The hedgehog gene family in Drosophila and vertebrate development

Although the role of signalling factors in organising cell pop ulations in developing embryos has long been recognised, it is only fairly recently that the molecular nature of these signals has begun to be elucidated. Some of the most notable examples to date are the various proteins found to mimic the mesoderm inducing capacity of cells of the vegetal hemisphere of early Xenopus embryos. These include members of the FGF (Slack et al., 1987) and TGF (Green et al., 1990; Kimmelman and Kirshner, 1987) growth factor families; in addition, members of the Wnt family of growth factor-like proteins have been implicated in this process (Christian et al., 1992; Smith and Harland, 1991). While the genes encoding these various protein families have been highly conserved at the structural level throughout evolution, few similarities in their deployment during the embryonic development of species from different phyla have been reported. One possible exception is provided by the Wnt-1 gene and its Drosophila orthologue, the segment polarity gene wingless. Activity of Wnt-1 in the mid-brain of vertebrate embryos appears to be required for the expression of the Engrailed genes (McMahon et al., 1992), a regulatory relationship that recalls the interaction between wingless and engrailed-expressing cells in the developing Drosophila embryo (discussed below).

The recent molecular characterisation of the segment polarity gene hedgehog, (Lee et al., 1992; Mohler and Vani, 1992; Tabata et al., 1992; Tashiro et al., 1993) has led to the discovery of a new family of putative signal-encoding genes in various vertebrate species that are highly homologous to the Drosophila gene (Echelard et al., 1993; Krauss et al., 1993; Riddle et al., 1993; Roelink et al., 1994). The deployment of one of these, Sonic hedgehog (Shh), in embryos of several different species presents some striking parallels with that of its invertebrate homologue. The hedgehog gene family thus provides the first clear example of a conserved signalling factor that regulates analogous processes in species of different phyla.

The Drosophila hedgehog gene contains a 471 codon open reading frame (ORF) capable of encoding a polypeptide of Mr 52,147 (Lee et al., 1992). Hydropathy analysis identifies a highly hydrophobic region near the N terminus between residues 63 and 83. In vitro translation analysis suggests that this region may act either as a conventional signal sequence, leading to a secreted form of the protein, or as a membrane spanning domain, anchoring the protein in the cells in which it is expressed (Lee et al., 1992). The results of immunolocal isation analysis on fixed Drosophila tissues are consistent with both of these possibilities (Tabata and Kornberg, 1994; Taylor et al., 1993). Thus the properties of the Hh protein may implicate it in either short or long range signalling.

Using a combination of reduced stringency hybridisation and polymerase chain reaction, we have identified a number of hh-related genes in the genomes of several vertebrate species including mouse (Echelard et al., 1993), chick (Riddle et al., 1993), Xenopus (J-P. C. and P.W.I. unpublished results) and zebrafish (Krauss et al., 1993). The proteins encoded by these genes show a high degree of sequence identity both within and between species which is reflected at the functional level by the ability of the zebrafish Shh gene to activate the Drosophila hh signal transduction pathway (Krauss et al., 1993; M.J.F. and P.W.I., in preparation).

Alignment of the predicted amino acid sequences of the Drosophila Hh protein with those of the mouse Dhh, Ihh and Shh proteins and the chick and zebrafish Shh proteins reveals several interesting features of the hh family (see Fig. 1). Like the Drosophila protein, all the vertebrate proteins possess an amjno-terminal hydrophobic region of approximately 20 residues; however, the initiation codon is located immediately upstream of this region, in contrast to Drosophila Hh which initiates some 60 residues upstream of this region. Thus it is likely that in vertebrates this sequence acts exclusively as a signal peptide sequence giving rise to secreted and not membrane spanning proteins. Interestingly, the Drosophila gene has a second ATG at a similar position, raising the pos sibility that it generates different forms of the protein via the control of translational initiation (Lee et al., 1992).

Sequence conservation between the proteins is highest in their amino-terminal ends; indeed, from position 85 (immedi ately after the predicted shared cleavage site) to position 249, 62% of the residues are completely invariant among the indicates that lhh and Shh are more closely related to each other (90% amino acid identity) than to Dhh (80% identity). Comparison of Shh between species reveals a 99% identity between mouse and chick and 94% identity between mouse and fish in the same region. Conservation falls off rapidly after residue 266, apart from a short stretch at the C terminus.

During the early stages of its development, the Drosophila embryo is subdivided into a series of repeating units, the parasegments. This subdivision is marked by the activation of the segment polarity genes wingless and engrailed in a series of discrete bands of cells along the anteroposterior axis of the embryo. Each wg domain abuts an adjacent en domain and these interfaces define the parasegment boundaries. Genetic studies have shown that parasegment boundaries have special properties, acting as sources of signals that organise the pat terning and polarity of the cellular fields which they define (reviewed by Ingham and Martinez Arias, 1992). One of these signals is encoded by wg itself: in the absence of wg activity, en expression is lost from neighbouring cells (Di Nardo et al., 1988; Martinez Arias et al., 1988) and the positional specifi cation of all the cells in each parasegment is disrupted, each cell now adopting a similar fate; this effect is clearly mani fested at the end of embryogenesis in the cuticular pattern secreted by the epidermal cells.

Several lines of evidence indicate that the signal produced by en-expressing cells is encoded by hedgehog. Like wg, hh activity influences the development of the entire parasegment and embryos homozygous for loss of function hh alleles display a phenotype very similar to that seen in wg mutants. In the absence of hh activity, wg transcription is activated normally, but disappears rapidly after gastrulation (Ingham and Hidalgo, 1993). Thus one of the principal functions of hh is to maintain the transcription of wg in the cells of neighbouring paraseg ments. Notably, the maintenance of wg is restricted to a single row of cells immediately apposed to those expressing hh. This characteristic suggests that the range of hh activity is extremely limited, perhaps even contact dependent; alternatively, it could be that only these cells are able to respond to the hh-encoded signal. This latter possibility can however, be ruled out since in transgenic embryos in which hh is expressed ubiquitously, tran scription of wg is activated ectopically (Ingham, 1993; Tabata and Kornberg, 1994; see Fig. 2). Significantly, this ectopic acti vation is limited to a subset of cells in each parasegment, imme diately anterior to those that normally express wg. The capacity of cells to express wg in response to the hh signal depends upon the activity of the sloppy paired (sip) gene, a transcription factor belonging to the forkhead related family. Activity of sip is necessary but not sufficient for wg transcription, the sip expression domain defining an equivalence group of “wg competent” cells (Cadigan et al., 1994). Thus in normal devel opment, hh acts to trigger expression of wg in a subset of the cells of this equivalence group, thereby restricting its expression to the parasegment boundary.

The importance of the restricted range of hh activity is illus-trated by the pattern defects that ensue when it is overex pressed. Expansion of the wg domain results in the ectopic induction of en (Tabata and Kornberg. I 994) (Fig. 2). The interface between these eclopically located en-expressing cells and their anterior neighbours in turn induces the formation of an additional segment border in each parasegment and this is accompanied by the elimination of certain denticle types and their replacement by others with reversed polarity. These effects mimic precisely the phenotype of mutations of the another segment polarity gene named patched (pte) (Martinez Arias et al., 1988: Fig. 2). This finding could suggest a role for pie in restricting the range of the Hh protein and indeed, Hh is much more widely distributed in pie mutant embryos than in wild type (Tabata and Kornberg, 1994; Taylor et al., 1993). Notably, however, activation of wg is rendered independent of hh activity in the absence of pie function (Ingham and Hidalgo, 1993; Ingham et al., 1991) suggesting instead that the normal role of pie is to suppress the/,/, signalling pathway, leaving it constitutively active in the absence of pie. Since pie encodes an integral membrane protein (Hooper and Scott. 1989; Nakano et al., 1989), one possibility is that the Plc and Hh proteins interact at the cell urface, the latter inactivating the former and hence triggering the pathway that controls wg tran scription. Despite this close functional relationship, no homologue of the ptc gene has yet been identified in any ver tebrate species.

One of the best characterised sources of signalling activity in developing vertebrate embryos is the notochord, the derivative of the axial mesoderm. Several processes have been associated with the inductive properties of this tissue including the induction of specialised ventral neural cells that form the floor plate (Placzek et al., 1990; van Straaten et al., 1989). the pec ification of neuronal differentiation (Placzek et al., 1991; Yamada et al., 1991) and the induction of paraxial mesoderm to form scleretome (Dietrich et al., 1993; Koseki et al., 1993; Pourquie et al., 1993).

Evidence for these interactions comes principally from experimental manipulations of developing mouse and chick embryos; in embryos of both species, ablation of the notochord results in a failure of floor plate and motor neuron differen tiation, whereas grafting of notochord to ectopic locations in chick embryos results in the induction of ectopic floor plate and motor neurons in close proximity to the graft. Since notochord is closely apposed to floor plate cells both in normal development and in the experimentally manipulated embryos, it has been suggested that the inductive signal must be contact dependent (Placzek et al., 1990), a conclusion supported by the results of in vitro studies (Placzek et al., 1993). Motor neuron differentiation, by contrast, depends upon diffusible factors that act in a contact independent manner (Yamada et al., 1993) and which emanate both from the notochord and the floorplate cells induced by the notochord. Thus the patterning of the neural tube in amniotes can be seen in terms of a sequence of inductive interactions, in which one signalling centre, the notochord, induces another, the floorplate, the activity of which alone can pattern the ventral half of the neural tube. We have found that the putative signal encoding hh family gene, Sonic hedgehog, is expressed in both the axial mesoderm and the floorplate of mouse and chick embryos (Echelard et al., 1993; Riddle et al., 1993), thus implicating it in at least some of the signalling activities associated with these tissues. Moreover, the spatiotemporal expression pattern of Shh is remarkably similar in zebrafish embryos (Krauss et al., 1993) suggesting that the molecular basis of mid-line signalling may be conserved between fishes and amniotes.

Expression of Shh is first detectable during gastrulation stages of each species: in the fish embryo, transcripts are restricted to the inner cell layer of the embryonic shield, the equivalent of the amphibian organiser, while in chick, expression is detectable in the homologous structure, Hensen’s node (Figure 3B,C). A slight difference is apparent in the mouse at this stage, where expression can first be detected in the midline mesoderm of the head process that arises from the node, though not in the node itself (Figure 3A); however, expression is detectable in the node soon thereafter.

Extension of the body axis of embryos of each species is accompanied by an extension of the Shh expression domain. In the zebrafish, by 9.5 hours of development, the Shh expression domain constitutes a continuous band of cells that extends from the tail into the head, the anterior boundary of expression being positioned in the centre of the animal pole anterior to the presumptive midbrain. In the mouse and chick, expression similarly extends rostrally from the node, although expression appears limited to the level of the midbrain. Whilst the early phase of Shh expression is restricted to the midline mesoderm a new phase of expression in the overlying neu roectoderm is initiated during early somitogenesis. In the mouse, neural expression is first seen at around the 8 somite stage when it is initiated at the ventral midline of the midbrain, above the rostral limit of the head process. Expression extends rapidly both rostrally, into the forebrain, and caudally into the hindbrain and spinal cord. In the chick, neural expression of Shh is initiated at the 7-8 somite stage and, in contrast to the mouse embryo, appears simultaneously along almost the entire length of the neural fold. In zebrafish, Shh expression is apparent in the embryonic CNS at the 5 somite stage extending from the tip of the forebrain caudally through the hindbrain and rapidly extends caudally along the length of the neural keel. Expression in each species is restricted in the hindbrain and spinal cord to the ventral midline, whilst in midbrain and forebrain, it extends more laterally. Up to the mid-brain forebrain boundary the expressing cells correspond to the morphologically identifiable floorplate; the rostral extension of the Shh domain suggests that the ventral forebrain may be functionally homologous to the floorplate in all vertebrates.

The spatiotemporal expression pattern of Shh together with the strong conservation of this pattern during vertebrate evolution provides good circumstantial evidence implicating Shh in the induction of floorplate and/or motor-neuron differ entiation. In line with this possibility, overexpression of Shh in fish, frog or mouse embryos is sufficient to induce ectopic expression of the floorplate markers axial/HNF3, F-spondin as well as Shh itself (Echelard et al., 1993; Krauss et al., 1993; Roelink et al., 1994; J.-P.C. and P.W.I., unpublished data). Furthermore, in vitro assays have shown that the rat Shh ortho logue, vhhl is capable of inducing floorplate and motorneu ron differentiation in neural tube explants (Roelink et al., 1994).

Despite the strong similarities between the initial phases of Shh expression an interesting difference arises after its induction in the ventral CNS. Whereas in chick and mouse, expression persists in the notochord at least until the end of somitogenesis, in fish, mesodermal expression begins to fade away soon after transcription is activated in the floor plate (Fig. 3G-I). This down-regulation proceeds, like the CNS induction, in a rostral to caudal sequence, coinciding with the changes in cell shape that accompany notochord differentiation. Thus by the 22 somite stage, while Shh expression is maintained at high levels throughout the ventral CNS, expression in the mesoderm is restricted to the caudal region of the notochord and to a bulge of undifferentiated cells in the tail bud. Although the signifi cance of this difference is unclear it could reflect a divergence in the mechanisms of CNS patterning between fish and amniotes. One possibility is that floorplate induction represents the original function of Shh in vertebrates and that subse quently it has been recruited to an additional midline signalling role, including secondary motorneuron induction, in amniotes. Certainly, the presence of a floorplate in the nerve cord of cephalochordates (Lacalli et al., 1994) implies an ancient origin for floorplate induction, predating the vertebrate radiation. By contrast, whereas signals from both the floorplate and notochord have been implicated in motorneuron differen tiation in chick and mouse embryos, the differentiation of primary and secondary motorneurons appears to be indepen dent of any floorplate-derived signal in zebrafish. This conclu sion is based upon studies of the cyclops mutation, in which floorplate differentiation is blocked but motorneuron differen tiation is unaffected (Hatta, 1992); because of the rapid decay of Shh transcripts in the notochord, such embryos are devoid of all midline Shh expression at the time of motorneuron differ entiation, a situation that contrasts with the persistent expression of Shh in both floorplate and notochord in amniotes at the equivalent developmental stage. Thus, whereas Shh is capable of inducing motorneurons and is expressed at the appropriate time and place in amniote embryos, it appears dis pensable for their differentiation in the fish. The persistent expression of Shh in the floorplate of fish embryos may reflect some other function in this tissue or it may simply be redundant. Clearly, mutations of Shh in fish and mouse will be required to resolve these issues.

In addition to its expression in axial midline structures, Shh is transcribed in a cluster of posterior mesenchymal cells in the limb buds of mouse and chick embryos (Echelard et al., 1993; Riddle et al., 1993; Fig. 4). The temporal and spatial pattern of Shh expression in these structures suggests a close association between the gene and the organising activity possessed by posterior mesenchymal cells that constitute the so-called zone of polarising activity or ZPA. Transplantation of cells from the Shh-expressing region of the limb bud to its anterior margin has long been known to result in the dupli cation of digits with reversed polarity. This phenomenon has been interpreted in terms of the ZPA acting as a source of a morphogen, a diffusible signal, different levels of whose activity would act to instruct cells to differentiate appropri ate to their position within the developing limb field. The pattern duplicating activity of the ZPA can be reproduced by overexpression of Shh in cells at the anterior limb margin (Riddle et al., 1993; Fig. 5) strongly suggesting that Shh rep resents the molecular basis of the ZPA. Notably, Shh is similarly expressed in the posterior mesenchyme of the pectoral fin buds in fish embryos (Krauss et al., 1993; Fig. 4C), suggesting that the same patterning mechanism operates in these homologous structures.

Since the number and character of duplicated structures caused by ectopic Shh expression seems to vary as a function of the level of its activity, one possibility is that Shh protein itself acts as a morphogen. Alternatively, like its postulated floorplate inducing activity in the notochord, Shh may act at short range in the limb, inducing the expression of another sig-nailing molecule or molecules in neighbouring cells. One po sible candidate for such a molecule is the TGF family member BMP2; the gene encoding this protein is initially tran scribed in a restricted domain in the posterior limb mes-enchyme (Francis et al., 1994) that overlaps and surrounds the Shh-expressing cells (Fig. 6). Moreover BMP2 transcription is first detectable just after the onset of Shh expression (R.J and C.T.. unpublished results) and can be induced ectopically in the anterior half of the limb bud both by ZPA grafts (Francis et al., 1994) and by ectopic Shh expression (R.J., E. Laufer and C.T. unpublished results). While these observations are con sistent with a role for Shh in inducing BMP2 expression, pre senting BMP2 as a possible effector of Shh activity in limb pat terning, functional studie have o far failed to establish such a role for BMP2 (Francis et al., 1994). Remarkably, however, a similar relationship between hh and the Drosophila BMP2 homologue decapemap/egic (dpp) appears to underlie the pat terning of imaginal discs, the fly equivalent of limb buds.

The limbs or appendages of holometabolous insects develop from imaginal discs, simple epithelial cell sheets whose primordia arise at the parasegment borders of the developing embryo (Bate and Marinez Arias, 1991). Thisorigin means that each disc incorporates and propagates portions of the cell pop ulations that define the parasegmental borders, their progeny forming distinct polyclonal lineages that subdivide the appendages into developmental compartments. The posterior compartment of each disc is thus characterised by the expression of hh (Lee et al., 1992; Tabata et al., 1992), whereas pie is expressed in cells of the anterior compartment (Phillips et al., 1990; see Fig. 7).

The function of hh in imaginal disc development was first analysed by Mohler (1988) using genetic mosaic techniques to remove the activity of hh from cells in different regions of the discs. These experiments demonstrated a requirement for hh activity in posterior compartment cells for the correct develop ment of genetically wild-type cells in the neighbouring anterior compartment. We have inve tigated further this aspect of hh function using transgenic animals carrying an HS-hh construct to induce transient ectopic expression of hh in the anterior com partments of the wing discs. Such ectopic expression results in the duplication of anterior wing structures with mirror image symmetry (see Fig. 8) an effect that shows a striking analogy to the digit duplications induced by ZPA grafts or ectopic Shh expression in vertebrate limbs (compare with Fig. 5). The same kjnds of duplications have also recently been reported by Basler and Struhl (1994), who used the “flip-out” technique to generate clones of cells expressing hh constitutively.

In some cases, ectopic hh activity results in duplication of only the most anterior structures, such as the wing margin and veins I and n, (Fig. 8B), whereas in other instances, differentiation of the anterior margin is almost completely suppressed, being replaced by veins 11 and Ill (Fig. 8C). As in the case of the chick limb, these variable effects could be indicative of a role for hh as a morphogen, different pattern elements being specified by different thresholds of hh activity. Several linesof evidence suggest, however, that in the imaginal disc. as in the embryo, hh acts in the wing to regulate the transcription of another signal-encoding gene.

Expression of dpp, which is absolutely required for normal wing morphogenesis (Po akony et al., 1991; Spencer et al., 1982), is restricted to a narrow band of cells that runs along the antero-posterior compartment boundary of the wing disc (Blackman et al., I 99 I; Masucci et al., I 990; see Fig. 7), closely apposed to thehh-expressing cells of the posterior com partment. In discs in which hh has been ectopically activated, dpp is similarly inappropriately expressed (Basler and Struhl, 1994; M.J.F. and P.W.L in preparation; see Fig. 7). implying the latter to be a target of hh activity. Ectopic expression of dpp is similarly induced in imaginal discs from animals with reduced activity of ptc (Capdevila et al., 1994; M.J.F and P.W.I. in preparation); thus as in the embryo, over-expression of hh has the same effect as the reduction or removal of ptc activity, suggesting that the same signalling mechanism acts to regulate dpp and wg at different stages of development. Thus in both cases, hh appears to act to regulate the source of other signalling molecules.

The parallels between the expression and function of hh family genes in Drosophila and vertebrate development are indeed striking. In the Drosophila embryo, hh acts as a localised signal that organises the patterning of each parasegment at least in part by regulating the expression of another signal-encoding gene wg. Ectopic expression of hh causes inappropriate acti vation of wg which in tum induces the expression of en in the middle of each parasegment; the result is a duplication of pattern elements and reversal of polarity that is reminiscent of the polarity reversals and ectopic differentiation induced by notochord grafts in chick embryos. Intriguingly, we have found that a close relative of hh, the Shh gene is expressed in the developing notochord, the activity of which is likely to be responsible for the inducing properties of this tissue. Thus molecules that have been highly conserved through evolution are deployed in different phyla to effect similar processes in the patterning of secondary fields.

The expression of hh family genes in the developing limbs of vertebrates and insects provides a yet more striking example of such functional similarity. In both cases, a member of the hh family is expressed in the posterior half of the limb pri mordium - and in each instance, its ectopic expression results in the duplication of pattern elements. Moreover, in both cases, activity of Shh and hh appears intimately associated with the expression of closely related members of the TGF family, namely BMP2 and dpp. Whereas functional analysis of dpp has clearly implicated it in appendage morphogenesis, no such role for BMP2 has yet been established. Nevertheless, it is difficult to escape the conclusion that, despite their apparently inde pendent evolutionary origin, the limbs of vertebrate and inver tebrates may be patterned by very similar mechanisms. Whether the remarkable similarities in the deployment of hh genes in the development of deuterostome and protostome embryos reflects a common origin for these various patterning processes or an example of evolutionary convergence remains to be seen. The isolation of hh family genes and analysis of their expression in organisms of other phyla should provide important new insights into the origin of the signalling mech anisms that underlie pattern formation in all metazoa.

We are grateful to Ron·Blackman for making the dpp-lacZ reporter strain available to us. The authors’ work was supported by a Human Frontiers Science Programme grant to A.P.M., P.W.I. and C.T. and by the Imperial Cancer Research Fund (P.W. I.) and the National Institutes of Health (C.T.). M.J.F. is a C.J. Martin Fellow of the Aus tralian M.R.C.