud and hindlimb bud, respectively (Gibson-Brown et al., 1996; Simon et al., 1997). T-box genes are characterised by the presence of a highly conserved motif (T-box) that encodes a 180 amino acid DNA-binding domain, the T-domain. The prototype of this family is the mouse T protein (Kispert and Herrman, 1993; Kispert et al., 1995) and its Xenopus homologue, Brachyury (Conlon et al., 1996). Members of this family are expressed throughout development and are important in tissue specification, morphogenesis and organogenesis (Papaioannou and Silver, 1998, Chapman et al., 1996; Chapman and Papaioannou, 1998). In addition, an Otx-related, paired-type homeobox transcription factor, Ptx1 (also named P-Otx) has been described in the mouse that is expressed in the developing hindlimb and not the developing forelimb (Shang et al., 1997; Szeto et al., 1996).

The chick embryo is a powerful system for embryological manipulation. To examine the possible functional roles of these genes in limb-type specification, we cloned the chicken homologues of Ptx1, Tbx4 and Tbx5 and, in the process, also isolated homologues of Tbx2 and Tbx3. As a first step toward evaluating the possible roles of these genes, we exploited the ability of FGF-soaked beads to induce ectopic limbs of wing, leg and intermediate identity and investigated the respective gene expression profiles. Our data are consistent with a role of cTbx4, cTbx5 and cPtx1 in limb-type specification.

Fertilised White Leghorn chicken embryos were obtained from SPAFAS, Norwich, CT. Eggs were incubated at 37.5°C. RNA was extracted from whole embryos, dissected wings and dissected legs of stage 22-23 (Hamburger and Hamilton, 1951). Chick embryos using the Trizol reagent (Gibco, BRL) following the manufacturers instructions. 36-48 limbs were homogenised with 0.8 ml Trizol and the isolated total RNA resuspended in water. 5 μg of total RNA was used in a first-strand synthesis reaction using Superscript reverse transcriptase (Gibco, BRL) following the manufacturers protocol. The reaction was diluted in 400 μl 10 mM Tris pH 8, phenol extracted and ethanol precipitated. The DNA was resuspended in 10 μl of 10 mM Tris pH 8 and used as template in degenerate PCR reactions (50 μl PCR reaction containing 67 mM Tris pH 8.8, 4 mM MgCl₂, 16 Mm (NH₄)₂SO₄, 33 μg/ml BSA, 400 μM NTPs, 10 μM primer, 2 μl template, 1 μl Taq polymerase (Boehringer Mannheim). The reaction was run for 40 cycles of 94°C, 1 minute; 40°C, 2 minutes; 60°C, 3 minutes. Degenerate primers were designed from an alignment of T-box sequences: TbxFwd: CACATCGATGTAC/TATICAC/TCCIG-AC/TA/TC/GICC; TbxRev: CACCTCGAGTA/GTC/GA/GTTC/GT-GA/GTAIC/GCIGTIAC (I=deoxyinosine). PCR products of the expected size were gel purified (Qiaquick columns, Qiagen) and cloned into the pGEM T vector (Promega, Madison, WI) following the manufacturers ‘ instructions. The resulting clones were sequenced (Sequenase kit, Amersham) and were compared to sequences in GENBANK using the BLAST network service, National Center for Biotechnology Information (NCBI). The chick sequences were most closely related to the mouse (U15566) and human (S81264/U28049) Tbx2 clones. We used a cTbx2 clone isolated from the degenerate PCR screen (cTbx2-13) and two mouse probes, mTbx1 and mTbx3 (a gift from Javier Capdevilla, Bollag et al., 1994) that span the T-box of each gene to screen a random primed plasmid library (Ausubel et al., 1987) prepared from chick stage 22 limb RNA (Johnson and Tabin, 1997). Filters were hybridised in 30% formamide; 10% dextran sulphate; 2× SSC, 1% SDS at 42°C overnight and washed in 2× SSC, 0.1% SDS at 55°C. Positive clones were sequenced manually (Sequenase, Amersham) and by automated sequencing (carried out by the Biopolymer Facility, HHMI, HMS using an ABI 373 automated sequencer), identifying partial clones for cTbx2, cTbx3 and cTbx5. Clones of cTbx4 were isolated in a separate screen. From the available mouse (mTbx4) and amphibian (NvTbx4) sequences primers 5TB-8Cla: GAGATCGA TAT/C AAA/G TTT/C GCI GAT/C AAT/C AAA/G TGG and 3TB-7Xho: CACCTCGAG CC C/TTT IGC A/GAA IGG G/ATT G/ATT T/CTC were designed and employed in degenerate PCR . Using the resulting PCR product as a probe we screened an oligo(dT)-primed chick limb (stages 18-24) Lambda Zap cDNA library (gift from Susan Mackem, NIH) as described above. Four clones were isolated and confirmed by sequencing.

To isolate chick homologues of the Ptx1/P-OTX gene, we screened an oligo(dT)-primed library (as described above) with a mouse probe generated by PCR. Based on the sequence of mPtx1/P-OTX (Lamonerie et al., 1996; Szeto et al., 1996), we designed primers OTHinf: CCAGCGAATCGTCCGACGCT and OTXho: CAAGGCTCGAGTTGCACGTG. A 720 bp fragment PCR

fragment (nucleotides 167-887) was generated using the One-Step RT-PCR kit (Life Technologies-a kind gift of Dr Jun Lee) following the manufacturers instructions and total RNA, extracted from a single E14.5-15 mouse head, as template. PCR conditions were 55°C, 30 minutes, 94°C, 2 minutes for the first-strand synthesis, followed immediately by PCR amplification, 94°C, 15 seconds; 55°C, 30 seconds; 72°C, 90 seconds, 40 cycles, 72°C, 5 minutes. A single band of the correct size was cloned into the pGEM T vector (Promega) following the manufacturers instructions. Sequencing confirmed that the clone contained a fragment of the mouse Ptx1/P-OTX gene.

Whole-mount in situ hybridisation was carried out essentially as described (Riddle et al., 1993). Probe templates were: cTbx2-15f (SalI, T7 RNA polymerase), cTbx3-06d (SalI, T7 polymerase), cTbx5-08c (SalI, T7 polymerase), cTbx4-R1-Xba (EcoRI, T3 polymerase), cPtx1-OT17 (PstI, T7 polymerase). All the probes span large portions of the respective open reading frames including the T-box/homeobox of each gene. Section in situ hybridisation was carried out as previously described (Bao and Cepko, 1997).

Affi-Gel beads (150-300 μm; Bio-Rad) were washed in phosphate-buffered saline and then soaked in a solution of FGF4 (1 μg/μl, a gift from Vickie Rosen, Genetics Institute, Cambridge, MA) for at least 1 hour before use. All operations were done on stage 14-16 embryos. Using tungsten needle, a narrow cut was made in the surface ectoderm over the lateral plate mesoderm of the flank. With forceps, an FGF4-soaked bead was placed into direct contact with the flank mesenchyme by inserting the bead through the cut and by tucking it under the ectoderm. Embryos were harvested after incubation for either 1.5 or 8 days following implantation of a bead.

The distal tip of the wing and leg buds were removed and transplanted essentially as described (Kieny, 1964). The distal 1/3 of either a wing or leg bud from a stage 22-23 embryo was removed using a tungsten needle. In most cases a reciprocal transplant procedure was carried out on a single embryo such that the distal 1/3 of the wing bud was transplanted onto the proximal stump of the leg bud and vice versa. The grafted tissue was held in place using staples made from small strips of platinum wire (Goodfellow, Cambridge, England). Wing- and leg-cell implants were done by making a deep incision into either the distal, wing or leg bud and removing a piece of tissue (with ectoderm intact), approximately 50 μm³. The removed tissue was replaced with a similar-sized piece from either the leg or wing.

Prior to removal of the AER, 3-5 μl of dilute (∼0.1%) Alcian Blue (Aldrich) solution was applied externally to improve visibility of AER. Small incisions were made between the ridge and underlying mesenchyme using a tungsten needle, being careful not to damage the mesenchyme cells. By making a series of such incisions the AER was removed.

Embryos were fixed overnight in 4% paraformaldehyde /phosphate-buffered saline at 4°C, washed thoroughly in phosphate-buffered saline and eviscerated. The skeletal elements were stained in a fresh solution of Alcian Blue 8GX (C.I. 74240; Sigma) (20 mg in 70% ethanol/ 30% glacial acetic acid) for 6-12 hours. Embryos were then brought through an ethanol series into distilled water and then cleared in 0.5% KOH and a stepwise series of 0.5% KOH/glycerol solutions into 100% glycerol. Cartilage stains deep blue while calcified tissues do not take up the stain.

To identify chick genes differentially expressed in the forelimb or hindlimb buds, we screened cDNA libraries for clones related to the T-box and Ptx gene families. Multiple T-box cDNA isolates were sequenced. An alignment of the deduced T-domain sequences for cTbx2, cTbx3, cTbx4 and cTbx5 with the respective mouse and human sequences (Agulnik et al., 1996; Bamshad et al., 1997; Basson et al., 1997) indicated a remarkably high degree of conservation for each gene (Fig. 1A). The cTbx2 T-domain is 100% conserved between chick, mouse and human at the amino acid (a.a.) level (182 of 182 a.a.) and the T-domain of cTbx3 shares 95% (173

of 182 a.a.) and 97.8% (178 of 182 a.a.) identity to mouse and human, respectively. The T-domain of cTbx5 is 96.7% (178 of 184 a.a.) and 97.8% (180 of 184 a.a.) conserved to the mouse and human sequence, respectively. For cTbx4, only a mouse sequence is available for comparison and this has 99.4% identity to the deduced chick peptide differing only by the presence of an extra alanine residue (Fig. 1A, a.a. residue 68). A phylogenetic tree constructed based on the alignment of T-domain sequences confirmed the assignment of each chick gene as an orthologue of the mouse and human genes (Fig. 1B). The Tbx2/3/4/5 genes constitute a subfamily within the larger T-box-containing gene family (Bollag et al., 1994). Within the subfamily, Tbx2 and Tbx3 are most similar to one another whilst Tbx4 and Tbx5 are more closely related (Bollag et al., 1994 and Fig. 1B). The crystal structure of the Xenopus Brachyury(T) T-domain (Muller and Herrman, 1997) confirmed previous suggestions that the T-protein binds DNA as a homodimer (Papapetrou et al., 1997) and identified specific amino acid residues within the T-domain as important in DNA-binding and dimerisation. These residues are almost completely conserved (Fig. 1A).

In a separate screen, clones of the chick Ptx1 (cPtx1) gene were isolated. Comparison of the complete deduced amino acid sequence with sequences of mouse Ptx1 (P-Otx) and human PTX1 (backfoot) (Semina et al., 1996, 1997; Shang et al., 1997) confirmed that the isolates represented a Ptx1 orthologue (Fig. 1C). Within the paired-type homeodomain, the amino acid sequence is 100% conserved and, outside the homeodomain sequence, conservation is generally maintained.

To gain insight into their possible functions, the spatial expression patterns of cTbx2, cTbx3, cTbx4 and cTbx5 and cPtx1 were first examined by whole-mount in situ hybridisation from the early stages of limb formation through to a time at which the limb buds are morphologically distinct (stages 17-29).

At stage 17, expression of cTbx5 is detected in the developing wing-bud mesenchyme (Fig. 2A) and, by stage 23, high levels of cTbx5 expression are detected there. At no stage is expression detected in the leg bud (Fig. 2B). There is also transient expression of cTbx5 in the flank, extending from the posterior of the wing bud to 2/3 of the interlimb flank region, from stage 22-24 (Fig. 2B and data not shown). Section in situ hybridisation analysis of the wing bud at stage 23 shows cTbx5 expression throughout the wing mesenchyme, although at lower levels distally and dorsally (Fig. 2B). No expression was detected in the ectoderm or AER but high levels of expression were detected in the flank immediately under the wing bud (Fig. 2B inset). By stage 29, the distinct skeletal elements of both the wing and the leg are beginning to form. At this stage transcripts of cTbx5 are detected throughout the limb mesenchyme, with high levels interdigitally and lower levels proximally and in areas of condensing cartilage (Fig. 2C).

In contrast to cTbx5, cTbx4 is expressed throughout the early leg-bud mesenchyme but is not detected in the wing bud (Fig. 2D-F). At stage 29, however, a very low level of cTbx4 expression is detected in the wing while high levels of transcripts are detected throughout the leg bud mesenchyme except at the very proximal and distal extremes (Fig. 2F). By section in situ analysis, transcripts are detected throughout the leg bud mesenchyme at stage 23 and, similarly to cTbx5 in the wing, expression of cTbx4 appears lower in the dorsal and distal extremes of the leg bud. Again, high levels of expression were detected in the flank mesenchyme directly underneath the leg bud (Fig. 2E inset). Unlike cTbx5, no expression of cTbx4 was detected in the interlimb flank during stages 18-29 (Fig. 2D-F and data not shown).

The expression pattern of cPtx1 in developing limb is very similar to cTbx4. At stage 18 through stage 24, cPtx1 is restricted to the leg-bud mesenchyme (Fig. 2G-I and data not shown). In later stage embryos, cPtx1 expression is detected in cells at the most proximal extent of the leg and in the flank of the embryo at the level of the leg bud (Fig. 2I and data not shown). No expression of cPtx1 is detected in the wing bud or in the ectoderm of the leg bud at any of the stages analysed. Section in situ hybridisation analysis at stage 23 confirmed that transcripts are present throughout the leg mesenchyme but are not present in the ectoderm or AER (Fig. 2H inset). Similar to cTbx4, high levels of expression are detected in the flank underneath the leg bud. However, unlike cTbx4, cPtx1 is expressed at high levels in the dorsal and distal extremes of the limb bud and at lower levels ventrally.

To understand when these genes may be first acting in limb formation, we examined their expression at prelimb bud stages. cTbx5 and cTbx4 are, respectively, first detected in the presumptive wing- and leg-forming regions of the lateral plate mesoderm at stage 15 (data not shown). cPtx1 expression is first detected at stage 14 in an area that extends from the level of the last somite to the tip of the tail and includes the prospective leg-forming regions (Fig. 2J). Therefore, transcripts for all three genes are present in the presumptive leg and wing mesenchyme many hours prior to overt outgrowth of the limb and appear to identify the area of the lateral plate mesoderm destined to contribute to the limb. As such, they provide useful early markers to distinguish the presumptive wing- and leg-forming regions of the lateral plate mesoderm.

Both cTbx2 and cTbx3 are initially expressed throughout the flank, lateral plate mesoderm at stage 18 (Fig. 3A,D). Expression levels are higher in the anterior and posterior domains of the nascent bud. At stage 23, cTbx2 is expressed in two proximodistal bands of cells, one in the anterior and the other in the posterior of the limb, excluding approximately the medial two-thirds of the limb-bud mesenchyme (Fig. 3B). Section in situ hybridisation analysis indicates that cTbx2 is excluded from the most distal limb bud (Fig. 3B and inset). At stage 29, cTbx2 is still expressed in anterior and posterior domains but transcripts are not detected in the hand and foot plate (Fig. 3C). cTbx3 is expressed in a similar pattern. cTbx3 is expressed in two proximodistal bands of cells in the anterior and posterior of the limb bud mesenchyme (Fig. 3E). Section in situ hybridisation reveals that mesenchymal expression is restricted to proximal and dorsal parts of the limb bud (Fig. 3E inset). However, there are some important distinguishing features of the cTbx3 expression pattern. Both the anterior and posterior bands of cTbx3 expression are broader than those of cTbx2. Furthermore, the posterior domain of expression within the wing bud is broader than the posterior domain of cTbx3-expressing cells in the leg bud. At stage 29, cTbx3 anterior and posterior expression is maintained (Fig. 3F). A broad posterior band of expressing cells extends into the most distal tip of the wing bud. In the leg, however, expression does not extend distally into the footplate but remains restricted proximal to the most posterior digit. A low level of cTbx3 expression is also observed interdigitally in both the wing and the leg.

Classical experiments have provided information about some of the properties of limb-type specificity which can be used for evaluating potential molecular players in this process. For example, limb-type identity is specified by stage 14 (Stephens et al., 1989) and limb-type identity is independent of the limb ectoderm (Zwilling, 1956a,b). In agreement with these observations, cTbx5, cTbx4 and cPtx1 are expressed in the limb primordia by around stage 14 and expression of each gene is limited to the limb mesenchyme (Fig. 2 and data not shown). Since all three genes are expressed prior to the expression of sonic hedgehog (shh) in the zone of polarising activity (ZPA) and prior to the formation of an apical ectodermal ridge (AER), the initial control of their expression must be independent of these signalling centres. Once these signalling centres form, a positive feedback loop is set up between FGFs produced in the AER and shh in the ZPA (Laufer et al., 1994; Niswander et al., 1994). Surgical removal of the AER removes the source of FGFs and results in a rapid down-regulation of shh. To test the potential role of shh and FGFs in the maintenance of cTbx5, cTbx4 and cPtx1, we investigated their expression after AER removal. Expression of cTbx5, cTbx4 and cPtx1 is maintained throughout the limb bud up to 24 hours after the removal of the AER (Fig. 4A-F), indicating that maintenance of cTbx5, cTbx4 and cPtx1 expression is independent of shh and of signals from the AER.

It has also been shown that distal wing or leg tissue grafted to a leg-bud or wing-bud stump retains its original identity and develops, respectively, into distal wing or leg structures (Kieny, 1964). We tested whether expression of cTbx5, cTbx4 and cPtx1 is maintained in such grafts. The distal third of wing and leg buds were transplanted at stage 22-23 and experimental embryos were analysed 12 and 24 hours after the surgery. In all examples analysed, distal wing tissue grafted to a leg stump maintained expression of cTbx5 (Fig. 4G) and grafted leg tissue maintained the expression of cTbx4 (data not shown) and cPtx1 (Fig. 4H). In all examples tested, the grafted tissue remained negative for cTbx5 or cTbx4/cPtx1 expression appropriate to the host site (data not shown). The ability of grafted tissue to retain the limb-type identity of its original environment was also tested with smaller cell-implants of wing and leg tissue. Prospective thigh mesoderm transplanted to the distal wing bud is capable of forming ectopic toes in the wing (Saunders, 1957). In similar cell transplant experiments, analysed 24 and 48 hours after the cell transplant, the grafted cells maintain gene expression corresponding to their origin (Fig. 4F and data not shown).

In summary, cTbx5, cTbx4 and cPtx1 are expressed and maintained in normal development and following surgical manipulation in patterns which suggests that these genes may play important roles in the specification of limb-type identity in the chick.

In order to gain insight into the possible roles of cTbx4, cTbx5 and cPtx1 in the developing limb we exploited the ability of an exogenous source of fibroblast growth factor (FGF) to induce ectopic limb buds. Beads soaked in FGFs, applied to the interlimb lateral plate mesoderm at stages 13 to 17, are capable of inducing ectopic limb outgrowths. Significantly, beads applied at levels closer to the presumptive wing bud generally give rise to wing-like structures whereas beads closer to the presumptive leg bud, generate ectopic leg-like outgrowths (Cohn et al., 1995, 1997; Crossley et al., 1996; Ohuchi et al., 1995; Vogel et al., 1996). We applied FGF4-soaked beads to the interlimb lateral plate mesoderm of stage 15 embryos at various axial levels and analysed the embryos after 2 and 8 days incubation. The early harvests were analysed by whole-mount in situ hybridisation with probes for cTbx4, cTbx5 and cPtx1 whereas the skeletal elements and featherbud patterns of later harvests were examined histologically. Some embryos destined for late stage harvest were photographed in ovo after 2 days so that the position of the early ectopic bud could be compared to the structures that formed.

When an FGF4-soaked bead is placed close to the presumptive wing bud (approximately at the level of somite 21 and 22), an ectopic outgrowth is visible after 2 days, adjacent to the normal wing (Fig. 5A-C). cTbx5 is most often expressed throughout such ectopic outgrowths (Fig. 5A; Table 1). However in some examples cTbx5 is expressed only in the more anterior of the ectopic bud (Fig. 5B). Single and two probe in situ hybridisation with probes for cTbx5 and/or cTbx4/cPtx1 indicated that the cells in the posterior of the ectopic limb that did not express cTbx5, expressed cTbx4 or cPtx1 (Fig. 5C and data not shown). Conversely, when an FGF4-soaked bead is placed close to the presumptive leg mesenchyme (approximately at the level of somite 24/25), the resulting ectopic outgrowth forms adjacent to the normal leg bud (Fig. 5E-G). In these examples, commonly the ectopic limb buds are composed of cells expressing cTbx4 and cPtx1 (Fig. 5G; Table 1). In some cases however the outgrowth expresses cTbx5 in the anterior portion and cTbx4/cPtx1 through the remainder of the limb bud (Fig. 5E,F; Table 1, and data not shown). Placement of an FGF-soaked bead centrally between the prospective wing- and leg-forming regions usually gives rise to ectopic limb buds that arise from cells in the middle of the interlimb flank (Cohn et al., 1995, 1997; Crossley et al., 1996; Vogel et al., 1996). These ectopic outgrowths often contain separate domains expressing cTbx5 and cTbx4/cPtx1 (Fig. 5D; Table 1 and data not shown). Interestingly the rostral-caudal boundary of cells that can be induced to ectopically express cTbx5 or cTbx4/cPtx1 is not fixed, but varies across a relative wide region within the interlimb flank (compare panels in Fig. 5).

As a further test of the roles of cTbx5, cTbx4 and cPtx1, we examined and compared ectopically induced limb buds for morphology after 8 days or for expression patterns after 2 days incubation, in parallel experiments. Those incubated for 8 days were first observed 2 days after bead implantation and a photographic record made of the location of the ectopic limb bud for comparison to those analysed by in situ hybridisation.

An anterior ectopic limb buds that formed adjacent to the normal wing bud, are generally wing-like, containing characteristic wing digits (Fig. 6A and inset; Table 2) and more proximal wing bones (not shown). In some cases, only the proximal and medial wing bones are present (data not shown). In a rare case, the ectopic limb bud developed into a structure with wing-like character proximally, consisting of a humerus, radius and ulna but, the single digit that formed is clearly toe-like (Fig. 6B).

Limbs that develop from the mid-flank almost always have a mosaic phenotype and with some skeletal elements characteristic of a wing and some of a leg (Fig. 6C,E; Table 2). Some limbs contain wing proximal elements (humerus and ulna) but distally, in the anterior, a wing digit forms while in the posterior, a leg digit 1 and 2 are obvious (Fig. 6C). Other examples have leg proximal elements (femur and tibia) while distally again the identity of the elements are split with anterior wing digits 4 and 3 and posterior leg digits 1 and 2. As previously described (Cohn et al., 1995; Crossley et al., 1996; Vogel et al., 1996), all ectopic skeletal elements form in mirror image of the normal wing and leg skeletal elements (Fig. 6C,E).

Posterior ectopic limb buds most often give rise to leg-like structures (Fig. 6D; Table 1), fused to the normal leg at the hip joint. Consistent with our observation that posterior ectopic limb buds occasionally contain small anterior areas of cTbx5 expression (Fig. 5E,F and data not shown), some such buds gave rise to mostly leg-like structures that contain a single anterior wing digit 4 and/or a ridge of featherbuds on their anterior surface (data not shown).

The most striking aspect of cTbx5, cTbx4 and cPtx1 expression is that each gene is highly restricted to either the wing or leg bud. The localisation of transcripts of these genes exclusively to the mesenchyme and not the ectoderm is consistent with evidence that the property of limb-type identity resides in the limb mesenchyme and is independent of the limb ectoderm (Zwilling, 1956a,b). All three genes are expressed in the prospective limb mesenchyme prior to the overt formation of a limb bud. The onset of expression of all three genes is coincident with the time at which the limb fields gain the ability to differentiate independently into either a wing or a leg when grafted to a host coelom (Stephens et al., 1989). Furthermore, the specific limb-type expression pattern of cTbx5 and cTbx4/cPtx1 is maintained in limb tissue grafted to a new location, consistent with the ability of such grafts to maintain their limb-type identity in a different environment.

Additional evidence supporting a role for these genes in limb-type specification is provided by the analysis of ectopic limbs induced by FGF. The type of ectopic structure that forms is related to the rostrocaudal position at which the source of FGF is applied (Cohn et al., 1995, 1997; Crossley et al., 1996; Vogel et al., 1996 and reported herein). A bead placed adjacent to the wing commonly induces an ectopic wing-like structure, while a bead placed adjacent to the leg commonly induces an ectopic leg-like element. We observe that the type of gene induced in the ectopic outgrowths correlates well with the structures that generally develop from such outgrowths. Anterior, ectopic outgrowths contain cTbx5-expressing cells, posterior ectopic outgrowths contain cTbx4/cPtx1-expressing cells and outgrowths from the middle of the flank have mosaic wing/leg identity. Although these genes are expressed exclusively in the mesenchyme, they appear to have an effect on the formation of ectodermal derivatives. In mosaic buds, wing-like featherbud patterns are apparent on the anterior surface while the posterior ectoderm has more leg-like features.

Candidates for endogenous FGFs involved in limb bud initiation include FGF8, which is transiently expressed in the intermediate mesoderm, adjacent to the wing and leg fields, just prior to limb-bud outgrowth (Crossley et al., 1996; Vogel et al., 1996) and FGF10, which is expressed in lateral plate mesoderm at stage 14/15 (Ohuchi et al., 1997). In Xenopus, a regulatory relationship between the T-box gene Brachyury and embryonic FGF (eFGF) has been described (Schulte-Merker et al., 1994). A similar regulatory mechanism may exist in the lateral plate mesenchyme between FGFs and cTbx5, cTbx4 and cPtx1. All three genes are initially expressed approximately coincident with the time at which FGF10 and FGF8 are first detected in the lateral plate mesoderm and intermediate mesoderm, respectively. It remains unclear what regulates the differential induction of cTbx5 or cTbx4/cPtx1 in anterior or posterior positions in response to FGF. Hox genes are expressed at defined rostrocaudal levels in the lateral plate mesoderm and their expression patterns in ectopic limbs change in response to FGF (Cohn et al., 1997). However, the border between cTbx5- and cTbx4/cPtx1-expressing cells in FGF-induced ectopic limb buds is not fixed on the rostrocaudal axis. Thus the molecular mechanism that specifies which of the limb-identity genes will be activated remains unclear.

Direct evidence for a role of TBX5 in the formation of the forelimb come from mutations in the human TBX5 gene (Basson et al., 1997; Li et al., 1997) which cause Holt-Oram Syndrome (HOS), a pleiotropic, autosomal dominant, developmental disorder in man. The syndrome is characterised by skeletal abnormalities in the upper limbs and cardiac septation defects. The skeletal abnormalities can vary from a triphalangeal thumb to a range of reduction deformities which in the most severe cases include phocomelia, an almost complete absence of limb elements.

Mutations in human TBX3 cause Ulnar-Mammary Syndrome (UMS), a pleiotropic disorder affecting limb, apocrine gland, tooth and genital development (Bamshad et al., 1997). Mutations characterised are predicted to cause haploinsufficiency of the TBX3 protein. This suggests that, similarly to TBX5, critical levels of this transcription factor are required for normal morphogenesis. Limb malformations in UMS are typically deficiencies, fusions or duplications of the posterior elements of the upper limb, including the ulna, metacarpals and phalanges (Bamshad et al., 1996, 1997). In contrast, limb malformations in HOS are most often observed in the anterior elements of the upper limb, primarily the radius, carpal and ulna bones. Thus TBX3 and TBX5 may play complimentary roles, patterning posterior and anterior elements, respectively. cTbx3 is expressed in both posterior and anterior domains and, in the chick at least, the posterior domain correlates well with the position of the cells fated to form the more posterior elements. The cells in the anterior domain are not fated to form any of the skeletal elements and indeed many undergo apoptosis (Saunders, 1966), which may explain the lack of a visible phenotype in the anterior of the limb. Deformities observed in Ulnar-Mammary Syndrome indicate a role for TBX3 in forelimb formation whereas hindlimbs are mostly unaffected (Bamshad et al., 1996), consistent with the clear distinction between the chick wing and leg expression domains of cTbx3 at later stages (Fig. 3F). In the developing wing, a much broader domain of expression for cTbx3 is evident compared to the leg, where expression extends into the hand plate but is absent in the foot plate.

Our results strongly support the hypothesis that Tbx5, Tbx4 and Ptx1 are likely to have important functions in limb-type specification. Other, independent studies of cTbx4 and cTbx5 in normal and ectopic chick limbs have reached identical conclusions (Ohuchi et al., 1998; Isaac et al., 1998; Papaioannou et al., 1998). Further functional studies will help determine where in the hierachy of molecular regulatory pathways these transcription factors act to establish limb identity.

We thank Dr Vicki Rosen (Genetics Institute, Cambridge, MA) for the kind gift of FGF4 protein. We also thank the labs of Drs Cheryl Tickle and Juan-Carlos Izpisua-Belmonte for communicating their results prior to publication. M. L. was funded by a postdoctoral fellowship from the Human Frontiers Science Program Organisation (HFSPO). C. T. is supported by a grant from the NIH.