Lhx2, a vertebrate homologue of apterous, regulates vertebrate limb outgrowth

The limbs of fruit flies and vertebrates are not only morphologically quite distinct, but also their development proceeds very differently. In Drosophila, ectodermal cells destined to give rise to limbs are set aside in the embryo. These cells divide in the larval stages to form invaginated epithelial sacs termed imaginal discs. During metamorphosis this epithelium unfolds and grows outward to form the adult limb (Cohen, 1993). In contrast, vertebrate limbs develop through a continued localised proliferation of the lateral plate mesoderm giving rise first to limb buds and ultimately to the adult limb (Tickle and Eichele, 1994).

The first hint that limb outgrowth in these diverse species might after all be based upon a similar genetic foundation was the finding that a gene known to be involved in limb outgrowth in Drosophila, distal-less (Cohen et al., 1989; Cohen, 1990) was also expressed in the distal regions of mouse and newt limbs (Beauchemin and Savard, 1992; Dollé et al., 1992). More recently, distal-less has been found in the appendages of a wide range of different species across several phyla (see Popadic et al., 1996 and references therein). Subsequently, many other genes that play a role in patterning the Drosophila appendages have been identified and shown to play a similar role in the vertebrate limb. These similarities are particularly clear for the genes acting along the anteroposterior and proximodistal axes (for reviews see (Lawrence and Struhl, 1996; Shubin et al., 1997). In contrast, homologies in the genes involved in specifying the dorsoventral (DV) axis are less obvious.

Two of the vertebrates genes that specify DV identity, Wnt-7a (Parr and McMahon, 1995; Yang and Niswander, 1995) and En-1 (Loomis et al., 1996), do not appear to have functional equivalents in Drosophila. A third gene, Lmx-1 (Riddle et al., 1995; Vogel et al., 1995), has been proposed to be a homologue of the Drosophila apterous gene, since not only do they share sequence similarity but also both genes are sufficient to specify dorsal cell fate (Blair, 1993; Diaz-Benjumea and Cohen, 1993; Williams et al., 1993; Blair et al., 1994). Importantly, besides determining dorsality, apterous is also necessary for wing outgrowth and contributes to the positioning of the wing margin at the dorsoventral boundary through regulating the expression of the fringe and Serrate genes (Irvine and Wieschaus, 1994; Kim et al., 1995). It appears that Lmx-1 does not share this activity since mis-expression of the Lmx-1 gene has no effect on limb outgrowth, or on apical ectodermal ridge (AER, a key structure required for limb outgrowth) (Summerbell et al., 1973; Todt and Fallon, 1984) formation.

In this paper we show that Lmx-1 is not only sufficient, but also necessary, to specify dorsal cell fate, since down-regulation of Lmx-1 activity results in limbs lacking dorsal-specific structures. We also demonstrate that Lmx-1 is neither able to induce R-fng when mis-expressed in the limb or the flank of the embryo, nor Drosophila fringe when ectopically expressed in flies. Further, we describe the cloning of another LIM-homeodomain gene, Lhx2, that shows higher sequence conservation with apterous and, unlike Lmx-1, when mis-expressed in the flank of the embryo induces ectopic R-fringe expression. Lhx2 is also able to induce fringe and Wingless (downstream targets of apterous) expression in Drosophila. Consistent with these data, down-regulation of Lhx2 activity causes a down-regulation of genes required for the outgrowth of the limb along its proximodistal axis and consequently results in arrested limb outgrowth. Finally, unlike Lmx-1, Lhx2 does not specify dorsal cell fate.

Our data suggest that Lhx2 could be a bona fide vertebrate homologue of apterous yet, in vertebrates, it does not specify dorsal cell fate. This raises the question of whether in vertebrates, unlike in Drosophila, limb outgrowth can be dissociated from the establishment of the DV axis and brings into focus the questions of whether or not vertebrate and Drosophila limbs can be considered to be homologous structures and how limbs in flies and vertebrates have evolved to be so different.

The open reading frame of the Drosophila apterous gene was used to screen a stage 20-22 HH cDNA library following standard procedures (Sambrook et al., 1989). The positive clones were sequenced with an automated sequencer.

Whole-mount in situ hybridisation was carried out as described (Wilkinson, 1993) with some minor modifications (Izpisúa Belmonte et al., 1993). Sectional in situ hybridisations were performed as described previously (Izpisúa Belmonte et al., 1991b). The probe used for Lhx2 (700 bp) encompasses the homeobox and the second LIM domain. The Wnt-3a probe (362 bp) was a kind gift of A. MacMahon and C. Tabin. The remaining probes have been described elsewhere: Wnt-7a (Dealy et al., 1993), Eng-1 (Logan et al., 1992), hoxd-13 (Izpisúa Belmonte et al., 1991a), Eph-7a (Kengaku et al., 1998), Fgf-10 (Ohuchi et al., 1997), Msx-1 (Robert et al., 1989), Shh (Ogura et al., 1996), Serrate-2 (Myat et al., 1996), Lmx-1 (Vogel et al., 1995), Fgf-8 (Vogel et al., 1996) and R-fng (Rodriguez-Esteban et al., 1997). apterous antibody staining was done according to Capovilla et al. (1994). In some cases the embryos were dehydrated in 30% sucrose, embedded in gelatin, frozen and sectioned with a cryostat. Cartilage and muscle staining were performed as described (Vogel et al., 1996).

Chicken embryos, obtained from either MacIntyre Poultry (San Diego, CA) or SPAFAS (Norwich, CT), were infected with RCAS(BP)A viruses containing the following constructs. Eng-RD/Lmx-1-HD was constructed by fusing the Drosophila engrailed repressor domain (aa 1-299) with the homeobox of the Lmx-1 gene (aa 1-60 in Fig. 3). Eng-RD/Lhx2-HD was constructed similarly by fusing the engrailed repressor domain with the homeobox of Lhx2 (as shown in Fig. 3). The Lhx2 viral vector was constructed by inserting the open reading frame of the Lhx2 into RCAS (BP)A. VP16-AD/Lhx2-HD was constructed by fusing the VP16 transactivation domain 78 aa (Kliewer et al., 1992) with the Lhx2 homeodomain. Virus preparation and injections were performed as described (Morgan et al., 1992). 408 embryos were injected with the Eng-RD/Lmx-1-HD construct and 16% of them showed an abnormal phenotype; 386 embryos were injected with the Lhx2 construct, and 22% of them showed ectopic expression of R-fng in the flank; finally, 580 embryos were injected with the Eng-RD/Lhx2-HD construct and 16% exhibited arrested limb outgrowth, which varied with the extent of the infection.

The following fly transformants were used: 71B-Gal4 (Brand and Perrimon, 1993), ap-Gal4 (Calleja et al., 1996), ptc-Gal4 (Bloomington Drosophila Stock Center), C5-Gal4 (Yeh et al., 1995), fringe-lacZ (Irvine and Wieschaus, 1994), UAS:apterous and UAS:hLhx2 (D. E. R. and J. B., unpublished). The UAS:cLmx-1 construct was created by inserting the full-length cLmx-1 cDNA into the pUAST vector (Brand and Perrimon, 1993). This construct was transformed into yw flies following standard procedures (Rubin and Spradling, 1982). For each UAS responder, at least two independent lines were used.

Whilst the ectopic overexpression of Lmx-1 indicates that this gene can induce dorsalisation of the ventral side of the limb bud (Riddle et al., 1995; Vogel et al., 1995), this experiment cannot determine whether this is the normal role of Lmx-1 on the dorsal side of the vertebrate limb, nor does it show if this is the only factor that specifies dorsal cell fate. To start addressing these issues in the chick limb we generated a chimaeric construct between the repressor domain of engrailed (kindly provided by J. Jaynes, Thomas Jefferson University, Philadelphia) and the homeodomain of Lmx-1 (Fig. 3). Since LIM homeodomains have been shown to act as transcriptional activators (German et al., 1992; Bach et al., 1995; Wang and Drucker, 1995; Szeto et al., 1996; Bach et al., 1997), this chimaeric protein, when in excess, would be expected to compete for and repress the genes normally activated by Lmx-1.

During normal limb development, the dorsal surface of the wing has a higher feather density than that of the ventral side (Fig. 1a,b). In the Eng-RD/Lmx-1-HD infected wings the feather density on what would normally be the dorsal side of the limb is greatly reduced (Fig. 1c). Furthermore, the limb is generally straighter and in a few cases a reduction in the thickening of the ulna was observed. In this case, the distal tip of digit 2 was missing. The dorsal surface of the tarsometatarsus and toes of the chick leg is covered by large scales or scuta, whilst the ventral surface has small scales or tuberculae (Fig. 1d,e). In the Eng-RD/Lmx-1-HD infected leg, the large scales of the dorsal side of certain toe areas were not present and, instead, tuberculae-like structures were observed (Fig. 1f). Digits showing this loss of dorsal-type integument also display a reduction of the claws (Fig. 1f), a structure of dorsal origin. Furthermore, instead of the normal dorsal bending of the leg an abnormal ventral curvature of the whole limb was observed (Fig. 1g). In addition, the digits of the infected legs have a more cylindrical morphology than the dorsally flattened digits of control limbs. Sectioning of the infected limbs shows that whilst the ventral muscles and tendons of the infected limbs are normal, the dorsal muscles are smaller or absent; particularly, those associated with digit 4 (interosseus, IOD) and digit 3 (extensor medius brevis; EMB). In some instances, the flexor indicis (FI), a ventral muscle associated with digit 2, was also present on the dorsal side of the limb (Fig. 1h,i). Overall, the changes we observed were a lack of dorsal structures, and only in some cases were additional ventral structures present on the dorsal side of the infected limb buds.

To determine whether these phenotypes are related to changes in ectodermal or mesodermal cell fate, several molecular markers were analyzed at different time after mis-expression of the engrailed repressor domain/Lmx-1 (Eng-RD/Lmx-HD) homeodomain fusion construct. In situ hybridisation of genes involved in the outgrowth of the limb along its proximodistal and anteroposterior axes, Lhx2, hoxd-13, Fgf-10, Msx-1 and Shh (mesenchyme, Fig. 2e-i) and R-fng, Wnt-3a, Fgf-8 (ectoderm, Fig. 2j-l) and Serrate-2 (data not shown), showed that there is no alteration in their spatiotemporal pattern of expression. Similarly, mis-expression of the Eng-RD/Lmx-1-HD chimera did not affect the three genes known to be involved in the patterning of the limb along its dorsoventral axis. Neither Wnt-7a, Lmx-1 transcripts (normally localized to the dorsal ectoderm and mesoderm, respectively (Dealy et al., 1993; Riddle et al., 1995; Vogel et al., 1995) nor En-1 (normally present in the ventral ectoderm; Logan et al., 1992) transcripts distribution was altered (Fig. 2a-c). However, in situ hybridisation to detect Eph-7A mRNA, a gene expressed in the dorsal mesoderm (Araujo and Nieto, 1997), showed down-regulation in the dorsal side of the infected limb buds (Fig. 2d). This suggests that the reduction of dorsal character observed after mis-expression of the Eng-RD/Lmx-1-HD construct is mostly attributable to changes in mesenchymal cell fates along the dorsal-ventral limb axis.

Taken together, these results suggest that Lmx-1 is not only sufficient, as was suggested previously (Riddle et al., 1995; Vogel et al., 1995), but also required to specify dorsal cell fate.

Whilst the sequence similarity (44% identity in the LIM domains and 42% in the homeodomain) prompted us and others (Riddle et al., 1995,_;Vogel et al., 1995) to suggest that the chick Lmx-1 gene might be the vertebrate homologue of the Drosophila apterous gene, we have now cloned a second LIM/homeodomain gene Lhx2 that shows significantly greater sequence similarity to Drosophila apterous (56% identity in the LIM domains and 96% in the homeodomain; Fig. 3). Lhx2 appears to be the chick homologue of a rat gene called LH2 (Xu et al., 1993) and a human gene (Lhx2), with which it shares 93% aa identity in the homeodomain and 57% aa identity in the LIM domains, respectively.

We have characterized the expression of Lhx2 during chick embryogenesis (Fig. 4). Lhx2 is expressed in the developing limb buds, nervous system, optic vesicles and nasal placodes, amongst other tissues. The expression pattern in these areas is very similar to that seen for apterous in Drosophila (Cohen et al., 1992). Lhx2 transcripts are first detected in the presumptive limb mesoderm at around stages 14-15HH, coincident with the first signs of limb budding. As limb bud outgrowth proceeds, Lhx2 mRNA is confined to the cells of the progress zone underlying the distal ectoderm, where it remains until the latest stage examined (stage 32, Fig. 4D-S). Lhx2 transcripts were never detected in the ectoderm of the limb bud. In contrast to Lmx-1 and apterous, which are dorsally restricted (Fig. 4A-C), the expression pattern of Lhx2 shows no DV asymmetry (Fig. 4D-S).

The sequence and expression data described above suggest that Lhx2 could be a vertebrate homologue of the Drosophila apterous gene. To determine whether the homology between these two genes extends to functional similarity we performed ectopic experiments in both the chick and fly embryos.

We first mis-expressed chick Lhx2 in the chick limb primordia using retroviral technology. Whilst no significant morphological perturbations were observed, in situ hybridisations indicate that infection in the flank (but not in the limb bud) leads to ectopic R-fng expression (Fig. 5A). The ectopic expression of R-fng, which was restricted to the flank ectoderm, was not accompanied in any case by the formation of an ectopic AER. This is supported by the absence of Fgf-8 expression (data not shown). Similar experiments with Lmx-1 did not induce ectopic fringe expression (Fig. 5B). These results suggest that Lhx2, but not Lmx-1, may play a role in regulating R-fng expression and that some other factor(s), expressed in the limb primordia but not in the flank, cooperate(s) with Lhx2 in directing AER formation and hence limb outgrowth.

To support our proposal that Lhx2 might be the vertebrate homologue of apterous we misexpressed apterous, Lhx2 and Lmx-1 on the ventral side of the Drosophila wing imaginal disc (using the 71B-GAL4 driver; Fig. 5C) (Brand and Perrimon, 1993) and probed for changes in the expression of fringe and Wingless, downstream targets of apterous. Following mis-expression of the three UAS responders, the expression of the fringe gene was monitored using a lacZ enhancer detector inserted in the fringe locus (Irvine and Wieschaus, 1994). Fig. 5D shows the wild-type dorsal restriction of fringe during late third instar larva. Following mis-expression of apterous (Fig. 5E) and Lhx2 (Fig. 5F), but not Lmx-1 (Fig. 5G), the fringe-lacZ marker is ectopically activated on the ventral side of the imaginal disc. Furthermore, the expression of Wingless, which is a well known marker of the wing margin, was also monitored using a wglacZ enhancer detector present in the CyOwglacZ balancer chromosome. Fig. 5H shows the wild-type expression of Wingless along the DV compartment boundary during third instar larva. After mis-expression of the UAS constructs, the expression of the Wingless-lacZ marker is ectopically extended toward the ventral compartment only with apterous (Fig. 5I) and Lhx2 (Fig. 5J), but not with Lmx-1 (Fig. 5K). This agrees with the fact that in these crosses only apterous and Lhx2 cause ablation of wing tissues in adult flies (Fig. 5M,N). In contrast, Lmx-1 does not disturb the wing morphology at all and only causes the appearance of 1-3 ectopic bristles in the vein 3 (Fig. 5O). We know that this UAS:Lmx-1 transgene is functional because it causes lethality with other GAL4 drivers such as apterous-GAL4 and patched-GAL4 (data not shown). In addition, when the C5-GAL4 wing driver (Yeh et al., 1995) was used, apterous and Lhx2 caused defects in the wing margin, whereas Lmx-1 only produced venation anomalies (data not shown). Thus, like apterous, Lhx2, but not Lmx-1, plays a role in regulating fringe and Wingless expression, and hence wing outgrowth.

Since overexpression of Lhx2 has no phenotype in developing chick limbs, we sought to repress Lhx2 function by using a similar strategy to that which we describe for Lmx-1 above. We mis-expressed a dominant negative Eng-RD/Lhx2-HD hybrid construct in the chick limb primordia at stages 8-12 (see Fig. 3). Infected limb buds failed to develop a normal AER and consequently limb outgrowth was arrested (Fig. 6A-C). In most cases, the truncations occurred in the most distal element (showing an absence of digits and reduction in the size of the radius and the ulna), but in some cases, truncated wings at the level of the humerus were observed (Fig. 6D-G).

Since the Lhx2 dominant negative chimaera is able to perturb AER formation, we would expect that this is preceded by changes in gene expression, both in the ectoderm and in the underlying limb bud mesoderm. In situ hybridisation of embryos infected with the Eng-RD/Lhx2-HD hybrid construct using riboprobes for the En-1, Wnt-7a and Lmx-1 genes showed, just as was seen with the Eng-RD/Lmx1-HD construct, no alteration in their expression pattern (Fig. 7a-c). Furthermore, and contrary to the effects observed with the Eng-RD/Lmx-1-HD construct, mis-expression of the Eng-RD/Lhx2-HD construct did not alter the transcript distribution of Eph-7a (Fig. 7d). In situ hybridisation of embryos using probes for mesodermal genes involved in the outgrowth of the limb hoxd-13 (Fig. 7e), Msx-1 (Fig. 7f), Shh (Fig. 7g), Fgf-10 and NF-kB (data not shown) showed a down-regulation in their transcript levels. None of these changes in gene expression were observed after mis-expression of the Eng-RD/Lmx-1-HD construct (see Fig. 2).

However, and contrary to the effects observed with the Eng-RD/Lmx-1-HD construct, in situ hybridisation after mis-expression of the Eng-RD/Lhx2-HD chimaera using probes for R-fng indicate that R-fng expression is indeed abolished in the infected limbs (Fig. 7h). Furthermore, transcripts for other ectodermal genes involved in limb outgrowth, Fgf-8 (Fig. 7i), Wnt-3a (Fig. 7j), and Serrate-2 (Fig. 7k) are also absent or downregulated.

The limb truncations observed following mis-expression of the Eng-RD/Lhx2-HD chimaera raise the concern that the fusion construct might repress genes other than those normally regulated by Lhx2. Very recently a second Lhx2 gene (Lhx2B) has been cloned (Nohno et al., 1997). The expression of Lhx2B is confined to the most anterior region of the limb bud, whereas Lhx2 is expressed in both the anterior and posterior mesoderm of the developing limb. Since the most frequent phenotype observed, following mis-expression of dominant negative forms of Lhx2, is the loss of posterior limb elements, it seems reasonable to conclude that these phenotypes are likely to be a result of the down-regulation of Lhx2 rather than interference with the function of Lhx2B. However, it is of course possible that the activity of both genes is down-regulated.

To further understand the properties of Lhx2 we made a second fusion construct, this time with the activation domain from VP16. Similarly to the Eng-RD/Lhx2-HD construct, mis-expression in the flank at stage 16 induced ectopic R-fng expression (data not shown). Mis-expression of this VP16-AD/Lhx2-HD chimaera in the limb primordia (stages 8-12) gave no abnormal phenotype.

As discussed above, apterous plays several roles during Drosophila wing development. Its role in wing margin formation and wing outgrowth is thought to be realized through the activation of fringe and Serrate expression (Irvine and Wieschaus, 1994; Kim et al., 1995; De Celis et al., 1996; Panin et al., 1997). Indeed, it has been suggested that apterous may serve directly as a transcriptional regulator for the fringe gene (Irvine and Wieschaus, 1994). Additionally, apterous acts as a selector gene specifying the dorsal wing disk compartment, although the exact molecular pathway is not yet established (Blair, 1993; Diaz-Benjumea and Cohen, 1993; Williams et al., 1993; Blair et al., 1994).

Our data indicate that in vertebrates these functions are executed by at least two proteins. We have shown that expression of Lmx-1 in the dorsal mesoderm is both necessary (this work) and sufficient (Riddle et al., 1995; Vogel et al., 1995) to define dorsal cell fate, but it plays no role in regulating gene expression required for limb outgrowth. Lhx2 on the other hand, which based on sequence comparison is more similar to Drosophila apterous, does not appear to specify dorsal cell fate and, moreover, is expressed with no dorsoventral asymmetry.

However, Lhx2 does appear to perform the second role of apterous, i.e. it regulates gene expression involved in AER formation and hence limb outgrowth. Furthermore, mis-expression of vertebrate Lmx-1 in Drosophila wing imaginal discs is neither able to rescue apterous mutant flies (lacking wings; data not shown), nor to induce fringe expression in the imaginal disc. Taken together with the fact that vertebrate Lhx2 is able to rescue apterous mutant flies (i.e. generating normal fringe and Wingless (downstream target of apterous) expression, wing morphology and other phenotypes due to lack of apterous (this work and our own unpublished results), it would appear that Lhx2 is functionally closer to Drosophila apterous than Lmx-1.

Not only does the appearance of Lhx2 transcripts precede AER formation, but they are also maintained throughout limb bud development. Several lines of evidence suggest that Lhx2 is involved in both induction and maintenance of the AER, and hence in limb outgrowth. First, ectopic Lhx2 expression induces R-fng expression, Fgf-8 and Wnt-3a, genes involved in ridge formation. Second, surgical removal of the AER results in loss of Lhx2 expression (our own unpublished results). Third, Fgf-8 expression maintains Lhx2 expression. Fourth, down-regulation of Lhx2 activity perturbs AER formation and results in arrested limb outgrowth at different stages of development. Thus, Lhx2 may not only be required for continuous limb outgrowth, but may also have a role in positioning the AER at the dorsoventral limb boundary.

Like apterous in Drosophila, the role of Lhx2 in AER formation could be mediated through R-fng. It is important to note, however, that whilst in Drosophila apterous is expressed in the same tissue as fringe, in vertebrates Lhx2 is expressed in the mesoderm. Since it is unlikely that Lhx2 directly regulates R-fringe transcription in the ectoderm, there must be an indirect mechanism that allows Lhx2 to activate R-fringe expression on the ectoderm. This could be achieved via diffusible factors present in the progress zone (i.e. Fgf-10, a gene able to induce a complete limb and initiate the cascade of ectodermal gene expression required for limb outgrowth (Ohuchi et al., 1997). By analogy, it would seem possible that in Drosophila apterous does not directly regulate fringe expression either, and thus may use a similar indirect mechanism.

The fact that Lhx2 expression in the flank induces R-fng expression, but does not result in AER formation or limb outgrowth, suggests that R-fng expression in the flank is not sufficient to induce an AER. Previous data suggest that the AER develops at the DV boundary in the limb bud where cells expressing R-fng are adjacent to cells lacking R-fng (Laufer et al., 1997; Rodriguez-Esteban et al., 1997). However, our data and those reported in the mouse (Forbes et al., 1997) suggest that some other factor(s) expressed in the limb primordia but not in the flank, such as Fgf-10 (Ohuchi et al., 1997), NF-kB (Bushdid et al., 1998; Kanegae et al., 1998) or Msx-1 (see Tickle and Eichele, 1994 for a review), cooperate(s) with Lhx2 in directing AER formation and driving limb outgrowth.

Finally, it would seem likely that in Drosophila dorsal restriction of fringe expression is achieved simply by the dorsal expression pattern of apterous. In the vertebrate limb R-fringe expression is also dorsally restricted. However, since Lhx2 is expressed symmetrically on both sides of the dorsoventral limb boundary, some other contribution is needed to define the R-fringe expression domain. We and others have shown previously that En-1 represses R-fringe expression (Laufer et al., 1997; Rodriguez-Esteban et al., 1997). Since En-1 is expressed in the ventral ectoderm at the time that R-fringe expression is induced by Lhx2, it seems likely that the expression domain of R-fringe is defined by the combination of Lhx2 and En-1.

Current models of Drosophila wing development propose that formation of the wing margin and subsequent wing outgrowth along the proximodistal axis depend upon the previous establishment of a boundary at the interface of the dorsal and ventral compartments. This view is based on a number of different studies (Bryant, 1970; Garcia-Bellido et al., 1973, 1976; see also Lawrence and Struhl, 1996, for a review), but primarily on the fact that apterous not only specifies dorsal cell fate (Blair, 1993; Diaz-Benjumea and Cohen, 1993; Williams et al., 1993; Blair et al., 1994) but also induces the dorsal expression of the fringe and Serrate genes which, in turn, induce initiation of wing outgrowth at the dorsoventral boundary (Irvine and Wieschaus, 1994; Kim et al., 1995). Thus, these two processes cannot be readily dissociated since they are controlled by a single molecule.

In vertebrates, however, whilst Lhx2 induces R-fringe expression, it does not specify dorsal cell fate. Since the two functions are dissociated, it raises the question of whether or not limb outgrowth is dependent upon previous dorsoventral patterning. Many different experiments indicate that disruption of limb outgrowth correlates with perturbation of the dorsoventral axis. Mis-expression of R-fringe on the ventral side of the limb induces ectopic ventral outgrowths (Laufer et al., 1997; Rodriguez-Esteban et al., 1997); mis-expression of En-1 on the dorsal side of the limb induces ectopic dorsal outgrowths (Laufer et al., 1997; Rodriguez-Esteban et al., 1997) and mis-expression of either Wnt-7a or Lmx-1 on the ventral side of the limb can cause the appearance of ectopic digits (Riddle et al., 1995; Vogel et al., 1995). In En-1 minus mice the AER is expanded ventrally, associated with the later formation of ectopic posterior digits (Loomis et al., 1996). In mice lacking Wnt-7a, although an AER is present, it appears to be abnormal since Fgf-4 (normally expressed in the posterior region of the ridge) is absent, accounting for the subsequent loss of posterior skeletal elements (Parr and McMahon, 1995; see also Cygan et al., 1997).

Finally, the correlation of DV patterning and limb outgrowth can also be observed in two naturally occurring chick mutants. The developing limb buds of eudiplopodia mutant chicks have an ectopic AER on the dorsal side of the limb bud and the resulting ectopic outgrowth has a bi-dorsal character (Goetinck, 1964; Fraser and Abbott, 1971). In chick limbless mutants bi-dorsal limb buds fail to develop an AER and outgrowth is arrested (Prahlad et al., 1979; Fallon et al., 1983).

Together, these data strongly suggest that normal limb outgrowth requires the asymmetric expression of some of the molecules that play a role in dorsoventral patterning and that limb outgrowth is intimately linked to DV patterning (Grieshammer et al., 1996; Noramly et al., 1996; Ros et al., 1996). However, this does not mean that the molecular events that initiate limb development, at earlier stages, do not precede the establishment of dorsoventral polarity. Indeed, elegant tissue recombination and fate-map experiments (Kieny, 1971,_;Altabef et al., 1997; Michaud, 1997), indicate that signals from the limb field mesoderm (before dorsoventral polarity is determined), commit the overlying ectoderm to form an AER. This suggests that limb outgrowth is initiated at very early stages, but that at later stages dorsoventral asymmetry is needed to fine-tune the positioning of the AER, accounting for the perturbations in limb outgrowth that result from disturbed dorsoventral patterning. It is interesting that some recent experiments suggest that Drosophila wing development may similarly be initiated prior to the interactions between the dorsal and ventral compartments that establish the wing margin (Klein et al., 1998). In conclusion, whilst it is likely that the initiation of limb outgrowth in vertebrates is independent of DV signalling (see (Zeller and Duboule, 1997) for a review), it seems clear that continued limb development requires dorsoventral asymmetry.

Whilst it is evident that at the morphological level, both during embryogenesis and in the adult stage, the limbs of Drosophila and vertebrates are quite distinct, the similarities at the molecular level, during early embryogenesis, suggest that they are evolutionarily related. Evidence that genes controlling the development of Drosophila and vertebrate limbs pre-existed as a functional genetic program (directing pattern formation) is now apparent from the similar and recurring expression patterns of several genes. Nonetheless, despite the genetic similarities between vertebrate and invertebrate limbs, the genetic modus operandi seems in some instances to be different. It seems likely that these differences hold the key to understanding how evolution has used conserved genetic programs to build novel structures.

We are most grateful to Dr James B. Jaynes for the engrailed repressor construct, Drs Randy Johnson and Alfonso Martinez-Arias for discussions and sharing results prior to publication, and to all members of the laboratory for comments and technical advice. We especially appreciate the invaluable help of Lorraine Hooks in assembling the manuscript. This work was supported by HFSPO and NIH research grants to J. C. I. B. and J. B.; an HFSPO Postdoctoral Fellowship to J. W. R. S., a Pew Postdoctoral Fellowship to D. E. R. L., a National Science Foundation Undergraduate Fellowship to J. D. L. P., and a grant from the G. Harold and Leila Y. Mathers Charitable Foundation to J. C. I. B., who is a Pew Scholar.

Whilst this manuscript was under revision, the knockout of Lmx1b was reported by Chen et al. (1998). The phenotype of these mutant mice entirely supports our conclusions that Lmx1 is necessary for dorsal limb specification. This report thus also provides support for the validity of dominant-negative approaches to study chick limb development.