Until now, the signals that control the development of the legs in insects and vertebrates have been thought to be different. But new work reveals similarities, which might have evolutionary implications.
We live in a three-dimensional world and, not surprisingly, the development of many animals — including ourselves — depends on the establishment of three body axes. Two of these, the head-to-tail (anterior–posterior) and back-to-front (dorsal–ventral) axes, are established in the egg before fertilization or, in some cases, as an immediate result of fertilization. In contrast, the proximal–distal axis, which extends from the base of each appendage (where it attaches to the body) to its tip, must be generated from scratch every time an animal grows a leg, arm or wing. How this axis is established has been the subject of debate for several decades, and work on leg and wing development in the fruitfly, Drosophila melanogaster, has provided many insights, at least for insects. Papers on page 781 of this issue1 and in a recent issue of Science2 take our understanding a step further, by demonstrating an unexpected role for signalling through a protein called the epidermal-growth-factor receptor in fruitfly leg development.
Fly legs emerge from flat discs of tissue that are set aside during embryonic development3. At this stage two secreted signalling molecules, Wingless (Wg) and Decapentaplegic (Dpp), are needed to set up the proximal–distal axis3. These molecules are expressed as two opposing wedges in the disc, one ventral and the other dorsal (Fig. 1a). Where they meet, at the centre of the disc, will become the distal tip of the leg.
Not only are both signals required in the formation of the proximal–distal axis3, but additional axes can be generated experimentally by creating new sites at which Wg and Dpp intersect. For example, bifurcated legs (which have an additional proximal–distal axis) can be generated by forcing just a few ventral cells, which normally express just Wg, to express Dpp as well4. Importantly, the additional legs generated in this way are composed mainly of wild-type cells, suggesting that the fates of these cells were changed by the Dpp- and Wg-expressing cells. Without this new source of Dpp and Wg, the wild-type cells would have contributed only to the normal legs.
Two models have been proposed to explain this result. The first is based on an analogy with Spemann and Mangold's classic demonstration5 of the existence of an 'organizer' — a group of cells that influence the behaviour of neighbouring cells — during amphibian development. According to this view, the intersection between the Wg and Dpp signals creates a distal 'organizer' in the leg disc4. It was proposed that this organizer is the source of a third signal that causes neighbouring cells to become distal; the farther a cell is from the organizer, the more 'proximal' its fate would be.
The alternative model posits that Wg and Dpp cause nearby cells to become distal directly, without inducing a third signal6. So, cells that receive the highest levels of both Dpp and Wg, such as those at the centre of the disc, would become distal, whereas cells receiving successively lower levels of both signals would take on successively more proximal fates.
Either model could account for the generation of bifurcated legs: a new source of Dpp and Wg would be expected to create a new proximal–distal axis whether indirectly (by creating an organizer) or directly. However, one limitation of the organizer model when it was originally suggested was that there was no candidate for the signal from the hypothesized organizer. By contrast, significant experimental support for the direct model came from showing that Dpp and Wg activate two genes along the proximal–distal axis directly, without inducing a third signal6. These genes, Distal-less (Dll) and dachshund (dac), are required for the generation of distal and intermediate regions, respectively3 (Fig. 1b). Thus, largely because of this work6, the prevailing theory has been that the combined and graded activities of Wg and Dpp directly produce the leg's proximal–distal axis.
But it takes more than Dll and dac to make a leg. After these genes have been turned on, several others, also required for distal leg fates, begin to be activated in subsets of the Dll-expressing leg region3. Although these genes cross-regulate each other7, it was not known whether they, too, would be targets for Wg and Dpp. Campbell1 and Galindo et al.2 now show that Wg and Dpp are not required at the time when these genes are first turned on. Accordingly, whereas removing either Dpp or Wg function early in development truncates the legs, later removal has only minor effects on leg development2.
If Wg and Dpp do not regulate the expression of these distal genes, what does? During vertebrate limb development, proteins from the fibroblast-growth-factor family are expressed at the distal edge of the growing limb. These proteins drive the formation of the proximal–distal axis by promoting the proliferation of underlying cells. They do this by activating receptors on the cell surface that are enzymes (receptor tyrosine kinases, RTKs), which transduce the signal through a well-characterized pathway to the cell nucleus8. Although fibroblast growth factors do not seem to be involved in fly limb development, mutations in another RTK, the epidermal-growth-factor receptor (EGFR), cause the loss of the distal tip of the fly leg9. Campbell1 has found that a gradient of EGFR signalling is likely to be at work, because a stepwise reduction in EGFR activity is correlated with a gradual loss both of leg structures and of leg-disc marker genes in a distal-to-proximal direction.
But which EGFR-activating proteins (ligands) are involved here? Both papers1,2 suggest the involvement of Vein, which is expressed in cells at the centre of the disc, precisely where the Dpp and Wg signals intersect (Fig. 1c,d). This ligand also seems to be a target of Wg and Dpp but, once activated, no longer needs these signals for its expression2. Yet removing Vein activity has only mild effects, if any, on the leg, so other ligands must also be involved. Indeed, Campbell shows that Rhomboid, which is required to process a different class of EGFR ligands, is also expressed in distal cells. In fact, it is not until Vein, Rhomboid and one of its relatives, Roughoid, are all simultaneously removed that a marked loss of distal structures occurs1. So, although the specific ligands that Rhomboid and Roughoid activate are not known, the 'third signal' from the hypothesized distal organizer is likely to be a cocktail of EGFR ligands that are functionally redundant.
Thus, as often occurs in biology, both the direct and the organizer models seem to be correct, and each provides only part of the solution to how the proximal–distal axis is generated. The activity of Dpp and Wg apparently directly regulates Dll and dac, which define broad domains along the proximal–distal axis6. But the new papers1,2 provide strong evidence for a distal organizer, induced by Dpp and Wg, that is the source of EGFR ligands, which in turn works at a distance to help pattern the distal part of the leg.
The region of the leg that is patterned by this activity gradient — the tarsus — seems to be evolutionarily ancient10. This fact, together with the involvement of RTK pathways in both vertebrate and invertebrate1,2 leg development, makes it tempting to draw comparisons between the evolutionary origins of these limbs. However, the embryological differences are too plentiful to propose that the limbs are directly related to each other. Rather, the many developmental parallels, including RTK signalling, are more likely to reflect conserved regulatory relationships between signalling pathways that work together in many places in development. Alternatively, it is possible that vertebrate and invertebrate limbs were both derived from some type of outgrowth, perhaps not a limb11, that grew from the body of a common ancestor. If this outgrowth had a primitive proximal–distal axis, it might have used many of the same genes and pathways that are active in the limbs we know today.
Beyond these evolutionary issues, other questions arise from this work. What are the EGFR ligands that Rhomboid and Roughoid process, and why are several different ligands required? Is activation of the EGFR pathway enough to generate an extra proximal–distal axis, or are other signals also required? To what extent is the regulation of the tarsal genes by the EGFR pathway direct and dependent on different thresholds of EGFR activity, as opposed to the result of cross-regulation between these genes? We look forward to the next round of studies to provide an even more complete description of how to make a leg.
Campbell, G. Nature 418, 781–785 (2002); advance online publication, 24 July 2002 (doi:10.1038/nature00971).
Galindo, M. I., Bishop, S. A., Greig, S. & Couso, J. P. Science 297, 256–259 (2002).
Morata, G. Nature Rev. Mol. Cell Biol. 2, 89–97 (2001).
Campbell, G., Weaver, T. & Tomlinson, A. Cell 74, 1113–1123 (1993).
Spemann, H. & Mangold, H. Arch. Microsk. Anat. EntwMech. 100, 599–638 (1924).
Lecuit, T. & Cohen, S. M. Nature 388, 139–145 (1997).
Kojima, T., Sato, M. & Saigo, K. Development 127, 769–778 (2000).
Capdevila, J. & Izpisua Belmonte, J. C. Annu. Rev. Cell. Dev. Biol. 17, 87–132 (2001).
Clifford, R. J. & Schupbach, T. Genetics 123, 771–787 (1989).
Casares, F. & Mann, R. S. Science 293, 1477–1480 (2001).
Shubin, N., Tabin, C. & Carroll, S. Nature 388, 639–648 (1997).