The Vestigial gene product provides a molecular context for the interpretation of signals during the development of the wing in Drosophila

The development and patterning of structures such as teeth, limbs, heart or liver are complex processes that rely on the spatial and temporal orchestration of cell interactions. These in turn rely on signalling systems but, in the last few years, it has become clear that a specific structure is not the outcome of specific signalling events. On the contrary, the available information suggests that there is a limited number of signalling systems which often work through a small number of effectors to elicit a great variety of responses. The specificity of the outcome resides on the effectors available to the signalling system through the history of the cell. For example, in vertebrates, Sonic Hedgehog (Shh) is involved in the development and patterning of the limb, the specification of the floor plate, the determination of neuronal cell fates and other processes (Riddle et al., 1993; Echelard et al., 1993). For each of these functions, Shh controls specific sets of genes. However, it always does it through the same signal transduction pathway and therefore something has to confer a particular interpretation of these signals in a specific manner: tooth, limb, heart or liver.

As a group of cells develop into a particular structure, something has to keep the ‘group identity’ constant. The wing of Drosophila provides a useful system to unravel the molecular basis of this notion of identity. Developmental studies have shown that the vestigial (vg) gene encodes a nuclear protein that plays a central role in the development of the wings in Drosophila: loss of vg results in the failure of wings to develop (Lindsley and Zimm, 1992) and ectopic expression of vg leads to the development of ectopic wing tissue (Kim et al., 1996). These experiments suggest that vestigial can be viewed as a ‘wing selector’ gene i.e. a gene that enforces wing developmental pathway on cells (Kim et al., 1996).This notion is supported by the observation that ectopic expression of vestigial can rescue the loss of wing caused by mutations in genes involved in the regulation of wing development such as apterous, Suppressor of Hairless or wingless (Klein and Martinez Arias, 1998a).

The expression of vg during wing development is regulated through two enhancers called the second intron or boundary enhancer (vgBE), and the quadrant enhancer (vgQE) (Williams et al., 1994; Kim et al., 1996; and Fig. 1). These names reflect the patterns of expression directed by these regulatory regions: the vestigial Boundary Enhancer (vgBE), a thin stripe over the prospective wing margin, and vestigial quadrant enhancer (vgQE), a pattern in four quadrants that are complementary to the vgBE and which fill in the developing wing blade. Both, vgBE and vgQE, act as integrators of signalling systems that drive wing development and, in this manner, these regulatory regions determine the tempo and the mode of wing development (Williams et al., 1994; Kim et al., 1996).

However, the function of vg and, more specifically, how it endows cells with identity is not clear. One possibility is that Vestigial regulates patterns of proliferation and gene expression that are characteristic of wing tissue. However, there is no direct evidence for this.

Here we study the regulation of the vg gene. We show that Vestigial is required for its own expression and that the three signalling systems known to play a major role in wing development, Notch, Wingless and decapentaplegic (dpp), not only act through the known enhancers of vg, but also show synergistic interactions with the vg gene product. These observations lead us to suggest that vestigial specifies identity in part by providing a molecular context in which different signalling events act to trigger the development of a wing. Perhaps by responding to this signalling system, Vestigial targets factors onto specific genes required for the proliferation and patterning of the wing. Our results further show that vg plays an important role together with, rather than simply downstream of, wingless in the patterning of the wing blade along the dorsoventral axis. The experiments suggest that the main function of wg is to enforce gene expression rather than initiate it.

The Ser mutants used in this study are Ser^RX106 (Speicher et al., 1994) and Ser^94C (Couso et al., 1995). Expression of vg at the DV boundary was detected using the vestigial Boundary Enhancer described in Williams et al. (1994) and referred to here as vgBE (for Boundary enhancer). The vg-quadrant enhancer is described in Kim et al. (1996, 1997b) and is referred to here as vgQE. For the expression of wg in the developing discs, we sometimes used a P-lacZ insertion in the wg gene on a CyO chromosome (Kassis et al., 1992).

Lines allowing us to overexpress the dominant negative UAS-wg carry a wingless gene with a stop codon at codon 398 of the wingless coding region (Zecchini and A. M. A., unpublished data). This C-terminal deletion mimics a mutation in Xenopus Wnt (Hoppler et al., 1996) and causes dominant negative effects in the embryo and the adult; details of the construct and the flies themselves are available upon request.

Ectopic expression of the different genes was achieved through the GAL4/UAS system of Brand and Perrimon (1993). The UAS-vg constructs were kindly provided by Sean Carroll (Kim et al., 1996); UAS-wg by K. Gieseler and J. Pradel; UAS-Nintra and UAS-Dl by L. Seugnet and M. Haenlin. The UAS-GFP is described in Yeh et al. (1995).

The expression of the different UAS constructs was driven in the imaginal discs with various GAL4 inserts. patched Gal4 (ptc GAL4) expresses UAS-X in a stripe along the AP boundary of the discs (Speicher et al., 1994). In the third instar, decapentaplegic Gal4 (dpp GAL4) (Wilder and Perrimon, 1995) is expressed in a similar pattern to ptc GAL4, although the expression is weaker over the ventral side (Klein and Martinez Arias, 1998a). The vgBEGal4 is described in Neuman and Cohen (1996a).

Stocks carrying various GAL4 and UAS combinations in wild type and mutants were generated. All stocks were balanced over the SM6a-TM6b compound balancer, which allowed the identification of larvae of the correct genotype because of the dominant larval marker Tb (Lindsley and Zimm, 1992). Details of the stocks as well as the stocks themselves are available upon request. In the case of ap mutations, stocks were established with a CyO balancer carrying a P-lacZ insertion in wg and the mutant discs were checked for the absence of the wg expression pattern. In the case of some experiments involving ap mutants, mutant discs were identified by the aberrant morphology of the wing disc and the absence of the CyO wg-lacZ balancer.

The Distalless antibody was a gift from S. Cohen and S. Bray and stainings were performed according to standard protocols. In situ hybridizations were performed as described in Tautz and Pfeifle (1989). The X-Gal staining is described in Ashburner (1989). The fluorescence of the green fluorescent protein (GFP) was detected by using the FITC-filter sets on a Zeiss Axiophot microscope.

The expression of vg during wing development is regulated by two enhancer sequences located in the second and fourth introns of the gene (Williams et al., 1994; Kim et al., 1996). The second intron enhancer regulates expression of vg along the DV boundary (Williams et al., 1994) and is therefore called the boundary enhancer (vgBE). The enhancer located in the fourth intron is called quadrant enhancer (vgQE), is activated much later than the vgBE (see Fig. 1 and Kim et al., 1996) and regulates the expression of vg in the pouch away from the DV boundary. This enhancer is only active in the wing pouch of the wing disc and not in any other imaginal disc.

The development and patterning of the wing depends on a series of interdependent regulatory loops that define a sequence of spatial patterns of gene expression. For this reason, it is important to determine both spatial and temporal aspects of the inputs on different genes (Klein and Martinez-Arias, 1998a). We do this below by focusing on the regulation and function of the two main enhancers of the vg gene.

The vgBE is activated first during the second instar and its expression pattern is very similar to that of the Vestigial protein at this stage (Klein and Martinez Arias, 1998b) suggesting that, in these early stages, the vgBE is responsible for the complete pattern of vg expression (also see below). Deletion analysis of the enhancer reveals two regions essential for its activity: a binding site for Suppressor of Hairless and sequences contained in the first 80 base pairs of the enhancer (Williams et al., 1994; Kim et al., 1996, 1997b).

In an attempt to map the nature and timing of the inputs into this enhancer, we have compared during wing development the activity of the wild-type vgBE and of deletions of the two essential regions (Fig. 2). To do this, we used lacZ fusions and monitored their expression in the developing wing disc (Williams et al., 1994; Kim et al., 1996, 1997b). We find that at the end of the second instar, lacZ expression from the wild-type vgBE outlines a horseshoe over the ventral region of the wing disc, with weak expression in the ventral anterior region where it overlaps with the expression of wingless (Figs 1A, 2A,L). lacZ expression increases in this region at the beginning of the third instar (Figs 1B, 2B,C). This increase does not occur in Su(H) mutants (data not shown), or in wg mutants (Klein and Martinez Arias, 1998a).

The activity of the enhancer deleted for the Su(H)-binding site is different from the wild type. The activity of this enhancer is never initiated over the ventral region of the disc, where the wing primordium is established and remains absent during later stages (Fig. 2E-G). This result demonstrates that the activity of Notch is required not only for the maintenance, but also for the initiation of the expression of vg through the vgBE (Klein and Martinez-Arias, 1998a; Fig. 2).

The activity of the enhancer deleted from 0-80 is similar to that of the wild-type enhancer early on (Fig. 2I), but it never acquires the high levels of activity in the anterior ventral region. As the wing blade develops, a line of faint activity can be detected over the DV boundary (Fig. 2J), but it fades quickly and, by the end of the third instar, in most discs there is no activity over the developing wing blade (Fig. 2K). This enhancer still shows some activity in the flanking notal regions of the wing disc in wild-type and Su(H) mutant discs (Fig. 2H,K). This indicates the existence of additional inputs in the regulation of the vgBE in the notal region (see also Kim et al., 1997a).

Our results (and see also Klein and Martinez Arias, 1998a) suggest that the cells in which the expression of the vgBE is upregulated at the end of the second instar represent the anlage of the wing and require Notch/Su(H) signalling. These cells are located at the DV interface, on the domain of wg expression and overlap the expression on nubbin. Our suggestion that these cells represent the primordium of the wing pouch can explain why a deletion of the vgBE, as in the vg^83b27 allele, results in the abolition of the development of the wing pouch since, in this mutant, the anlage would never be defined.

We have recently demonstrated that, in addition to the activity of Notch, the activity of wg is required for the initiation of vg expression since, in second instar larval discs of wg mutants, the activity of the vgBE is absent over the region of the wing anlage (Klein and Martinez Arias, 1998a). The possibility that Wingless acts on the vgBE is supported by the observation that a dominant negative version of DFrizzled2, a receptor for Wingless, reduces the activity of the vgBE (Zhang and Carthew, 1998). Altogether these results raise the possibility that Wingless has a direct input on the vgBE. Consistent with this we find that the first 80 bp of the vgBE, which are required for the full activity of the enhancer, contain two putative TCF-1-binding sites that are associated with Wingless signalling (Fig. 3A,B).

To test further the relationship between wingless and vg early in wing development, we monitored the influence of Notch and Wingless signalling on the activity of the vgBE. In contrast with ectopic expression of wg, which never leads to ectopic expression of the vgBE (Fig. 3C), ectopic expression of Delta with ptcGAL4 leads to ectopic expression of the vgBE, but only within the developing wing blade and where the levels of ectopic expression of Delta are high, near the AP compartment boundary (Fig. 3D). Interestingly, an effect of Wingless on this enhancer can be detected when Wingless is coexpressed ectopically with Delta. Although Wingless alone has no effect on this enhancer, coexpression of Delta with Wingless extends the realm of activity of the vgBE into regions where expression of Delta on its own has no detectable effects regions (Fig. 3E and see also Klein and Martinez Arias, 1998a).

These results suggest that the major effect of Wingless on the vgBE is to collaborate with Notch/Su(H) signalling during the early stages of wing development. The failure to observe effects of ectopic expression of Wingless on this enhancer might be related to the absolute requirement for Notch signalling in this function of Wingless. Since, during the early stages of wing development, Notch signalling is restricted to the DV interface (see Klein and Martinez Arias, 1998b), the effects of Wingless could only be observed in this region and are highlighted by the loss of wg function or the loss of Wingless responding sites in the enhancer. Extension of this domain by activating Notch signalling ectopically allows the visualization of the effects of ectopic Wingless. It is likely that the residual activity of the vgBE^Δ0-80 observed along the DV boundary during early stages of wing development reflects the input of Notch signalling without the enhancement of Wingless signalling.

The activity of the vgQE can be detected first at the beginning of the third instar, several hours after the upregulation of the vgBE, when it closely outlines the realm of the growing wing (Fig. 1E-H and see Kim et al., 1996). This enhancer is only expressed in the growing wing blade and thus provides a unique and most specific marker for wing blade tissue.

A variety of experiments have shown that the vgQE receives a negative input from Notch signalling and a positive one from Dpp (Kim et al., 1996 and see below). The presence of a E(spl)-binding site in the sequence of the vgQE led to the suggestion that this suppression by Notch is mediated by the E(spl) protein. However, we do not find a strong suppression of the activity of the vgQE if UAS-m8 (E(spl)) is ectopically expressed with dppGal4 (data not shown), suggesting that the effect of Notch requires other mediators. In our experiments, we find that, although the vgQE is suppressed in the domain of Notch activity (Fig. 5D), Notch signalling plays a non-autonomous role in its activation. For example, the vgQE is never active in Serrate (Ser) mutants in which wing development initiates normally but is aborted very early (Klein and Martinez-Arias, 1998b; Fig. 4A). Expression of UAS-Dl under the control of dppGal4 in Ser mutants rescues the wing pouch and leads to the activation of the vgQE (Fig. 4B). Interestingly, this activity arises in regions devoid of Notch signalling (compare Fig. 4B and C). This result suggests that Notch signalling influences the activity of the vgQE in two ways: it represses the activity of the vgQE autonomously but it is also required for its activity in a non-autonomous way.

It is clear that the activity of the vgBE is required for the activation of the vgQE (Kim et al., 1996; T. K. and M. A. M. unpublished data), but little is known about how this interaction takes place. The observation that vgQE is activated in Ser mutant discs in which UAS-vg is expressed ectopically suggests that the activation of the vgQE is mediated by Notch signalling through the activity of Vestigial on the vgQE (Fig. 4E,F). However, in this experiment, expression of wg is also restored (Fig. 4F) and this raises the possibility that activation of the vgQE is mediated through the presumed ‘organizing’ activity of Wingless (Zecca et al., 1996; Neumann and Cohen, 1996b). This is probably not the case since we find that ectopic expression of UAS-wg alone does not lead to the activation of the vgQE in Ser mutants (Fig. 4D). The inability of Wingless to activate the vgQE in this experiment is not due to a general insensitivity of the cells to Wingless signalling, since UAS-wg is always capable of inducing hinge fate ectopically (Klein and Martinez-Arias, 1998a).

These results clearly demonstrate a requirement for vg in the activation of the vgQE. However, clonal analysis has shown that vg acts cell autonomously (Kim et al., 1996) and therefore, in the wild type, the non-autonomous effects of Notch on the vgQE must be mediated by another, diffusible molecule(s) which is under control of Notch signalling. A number of studies suggest that Wingless has an influence on the expression of vg in the wing pouch and that its expression at the wing margin is under control of Notch signalling (Zecca et al., 1996; Neumann and Cohen, 1996b). Therefore, it is possible that Wingless is mediating the non-autonomous effect of Notch on the vgQE. It might be that Wingless acts by acting on the vgQE to elevate and maintain the levels of vg expression that had been induced by Vestigial through the vgBE. In agreement with this proposal, we find that the activity of the vgQE is elevated in response to the ectopic expression of UAS-wg with dppGal4 (Fig. 5C,H) and that expression of a dominant negative Wingless molecule (see Materials and Methods) suppresses the activity of the vgQE and reduces the size of the wing pouch (Fig. 5J).

Altogether our results suggest that the upregulation of vg expression in response to wg is mediated by the vgQE. Our conclusion is supported by the existence of several putative TCF-1 binding sites in the vgQE (data not shown). However, the effects of Wingless are always restricted to the normal domain of vg expression (compare Fig. 3F,K), in agreement with the results presented above that Wingless alone is not sufficient to initiate ectopic expression of vg through the vgQE. These effects are likely to be mediated by Vestigial itself and, as indicated above, the role of Wingless on this regulation is to maintain and modulate the levels of activity of the vgQE. To test this possibility, we have expressed vg under the control of dppGAL4 and observed that this leads to elevated activity of the vgQE and, in addition to weak, but reproducible amounts of ectopic activity of the vgQE in the notal region (arrows in Fig. 5F). No ectopic expression of wg is observed in the dorsal side of the wing disc in these experiments and therefore the ectopic activation of the vgQE in the dorsal is probably independent of Wingless signalling (Fig. 5L). Although this does not rule out the existence of regulatory intermediaries, it does suggest that Vestigial has an effect on the activity of this enhancer.

Consistent with our conclusion that Wingless enhances the effects of Vestigial, we find that coexpression of Wingless and Vestigial with dppGAL4, which leads to the ectopic induction of pouch and hinge fate in the notal regions (Klein and Martinez-Arias, 1998a), triggers a widespread and stable expression of the vgQE throughout the wing disc (Fig. 5G). This synergistic effect is more clear in the case of the leg discs where the vgQE enhancer is not active during normal development. Expression of wg alone in the leg disc does not activate the vgQE and, similar to the wing disc, vg alone leads to some weak activation of this enhancer (data not shown). However, their combined expression leads to very robust and widespread expression of this enhancer (Fig. 5M).

These results suggest that the activity of wg is required to enhance the activity of the vgQE which is initiated by the Vg protein itself. They further suggest that the role of Notch in the regulation on the activity of the vgQE is twofold. First, it activates the expression of Vestigial (via the vgBE) and of wg along the DV boundary. This sets up the conditions for the activation of the vgQE through Wingless and Vestigial as the wing pouch begins to grow. Second, at the DV boundary, Notch signalling leads to a suppression of the activity of the vgQE but maintains the expression of wg and of the vgBE.

To confirm this conclusion, we have monitored the influence of coactivation of Wingless and Notch signalling on the activity of the vgQE enhancer. We have shown previously that the definition of the size and the fate of the wing pouch requires the combination of Wingless and Notch signalling on the vgBE (Klein and Martinez-Arias, 1998a and this work). We now find that ectopic expression of Dl or wg alone are not sufficient to activate the vgQE (Fig. 5C,D; see above) but that the coexpression of both results in the extension of the vgQE into the notum as it has been previously shown for vg expression by antibody staining (Fig. 5E; Klein and Martinez-Arias, 1998a).

Multiple experiments have shown that dpp provides an important input into the growth of the wing (see e.g. Zecca et al., 1995) but appears not to play a role in the initiation of wing development since wings derived from dpp mutants that affect its function during wing development, have a normal margin (see for example Zecca et al., 1995). However, the vgQE contains functional Dpp response sites (Kim et al., 1997a) and responds to Dpp (Kim et al., 1996). These observations, together with those of other inputs on the expression of vg, have led to the suggestion that, as the wing grows, it integrates inputs from both wg and dpp (Kim et al., 1997a).

To test this integration directly, we have created a complete overlap of the activities of wg, dpp and vg by activating UAS-wg and UAS-dpp with vgBEGAL4 (Staehling-Hampton and Hoffmann, 1994; Neumann and Cohen 1996a and Fig. 6A). Ectopic expression of wg alone in this manner leads to ectopic expression of the vgQE in the posterior region of the domain of vgBE expression (Fig. 6B), over a region in which dpp is normally expressed (compare Fig. 6B with A and G). Interestingly, elevation of wg expression achieved in the area of the DV boundary does not enlarge the wing pouch (compare Fig. 1G and 6B). This suggests that, although wg is required for the correct growth of the pouch, it is not a limiting factor. Expression of dpp under the same conditions leads to an enlargement of the wing blade along the AP axis, with the vgBE as a reference but without extending it into the notum (Fig. 6C).

Coexpression of wg and dpp leads to a pronounced extension of the wing blade towards the notum (Fig. 6D). In some instances, we can observe the development of up to eight wing blades, all with a common origin over a point ventral to the notum (Fig. 6E,F). All these wings express the vgQE and each of them has a defined margin (Fig. 6F). It is possible that these multiplets are produced by splitting of an initial primordium. It is the case that the expression driven by vgBEGAL4 is discontinuous (see Fig. 6A) and therefore leads to the generation of several foci of overlap of the activities of wg and dpp, which appear to act as foci for growth and patterning.

In the wild type, the activity of the vgQE is initiated at the intersection of wg and dpp expression and radiates from this focus (see Fig. 1 and Kim et al., 1996). The results presented here indicate that this overlap determines important parameters of the morphogenesis of the wing. For example, it is possible that the distance from this focus – perhaps defined by the range of diffusion of the two molecules – defines a threshold that contributes significantly to the determination of the size and shape of the wing blade. By tampering with these overlaps, one can alter the shape of the wing (Fig. 6) or, by generating a series of them, trigger the development of multiple wings from one primordium.

A variety of experiments have suggested that the expression of wg at the wing margin is a source of Wingless which, in a concentration-dependent manner, determines patterns of gene expression (Zecca et al., 1996; Neumann and Cohen, 1996b). One of these targets is the vg gene and here we have shown that, in the wing blade, these effects are likely to be targeted through the vgQE. However, here we have shown that Vestigial can alter the spatial activity of the vgQE without altering the endogenous pattern of wg expression. Since vg is itself expressed in a gradient similar to that generated by the diffusion of Wingless (Williams et al., 1991; Zecca et al., 1996; Neumann and Cohen et al., 1996), and displays synergistic interactions with Wingless (Figs 2, 3), this raises the possibility that Vestigial operates together with, rather than downstream of, Wingless in the regulation of gene expression within the wing pouch. To test this, we have studied the effects of Vestigial on another target of Wingless signalling: Distalless (Dll).

In the wild-type wing, Dll is expressed in the wing pouch in a circle of cells that lies within that defined by the expression of vestigial (Zecca et al., 1996; Neumann and Cohen, 1996b and Fig. 7A). We find that expression of vg with dppGAL4 elevates the levels of Dll expression within the developing wing blade and triggers ectopic expression of Dll outside this area (Fig. 7B). This differs from the effects of ectopic expression of wg, which are always restricted to the developing wing blade and never induce ectopic expression of Dll (Zecca et al., 1996; Neumann and Cohen, 1996b and Fig. 7C). Coexpression of both, wg and vg, results in synergistic effects similar to those described before for other targets and an increase in the area of ectopic expression (Fig. 7D).

The effects of vg are independent of wg (Fig. 7E), since they are achieved even in ap mutant where wg is never expressed along the DV boundary (Fig. 7F-H and see also Ng et al., 1996). In these mutants, there is no Notch/Su(H) signalling, but ectopic expression of vg can rescue the loss of the wing pouch caused by the loss of ap function (Klein and Martinez Arias, 1998a). Although vg is required for wg expression, there is not much wg expression in the rescued ap mutants. However, there is a clear stripe of Dll expression that correlates with the expression of vg.

These results, together with our previous observation that Vestigial can contribute to the expression of wg at the wing margin, indicate that Vestigial can indeed regulate gene expression in the wing blade. In addition, the correlation between levels suggests that different concentrations of Vestigial elicit different effects and that, in some instances, these effects seem to be independent of the levels of Wingless. Thus Vestigial might act in concert with Wingless, and not simply as an effector of wg, in the regulation of gene expression.

In our experiments, we notice that very high levels of expression of Vestigial, or long-term exposure to vestigial expression, results in the cessation of proliferation, loss of gene expression and, eventually, in cell death (T. K. and A. M. A., unpublished data). This might account for the small of wings that are often visible when vg is overexpressed in the developing wing and would correlate with the zone of non-proliferation that appears at the wing margin, where the levels of vg expression are elevated.

The vestigial gene plays a central role in the development of the wing of Drosophila. Previous studies have shown that the spatial and temporal modulation of vg expression through specific enhancers is a driving force in the development and patterning of the wing (Williams et al., 1994; Kim et al., 1996, 1997; Neumann and Cohen, 1996a,b; Klein and Martinez Arias, 1998a,b). This regulation is exerted by three major signalling pathways, Notch/Su(H), Wingless and Decapentaplegic, acting differentially on two collections of enhancers: one early acting (vgBE) and one later acting (vgQE). Here we have shown that the vg gene product itself is an important input in the regulation of vg gene expression through the vgQE and that, in this process, Vestigial acts synergistically with the three signalling pathways that regulate its expression to drive the development of wing blade tissue. This result might provide an answer as to why the activity of the vgBE is required for the activation of the vgQE. The finding explains also the non-autonomous requirement of the Notch signalling cascade, since it is required for the activation of the vgBE.

We have also shown that, during early stages of wing development, the expression of vg is controlled by the vgBE and that Notch signalling is not only required for the maintenance of its activity but also for its initiation. Furthermore, we have provided some evidence to suggest that the activity of Wingless is required for the upregulation of this enhancer. These results add support to, but do not prove, the suggestion that wing development is initiated by the cooperation of Notch and Wingless signalling on the vg gene (Couso et al., 1995; Klein and Martinez Arias, 1998a).

The existence of an input of Wingless on the vgBE would be consistent with the existence of two putative TCF-1 binding sites in the first 80 bp of this enhancer. Deletion of this region of the vgBE substantially reduces its activity over the domain of wg expression. Further evidence for a participation of wg in the activity of the vgBE is provided by the observation that a dominant negative form of the Wingless receptor Dfrizzled2 leads to a strong reduction of the activity of the vgBE (Zhang and Carthew, 1998).

These results contrast with the report that loss of wingless function does not affect the activity of the vgBE (Neumann and Cohen, 1996a,b). This difference might reflect different ways of carrying out the experiments. The published results are based on clonal analysis, which might not be suitable for the study of the vgBE: our results suggest that the loss of wg activity reduces dramatically the activity of the enhancer but does not abolish it (Klein and Martinez Arias, 1998 and this work). This means that, taking into account the perdurance of lacZ and the limitations imposed by the need of the clones to survive, loss of Wingless in clones might affect the endogenous enhancer but have no observable effect on the reporter. This caveat and our results would account for the observation that large clones of wingless mutant cells do not affect the expression of a vgBElacZ but display nicks without affecting the size of the wing (see Neumann and Cohen, 1996a): it is likely that in these clones the activity of the endogenous enhancer is affected in a manner that is not reflected by the lacZ fusion.

Once vg expression is initiated in the wing through the vgBE, our results suggest that it is maintained in the growing pouch by a synergistic activity of Wingless and Vestigial on the vgQE. This model predicts that loss of the vgBE will result in the loss of the wing, and that loss of the vgQE should lead to small wings with a well-defined margin. The first prediction is fulfilled by the vg^83b27 mutation, which removes the vgBE and abolishes the development of the wing (Williams et al., 1994).

There are no mutations that remove the vgQE without affecting all functions of vg and which would allow us to test whether or not the growth of the wing requires the expression of vg through the vgQE. However, support for a requirement of the vgQE in the growth of the wing blade is provided by the clonal analysis, which demonstrates the requirement of vg activity for the survival of the cells in the wing pouch and by a recent report on the effects of dominant mutations in the vg gene (Kim et al., 1996; Simmonds et al., 1997). The dominant mutation vg^W leads to a very small wing pouch, which still contains a wing margin (Simmonds et al., 1997). Indeed detailed analysis of this mutation shows that, whereas vg^W mutant flies have vg function from the vgBE, the vgQE in completely inactive (Simmonds et al., 1997). Our results predict that the loss of the activity of the vgQE should lead to the loss of the wing pouch, but not of the margin, as is observed in the vg^w mutant. Interestingly, the presence of the vg^w mutation causes extensive cell death at the early third instar (Simmonds et al., 1997), at the time of extensive cell growth in the wing pouch. This cell death is probably due to the loss of vg expression in descendants of cells that are determined at the DV boundary and which come to lie outside the domain of the vgBE due to cell division. These cells would normally maintain the expression of vg (and therefore their fate) through the activation of the vgQE, which is inactive in the presence of vg^w. Several other recently reported results are in agreement with our conclusion: the clonal analysis suggests an autonomous requirement for vg by the cells of the wing pouch for survival (Simpson et al., 1981; Kim et al., 1996) and reduction of wg function results in the reduction of the size or complete loss of the pouch with vg expression reduced to the DV boundary (Couso et al., 1995; Zecca et al., 1996; Neumann and Cohen, 1996a,b).

Our results highlight some features of the process of wing development in Drosophila. In particular they highlight interactions between the vgBE and the vgQE which can explain how the wing grows and is patterned. During wing development, the vgBE has two well-defined roles. First, it is required to express vg in the primordium for the wing blade at the end of the second larval instar. Second, vgBE-mediated expression of vg in the primordium is a prerequisite to initiate the activity of the vgQE which will drive vg expression in the developing wing blade outside the domain of the vgBE. Thus, failures in the activity of the vgBE will result in failures in the activation of the vgQE and thus in the early stages of the development of the wing blade. Later expression of vg from the vgQE is likely to be sufficient to maintain the expression and function of vestigial (and identity) in the growing blade.

These interactions suggest a framework to consider how the wing blade develops and, in particular, about the interactions observed between the DV interface and the wing blade during the growth of the wing (Fig. 8). The primordium for the wing is defined, in the late second instar, precisely at the DV interface through interactions between Notch and Wingless signalling (Fig. 8; for further details see Klein and Martinez Arias, 1998a and Fig. 8A-C). These local activities are channelled through the vgBE and can account for the different mutant phenotypes and the regulatory and genetic interactions between these molecules.

In the early stages of the third instar, the primordium expands (Klein and Martinez Arias, 1998a,b), begins to grow and cells that come to lie outside the DV interface maintain the expression of vg through the activity of the vgQE (Fig. 8D). Clonal analysis and studies of cell proliferation (Milan et al., 1996) have not revealed any specific pattern to this phase of disc growth. However, it is clear that the DV interface and the AP boundary play some role in this process. Our results indicate that this role is likely to be the maintenance of the expression of vg through the vgQE (Fig. 8D,E) by the convergent effects of Vestigial, Wingless and Dpp. However, although vg as well as wg are required for the proper growth of the wing pouch, overexpression of these genes along the DV boundary does not lead to a larger pouch size (T. K. and A. M. A., unpublished data; see above), suggesting that the role of these genes during cell proliferation is more permissive rather than instructive.

At the beginning of the third instar, the wing primordium is subdivided into a proximal region, which will develop as hinge, and a distal one, which will develop as blade and upon which the margin will define the most distal point (Klein and Martinez Arias, 1998a, Klein et al., 1998). These subdivisions seem to occur in a sequence and as the wing blade grows they continue as revealed by the expression of Dll in a subset of cells within the wing blade. The expression of Dll requires Vestigial and is likely to define a domain within which high levels of Wingless signalling will induce bristle formation. Consistent with this, loss of Dll function results in the loss of bristles in the wing margin (Gorfinkiel et al., 1997). Dll delimits the distalmost region of the leg disc and its expression within the wing blade might reflect the homology that has been suggested to exist between the wing and the leg discs and highlights a clear proximodistal organization of the wing disc (Klein et al., 1998).

Analysis of the vg¹ and Ser mutant phenotypes suggests that the activity of vg in the initiation of wing development is only required for a limited period of time i.e. it is stabilized very quickly (Klein and Martinez Arias, 1998a,b). In both mutant situations although vg expression is initiated normally, it is rapidly lost and the mutant flies develop a small winglet, which is not observed in ap mutant flies or flies mutant for vg null alleles (T. K. and A. M. A., unpublished data). Therefore the short-lived activity of vg in these mutants is sufficient for the stable commitment of cells to the wing pouch fate. It is likely that wg plays a very important role in this commitment, which is reflected in its synergistic effects with Notch.

The contributions of the DV and AP axes to the patterning of the wing are becoming clear. However, little is still known about what triggers the foci of proliferation that drive the growth of the wing blade or how the interactions that we have described here control the final size of the wing. Our results here suggest a more indirect role of wg and vg in cell proliferation than previously suggested. Although the loss of function of each of these genes has a significant influence on the cell proliferation in the wing, the overexpression of both with the vgBEGal4 line do not lead to the increase of the wing pouch of the late third instar (see above). This suggests a more permissive role of the two genes in cell proliferation, probably to maintain the identity of the wing pouch. Studies that suggest that the proliferation is induced by local clues and is a cell autonomous property (Milan et al., 1996) support this conclusion.

vestigial encodes a nuclear protein with an unknown molecular function, but which can trigger the development of wing-tissue-specific properties upon developmentally unrelated groups of cells (Kim et al., 1996; Klein and Martinez Arias, 1998a and this work). Although there is no evidence that Vestigial interacts directly with DNA, our results suggest that it can influence patterns of gene expression and that this is probably the way in which it exerts its function during development.

It is possible that the role of Vestigial is to ‘select’ a combination of downstream genes that will lead to the development of wing tissue. In this situation, the interactions that we observe between Vestigial and the different signalling pathways might be essential in channelling the nuclear effectors of these signalling pathways to the promoters of genes that provide the identity of the wing. For example, wingless (Struhl and Basler, 1993; Wilder and Perrimon, 1995), dpp (Lecuit and Cohen, 1997) and Notch (Couso et al., 1994) are also required for the development and patterning of legs; however, if vg is expressed in cells from the leg discs, their development is altered and they develop wing tissue. This suggests that Vestigial provides an essential molecular context for wing development.

Pattern formation is often described in terms of ‘positional information’. In this paradigm, the informational input is thought to be universal, but its ‘interpretation’ is cell type specific (Wolpert, 1969). The widespread and recurrent use of a few signalling pathways in different developmental systems could be viewed as the proof of the universality of positional information, at least with regard to some of its elements. Our results suggest that Vestigial is an example of a key element in the interpretation of this information: wherever it is expressed, it channels the output of these signalling pathways into wing tissue. In some sense, Vestigial is an example of a molecule directly involved in the interpretation of positional information. It is likely that in different organs and structures, there are molecules that perform a similar role to that of Vestigial during wing development in Drosophila. Furthermore, the architecture of regulatory networks allows some flexibility as to the position and molecular nature of these molecular interpreters in the signalling networks.

Work during the last few years has begun to clarify the role of Vestigial during wing development (Williams et al., 1993, 1994; Kim et al., 1996, 1997a,b; Couso et al., 1995; Klein and Martinez Arias, 1998a). Analysis of epistatic relationships between genes involved in wing development has shown that vestigial does not lie at the bottom of a linear hierarchy (see e.g. Neumann and Cohen, 1996b; Zecca et al., 1996), but that it is a central element of a dynamic molecular network in which Wingless, Notch and Vestigial interact in a sequence of interdependent regulatory steps that set up the wing primordium development (Klein and Martinez Arias, 1998a,b and this work). The results presented here suggest that, during the growth of the wing, these interactions continue with the addition of one more element to the network, Dpp, and with Vestigial continuing to play a central role in the control of gene expression.

During the third larval instar, the expression of vg, like that of Dll, is regulated by Wingless (Kim et al., 1996; Zecca et al., 1996; Neumann and Cohen, 1996b). However, we have shown here that Vestigial can itself regulate the expression of targets of Wingless, even in the absence of wg expression and that Vestigial, but not Wingless, can rescue the loss of wing development that results from the loss of Notch signalling (see above and Klein and Martinez Arias, 1998a). Since Vestigial itself is distributed in a gradient across the developing wing (Williams et al., 1991; Zecca et al., 1996; Neumann and Cohen, 1996b), these observations suggest that Vestigial might play a more instructive role in the development and patterning of the wing.

The results that we have presented here and elsewhere (see Klein and Martinez Arias, 1998a), raise questions about the proposal that Wingless acts as some sort of organizer during the development of the wing blade (Diaz-Benjumea and Cohen, 1995; Neumann and Cohen 1996b, Zecca et al., 1996) but they also provide some insights as to what might be the actual function of Wingless. In the instances that we have tested, it seems that Wingless cannot elicit de novo gene expression alone and always acts synergistically with other signalling systems that direct gene expression during wing development. In particular, the function of wg appears to be to stabilize and maintain patterns of gene expression rather than direct patterns of gene expression. This function is very similar to the one that it plays in the embryo where Wingless does not initiate gene expression but rather reveals and maintains patterns of gene expression that have been established by other regulatory events (see Martinez Arias, 1998 for a review). Interestingly, a similar role for other Wnt proteins can be inferred from experiments that test their potential during the specification and patterning of the mesoderm in Xenopus (Cui et al., 1996; Hoppler and Moon, 1998).

We want to thank S. Carroll for the gift of the vg antibody, several vg enhancer-lacZ constructs and many useful discussions. We also want to thank S. Bray and S. Cohen for providing the anti-Dll antibody and K, Brennan, J. Modolell and N. Lavery for comments on the manuscript. This work was supported by The Wellcome Trust and partly by the Deutsche Forschungs-Gemeinschaft.