Different mechanisms initiate and maintain wingless expression in the Drosophila wing hinge

Members of the Wnt family of secreted glycoproteins are involved in numerous developmental events in many organisms, from the nematode C. elegans to mammals. Among functions provided by Wnt proteins are embryonic patterning, cell fate specification, cell polarity, cell proliferation and pattern-organising activity (Cadigan and Nusse, 1997).

In Drosophila the best characterised Wnt gene is wingless (wg), one of the first members of this gene family to be identified (Sharma and Chopra, 1976; Baker, 1987; Rijsewijk et al., 1987). Wg function is required throughout development in a wide range of patterning events at different times and in different tissues. These include the development of embryonic epidermis, the head, the CNS, midgut, the heart, muscles and malpighian tubules (reviewed by Cadigan and Nusse, 1996). In imaginal discs, wg is expressed in a very complex pattern (Couso et al., 1993), and one interesting feature is that distinct enhancers control the different expression domains; enhancers that are themselves activated by different signalling pathways. In ventral discs (legs and antenna), wg is expressed in a sector that corresponds to ventral/anterior cells. The secreted protein Hedgehog activates the enhancer that drives this expression domain, which defines ventral fate and provides organising activity for the development of the proximodistal axis (Struhl and Basler, 1993; Basler and Struhl, 1994; Díaz-Benjumea et al., 1994). In the wing disc wg shows a very dynamic pattern of expression. In second instar larvae, wg is expressed in a ventral/anterior sector in a pattern similar to that displayed in the leg. This early expression is required for the specification of wing fate (Morata and Lawrence, 1977) and is under the control of Hedgehog signalling pathway (J. D.-B., unpublished). Unlike in the leg, this expression later fades away and wg starts to be expressed in a wide stripe that corresponds to the presumptive wing margin. The Notch signalling pathway also controls this enhancer. In this way, Wg is involved in the specification of the wing margin, which is required for the promotion of cell proliferation and patterning of wing cells (Phillips and Whittle, 1993; Díaz-Benjumea and Cohen, 1995; Zecca et al., 1996; Neumann and Cohen, 1997). In the mesothorax, wg is expressed in an anterior/posterior stripe, and is required for the specification of dorsocentral bristles. The enhancer that positions this stripe seems to be controlled by the GATA protein Pannier and the zinc-finger protein U-shaped (García-García et al., 1999). wg is also expressed in two concentric rings that surround the wing pouch. The inner ring (IR) is activated in early third instar, and the outer ring (OR) is activated in late third instar. These two rings define the wing bases (also called wing hinge), which is the region that joins the wing and thorax. This region has a complex structure that is required for the fluttering of the wing. In some wg mutants, this region is deleted and the wing is joined directly to the thorax. The enhancer that drives wg expression in the IR has been identified within a 1.2 kb DNA fragment located about 9 kb 5′ of the wg promoter (Neumann and Cohen, 1996). This fragment is sufficient to direct reporter gene expression in the IR. wg alleles that specifically affect this enhancer have been characterised as spade mutants (Couso et al., 1994; Tiong and Nash, 1990).

We have analysed the mechanisms that drive wg expression in the IR. We present evidence that indicates that the genes vestigial (vg), rotund (rn) and nubbin (nub) are required, and that wg expression in the IR is driven by two independent mechanisms. The first initiates wg expression in early third instar larvae, and depends on cell interactions between vg-expressing and vg-non-expressing cells. The second mechanism is required for the maintenance of wg expression, and depends on an autoregulatory loop that requires the function of the genes homothorax (hth) and rn. hth expression in the IR seems to be controlled by Wg signalling, but rn expression depends on a signal secreted by the vg-expressing cells. Thus, wg expression in the IR is not maintained by lineage. We also present evidence that indicates that Wg function in the IR promotes the patterning of the hinge by generating different domains, which are defined by different combinations of gene expression.

The following fly strains were used: spd^flag and spd-lacZ (Neumann and Cohen, 1996); rn^Δ2.2, rn-lacZ, UAS-rn and rn-GAL4 (St Pierre et al., 2002); nub¹ and nub² (Ng et al., 1995); vg^83b27R (Williams et al., 1993); N^ts (Shallenberger and Mohler, 1978); hth-lacZ (Rieckhof et al., 1997); UAS-vg and vg^QE-lacZ (Kim et al., 1996); UAS-hth (Casares and Mann, 1998); dpp-GAL4 (Wilder and Perrimon, 1995); ap-GAL4 (Calleja et al., 1996); and en-GAL4 and wg-lacZ (Kassis et al., 1992).

To induce clones of ectopic expression, y hs-FLP122; Ac>y⁺>GAL4 UAS-GFP females were crossed either with UAS-vg or UAS-rn males. Embryos were collected after 24 hours and heat shocked at 34.5°C for 12 minutes at 36±12 hours of development.

To induce loss-of-function clones, embryos from the appropriate crosses were collected for 24 hours and heat shocked at 37°C for 1 hour at 36±12 hours of development. The genotype examined were: for rn clones, y hs-FLP122; rn^Δ2.2 FRT[80] / Ubi-GFP FRT[80]; for nub clones, y hs-FLP122; nub¹ FRT[40A] / Ubi-GFP FRT[40A]; and for hth clones ,y hs-FLP122; hth^P2 FRT[80] / Ubi-GFP FRT[80].

Ac>y⁺>GAL4 UAS-GFP / UAS-FLP; rn-GAL4 / rn-lacZ larvae were generated, dissected and stained with anti-β-galactosidase for analysis under the confocal microscope. This experiment lineage-tagged cells that expressed rn at any time during the development of the disc. FLP recombinase is expressed in cells expressing rn-GAL4, and mediates excision of the flip-out y⁺ cassette from the inactive construct to generate an active Ac>>GAL4 transgene that will express UAS-GFP. After excision of the cassette, GAL4 expression is regulated by the actin promoter and is clonally inherited in all the progeny of rn-GAL4-expressing cells (Struhl and Basler, 1993; Weigmann and Cohen, 1999). The pattern of GFP expression is compared with the pattern of β-galactosidase expression that shows the final pattern of rn expression.

Discs were dissected in PBS + 0.1% Tween and fixed with 4% formaldehyde in PBS for 20 minutes at room temperature. Standard protocols for immunostaining were followed. The antibodies used were: mouse anti-Dll (Vachon et al., 1992); mouse anti-Nub (Ng et al., 1995); rabbit anti-Vg (Williams et al., 1991); mouse anti-Wg (Brook and Cohen, 1996); and rabbit anti-β-galactosidase (Cappel).

The adult Drosophila wing is formed by a continuous monolayer of epidermal cells that folds to form the dorsal and ventral surfaces of the wing pouch. The two surfaces contact at the margin of the wing and extend proximally through the wing hinge to the dorsal notum and the ventral pleura. In the presumptive wing region of the wing disc, wg is expressed in a narrow stripe of cells that run all along the wing margin and in two rings that surround the wing pouch (Fig. 1A,B). We have examined the phenotypes and wg expression in several mutants in which the wing hinge is deleted.

The effects of removing wg expression in the IR can be observed in spade (spd) mutants (Tiong and Nash, 1990; Couso et al., 1994). spd mutations are a type of wg alleles that specifically removes wg expression from the IR, with no effects on other expression domains (Fig. 1C). In spd^fg wings, the hinge region is deleted, and the wing pouch appears directly joined to more proximal cells. In these wings, both wg-expressing cells and surrounding cells are missing. It has been shown that this phenotype is not caused by cell death is rather a consequence of underproliferation in this region, suggesting that one of the functions of Wg in the IR is to promote local cell proliferation (Neumann and Cohen, 1996).

The rotund (rn) gene is a member of the Krüppel family of zinc-finger encoding genes (St Pierre et al., 2002). Among other phenotypes, rn mutations delete the wing hinge and remove wg expression from the IR (Fig. 1D).

The nubbin (nub) gene encodes a member of the POU family of transcription factors (Ng et al., 1995). In strong nub mutations wings are vestigial, but phenotypic analysis of weaker alleles shows that the wing hinge is deleted and the expression of wg in the IR is missing (Fig. 1E). We examined the hinge phenotype of the triple mutant spd^fg nub²; rn^Δ2-2 and it is similar to the phenotype of all of them, suggesting that the main cause of the phenotype is the lack of wg expression in the IR (data not shown).

The gene vestigial (vg) encodes a nuclear protein with no homology with other identified families of nuclear proteins (Williams et al., 1991). Based on its interaction with scalloped (sd), a transcription factor with a TEA/ATTS DNA-binding domain, it has been suggested that the function of Vg is to mediate transcriptional activation by Sd (Paumard-Rigal et al., 1998; Simmonds et al., 1998). vg expression in the wing is regulated by two separate enhancers: the boundary enhancer (BE) and the quadrant enhancer (QE). The BE is activated by the Notch signalling pathway and drives vg expression at the dorsal/ventral boundary in middle/late second instar larval stage. The QE is activated by the combined action of Wg and Dpp (a TGFβ homologue), and drives vg expression in the rest of the wing pouch from early third instar larval stage (Williams et al., 1994; Kim et al., 1996; Neumann and Cohen, 1997).

We examined the expression patterns of vg, rn and nub. In mature wing discs vg, rn and nub are expressed in three concentric domains, with the Vg domain the smallest one. At this stage the wing hinge is lined with several anterior/posterior folds (Fig. 2K). The boundary of vg expression coincides with the distal-most fold of the disc (Fig. 2E,K). The Rn domain is slightly broader and its boundary coincides with a second fold in the disc (Fig. 2F,K). The Nub domain contains the Rn domain and coincides with the third fold in the disc (Fig. 2G,K). The IR domain corresponds to the proximal-most area of the Rn domain (Fig. 2F inset, K).

We next examined the expression of these genes in early larval development. In middle/late second instar larvae the expression domains of vg, rn and nub in the presumptive wing pouch are slightly broader than the vg domain (Fig. 2A,B). The rest of the cells of the disc, which do not express nub, express the gene teashirt (tsh) (Ng et al., 1996; Fasano et al., 1991). wg is expressed only in a stripe of cells that corresponds to the presumptive wing margin (Fig. 2C). In early third instar larvae, wg starts to be expressed in the IR (Fig. 2D). This expression domain corresponds to cells that express rn and nub but do not express vg. wg expression in the IR promotes the growth of the hinge (Neumann and Cohen, 1996) and, in third instar larvae, gives rise to the expression patterns described above for vg, rn and nub. At this stage, the cells that express the IR enhancer are located at the limit of the domain 3 (Rn + Nub), and are several cells away from the boundary of vg expression (Fig. 2E). To investigate how the IR enhancer would be expressed in spd^fg discs, we examined lacZ expression in transgenic flies carrying lacZ driven by the DNA fragment that contain the IR enhancer (Neumann and Cohen, 1996). In wild-type discs the spd-lacZ construct is activated in a ring around the wing pouch that corresponds to the IR but also in the wing margin (Neumann and Cohen, 1996). In spd^fg discs of third instar larvae the spd-lacZ expression remains adjacent to the Vg domain as a consequence of local underproliferation, and the Rn and Nub domains, which contain the cells that express the spd-lacZ, are only slightly broader than the Vg domain (Fig. 2H-J).

As vg, rn and nub are expressed in the wing pouch in very similar domains, we wanted to determine whether they could be placed in a regulatory cascade. vg^83b27r is considered a null allele and produces flies with no wings (Williams et al., 1993). We observed that in vg^38b27r wing discs expression of wg, rn and nub was not detected in the wing pouch, although other domains of expression in notum or legs were not affected (Fig. 3A-C). We observed the same result in earlier discs, suggesting that Vg is required to initiate the expression of wg, rn and nub in the wing pouch.

To confirm these results and to test whether vg expression is sufficient to activate wg, rn and nub, we made use of the UAS/GAL4 system (Brand and Perrimon, 1993) to ectopically express vg. vg misexpression in the notum does not activate neither rn nor nub, so we tested other discs. In the eye discs of wild-type larvae, Rn and Nub are not expressed, and in the antenna discs they are expressed in the same pattern as in the leg, with Rn in a broad ring and Nub in several narrow rings (Fig. 3B,C). In dpp-GAL4/UAS-vg eye/antenna discs, Rn and Nub are detected both in the eyes and in the antennae in the Dpp domain (Fig. 3E,F).

We also examined the expression of Vg and Rn in nub¹ discs and the expression of Vg and Nub in rn^Δ2-2 discs. In both cases we did not detect changes in the expression patterns of these genes (data not shown). Ectopic expression of either rn or nub did not drive the expression of the other gene (data not shown). From these results, we conclude that Vg is necessary and sufficient to activate the expression of rn and nub in the wing pouch.

Although, in late second instar larvae Vg, Rn and Nub are found in almost coincident domains, in third instar the Rn and Nub domains include two sectors that do not express vg (Fig. 2K, domains 2 and 3). This observation suggests either that the activation of rn and nub is mediated by a non-autonomous mechanism, or that although Vg is required to initiate rn and nub expression, Vg-independent mechanisms might be required to maintain them (see below).

In vg^83b27R larvae, the IR is missing, but vg misexpression in the eye/antenna disc is not able to activate wg expression. This suggests that a more sophisticated mechanism may drive the expression of wg in the IR. vg is not expressed in the hinge. When vg is misexpressed in the IR domain by any one of several different GAL4 lines, wg expression is lost (Fig. 4I and data not shown). This indicates that Vg, although required for the IR activation, represses the IR when both are co-expressed in the same cells.

To better assess the role of Vg in the activation of wg expression in the IR, we made clones of vg-expressing cells in the different expression domains of the wing hinge (Fig. 2K). These clones behave differently depending on the domain where they appear. In domains 2 (Nub + Rn) and 4 (Tsh), vg-expressing clones did not induce wg expression. In domain 3 (Nub), clones of vg-expressing cells produce a non-autonomous expression of wg. In these clones, induced at 36±12 hours of development, we found three distinct results: first, wg is expressed in a line of cells surrounding the clone (Fig. 4A); second, the domain of cells expressing wg becomes broader (Fig. 4B); and third, wg-expressing cells form a ring that lies several cells away from the boundary of the clone (Fig. 4C). We conclude that the behaviour of these clones and their surrounding cells reproduces the normal development of wg expression in the IR. wg expression is first activated in cells abutting the Vg domain, and this ring of expression later moves away from the Vg domain. We propose that a signal from the vg-expressing cells induces wg expression in surrounding cells. As a consequence of Wg-promoted cell proliferation, the IR moves several cells away from the boundary of vg-expressing cells. Because in vg-expressing clones the cells that express vg are related by lineage, we must assume that some cells lose the expression of Wg and that this happens at least in the internal border of the IR (see Discussion).

To assess the role of Rn in this signalling, we examined the expression of rn in these vg-expressing clones. Clones of vg expression in domain 3 (Nub) activate rn expression. This expression is not limited to the cells of the clone, but the surrounding cells also express rn (Fig. 4D,E). Considering that Vg is a nuclear protein, a Vg-dependent signalling may be required to activate rn expression. The activation of rn in these clones seems to be restricted to domain 3 (Nub). One explanation could be that Nub is required. Nevertheless, rn expression was not altered in nub¹ discs, in which Nub protein is detected by antibody staining in only a few cells (Ng et al., 1995; Cifuentes and García-Bellido, 1997). We therefore think that the activation of rn by Vg-dependent signalling may be repressed in tsh-expressing cells (domain 4). Similar to the observed dynamics of wg expression, we found clones in which rn is only expressed within the cells of the clone, and clones in which rn is also expressed in the surrounding cells. It is interesting to note that only clones in which rn expression is non-autonomously activated show wg expression (Fig. 4F). These results suggest that in early third instar larvae, a Vg-dependent signalling pathway non-autonomously activates rn expression in neighbouring vg-non-expressing cells, and this makes these cells competent to activate the IR enhancer.

To determine whether Rn is by itself able to activate wg expression in cells that do not express vg, we examined the expression of wg in clones of rn-expressing cells. These clones, when induced in domain 3 (Nub), activate wg expression. Unlike vg-expressing clones, in these clones wg expression is restricted to cells of the clone (Fig. 4G,H). In addition, within the clones the only cells that express wg are those lying close to the Vg domain. This suggests that Rn, although required for wg expression, it is not sufficient. We think that when the IR was activated, some of these clones had a few cells, but only the cells that lie close to the Vg domain activated wg. After several cell divisions, some of these cells lie out of the IR domain but, because they express rn, they retain the expression of wg.

In view of these results, we propose a two-step model, in which Vg first non-autonomously activates rn expression both within the Vg domain and in surrounding cells. This generates two adjacent domains: vg/rn-expressing cells and rn-expressing cells. Cell interactions between these two groups of cells then drive the activation of the IR enhancer in rn-expressing cells.

The results presented above indicate that Rn and Nub are required for the activation of the IR and that this activation depends on a signal from vg-expressing cells. To identify in which cells Rn and Nub functions are required for IR activation, we performed genetic mosaic analysis of strong loss-of-function alleles of rn and nub.

rn^Δ2-2 is considered to be a null allele (St Pierre et al., 2002). rn^Δ2-2 homozygous flies are viable, and display as wing phenotypes the deletion of the wing hinge and a nick in the posterior proximal wing margin (St Pierre et al., 2002). Clones of rn^Δ2-2 cells in a rn^Δ2-2/+ background remove wg expression from the IR. This phenotype is strictly cell autonomous, indicating that Rn function is only required in the IR cells (Fig. 5A,B). rn clones that straddle the wing margin do not affect the expression of wg.

nub¹ is a viable and recessive strong loss-of-function allele that causes a dramatic reduction in wing size and the deletion of wing hinge (Fig. 1E) (Ng et al., 1995). We induced clones of nub¹ cells in an otherwise wild-type nub¹/+ background. These clones remove wg expression. The phenotype is strictly cell autonomous, indicating that Nub is only required in the IR expressing cells (Fig. 5C,D). It therefore appears that Rn and Nub, although expressed in vg-expressing cells, are not required in these cells.

As we have seen above, vg expression in second instar larvae drives the expression of rn and nub, two genes that are required both for the development of the wing hinge and for the activation of the wg IR enhancer. Clones of vg-expressing cells in the Nub domain activate the IR enhancer. These ectopic IR domains, although generated in cells that abut the clone boundary, end up several cells away. This behaviour reproduces the normal development of the IR, which is initially expressed at the boundary of vg expression, but later, as a result of Wg-induced cell proliferation, moves several cells away. These observations suggest that a signal coming from vg-expressing cells activates the IR enhancer in the surrounding cells.

To test this hypothesis, we generated a new artificial boundary of vg expression at an ectopic position, by removing vg expression in cells within the vg domain. vg mutant clones do not generate this situation, as they do not proliferate for more than a few cell divisions (Azpiazu and Morata, 2000). We therefore misexpressed the gene homothorax (hth) in the Vg domain. hth encodes a homeodomain protein of the Meis family (Rieckhof et al., 1997). In the wing, hth is expressed in two rings that overlap with the IR and the OR of Wg (Casares and Mann, 2000; Azpiazu and Morata, 2000), and it has been suggested that Hth interferes with Wg signalling (Abu-Shaar and Mann, 1998).

In dpp-GAL4/UAS-hth, the Wg-dependent vg expression was repressed in the Dpp domain (Fig. 6A, note that the N-dependent BE is not affected, indicating that Hth does not interfere with N signalling). In these discs, rn and nub expression are not affected (Fig. 6B,C), and two new stripes of wg expression appear (Fig. 6D-F, note that not all cells that express hth activate wg expression). These two new stripes of Wg are clearly seen in middle third instar larvae, and correspond to Dpp-expressing cells that abut the vg expression domain. In mature larvae the epithelia is folded, and only a stripe of Wg is detected (Fig. 6F).

To confirm that this ectopic expression of wg is driven by he IR enhancer, we examined Distal-less (Dll) expression and N function. Dll is normally expressed in the wing pouch but is not expressed in the IR. Its expression depends of wg expression in the wing margin (Neumann and Cohen, 1997). In dpp-GAL4/UAS-hth, Dll expression is missing in the Dpp domain, where Wg is being ectopically expressed (Fig. 6G). N is required for the expression of wg in the wing margin but is not required for the activation of the IR enhancer. N^ts is a thermosensitive allele which at a restrictive temperature (30°C) behaves as a strong loss-of-function allele. In N^ts; dpp-GAL4/UAS-hth, the expression of wg in the wing margin is missing, but neither the ectopic wg expression in the Dpp domain nor the wg expression in the IR is affected (Fig. 6H). This result suggests that both domains are independent of N signalling.

For final confirmation that cell interactions between vg-expressing and vg-non-expressing cells activate the IR enhancer, we tested whether hth misexpression was able to activate wg directly by examining wg expression in engrailed-GAL4/UAS-hth. In these discs, hth is expressed in the whole posterior compartment, but Wg is only detected in posterior cells that abut vg-expressing cells in the anterior compartment (Fig. 6I).

Once the IR enhancer is activated, Wg-induced local cell proliferation moves the IR several cells away from the boundary of vg expression. This cell proliferation generates three new domains defined by different combinations of gene expression. These are: cells expressing Nub+Rn; cells expressing Nub+Rn+IR; and cells expressing Nub (Fig. 2K). This raises the question of how wg expression is maintained far from the Vg boundary. To address this we examined the role of hth.

In second instar larvae, hth is expressed in the wing disc at low levels, in a pattern complementary to vg. In early third instar larvae, after wg is expressed in the IR, hth starts to be expressed at higher levels in two rings that overlap with the IR and the OR of wg expression (Fig. 7A). Based on experiments in which Wg signalling was compromised, it has been proposed that hth is a target of Wg signalling in the hinge (Casares and Mann, 2000). hth expression is missing in the spd^fg mutant (Fig. 7B). Clones of hth mutant cells prevent wg expression at the late third instar stage (Fig. 7D) (Casares and Mann, 2000). But hth clones do not block wg expression observed in early third instar larvae (Fig. 7E). These observations suggest that wg expression, although induced by a Vg-dependent signal, is maintained by a different mechanism that requires Hth.

Although wg-expressing cells in the IR move several cells away from the Vg domain during the growth of the hinge, the proximal limit of the IR always coincides with the border of rn expression. Clones of rn-expressing cells in domain 3 (Nub) maintain wg expression in the proximity of the IR (Fig. 4G,H). Taken together, these results suggest that Rn is also required to maintain wg expression. In this model, wg would maintain its own expression by an autoregulatory loop that requires Hth and Rn. hth expression depends on Wg, but rn expression depends on a signal from vg-expressing cells. When IR cells proliferate and drop out of range of Vg-dependent signals, they would lose rn expression and, as a consequence of this, wg and hth expression may also be lost. Thus, wg expression, although maintained by an autoregulatory loop, would not be maintained by lineage alone.

To confirm this model, we conducted a lineage-tracing experiment to examine if cells at the border of the Rn domain have a tendency to lose rn expression. We compared the pattern of rn expression in mature discs with the distribution of cells that express rn at any time during larval development (see Materials and Methods for details). When compared, we observed that the final domain of rn expression was smaller than the domain of cells that expressed rn in larval stages (Fig. 7C). This indicates that in development cells are dropping out of the Rn domain.

Interestingly, wg expression in the IR plays a role in maintaining vg expression. The sharp border of the Vg domain becomes less well defined in spd^fg wing discs, but is not affected in N^ts (30°) disc, in which both vg BE and wg expression from the wing margin have been removed (Fig. 7F-H). This implies that Wg from the IR helps to maintain the expression of the vg quadrant enhancer.

The very precise spatial and temporal control of gene expression plays an important role in pattern formation. One of the best-characterised examples of this is the development of the Drosophila embryo, in which sets of genes can be placed in a regulatory hierarchy, and have a very precise expression that is spatially and temporally regulated. In several cases, these genes are expressed at different times under the control of different enhancers, which in turn are activated by different developmental mechanisms (Pankratz and Jäckle, 1993). This precise control of gene expression is not restricted to early stages of development, and to some extent post-embryonic development introduces increased complexity, as pattern formation is associated with the control of cell proliferation (Day and Lawrence, 2000).

The wg gene has been exhaustively studied, and it has been reported that Wg plays multiple roles in the development of imaginal discs. In many cases, Wg has an instructive function, as its very precisely regulated expression is required for normal development (Klingensmith and Nusse, 1994; Martínez-Arias et al., 1999). One interesting feature is that an equally complex regulatory region controls the complex pattern of wg expression in the development of imaginal discs (Neumann and Cohen, 1996) (K. Johnson and J. P. C., unpublished).

wg is expressed in the wing in two rings that are required for patterning the wing hinge. This flexible region is required for wing flapping and for the movement of extension and flexion over the abdomen at rest (Snodgrass, 1935). We have examined the mechanisms that drive wg expression in the IR of the hinge. Our results indicate the expression of wg is regulated by two different mechanisms, one to initiate and the other to maintain its expression. These processes involve the genes vg, rn, nub and hth, and other as yet unidentified genes.

At middle second instar larvae, vg gene, as detected by Vg antibody staining, is expressed in the wing disc in a horseshoe-like domain. The centre of this domain corresponds to the presumptive wing pouch (Klein and Martinez-Arias, 1999; Williams et al., 1991). The results presented here indicate that Vg is required to activate the expression of rn and nub genes in the wing disc. This activation is restricted to the cells that will take wing fate and takes place in the cell that express vg, and also in the surrounding cells, suggesting that a Vg-dependent short-range signal activates rn and nub expression. At this time, the expression of nub and tsh in the wing disc are complementary and cover the whole disc (Ng et al., 1996) (Fig. 8).

The expression of these genes in a domain broader than the Vg domain creates a ring of cells that express rn and nub but not vg. We have presented evidence indicating that a signal from vg-expressing cells activates the wg IR enhancer in adjacent rn/nub-expressing cells. Unlike the activation of rn and nub, the activation of wg expression by the IR enhancer is repressed in cells that also express vg. So, the IR enhancer is activated only in cells that surround the Vg domain.

During the development of the disc, the position of the IR moves several cells away from the Vg domain. This implies either that the Vg-dependent signalling is able to activate the IR over a long range (Liu et al., 2000), or that a different, Vg-independent, mechanism maintains the IR.

When artificial Vg/Rn-Nub interfaces are generated experimentally, the IR enhancer is activated in rn-nub-expressing cells that abut the Vg domain (Fig. 6D-F). This ectopic IR is around four cells wide, indicating the active range of the signal that activates wg expression. Our results indicate that at distances greater than this, a Vg-independent mechanism maintains wg expression in the IR.

Several results presented here, plus others reported elsewhere, indicate that Wg signalling activates hth expression, which is in turn required to maintain wg expression. wg and hth are co-expressed in the IR and OR, and wg expression precedes hth expression. Furthermore, hth expression is missing in spd^fg discs, and wg expression is lost in hth mutant clones. Nevertheless, spd^fg discs show activation of the IR enhancer, as revealed by the spd-lacZ construct (Fig. 2H-J) and wg expression is not affected in hth mutant clones when observed in early third instar larvae (Fig. 7E). This indicates that Hth, while required to maintain IR activation, is not required to initiate wg expression.

The rn clonal analysis indicates that Rn is also required for wg expression. One interesting observation is that, when the IR moves away from the Vg domain, wg-expressing cells are always maintained at the limit of rn expression. rn is activated by a Vg-dependent signal. This implies that the activity range of the signal and the lifetime of the Rn protein together limit the domain of rn expression. So one explanation for why the IR is always maintained in the limit of rn expression is that, as a consequence of cell proliferation, cells drop out of the range of the Vg-dependent signalling. Thus, cells simultaneously lose the expression of both rn and wg. The result of the rn lineage-tracing experiment supports this prediction (Fig. 7E). Taken together, these results suggest that an autoregulatory loop involving Hth and Rn maintains wg expression. Although hth expression depends on Wg, rn expression depends on Vg, so wg expression in the IR is not maintained by lineage. wg autoregulation has been reported in embryo development (Hooper, 1994; Manoukian et al., 1995), and a negative mechanism of ‘self-refinement’ has been suggested in wing margin specification (Rulifson et al., 1996). However, in neither of these cases has a role been reported for Hth or Rn.

Wg-promoted cell proliferation generates a new domain between the IR and the Vg domain (Fig. 2K, domain 2). This indicates that at this stage Vg-dependent signalling is unable to activate wg expression in adjacent cells. Otherwise the IR would be expressed in the whole Rn domain. One explanation for this could be that there is a temporal window for the activation of wg, but vg-expressing clones induced in mid/late third instar larvae are able to activate wg (data not shown). Another explanation could be that a repressor is expressed in this domain (Liu et al., 2000). Clones of vg-expressing cells placed in this domain do not activate wg, which supports this explanation. In the experiment in which we prevented vg expression in the Dpp domain, the new stripe of ectopic Wg did not recognise this domain (Fig. 6D-F), suggesting that the proposed repressor may be a target of Vg signalling. One alternative explanation is that wg refines its own expression domain by repressing the Vg-dependent activation. This has been proposed for the expression in the wing margin (Rulifson et al., 1996), but does not seem to be the case here. In experiments in which third instar larvae carrying a thermosensitive allele of wg (wg^ts/wg^cx4) were reared at the restrictive temperature (16 hours at 30°C) and stained with Wg antibody, we did not detect changes in the pattern of wg expression in the IR (data not shown). However, we observed that the expression in the wing margin was widened (Rulifson et al., 1996).

Thus, the proximal and distal limits of the IR would be defined respectively by the limit of rn expression and by the limit of the expression of the proposed repressor. In summary, our results suggest that at least four different target genes are independently activated by one or more signals that emanate from vg-expressing cells: rn and nub are activated in second instar larvae; wg is activated in early third instar larvae (this activation requires the function of Rn and Nub and is repressed by Vg); and finally the repressor, which would be activated in middle third instar larvae.

One interesting observation that can be made our results relates to how the hinge is patterned. As a result both of local cell interactions and Wg-promoted cell proliferation, several domains, which are defined by different combinations of gene expression, are established (Fig. 8). The generation of these domains is, in part, a consequence of that the expression of these genes are not maintained by lineage, but also because there is not evidence of lineage restrictions. Thus, cells at the borders of both the IR domain and the Vg domain lose wg and vg expression, and fall into adjacent domains. However the expression in cells within a given domain, away from the border, must be more efficiently maintained by a phenomena similar to the reported community effect (Gurdon et al., 1993), because no holes are detected in the pattern of expression.

We thank Natalia Azpiazu, Sean Carroll, Steve Cohen, Richard Mann and the Bloomington Stock Centre for fly stocks. We are grateful to Sean Carroll for the Vg antibody, to Steve Cohen for Dll, Wg and Nub antibodies, and to Sonsoles Campuzano for critical reading of the manuscript. This work has been supported by a grant from the Ministerio de Educación y Cultura (PM97-0145) and an institutional grant from the Fundación Ramón Areces to the Centro de Biología Molecular-Severo Ochoa.