Expression and function of decapentaplegic and thick veins during the differentiation of the veins in the Drosophila wing

Cell differentiation involves the regulated expression of specific sets of transcription factors, such as the basic helix- loop-helix proteins of the MyoD family in muscle differentiation (Olson, 1990) or the proneural Achaete and Scute proteins during the establishment of neural precursors (Campuzano and Modolell, 1992; Ghysen et al., 1993). In many cases, the activity of specific signal transduction pathways directs the expression of regulatory proteins involved in the differentiation of cell types in precise patterns (Woods and Bryant, 1992; Greenwald and Rubin, 1992).

In the wing of Drosophila, two classes of structures, sensory organs and veins, differentiate in characteristic positions. The veins are formed by stripes of dorsal and ventral cells that are more compact and differentiate more pigmented cuticle than intervein cells. The veins correspond to clonal restrictions in the wing, suggesting that their positioning is related to the proliferation of the wing imaginal disc (Gonzalez-Gaitan et al., 1994; Milan et al., 1996). In wing discs of third instar larvae the expression of several genes is restricted to presumptive vein regions (Sturtevant et al., 1993; Heberlein et al., 1993), indicating that veins are already genetically specified. After eversion of the imaginal disc, dorsal and ventral wing surfaces adhere to each other and interactions between dorsal and ventral vein territories lead to the refinement of vein differentiation (Garcia-Bellido, 1977). At this stage the expression of different cell adhesion molecules becomes restricted to either vein or interveins (Fristrom et al., 1993; Murray et al. 1995).

The patterning and differentiation of the veins involves the activities of many genes, some of them components of different signal transduction pathways (Garcia-Bellido and de Celis, 1992). In the third instar wing disc the expression of veinlet (ve) is initiated in proximodistal stripes of cells that correspond to the presumptive veins (Sturtevant et al., 1993). The Veinlet protein appears to mediate an increase in the activity of the tyrosine kinase receptor, Torpedo (top; Sturtevant et al., 1993), and mutations in top itself, as well as in several genes belonging to the top pathway, result in the absence of veins (Diaz- Benjumea and Garcia-Bellido, 1990a; Diaz-Benjumea and Hafen, 1994). Conversely, mutations in which the activity of top or any of its downstream genes is increased, result in thicker and ectopic veins (Diaz-Benjumea and Garcia-Bellido, 1990a; Brunner et al., 1994). The formation of the veins also depends on the activity of the Notch pathway (Artavanis-Tsakonas et al., 1995), which restricts vein differentiation within broader ‘vein competent’ regions (de Celis and Garcia-Bellido, 1994b).

In addition to Notch and torpedo, the formation of veins requires the function of the gene decapentaplegic (dpp), as indicated by the phenotypes of particular dpp alleles (Segal and Gelbart, 1985). dpp is the best studied member of the Transforming growth factor β (TGFβ) superfamily in Drosophila (St. Johnston et al., 1990) and its function is required in processes as diverse as dorsoventral patterning in the embryo (Ferguson and Anderson, 1992; Wharton, et al., 1993), mesoderm specification (Staehling-Hampton and Hoffmann, 1994; Frasch, 1995), gut development (Panganiban et al., 1990) and growth and patterning of all imaginal discs (Posakony et al., 1991; Held et al., 1994). Recently, Dpp receptors that mediate most of the embryonic functions of Dpp have been isolated. They are encoded by the genes thick veins (tkv), saxophone (sax) and punt (pnt) (Xie et al., 1994; Brummel et al., 1994; Penton et al., 1994; Letsou et al., 1995; Ruberte et al., 1995), and correspond to transmembrane proteins with serine-threonine kinase activity that belong to the previously identified TGFβ type I (Tkv and Sax) and type II (Punt) receptors (Massague et al., 1994; Wrana et al., 1994). The genetic structure of dpp is complex, with large cis-regulatory regions that direct dpp expression in particular tissues (St. Johnston et al., 1990). Two classes of dpp alleles specifically affect the development of the imaginal discs; dpp^discalleles (dpp^d) affect the characteristic expression of dpp in anterior cells along the anteroposterior compartment boundary, and result in a failure in disc growth (Masucci et al., 1990; Blackman et al., 1991). Another class of alleles, short vein (dpp^s) affects the differentiation of the veins, causing incomplete veins which fail to reach the wing margin (Segal and Gelbart, 1985). Paradoxically, mutations that reduce the activity of the Dpp receptor tkv, result in opposite phenotypes, with the differentiation of ectopic vein tissue and thicker veins (Terracol and Lengyel, 1994).

In this paper the function of dpp and tkv during vein differentiation has been analysed, with particular emphasis on the relationships between them and between dpp/tkv and other signal transduction pathways affecting the same process. It is shown that dpp is expressed in the developing veins during pupal development, where it is required for the differentiation of the veins. dpp and tkv mutations interact with Notch and ve alleles, and alter the vein restricted expression of ve and the Notch ligand Delta. Finally, dpp is able to trigger the vein differentiation program when it is inappropriately expressed in the pupal wing, leading to ectopic expression of ve and Delta, and to the formation of ectopic veins. These results suggest that the differentiation of the veins is a progressive process that involves sequential and coordinate activities of Torpedo, Notch and Dpp signal transduction pathways.

The following alleles have been used. At the dpp locus the short vein alleles dpp^s22, dpp^s4and dpp^s8(Segal and Gelbart, 1985), at the tkv locus the hypomorphic, viable allele tkv¹and two lethal alleles, one null (tkv^IIB), the other antimorphic (tkv^IO) (Terracol and Lengyel, 1994; Nellen et al., 1994; Penton et al., 1994); at the Notch locus the loss-of-function allele fa^ndand the gain-of-function allele Ax²⁸; at the veinlet locus (also named rhomboid after the embryonic phenotype of its lethal alleles; Bier et al., 1990) the viable allele ve¹(Sturtevant et al., 1993) and the viable allele of vein (vn) vn¹(Simcox et al., 1996). Transgenic lines used were as follows. 3A (activated ras-2 under the regulation of the actin promoter; Bishop and Corces, 1988), rho30 (ve under hs promoter; Sturtevant et al., 1993), UAS-dpp (Staehling-Hampton and Hoffmann, 1994), UAS-tkv^* (a constitutively active form of Tkv; Lecuit et al., 1996), UAS-IMP (Sweeney et al., 1995), UAS-E(spl)mβ (de Celis et al., 1996a) and the GAL4 lines GAL-580 and GAL4-1348. Clones of Dpp- and Tkv^*- expressing cells were generated by FLP- mediated excision of a lacZ flip-out casette in flies of genotype FLP1.22/+; Ubx>lacZ>dpp/+ and FLP1.22/+; Ubx>lacZ>tkv^*/+ as described by Lecuit et al. (1996). Two reporter lines were also used: a construct in which an 8.5 kb EcoRI fragment of the short vein regulatory region (St Johnston et al. 1990) is fused to β-gal (shvLZ, a gift from J. Botas), and a lacZ insertion in the Notch locus (N^MLz) that reproduces the expression pattern of Notch (data not shown). The phenotypic analysis of genetic combinations was carried out in flies raised at 25°C, with the exception of those involving the UAS-dpp which were done at 18°C. Wings were mounted in lactic acid/ethanol (1:1) and photographed using a Zeiss axiophot microscope.

dpp^s22 clones were generated by X-ray-induced mitotic recombination. Larvae were irradiated (dose 1000 R; 300 R/min, 100 kv, 15 mA, 2 mm Al filter) 48-72 hours after egg laying (AEL). Clones were scored in males of genotype f^36a; dpp^s22/M(2)Z Ins[f⁺]30B. Mitotic recombination proximal to the f⁺insertion produces homozygous dpp^s22 mutant cells labelled with the cell marker forked (f). 23 dorsal, 14 ventral and 6 dorsal and ventral clones were analysed.

Whole-mount in situ hybridisation with digoxigenin-labelled DNA probes to imaginal discs were performed as described previously for both imaginal discs (Cubas et al., 1991) and pupal wings (Sturtevant et al., 1993). In situ hybridisation with digoxigenin RNA labelled probes was carried out using the same protocols, but the hybridisation step and washes were at 55°C. The following DNA probes were used: 2 kb EcoRI fragment from a dpp cDNA clone (St. Johnston et al., 1990), 1.4 kb EcoRI/ClaI from a tkv cDNA clone (Nellen et al., 1994), 3 kb EcoRI from a Dl cDNA clone (Vassin et al., 1987), 0.7 kb HindIII/NotI from c-mβ-14a (Delidakis and Artavanis-Tsakonas, 1992), 0.8 kb BglII/KpnI from a Notch cDNA clone (Kidd et al., 1986). dpp and ve RNA probes were synthesised from appropiate cDNA clones.

Immunocytochemistry with rabbit anti-β-galactosidase (Cappel), mouse monoclonal anti-Notch (Fehon et al., 1991) mouse anti-IMP (Sweeney et al., 1995) were carried out as described by Cubas et al. (1991). Secondary antibodies were from Jackson Immunological Laboratories (used at 1/250).

The effects of dpp and tkv mutations in vein formation prompted an analysis of their expression at different stages during wing development. In the wing imaginal disc, dpp and tkv expression patterns have no apparent relationship with presumptive vein territories. The expression of dpp is restricted to a narrow stripe of anterior cells abutting the anteroposterior compartment boundary (Fig. 1A; Sanicola et al., 1995), where dpp it is required for the growth of the disc (Posakony et al., 1991). The expression of tkv is wide-spread, but clear differences in levels of expression can be recognised. Minimal expression of tkv occurs in a central region of the disc that includes both the stripe of dpp-expressing cells and the dorsoventral boundary, whereas maximal expression is observed in more anterior and posterior domains of the presumptive wing blade, pleuras and thorax (Fig. 1B; Brummel et al., 1994). dpp and tkv expression in wild-type discs does not change during the first 8 hours after puparium formation (not shown), but subsequently becomes related to the differentiating veins (Fig. 1C and data not shown). dpp transcription appears in broad stripes of dorsal and ventral cells along the presumptive veins 12-16 hours after puparium formation (Fig. 1C), and latter becomes restricted to the developing veins (Fig. 1D,E,I-K). At approximately the same time tkv expression is modulated with respect to the veins, with minimal expression within the vein, and higher expression at both the anterior and posterior boundaries of every vein (Fig. 1F-H). The expression of dpp and tkv occurs in both the dorsal and ventral vein components. The relationship between the developing veins and expression of dpp and tkv suggests that they are involved in vein formation at the pupal stage. In addition, the coincidence between low levels of tkv expression and the regions where dpp is expressed in the wing disc and in the pupal veins (Fig. 1G,H) suggests that the expression of tkv is reduced in cells that produce or are exposed to Dpp. A similar relationship between dpp expression and lower levels of tkv transcription is also manifest in leg and eye-antenna discs (Brummel et al., 1994 and data not shown).

Viable dpp^s and tkv alleles cause loss and gain of vein tissue, respectively (Fig. 2A,D; Segal and Gelbart, 1985; Terracol and Lengyel, 1994). The expression of dpp and tkv in the third instar wing disc is not modified by dpp^sand tkv mutations that result in strong modifications of the vein pattern (data not shown), suggesting that the effect of these mutations on vein differentiation occur after the veins have been specified in the imaginal disc. In dpp^s combinations (dpp^s4/dpp^s8) there is a good correlation between the veins that are absent in the adult wing and the disappearance of dpp expression in the corresponding presumptive vein regions in the pupal wing (Fig. 2A,B). These observations are compatible with dpp^s mutations affecting cis-regulatory regions (St. Johnston et al., 1990) required to activate dpp in presumptive vein cells in the pupal wing. In addition, the absence of dpp always occurs symmetrically in both the dorsal and ventral components of the affected veins. The expression of tkv in dpp^s pupal wings remains at high levels in the regions of the developing veins where dpp is absent (Fig. 2C), indicating a causal relationship between the presence of Dpp and the reduction of tkv transcription in the developing veins. The expression of tkv in tkv viable heteroallelic combinations (tkv¹/tkv^IIB) is only modified in regions where ectopic veins differentiate (Fig. 2D,F). Interestingly, dpp is ectopically expressed in pupal wings of tkv mutants, mainly in the sites where ectopic vein tissue differentiates in tkv wings (Fig. 2E).

The effects of tkv mutations on dpp expression suggest a feedback mechanism in which Tkv represses dpp expression. This observation helps to explain the paradox that mutations in dpp eliminate veins while mutations in its receptor cause ectopic vein formation. Thus, the ectopic veins that differentiate in tkv mutants could be a consequence of ectopic Dpp acting on Tkv functional products present in cells with reduced Tkv activity. In addition, it appears that Tkv is also necessary to activate vein differentiation. Thus double mutant combinations between dpp^sand tkv result in strong dpp^sphenotypes in which the differentiation of tkv ectopic veins is suppressed, and, in addition, the loss of veins typical of dpp^sis strongly enhanced (Fig. 3B,E,F), reproducing the pattern of vein loss observed in stronger dpp^s heteroallelic combinations (Fig. 3C). This model is compatible with the observation that tkv heteroallelic combinations that result in further reductions in the levels of Tkv functional products (tkv¹and the antimorphic tkv allele tkv^IO) cause the elimination of veins (Fig. 3D; Terracol and Lengyel, 1994). The interpretation that Tkv has a dual function during vein differentiation, repressing dpp in intervein cells neighbouring the vein, and activating vein differentiation in vein cells, is reinforced by the behaviour of tkv mutant clones: they prevent vein differentiation autonomously, but induce vein differentiation in adjacent wild-type cells (Burke and Basler, 1996), consistently with the postulated repression of dpp by Tkv.

To confirm that the effects of dpp in vein differentiation are due to the localised expression of dpp in pupal veins and not to its earlier expression in anterior cells in the disc, a mosaic analysis of a strong dpp^sallele (dpp^s22) was carried out. It is expected that posterior as well as anterior dpp^s22clones would affect vein formation. Clones of Minute⁺ cells homozygous for the dpp^s22allele are of normal size and do not modify the overall patterning of the wing (Fig. 4A,B). When clones are restricted to only one wing surface (dorsal or ventral) both anterior and posterior clones have minimal effects on vein differentiation (Fig. 4D-F,H). However, anterior and posterior dpp^s22clones induced before the segregation of dorsal and ventral compartments consistently remove the distal regions of LII and cause gaps in LIII in anterior dorsoventral clones, and remove most of LIV and LV in posterior dorsoventral clones (Fig. 4A-C,G). The comparison of the dpp^s2 phenotype in dorsal, ventral and dorsoventral clones indicates that the presence of wild-type cells in either the dorsal or ventral wing surface is enough to implement vein differentiation in both surfaces, suggesting that dpp expression in one surface is sufficient to provide dpp function to the other surface.

The expression pattern of dpp and the phenotype of dpp^salleles in viable combinations or in genetic mosaics indicates that the localised expression of dpp during pupal development mediates vein differentiation. Thus it is expected that ectopic dpp expression or Tkv activation during this period would result in the differentiation of ectopic veins. This was tested using the GAL4 system (Brand and Perrimon, 1993) to drive ectopic expression of dpp and an activated, ligand independent, Tkv receptor (Tkv^*, Lecuit et al., 1996) in the pupal wing. In GAL4- 580/+; UAS-dpp/+ flies (580-dpp) raised at 18°C during larval development and at 25°C during pupal development, there is ectopic expression of dpp in the pupal wing in two broad domains, one anterior and other posterior (Fig. 5C). As a consequence, the expression of tkv is strongly repressed in the same territories (Fig. 5D), and the resulting adult wing differentiates ectopic vein tissue in the places that correspond to the domain of ectopic dpp expression (Fig. 5B). As expected, the phenotype of 580-dpp flies is temperature-sensitive, being weaker when pupal development occurs at 17°C (Fig. 5A). When dpp is expressed in all intervein cells during pupal development (using the line GAL4-1348, Fig. 6A), most wing cells differentiate as vein (Fig. 6B). Ectopic expression of activated Tkv also leads to the formation of ectopic veins, both using GAL4-580 (data not shown) and GAL4-1348 (Fig. 6C,F). The formation of ectopic vein tissue can also be induced in mosaics of cells that express Dpp or Tkv^* (Fig. 6G-I). Although in these experiment clones are not labelled, there is a good correlation between the time of clone induction and the size of the ectopic veins (Fig. 6G-I).

The activity of early-acting vein promoting genes might be required to activate dpp and tkv expression in the developing veins in the pupa. The pathway mediated by the tyrosine kinase receptor Torpedo is involved in the early positioning of vein territories in the imaginal disc, and the localised expression of its putative ligand Vein and the transmembrane protein Veinlet are important components of this function (Sturtevant et al., 1993; Sturtevant and Bier, 1995; Simcox et al., 1996; Schnepp et al., 1996). Thus, a regulatory mutation that removes ve expression in the developing veins (ve¹; Sturtevant et al., 1993) causes the loss of the distal ends of all longitudinal veins (Fig. 7A; Diaz- Benjumea and Garcia-Bellido, 1990b), and the genetic combination ve¹vn¹eliminates all longitudinal veins (Fig. 7J; Diaz-Benjumea and Garcia- Bellido, 1990b), presumably because of a strong reduction in Top activation in vein territories. Conversely, ectopic expression of ve in the disc and pupal wing (as in rho³⁰ heterozygous flies) results in the differentiation of ectopic and thicker veins (Fig. 7D; Sturtevant et al., 1993). The resulting phenotypes of genetic combinations between ve, dpp and tkv alleles are presented in Fig. 7. These results can be summarised as follows: ve¹ suppresses the differentiation of the thicker veins of tkv alleles and results in fewer veins in combination with both tkv and dpp^s (Fig. 7B,C). The ectopic veins typical of rho30 are suppressed in combination with dpp^s, even in regions where dpp^sdoes not show any obvious phenotype (Fig. 7E); conversely, the differentiation of thicker veins in rho30/+ is strongly increased in combination with tkv (Fig. 7F). The ve¹vn¹phenotype is not modified in combinations with either dpp^sor tkv (data not shown), but vein tissue differentiates when dpp is expressed ectopically in ve¹vn¹; 580-dpp flies (Fig. 7J-L). In pupal wings of this genotype, ve expression is absent in the wing blade (data not shown), indicating that Dpp activity is able to trigger vein differentiation even in the absence of ve. The interactions between dpp^s and tkv mutations and a transgenic line in which an activated version of D-ras2 (ras2^*) is expressed ubiquitously under the control of the actin promoter were also analysed. ras2^* flies differentiate ectopic veins (Fig. 7G), and in ras2^* pupal wings ve is ectopically expressed in the proximity of the veins (data not shown). As with rho30, the phenotype of ras2^* flies is suppressed in combination with dpp^s (Fig. 7H) and increased in combinations with tkv mutations (Fig. 7I). To further investigate the epistatic relationships between Top activity and Dpp signalling, a dominant negative Draf1 (raf^DN, a gift from E. Martin) and Tkv^* were expressed in the same cells using the GAL4 system (Fig. 6C,E). These flies (GAL4-1348/+; UAS-tkv^*/UAS-raf^DN)have a weaker phenotype of ectopic vein differentiation than their control siblings (GAL4-1348/+; UAS- tkv^*/+), suggesting that Top and Tkv activities co-operate in promoting vein differentiation.

The adult vein patterns observed in genetic combinations between ve, dpp^sand tkv are correlated with similar changes in the expression patterns of ve, dpp and tkv in mutant pupal wings (Fig. 8). Thus, dpp and tkv are not expressed in ve¹vn¹wings (Fig. 8A,C), indicating that their expression in vein territories is activated by Top. In rho30/+ and ras2^* pupal wings, dpp is expressed ectopically at the sites where ectopic ve occurs (Fig. 8B and not shown), suggesting that ve is involved in the control of dpp expression in the developing veins. These transcriptional effects are reciprocal: ve expression is eliminated in the affected veins of dpp^s wings (Fig. 8D), and ectopically expressed in tkv pupal wings (Fig. 8E), and in 580-dpp and 1348-tkv^* pupal wings (Figs 5E, 8F), indicating that dpp activity is involved in the regulation of ve expression in the developing veins. The phenotypic interactions between top/ve and dpp genetic variants, and the requirements of ve and dpp functions for dpp and ve expression respectively, suggest that dpp and ve transcription are linked to each other: Top function would be required to activate dpp, and subsequently Dpp activity would maintain the expression of ve in the veins.

The function of the Notch pathway is required for the veins to differentiate their normal thickness; mutations that reduce the activity of Notch result in thicker veins, whereas mutations that increase Notch activity cause the loss of veins (de Celis and Garcia-Bellido, 1994a,b). Notch function is also required during pupal development, since temperature shifts experiments at this stage with N^tsalleles alter vein thickness (Shellenbarger and Mohler, 1978). The relationships between Notch and dpp/tkv were analysed using genetic combinations between mutations that modify the activities of these pathways (Fig. 9). In combinations between dpp^sand Notch loss-of-function alleles, the dpp^sloss of vein phenotype is epistatic, and the vein thickening typical of the Notch allele is eliminated (Fig. 9A,B). Conversely, Notch gain-of-function alleles that lead to truncation of veins (Abruptex, Ax) result in very pronounced vein loss in combination with both dpp^s and tkv mutations (Fig. 9D-F). The thickening of veins caused by tkv is suppressed in combinations with Ax mutations (Fig. 9F), and strongly enhanced in combination with Notch loss- of-function alleles (Fig. 9C). Finally the extent of extra-vein differentiation caused by ectopic expression of Tkv^* (Fig. 6C) is strongly reduced when the Enhancer of split gene mβ is expressed in the same cells during pupal development (Fig. 6D), suggesting that Dpp/Tkv^* and the Notch downstream gene E(spl)mβ have antagonistic effects on common target genes during vein differentiation, a good candidate being the gene ve (Sturtevant and Bier, 1995).

The relationship between Notch and Dpp activities in vein development is also manifest in the effects of dpp mutations on the transcription of the Notch ligand Delta (Dl) and the Notch target gene E(spl)mβ. In normal pupal wings the expression of Dl and E(spl)mβ occurs in vein territories (J. de Celis, S. Bray and A. Garcia-Bellido, unpublished data). In dpp^smutations, Dl and E(spl)mβ expression are absent in the affected vein regions (not shown), and both Dl and E(spl)mβ transcription are activated when Dpp or tkv^* are ectopically expressed in pupal wings (Figs 5F, 8G-I and data not shown). These results indicate that the expression of Dl, and consequently the maintenance of Notch signalling and E(spl)mβ expression, depends on Dpp function during pupal development.

The differentiation of the veins is a progressive process, in which vein territories are first specified in the imaginal disc, and then vein differentiation proceeds during the pupal stage through the combined activities of Top/Ve, Notch and Dpp signalling pathways. The function of dpp during the generation of the venation pattern is first required in the imaginal disc to establish the position of the veins, partly through the regulation of the transcription factors encoded by the genes spalt and spalt-related (de Celis et al., 1996b). Later, dpp expression is activated in the veins, where it is required for their differentiation. The restricted expression of dpp and its receptor Tkv to vein regions depends upon early acting genes (top/ve and Notch) that define the position of the veins in the disc. In addition, the presence of Dpp itself in the veins is required for the maintenance of at least ve and Dl expression in these places, since either reduced or ectopic Tkv activation is associated with similar changes in the expression of both ve and Dl. Thus it appears that the activity of early acting genes promotes the expression of late acting ones, and subsequently they become linked through feed-back mechanisms that keep their expression restricted to the developing veins (Fig. 10).

The expression of dpp in anterior cells abutting the anterior- posterior compartment boundary directs growth and patterning of the wing disc (Posakony et al., 1991). dpp mutations that reduce or eliminate this expression (dpp^dalleles) produce smaller wings in which the veins are absent or incorrectly positioned (Gelbart, 1989; Masucci et al., 1990). In addition, the differentiation of the veins is directly affected by dpp late during pupal development, when new expression of dpp appears in the developing veins. Strong dpp^sheteroallelic combinations that remove most of the longitudinal veins do not affect the size of the wing or the expression of dpp in the imaginal disc. However they eliminate the expression of dpp in the affected presumptive vein territories, indicating a separation between the functions of dpp in the growth of the disc and in the differentiation of the veins. Mosaic analysis of dpp^dand dpp^salleles confirm this separation, since posterior clones of dpp^s22affect the differentiation of the posterior veins, while posterior clones of a strong dpp disc allele (dpp^d5) differentiate wild-type veins (unpublished observations). dpp^salleles map within a 10 kb region 5′ to the dpp transcription unit, whereas the dpp^d alleles map within a 20 kb 3′ region, and therefore the separate requirements of dpp can be attributed to distinct mechanisms involved in regulating its expression in the imaginal disc and in the pupal wing (St. Johnston et al., 1990).

The activity of Dpp is mediated by receptor complexes, one of its elements being the Tkv protein (Affolter et al., 1994; Brummel et al., 1994), whose expression in the pupal wing is related to the differentiating veins. The effects on vein differentiation of complete removal of tkv (Burke and Basler, 1996), the extreme absence of veins seen in combinations between dpp^sand tkv mutations, and the effects of ectopic activation of Tkv, indicates that Dpp directs vein differentiation through activation of Tkv. The extra vein differentiation observed in viable tkv heteroallelic combinations correlates with ectopic dpp expression at the sites where ectopic vein tissue develops, and requires dpp, as illustrated by the suppression of this phenotype in combination with dpp^salleles. Thus it appears that excess vein differentiation in hypomorphic loss-of-function tkv alleles is due to derepression of dpp in cells close to the vein, the ectopic Dpp product acting on residual Tkv functional receptors being responsible for the differentiation of thicker veins. The repression of dpp transcription by Tkv could participate in restricting dpp expression to the veins, and implies that some mechanism must ensure that this repression is not operative in the vein cells that express dpp. It is possible that the observed down- regulation of tkv expression in vein cells participates in generating the levels of Tkv activation necessary to activate vein differentiation, but insufficient to repress dpp expression. In addition, other elements such as additional Tkv ligands, or other Dpp receptors could participate in determining the two different outcomes of Tkv activation: vein differentiation in cells that express dpp, and dpp repression in adjacent cells (Fig. 10).

The correlation between sites of dpp expression and down- regulation of tkv in the imaginal discs and in the developing veins is maintained when the distribution of dpp is altered. Thus when dpp is ectopically expressed or when its expression is reduced, the expression of tkv is reduced or increased respectively. It is possible that the downregulation of tkv is a direct consequence of Tkv activation, following Dpp binding and activation of downstream transcription factors that would regulate tkv expression. Interestingly, a similar repression of torpedo expression occurs following Top activation, both during embryonic development and during the differentiation of the veins (Sturtevant et al., 1994), suggesting that mechanisms which regulate the levels of receptor synthesis are linked with receptor activity, and could be significant in modulating the activities of signalling pathways.

The establishment of vein territories in the imaginal disc requires the coordinate and antagonistic activities of Top and Notch pathways. In both cases, expression of at least some of their elements, namely veinlet and Delta, is restricted to the veins from the late third instar disc and throughout pupal development. The expression of dpp and tkv in vein territories depends, directly or indirectly on Top activity, because their transcription is not activated when Top activity is reduced (ve vn pupal wings). Once Dpp is established in the veins, local activation of Tkv in these cells is required both for the maintenance of ve and Dl expression and for the veins to differentiate. In experimental conditions in which dpp is expressed ectopically, vein differentiation proceeds in the absence of ve activity, suggesting that dpp is able to implement vein differentiation even when the levels of top activity are reduced. In addition, the extra vein differentiation produced as a consequence of ectopic Tkv activated receptor is corrected by the simultaneous presence of either a dominant negative DRaf protein, or the vein suppressing protein E(spl)mß, suggesting that vein differentiation genes are activated more efficiently by the coordinate activities of both Tkv and Top (Fig. 10).

The differentiation of the veins is a complex process that starts in the imaginal disc and continues during pupal development. It appears that signals that establish the position of the veins in the imaginal disc (top/ve and Notch), and late acting elements of the vein differentiation cascade (dpp) interact with each other during pupal development, as indicated by the existence of genetic interactions between mutant alleles affecting the activities of these signalling pathways. This work shows that an important outcome of the activity of early acting genes is the localised activation of dpp and tkv expression to the developing veins. dpp participates in the regulation of its own expression, presumably in a mechanism that involves up- regulation of tkv transcription in the cells that separate the vein from the intervein. Tkv activation within the vein directs the expression of ve and Dl, offering a mechanism to link the activities of different signal transduction pathways in the differentiation of the veins.

I thank Michael Ashburner, in whose laboratory this work has been carried out, for continuous support; E. Martin-Blanco, J. Botas, S. Cohen, P. Martin, M. Baylies, K. Basler, M. Hoffmann, J. Roote and S. Gonzalez-Crespo for DNA clones and flies, and S. Bray for help in preparing the figures. Constructive criticism of the manuscript from S. Bray, D. Gubb, M. Ruiz-Gomez and S. Russell is also acknowledged. J. F.de C. is postdoctoral fellow of Spanish C. S. I. C.