Connecting Hh, Dpp and EGF signalling in patterning of the Drosophila wing; the pivotal role of collier/knot in the AP organiser

The secreted proteins of the Hedgehog (Hh) family play a crucial role in the patterning of many structures in vertebrates and invertebrates, by providing differentiating cells and tissues with positional information (Ingham and McMahon, 2001). Hh is a central patterning signal in the Drosophila wing, which, like all adult appendages in the fly, derives from a monolayer epithelium known as an imaginal disc, which is subdivided by a lineage restriction into anterior (A) and posterior (P) compartments. Hh synthesised by P cells diffuses into the A compartment forming a short-range activity gradient that activates the localised expression of a number of target genes, including patched (ptc) and decapentaplegic (dpp, a TGFβ family member). By sequestering extracellular Hh, Ptc limits the diffusion and range of activity of Hh (Chen and Struhl, 1996). Dpp, a long-range morphogen, diffuses into both the A and P compartments, and acts to relay positional information from the AP organiser, controlling growth and patterning of the distal regions of the wing (Brook et al., 1996; Lawrence and Struhl, 1996). A similar relay mechanism has been proposed for anteroposterior patterning of the vertebrate limb. In this case, Sonic Hedgehog (Shh) initially acts at long range to prime the region of the limb competent to form digits, and later acts at short range to induce expression of bone morphogenic proteins (Bmps/TGFβ), the morphogenetic action of which specifies digit identity (Drossopoulou et al., 2000). In both the Drosophila wing and vertebrate limb, however, it remains unclear how Hh (Shh) exerts its direct patterning activity.

One morphological read-out of the AP patterning information in the Drosophila wing is the positioning of intervein and longitudinal L2 to L5 provein domains in larval imaginal discs (Sturtevant and Bier, 1995). Position of veins is prefigured by expression of rhomboid (rho) and activation of the EGF pathway in the central provein cells, which will differentiate as the adult vein cells (De Celis, 1998; Guichard et al., 1999; Sturtevant et al., 1993). Activation of blistered/Drosophila Serum Response Factor (bs/D-SRF) in complementary regions endows cells with an intervein fate (Montagne et al., 1996; Price et al., 1989; Roch et al., 1998; Sturtevant et al., 1993).

Hh short-range activity is responsible for patterning the central region of the Drosophila wing: specification of the L3-L4 intervein region, the anterior L3 vein and the posterior L4 vein (Mullor et al., 1997; Strigini and Cohen, 1997). We have previously shown that this activity is mediated by activation of the transcription factor Collier/Knot (Col/Kn), in a narrow stripe of cells along the AP boundary that we designate below as the AP organiser. The restricted domain of col expression, compared with that of dpp, established that dpp and col are induced by different levels of Hh activity, providing a paradigm for Hh morphogen properties (Vervoort et al., 1999). Col activates and/or maintains high levels of expression of BS and downregulates expression of EGFR in the AP organiser (Mohler et al., 2000; Vervoort et al., 1999), therefore preventing these cells from adopting a provein fate. Formation of posterior L4 vein requires activity of vein (vn), which encodes a diffusible neuregulin-like protein, one of the known EGFR ligands and activates the EGF pathway (Schnepp et al., 1996; Simcox et al., 1996). Like col, vn is transcribed in anterior cells along the AP border, consistent with mosaic analyses indicating that vn activity is provided by anterior cells (Garcia-Bellido et al., 1994). Together, these data led to the proposal that Hh patterns the central part of the wing via activation of col and vn in cells along the AP boundary (Mohler et al., 2000; Vervoort et al., 1999). col requirement for L4 vein formation has not been previously investigated, however, and the links between col and vn activity remain elusive. Moreover, the molecular mechanisms involved in positioning of the L3 vein remain to be firmly established.

In this report, we re-examine in depth the role of col activity in establishing the vein pattern and size of the wing. First, detailed morphological comparison of the hypomorphic and null col mutant wings allowed to precise the null mutant phenotype: a reduced size of the wing contributed by both reduction of the posterior compartment and lack of L3-L4 intervein, the loss of L4 vein, and a posterior shift in the position of L3 vein. At the molecular level, we show that, in addition to BS, col upregulates transcription of vn in the AP organiser, providing a first molecular interpretation for Col requirement for L4 vein formation. Col is also required for local upregulation of extramacrochaete (emc) and Notch (N), two genes involved in the control of cell proliferation, thus linking Hh signalling to cell division in the L3-L4 intervein primordium. Furthermore, we show that the recently reported downregulation of Dpp signalling in cells receiving high doses of Hh (Tanimoto et al., 2000) is mediated by col activity. By manipulating the levels of vn expression and Thickvein (Tkv, a type 1 Dpp receptor) activity, we provide evidence that modulation of Dpp signalling in the AP organiser is essential both for positioning L3 vein and formation of L4 vein. The integrative role of Col thus revealed a complex and previously underemphasised interplay between the Hh, Dpp and EGF pathways in determining the size and vein pattern of the Drosophila wing.

col¹ is classified as a null allele and kn¹ as an hypomorphic allele of col (Crozatier et al., 1999; Vervoort et al., 1999). The P[col5-cDNA] transgene allows to rescue the embryonic lethality (Crozatier et al., 1999), but not the adult wing phenotype of col¹ mutants. vn^M2 /TM3, Sb was provided by A. Garcia-Bellido.

The transgenic lines UAS-Vn, UAS-TkvDN (TkvDN corresponds to Tkv1DGSK, which lacks the GS box and kinase domain), dpp-Gal4 and ptc-Gal4 have previously been described (Haerry et al., 1998; Morimura et al., 1996; Simcox et al., 1996). For ectopic expression of vn and tkvDN in the col¹ genetic background, the ptc-Gal4, UAS-TkvDN and UAS-Vn transgenes were introduced into the col¹, P(col5-cDNA) chromosome by recombination.

Clones of col¹ mutant cells were generated by Flp-mediated mitotic recombination (Xu and Rubin, 1993) by heat-shock treatment at 37°C for 2 hours of first and second instar larvae of the following genotype: y, w, Hs-Flp/+; FRT-42D, ubi-GFP/ FRT-42D, sha^VAIS1, col¹. The FRT-42D, sha^VAIS1, col¹ chromosome was constructed by recombination. The cell marker was shavenoid (sha).

In situ hybridisation of wings of third instar larvae and pupae were carried out as described elsewhere (Sturtevant et al., 1993). Double labelling with two probes simultaneously was described elsewhere (Crozatier et al., 1996), using the alkaline phosphatase substrate kit 1 provided by Vector. RNA probes were synthesised from cDNA plasmids except for the col probe, which contains intronic sequences (Crozatier and Vincent, 1999). The following primary antibodies were used: rabbit anti-Col, anti-P-Mad (Persson et al., 1998) (provided by T. Tabata), anti-phospho histone H3 (from Upstate Biotechnology), rat anti-Ci (which recognises both Ci155 and Ci75, a gift of R. A. Holmgren), mouse monoclonal N (C17.9C6) from Hybridoma bank (a gift of Véronique Van de Bor) and mouse monoclonal anti-BS produced by Michael Gilmam at Cold Spring Harbor Laboratory.

We have previously reported that in col¹ mutant flies [col¹ is a null mutation; col¹ adults are obtained upon rescue of the embryonic lethality by a col transgene (Crozatier et al., 1999)], the wings completely lacked the intervein normally separating L3 and L4 veins. In col¹/kn¹ mutant wings (kn¹ is a hypomorphic viable mutation) (Diaz-Benjumea and Garcia-Bellido, 1990; Nestoras et al., 1997) the L3-L4 intervein was only reduced in size, with an occasional partial apposition of L3 and L4 veins (Vervoort et al., 1999) (Fig. 1A-C). Detailed inspection indicated that, in both hypomorphic and null col mutants, there is a 20% reduction in the overall wing size compared with wild type (Fig. 1D). To investigate this phenotype, we compared the surface area of each intervein in col¹, col¹/kn¹ and wild-type wings. Average values were obtained from measurements of 10 individual wings of each genotype (data not shown). In col¹/kn¹ wings, the size of the L3-L4 intervein was 66±6% smaller that of wild type, whereas the size of the other A interveins (anterior margin to L2 and L2-L3) was unchanged, indicating that the reduction in number of cells is specific to this intervein. Surprisingly, however, in col¹ mutant wings in which the L3-L4 intervein (and L4 vein, see below) were missing, the L2-L3 domain was 12±5% larger than in either wild-type or col¹/kn¹ wings. Finally, in both col¹ and col¹/kn¹ wings, the L4-L5 and L5 to posterior margin intervein sectors were smaller than in wild type, by 17±5% and 27± 4%, respectively. Taken together, our data indicated that the reduced size of co1 mutant wings was due to the reduced size (or absence) of both the L3-L4 intervein and posterior interveins (Fig. 1D). As col expression is restricted to the L3-L4 intervein (Vervoort et al., 1999), its activity on the regulation of cell proliferation in the P compartment must be cell non-autonomous. The larger size of the L2-L3 intervein in col¹ wings raised both questions of the identity of the single central thick vein observed in these wings and its position relative to the AP boundary. This vein showed the presence of campaniform organs and dorsal corrugations, which are specific for L3 vein (reviewed by Campuzano and Modolell, 1992; Milan et al., 1997). The col mutant phenotype therefore does not correspond to apposition of L3 and L4 veins, as first proposed (Vervoort et al., 1999; Mohler et al., 2000) but to loss of L4 vein and repositioning of L3 vein.

We have previously shown that BS, an intervein cell marker, is not expressed in the presumptive L3-L4 intervein in col¹ wing discs of third instar larvae (Vervoort et al., 1999). We asked whether this central area devoid of BS staining corresponded to presumptive L3 or L4 vein, or was composite and examined BS expression in 24 hours APF (after puparium formation) pupae; BS expression enables each longitudinal wild type provein to be visualised as a three- to four-cell wide stripe devoid of staining (Fig. 2A). In col¹ mutant wings, instead of L3 and L4 proveins, there was a single central stripe of cells, wider than the wild type L3, which did not express BS (Fig. 2B). A truncated stripe of BS-negative cells located near the presumptive hinge (arrow in Fig. 2B) correlated with the residual L4 vein observed in a proximal position in col¹ adult wings (arrow in Fig. 1C). This indicated that L4 vein was lacking and suggested that the central vein in col¹ mutant wings was a widened L3 vein. In order to confirm this, wings of wild-type and col¹ pupae at 28-30 hours APF were stained for Ci, which is expressed only by A cells (Blair, 1992) (Fig. 2C,D). At that stage, resolution of provein into vein has occurred and the col¹ central vein is morphologically distinguishable (arrowhead in Fig. 2D). All cells of this vein expressed Ci, indicating that it was entirely located in the anterior compartment, confirming its L3-like identity (L3m – m for mutant). However, although the posterior limit of Ci expression bisected the wild type L3-L4 intervein (Fig. 2C), it coincided with the posterior limit of col¹ L3m vein. This shows that the position of L3m vein is shifted towards the AP boundary, correlating with the increased size of the L2-L3m intervein of col¹ adult wings (see above). Taken together, morphological observations and size measurements of adult wings and immunostaining of pupal discs revealed three types of defects in col¹ mutant wings: a reduced number of cells, the lack of L4 vein and a posteriorwards shift in the position of a wider L3 vein.

The observation that col¹/kn¹ wings displayed both an intervein L2-L3 of normal size and a reduced L3-L4 intervein suggested that the loss of the L3-L4 intervein and the displacement of L3 vein observed in col¹ mutant were not necessarily linked. The specific loss or reduction of the L3-L4 intervein territory in col adult wings could be due to local programmed cell death and/or reduced cell proliferation. To distinguish between these possibilities, we stained wild-type and col¹ pupal wing discs with a probes for reaper (rp), a marker for apoptotic cells (Abrams et al., 1993), and an antibody against a phosphorylated form of histone H3 (H3P), which reveals mitosis. No difference in rp expression was detected between wild-type and col¹ mutant wings, which excluded a major contribution from apoptosis and suggested instead defective proliferation. In wild-type pupal wings, a wave of mitosis takes place in each intervein primordium between 15 and 21 hours APF (Schubiger and Palka, 1987). H3P staining of wild-type pupal wings 18-20 hours APF allowed the visualisation the proliferation of intervein cells along the proveins (Fig. 3A). Staining was specifically absent from the central region of either col¹ or col¹/kn¹ mutant wings (Fig. 3B and data not shown), correlating with changes from intervein to provein fate. However, the reduced size of the medial region of the wing already detectable at this early pupal stage probably reflects an earlier proliferation defect. Double staining with BS and propidium iodide (PI; not shown), which labels all nuclei, enabled us to count the number of rows of cells that separate the L2-L3 and L4-L5 intervein primordia in wild-type and col¹ larval discs. Ten independent measurements gave average numbers of 10.5 and 7.5 rows, respectively, indicating a significant reduction in col¹ mutants (not shown). We then looked at the expression of extramacrochaetae (emc), which encodes a helix-loop-helix (HLH) protein lacking a basic motif, and Notch (N), as both genes have previously been shown to be involved in the control of cell proliferation in the wing (De Celis and Garcia-Bellido, 1994; De Celis et al., 1995; Ellis et al., 1990; Garcia-Bellido et al., 1976; Garrell and Modolell, 1990). In third instar larvae, emc is expressed at a low level throughout the wing disc and at a higher level in two stripes of cells corresponding to the prospective A margin and the AP organiser (Baonza and Garcia-Bellido, 1999; de Celis et al., 1995) (Fig. 3C). Unmodified at the A margin, emc expression was completely lost from the AP organiser cells in either col¹ or col¹/kn¹ mutant discs (Fig. 3D and data not shown), showing that Col is required for emc transcription in the L3-L4 intervein primordium. Levels of N protein are high in intervein regions and low in presumptive vein territories in late third instar (De Celis et al., 1997). In col¹ mutants, N is downregulated in the L3m provein domain (data not shown). col requirement for emc and N upregulation in the AP organiser cells is consistent with the reduced cell number in the central region of col¹ mutant discs. This could not, however, account for the reduced number of cells in P interveins also observed in col mutant wings.

That Hh short-range patterning activity in the wing (Biehs et al., 1998; Mullor et al., 1997; Strigini and Cohen, 1997) is mediated by col was first revealed by hh overexpression experiments in which the anterior displacement of L3 vein observed in these conditions could be reverted by reducing the col dose (Vervoort et al., 1999). This was confirmed by the posterior shift of L3 vein in col¹ mutant wings (Fig. 2C,D). However, the molecular mechanisms involved remained unknown. We first examined the position of vein primordia in col¹ mutant discs, using rho in situ hybridisation, which labels the vein primordia (Sturtevant et al., 1993). In wild-type third instar wing discs, single rows of rho-positive cells marked the positions of L3 and L4 veins on either side of the AP boundary (Fig. 4A). In col¹ mutant discs, we observed two types of changes: first, whereas rho was only expressed in a few cells in the dorsal presumptive L4 vein, it was expressed in two to four rows of cells, rather than a single one, in the presumptive L3 vein (Fig. 4B). Second, the distance separating the L3 and residual L4 veins was reduced (Fig. 4B), suggesting a posterior shift, in addition to widening, of the presumptive L3m vein, compared with L3. We next directly compared rho and col expression, as col transcription, which is not modified in col mutant discs, marks the cells receiving high doses of Hh, thus enabling the AP boundary to be positioned. In wild-type third instar imaginal discs, rho-expressing cells flanked the col expression domain on both sides, with one to two rows of intercalary cells expressing neither gene (Fig. 4C). In col¹ mutant discs, the L3m rho and col expression domains partially overlapped, confirming that the position of L3m vein has been shifted posteriorwards by several cell diameters relative to wild type (Fig. 4C,D), as previously deduced from Ci labelling of 24 hours APF pupae (Fig. 2D). However, rho labelling did not abut the AP boundary, indicating that rho transcription remained repressed in the anterior cells, which express En (Blair, 1992) (Fig. 4I). The changes in position and number of rho-expressing cells in L3m and the lack of rho expression in L4 vein show that the abnormal vein pattern of col¹ adult wings reflects patterning defects occurring in third instar larvae (Fig. 4I).

The displacement of L3m rho-expressing cells in col¹ discs confirmed that Col mediates Hh activity in positioning the L3 vein. However, specification of L3 vein per se does not depend on col, but on activity of the homeobox-containing genes araucan (ara) and caupolican (cau) from the iroquois complex (iro-C) (Gomez-Skarmeta et al., 1996; Gomez-Skarmeta and Modolell, 1996). In order to investigate how the processes of vein specification and vein positioning were connected, we examined the expression domain of ara and col in wild type and col¹ third instar imaginal discs. In wild-type discs, ara and col expression partly overlapped, with ara expression extending anteriorly by about three rows of cells into the region corresponding to presumptive L3 vein (Fig. 4I). In col¹ mutant discs, the stripe of ara expression was both weaker and narrower. The posterior border of ara expression which is situated within the col expression domain and defined by En repression (Gomez-Skarmeta and Modolell, 1996) was not modified, but its anterior border was shifted by two to three rows of cells closer to the AP boundary (Fig. 4F,I). It has previously been reported that positioning of the anterior border of iro-C expression involved transcriptional repression by sal/salr, which are themselves targets of Dpp signalling (De Celis and Barrio, 2000). This regulation by sal/salr raised the possibility that the posterior shift of L3m vein position in col¹ wing discs could be due to a modified range of Dpp signalling. In that respect, the recent report that Dpp signalling was attenuated in the AP organiser because of downregulation of expression of Tkv (Funakoshi et al., 2001; Tanimoto et al., 2000) was particularly intriguing. Dpp signalling can be monitored in situ, using antibodies recognising the phosphorylated form of the Mothers against Dpp (p-Mad) protein (Persson et al., 1998). In wild-type discs, levels of p-Mad are low in the AP organiser and high in both A and P flanking cells (Tanimoto et al., 2000) (Fig. 4G). In col¹ mutant discs, a uniform staining of p-Mad was observed in the centre of the disc (Fig. 4H), indicating that downregulation of Dpp signalling in the AP organiser requires col activity.

Differentiation of posterior L4 vein specifically requires the activity of Vein (Garcia-Bellido et al., 1994). Genetic mosaic experiments indicated that this activity is provided by anterior cells close to the AP boundary where the vn transcripts accumulated in third instar larval discs (Simcox et al., 1996) (Fig. 5C). The absence of L4 vein in col¹ wings suggested that col and vn might genetically interact. To test for this possibility, we first looked for dominant genetic interactions between the two genes. Flies either homozygous for a weak col/kn allele, kn¹, or heterozygous for a strong allele of vn, vn^M2, formed a normal L4 vein. However, in kn¹/kn¹; vn^M2/+ flies, L4 vein was missing, except for its proximal- and distal-most parts, showing that vn and col cooperate to promote L4 vein formation (Fig. 5A,B). We then looked at vn expression in col¹ mutant discs. In the absence of col activity, a strong downregulation of vn transcription was observed in the AP organiser, except in the cells located near the presumptive hinge. vn transcription outside the wing pouch was not affected (Fig. 5C,D and data not shown). Together, these data indicate that col acts upstream of vn transcription in the AP organiser cells.

To ensure that col requirement for L4 vein formation was only through the upregulation of vn transcription in the AP organiser, we asked whether re-introducing vn expression would be sufficient to rescue of L4 vein in the absence of col. As expression of UAS-col under the control of a dpp-Gal4 driver in col mutant wings restores formation of L4 vein (data not shown), we expressed vn under control of the same driver. In contrast to col, vn expression did not restore formation of L4 vein, however. Instead, it led to the formation of a wider L3m vein (Fig. 5E,F). Accordingly, rho expression in larval discs was not restored in the L4 vein primordium, while strongly upregulated in the anterior L3m provein and over a distance of several cell diameters in the posterior wing margin (Fig. 5G,H). One possible reason for the failure of presumptive L4 vein cells to respond could be a limited range of diffusion of Vn as the P compartment is separated by three to four cell diameters from the source of Vn when expressed under control of the dpp-Gal4 driver, due to the transcriptional repression of dpp by anterior En (Alves et al., 1998; Strigini and Cohen, 1997). To rule out this possibility, we examined L4 vein formation in col¹ mutant clones located in the A compartment, at varying distances from the AP boundary. We observed a loss of vein L4 when col¹ mutant clones encompassed most, if not all, dorsal and ventral cells in the L3-L4 region. By contrast, L4 vein formed in cases where dorsal and ventral col¹ cells spanned three to four rows of cells along the AP boundary (Fig. 5I). These data indicated that Vn is able to induce L4 vein formation, even when expressed at a distance of three to four cell diameters. The inability of Vn to activate rho expression in the L4 vein primordium and restore L4 vein formation in col¹ mutants indicated that another signal regulated by col in the AP organiser is required, in addition to vn. We therefore analysed the possible role of Col-mediated downregulation of Dpp signalling (Fig. 4H,I).

In order to determine whether Col-dependent downregulation of Dpp signalling in the AP organiser was involved for L4 vein formation, we expressed a dominant-negative form of Tkv, TkvDN (UAS-Tkv1ΔGSK) (Haerry et al., 1998) in the AP organiser, using a ptc-Gal4 driver. The TkvDN protein has a dual effect when overexpressed: blocking Dpp signalling and trapping Dpp itself, thus limiting its diffusion (Haerry et al., 1998). Expressing TkvDN in the AP organiser resulted in smaller wings with a loss of L4 vein and anterior crossvein (Fig. 5J). The decrease in wing size has been attributed to the titration of Dpp molecules when high levels of TkvDN (or Tkv) are expressed, leading to reduced growth of the posterior wing compartment (Haerry et al., 1998) (see Discussion). More striking is the loss of the L4 vein, suggesting that this is the first vein affected when Dpp levels are reduced. That the loss of L4 vein is an early patterning defect was confirmed by rho in situ hybridisation on larval wing discs. In UAS-TkvDN/ptc-Gal4 discs, rho was not expressed in presumptive L4 vein (data not shown). Taken together, our data suggested that increased sequestering of Dpp in the AP organiser in either col mutant or UAS-TkvDN/ptc-Gal4 discs could modify the range of Dpp signalling and lead to both a reduced size of the posterior compartment and preferential loss of the L4 vein. Activation of Col expression by high doses of Hh thus appears to be an essential component of the interplay between Hh and Dpp signalling that specifies the provein fate of cells located on either side of the AP organiser.

The Drosophila wing primordium is subdivided into anterior (A) and posterior (P) compartments, with the AP boundary reflecting a true lineage restriction (Garcia-Bellido, 1975). How this discontinuity serves as a source of positional information, which patterns the imaginal disc and adult appendage, has been the subject of intensive investigation for over 25 years (Crick and Lawrence, 1975). The current model is a signalling-relay model where Hh secreted from the P cells diffuses into the A cells along the AP boundary and exerts a patterning function both independently and through the induction of dpp expression; Dpp in turn diffuses in both directions and acts at long-range to organise the pattern of more distal regions (Basler and Struhl, 1994; Chen and Struhl, 1996; Tabata and Kornberg, 1994; Mullor et al., 1997; Neumann and Cohen, 1997; Strigini and Cohen, 1997). We show here that activity of the Hh, Dpp and EGF signalling pathways in patterning of the Drosophila wing are coordinated by Col, a Hh target specific to the wing AP organiser (Fig. 6).

An unanticipated intricacy of the independent, versus Dpp-mediated, Hh patterning activity in the wing arose from comparison of the gradient of Dpp activity with dpp expression. It revealed that Dpp signalling is downregulated in the AP organiser and upregulated in both A and P flanking cells (Tanimoto et al., 2000). This downregulation of Dpp activity was shown to result from the localised transcriptional repression of tkv, which is itself due to activation of the transcriptional repressor Master of thick vein (Mtv) in response to Hh signalling (Funakoshi et al., 2001). We show here that Col activity is required for the downregulation of Dpp signalling in the AP organiser and link this regulation to Col requirement for positioning L3 vein and formation of L4 vein. While the observation that mtv transcription is downregulated in col mutant discs (data not shown) and tkv is upregulated in clones of col mutant cells (T. Tabata, personal communication) suggests that col may act upstream of mtv in the regulatory cascade, attenuating Dpp signalling in the AP organiser, the relative functions of these two genes remain to be examined in detail.

It has previously been proposed that Hh does directly control the position of L3 vein, although the molecular mechanisms of this control were not firmly established (Mullor et al., 1997; Stigini and Cohen, 1997). In both col (Nestoras et al., 1997) (this paper) and mtv mutant clones (Tanimoto et al., 2000), the position of L3 vein is shifted posteriorwards. That both col and mtv control the position of L3 vein suggested that this position is defined by Hh signalling through the modulation of Dpp signalling. It has previously been established that iro was required for rho activation in the L3 primordium and formation of L3 vein (Gomez-Skarmeta et al., 1996; Gomez-Skarmeta and Modolell, 1996). iro is activated by both Dpp and Hh signalling and its anterior border of expression is under control of sal/salr, a target of Dpp (De Celis and Barrio, 2000). We show that the patterns of col, iro and rho expression are intimately connected. We observed both an increased number of cells expressing rho and a posterior shift of the anterior border of iro expression in col¹ mutant discs. We interpret this posterior shift as reflecting a modified range of Dpp signalling relayed, at least in part, by sal/salr activity. The increased number of rho-expressing cells, for its part, indicated that Col is able to antagonise rho activation by iro in cells, which express both iro and Col. This correlates well with the wing phenotype – anteriorwards shift of the L3 vein, together with gaps in its distal region – which results from anterior extension of Col expression, in UAS-Col/dpp-Gal4 wing discs (Mohler et al., 2000) (M. C., unpublished). The distal gaps could reflect the complete absence of rho expression close to the DV border, because of the complete overlap between col and iro expression where iro expression is narrower. From col loss- and gain-of function experiments, we therefore conclude that the primordium of L3 vein corresponds to cells that express iro but not col (Fig. 4I). Col thus appears to play a dual role in defining the position and width of L3 vein: activating BS and repressing EGFR in the wing AP organiser cells, endows these cells with an intervein fate, while attenuating Dpp signalling indirectly positions the anterior limit of iro expression domain, and L3 vein competence anterior to the AP organiser (Fig. 6).

We have shown that Col regulates vn transcription in the AP organiser. This expression of vn is required for formation of L4 vein (Garcia-Bellido et al., 1994; Schnepp et al., 1996; Simcox et al., 1996). The loss of this vein in col¹ mutants could not, however, be rescued by expressing high level of Vn in the AP organiser, suggesting that a second signal dependent upon Col was also required. The specific loss of L4 vein was previously observed in conditions of reduced levels of Dpp signalling caused by ubiquitous expression of either Tkv or TkvDN (Haerry et al., 1998). Together with this, the col¹ wing phenotype and Col requirement for downregulation of Dpp signalling in the AP organiser suggested a role of Dpp signalling in formation of L4 vein. Indeed, expressing a dominant-negative form of Tkv (TkvDN) in the AP organiser resulted in L4 loss. At first sight, it may appear contradictory that either upregulation (in col mutants) or downregulation (by expressing TkvDN) of Dpp signalling in the AP organiser leads to the preferential loss of posterior L4 vein. In both cases, however, there is increased sequestering of Dpp in the AP organiser, which limits its range of diffusion and signalling in posterior cells (Haerry et al., 1998; Lecuit and Cohen, 1998). Therefore attenuation of Dpp signalling in the AP organiser and increased signalling in posterior flanking cells (Tanimoto et al., 2000) appears to be required in addition to Vn activity for formation of L4 vein. By modulating Dpp signalling and vn transcription in cells receiving high doses of Hh, Col thus links Hh short-range activity to both positioning of the anterior L3 vein and formation of the posterior L4 vein.

In both null and hypomorphic col mutants, the wing blades are 20% smaller than wild type, owing to a reduced number of cells, both in the L3-L4 intervein and the posterior compartment. Reduced proliferation in the posterior compartment is a cell non-autonomous effect that is probably linked to the decreased range of Dpp diffusion that results from the upregulation of Tkv in the AP organiser cells (Lecuit and Cohen, 1998; Tanimoto et al., 2000). By contrast, the reduced size or the lack of L3-L4 intervein in col¹/kn¹ and col¹ mutants, respectively, is an independent defect as Dpp signalling is up- rather than downregulated in this region. Defective cell proliferation in the central region of col wings in pupae could be linked to Col requirement for upregulation of emc and N, previously shown to be required for cell proliferation (Garcia-Bellido et al., 1976; De Celis and Garcia-Bellido, 1994; Baonza and Garcia-Bellido, 1999). Thus, the size reduction of col¹ mutant wings therefore reflects the cumulative effect of changes in Dpp signalling and decreased expression of emc and/or N in the AP organiser.

Col was initially characterised for its function in the segmentation of the embryonic head, acting upstream of Hh in this process (Crozatier et al., 1996; Crozatier et al., 1999). Its key role in mediating short-range Hh activity in the wing was therefore unanticipated. Col role in the wing illustrates the pivotal importance of tissue-specific mediators of Hh and Hh→Dpp signalling pathways in patterning different appendages. The concentration of Shh is the primary determinant of AP polarity in the vertebrate limb (Lewis et al., 2001; Yang et al., 1997). In this case, as in Drosophila imaginal discs, a signalling relay involving Bmps is implicated (Drossopoulou et al., 2000; Duprez et al., 1996). Col/Kn is the single Drosophila member of the family of COE transcription activators that includes products of the vertebrate ebf/olf-1/coe genes and C. elegans unc-3 (for a review, see Dubois and Vincent, 2001). Whether Col function in mediating tissue-specific interpretation of Hh signalling in Drosophila has some equivalent in vertebrates or corresponds to the co-option of Col in a highly derived patterning process remains to be investigated.

We are grateful to P. Blader, F. Roch, J. Smith and two anonymous reviewers for critical reading of the manuscript; to S. Cohen, A. Garcia-Bellido, K. Wharton and the Bloomington stock Center for fly strains; to R. A. Holmgren, M. Gilmam and T. Tabata for antibodies; and to J. de Celis, J. Modolell, F. Schweisguth and A. Simcox for plasmids. This work was supported by CNRS and Ministère de la Recherche (ACI Biologie du Développement). B. Glise was supported by a fellowship from the ‘Fondation pour la Recherche Médicale’.