The Drosophila Serum Response Factor gene is required for the formation of intervein tissue of the wing and is allelic to blistered

The Drosophila wing is derived from a single epithelial sheet, the wing pouch, which folds at the future wing margin into a dorsal compartment and a ventral compartment; these adhere to one another and form an epithelial bilayer (for reviews see Cohen, 1993; Fristrom and Fristrom, 1993). The wing is organized into two major cell types, vein and intervein. During formation of the wing at metamorphosis, intervein cells differentiate a specific cytoskeletal support, the trans-alar array, which allows the formation of tight connections between the dorsal and ventral cell layers (Tucker et al., 1986). After eclosion, intervein cells die leaving a double layer of transparent cuticle (Johnson and Milner, 1987). Vein cells are smaller than intervein cells, form a lumen after apposition of the dorsal and ventral layers, and remain alive after eclosion.

The vein-intervein network of the adult wing is organized in a stereotyped pattern (Waddington, 1940), which appears to be established by earlier positional information during the larval stages (Garcia-Bellido and Merriam, 1971; Zecca et al., 1995). Differentiation of veins and interveins occurs during metamorphosis (Fristrom et al., 1993). Vein-intervein patterning is a powerful model system for studying how a cell chooses between two possible fates, and many mutations displaying modifications in the vein pattern have been identified (Lindsley and Zimm, 1992). Most of these mutants have been classified according to their loss-of-function phenotypes (DiazBenjumea and Garcia-Bellido, 1990; Garcia-Bellido and de Celis, 1992): (1) mutations in ‘vein-promoting’ genes (e.g. veinlet or vein) lead to the disappearance of portions of veins;(2) mutations in ‘vein-suppressing’ genes (e.g. plexus or net) lead to the formation of ectopic veins; and (3) ‘thick veins’ mutants (e.g. Notch, Delta) result in enlarged veins. Some of these genes are locally expressed in veins or interveins. For instance, veinlet (which is allelic to rhomboid) is expressed specifically in the future vein territories in the third instar wing disc, and loss-of-function mutations lead to the disappearance of the distal portion of the longitudinal veins (Sturtevant et al., 1993). Genetically, veinlet interacts with the ventrolateral group genes (spitz, Star, Ras; Bier et al., 1990; Mayer and Nüsslein-Volhard, 1988) and is proposed to potentiate the EGF receptor pathway in the vein territories (Sturtevant et al., 1993).

In this study, we describe the expression and function of the Drosophila homolog of the vertebrate Serum Response Factor (DSRF) during the development of the wing imaginal disc. SRF proteins are members of the MADS box family of transcriptional regulators (Shore and Sharrocks, 1995). The mammalian counterpart (SRF) has been shown to be involved in induction of the c-fos promoter in response to serum growth factors (Treisman, 1994). We previously isolated the Drosophila homolog of the vertebrate SRF and found that the protein is expressed in a subset of tracheal cells and, during late larval stages, in the developing wing disc (Affolter et al., 1994). More recently, a P-element-induced loss-of-function mutation in Drosophila SRF (DSRF) was identified and called pruned¹ because of its embryonic and larval phenotype in which all terminal branches of the trachea are missing (Guillemin et al., 1996). We now report that DSRF protein accumulates in the third instar wing imaginal disc exclusively in nuclei of intervein cells and is strictly required for the formation of these highly specialized cells. In addition, we show that pruned¹ and other DSRF mutations are allelic to blistered, classical mutations which produce additional veins and blistered wings (Bridges and Morgan, 1919; Fristrom et al., 1994; Lindsley and Zimm, 1992). Furthermore, both DSRF and blistered null mutations cause a haploinsufficient ectopic vein phenotype. Our results demonstrate that DSRF plays a dual role during wing development; it is required for the restriction of veins to vein territories and for the terminal differentiation of intervein cells.

The pruned¹ allele was isolated in a screen for tracheal enhancer trap lines (Terminal-1, Samakovlis et al., 1996). It carries a P[lacZ, rosy⁺] element and corresponds to strain l(2)03267 of the Spradling collection (Karpen and Spradling, 1992). The insertion is located in the 5′ untranslated region of the DSRF transcription unit; pruned¹ behaves as a genetic null allele for the tracheal function (Guillemin et al., 1996). All blistered alleles were kindly provided by D. Fristrom (Fristrom et al., 1994 and unpublished results). Homozygous lethal excision alleles from a P-element insertion in slbo were generously provided by D. Eberl and L. M. Hall. The slbo transcription unit (Montell et al., 1992) maps approximately 8 kb upstream of the DSRF gene (Guillemin et al., 1996) and Southern blot experiments indicate that one of the deletions (DE1) removes at least the first exon of the DSRF gene (unpublished results). Other lethal excisions of the Pelement insertion in slbo, as well as the small deletions C606, C470 and D707 provided by E. C. Raff (Kimble et al., 1990), were used to analyze the contribution of the different genes deleted in the overlapping region of the Px² and Px⁴Plexate deficiencies towards the formation of the haploinsufficient wing phenotype.

Homozygous pruned¹ mitotic clones were generated essentially as described by Grieder et al. (1995). Adult wings were mounted as described by Sturtevant and Bier (1995) and photographed under bright-field optics at 40× magnification on a Zeiss Axiophot microscope. The veinlet, vein double mutant fly strain that was used is described by Diaz-Benjumea and Garcia-Bellido (1990).

Third instar imaginal discs were dissected as described by WagnerBernholz et al. (1991) and immediately fixed for 30 minutes on ice in phosphate-buffered saline (PBS), freshly adjusted to 2 mM EGTA, 0.5% NP40 and 3.7% formaldehyde. Discs were washed 3× in PBS and permeabilized for 30 minutes at room temperature in 0.05% Triton X-100, 0.05% NP40, 0.05% DOC and 2 mg/ml BSA in PBS. After washing 3× in PBS, non-specific binding was blocked in solution A (0.1% Triton X-100, 2 mg/ml BSA in PBS). Antibody detection was performed as described by Affolter et al. (1994), except that all the washing and incubations were performed in solution A.

For histochemical staining, third instar discs were fixed for 20 minutes at room temperature in 0.1 M Pipes (pH 6.9), 20 mM EGTA, 1 mM MgSO₄ and 0.75% glutharaldehyde. Adult wings were fixed for 20 minutes at room temperature on a rotating arm, with 1 volume of fixation solution and two volumes of heptane. β-galactosidase activity was detected according to the procedure of Bellen et al. (1989), requiring 2 hours of incubation for the imaginal discs and overnight incubation for the adult wings.

Template DNA for PCR amplification was prepared from homozygous blistered larvae or adult flies, as described by Wilson et al. (1989). To isolate DNA from homozygous lethal lines, blistered chromosomes were maintained over the T(2;3)L14,SM5-TM6B translocation (a gift of J. Botas) where the 2nd and 3rd chromosomes cosegregate, allowing the selection of non-tubby, homozygous blistered larvae. For amplification of coding sequences of the first exon, one oligonucleotide primed just upstream of the AUG initiator codon (5′GATTTCGGTTTCGGGGTTGGG-3′), the other in the first intron (5′-CCTTGGATCTCTACGGCTCC-3′). The second exon was amplified with oligonucleotides priming in the first intron (5′GTACTTACCACGATCCTCTGC-3′) and in the second intron (5′GTGACCTAGATGAGCGTGGC-3′). PCR fragments were ligated directly into a T-tailed vector (pGEM-T, Promega) and sequenced with the primers used for PCR amplification.

We analyzed the expression of DSRF during imaginal disc development with an anti-DSRF antibody (Affolter et al., 1994). DSRF protein was distributed in a specific patched pattern in the wing disc of third instar larvae (Fig. 1B). Comparison with the fate map of the wing imaginal disc (Bryant, 1975; Fig. 1A) indicated that the patched staining was restricted to the wing pouch and the hinge region. Within the developing wing pouch, DSRF protein appears to be absent from the future wing margin, and from four stripes of cells extending at a right angle from the wing margin towards the proximal wing regions. As the intervein tissue of the wing is separated into five areas by the wing veins, the expression domains of DSRF suggested that the gene is expressed exclusively in those cells that correspond to the future intervein tissue. To determine whether DSRF protein was indeed restricted to intervein territories, the expression pattern of DSRF was compared to the A101 marker. A101 contains a P[lacZ] insertion (Bellen et al., 1989) in the neuralized gene and expresses β-galactosidase in sensory organs along the anterior wing margin and vein L3; it was used previously to show that veinlet expression is restricted to the future vein tissue of the wing disc (Sturtevant et al., 1993). Double staining for DSRF and β-galactosidase confirmed that in the wing pouch the middle stripe of cells that does not express DSRF corresponds to the anterior wing margin, and that one of the perpendicular stripes corresponds to L3 (Fig. 1C). In the adult wing, intervein areas B and C flank vein L3 on each side (for intervein nomenclature, see Fig. 1F). Consistent with this arrangement, the future vein L3 is located between the second and third patches of DSRF expression (Fig. 1C).

A P[lacZ] insert in the DSRF gene called pruned¹ mimicks endogenous DSRF protein expression in the tracheal system of the embryo and larva (Guillemin et al., 1996). We found that the pattern of β-galactosidase activity in wing imaginal discs was also virtually identical to DSRF expression (compare Fig. 1D with B, and C). Although intervein cells die soon after emergence, it is possible to detect β-galactosidase activity in the wings of newly hatched flies (Hama et al., 1990). Staining of wings from flies carrying one copy of the P[lacZ] insertion in DSRF showed that β-galactosidase activity was detected exclusively in the intervein tissue of the adult wing (Fig. 1E). Thus, DSRF expression is confined to the intervein territory of the third instar wing imaginal disc and, upon eclosion, is detectable in all intervein cells.

The highly specific expression pattern of DSRF in the wing disc suggested that DSRF might be necessary for the development of intervein tissue. Homozygous pruned¹ mutants that carry a lethal P-element insertion in DSRF die during the larval period (Guillemin et al., 1996). To analyze the consequences of loss of expression of DSRF in the wing, we generated clones of homozygous pruned¹ cells by FLP-mediated recombination. In these experiments, the pruned⁺ chromosome carried a forked⁺ bristle marker. Thus, homozygous pruned¹ mutant clones could be identified by their forked⁻ bristle phenotype under high magnification. Vein cells were distinguished from intervein cells by their color (vein cells are darker) and by their size (vein cells are smaller) (Fristrom et al., 1994; GarciaBellido and de Celis, 1992).

When FLP-mediated recombination was induced during late third instar, we observed many small forked mutant clones (Fig. 2A). When recombination was induced during late second/early third instar, we observed fewer yet larger clones (Fig. 2B), and when induced even earlier (during embryogenesis or first instar), the resulting wings were dark and small, and looked like tubes (Fig. 2C).

Fig. 2D,E shows a close up of the dorsal and ventral layers, respectively, of the same wing. Two individual pruned¹ mutant clones (induced during late third instar in the intervein territory) are visible, one in the dorsal and the other in the ventral compartment. The mutant cells were smaller and darker than the surrounding intervein cells and resembled vein cells. In the case of the dorsal clone (right side, Fig. 2D), focusing on the underlying cells on the ventral side showed that the size of these cells was the same as that of the surrounding intervein cells (Fig. 2E). Similarly, the dorsal cells overlying the ventral clone (left side Fig. 2E) were unaffected (Fig. 2E). Because only the pruned¹ mutant cells were affected, we conclude that the DSRF gene behaves in a cellautonomous manner.

Mutant clones induced earlier in development, during early third instar or before, gave larger clones (as expected) and mutant cells were again smaller and darker than the surrounding intervein cells, thus displaying vein-like properties. In addition, these large mutant regions formed blisters that were generally limited to the mutant tissue (data not shown). Because blisters formed even when the opposite cell layer differentiated as intervein, it appears that the formation of tight connections between dorsal and ventral cells of the wing blade requires intervein identity in cells of both compartments.

Our results demonstrate that DSRF acts cell-autonomously and is essential for the formation of intervein cells; in the absence of DSRF, intervein cells fail to differentiate and acquire vein cell morphology.

blistered mutations lead to the formation of ectopic veins and blistered wings (Bridges and Morgan, 1919; Fristrom etal., 1994). The blistered locus is uncovered by Df(2R)Px² and Df(2R)Px⁴, which define the cytological region 60C5-6;60D1. Our previous mapping studies localized the DSRF transcription unit to the same chromosomal interval (Affolter et al., 1994). To analyze the relationship between these two genes, seven available blistered mutants (see Materials and Methods) were crossed to the pruned¹ DSRF loss-of-function mutant. The mutant chromosomes failed to complement each other (Table 1). When transheterozygous with pruned¹, the viable blistered alleles (bs10, bs11 and bs12) showed a wing phenotype that was stronger than the wing phenotype of the corresponding homozygous-viable blistered alleles (Fristrom et al., 1994; Fig. 3 and Table 1). The homozygouslethal bs13, bs14, 2R14 and 2R19 blistered alleles were lethal when transheterozygous with pruned¹ (Table 1). pruned¹ homozygotes die during second or third larval instar, and the same lethal period was observed for bs14 and 2R14 homozygotes (Fristrom et al., 1994), as well as for transheterozygous combinations with pruned¹ (Table 1). Homozygous 2R19 individuals died just at emergence, and the transheterozygotes with pruned¹ died during the same period.

Thus, we find that blistered mutations do not complement the pruned¹ DSRF loss-of-function mutation, suggesting that blistered alleles contain mutations in the DSRF gene.

To establish that the DSRF gene and blistered correspond to the same genetic locus, DSRF coding sequences from blistered alleles were cloned and sequenced. DSRF is a putative transcription factor of 450 amino acids containing a glutamine-rich region (residues 81 to 107) and an extended region of high sequence identity with mammalian SRF (residues 161 to 247), which includes the DNA binding MADS domain (residues 165 to 221) (Affolter et al., 1994). Oligonucleotides flanking the two first exons, which encode the N-terminal 278 residues, including the entire region of homology between DSRF and human SRF, were used to amplify blistered genomic DNA.

Point mutations were identified in three blistered alleles (see Fig. 4). In bs14, a nonsense mutation in codon 102 is predicted to give rise to a truncated protein lacking the MADS domain. The mutation in 2R19 leads to a stop codon just 3′ of the MADS-box coding sequence and eliminates an alpha-helix (αII) required for dimerization and interaction with accessory factors (Pellegrini et al., 1995). The viable bs11 mutation causes a glycine to glutamate substitution within the MADSbox (Gly204Glu). The glycine residue is conserved in all vertebrate SRF and yeast orthologs, and X-ray crystal structure determination of the human SRF MADS domain bound to DNA (Pellegrini et al., 1995) reveals that this glycine is centered in a tight turn joining an alpha-helix (αI) involved in DNA binding and a beta-strand (βI) required for dimerization. These two critical secondary structural elements might be prevented from folding into an optimal native tertiary structure due to steric hindrance by the Gly204Glu substitution within the tight turn.

There is a strong correlation between the severity of the different blistered phenotypes and the predicted molecular defects in the encoded DSRF proteins. The single amino acid change in the DSRF MADS-box in bs11 is sufficient to induce a homozygous wing phenotype, but causes neither haploinsufficiency (see below) nor larval lethality. Truncation of the DSRF protein (2R19, bs14) affects the function much more dramatically and causes a haploinsufficient wing phenotype and homozygous lethality. The most severe truncation, bs14, results in a tracheal phenotype similar to that of pruned¹ (see Guillemin et al., 1996).

In summary, blistered and DSRF mutations map in the same chromosomal location, lead to formation of ectopic veins when mutated (see also below), and do not complement each other. The DSRF loss-of-function mutation pruned¹ and the strong blistered alleles are homozygous lethal at the same developmental stages. These genetic results and the identification of point mutations in DSRF coding sequences from three blistered alleles demonstrate conclusively that the DSRF and blistered genes are the same.

Previous analysis have shown that a number of Plexate (Px) deficiencies delete the DSRF transcription unit (Affolter et al., 1994). Px deficiencies were isolated by virtue of their formation of ectopic veins in heterozygotes (Lindsley and Zimm, 1992). The finding that weak blistered mutations result in homozygous adult flies with ectopic veins prompted us to analyze the morphology of adult flies heterozygous for blistered/DSRF null mutations. Adult wings heterozygous for either a small deletion removing part of the DSRF gene (DE1; generously provided by D. Eberl and L. Hall; see Materials and Methods), the pruned¹ mutation, or several of the stronger blistered alleles, displayed ectopic veins similar to those seen in Plexate mutations (Fig. 5; see also figure legend). Although the haploinsufficient wing phenotypes due to DSRF loss-offunction mutations were somewhat weaker than the phenotypes generated by Plexate deficiencies, we noticed that both of them were more extreme in females and enhanced by cold (18°C) (data not shown; Lindsley and Zimm, 1992). To investigate whether the haploinsufficient wing phenotype observed in Plexate mutations was due primarily to the loss of DSRF function, we individually analyzed small deficiencies throughout the region deleted in both Px² and Px⁴ (see Materials and Methods). None of these smaller deficiencies showed a haploinsufficient wing phenotype when crossed over a wild-type chromosome (data not shown), or significantly modified the DSRF haploinsufficient phenotype in transheterozygous combination with pruned¹ (data not shown). This demonstrates that many or most aspects of the Plexate haploinsufficient wing phenotype are attributable to the removal of one copy of the DSRF locus.

As shown in the previous section, adult flies carrying only one functional copy of the blistered/DSRF gene develop additional vein tissue in the wing. Analysis of mitotic loss-of-function clones demonstrated that in the complete absence of blistered/DSRF function, intervein cells do not differentiate as such, but appear vein cell-like in morphology and in their inability to form tight connections with the apposed cells of the opposite compartment. Based on these phenotypes, one could postulate that the primary or even the sole function of blistered/DSRF during wing development is to suppress the activity of ‘vein-promoting’ genes in the intervein territory. To determine the relationship between these genes, we generated flies carrying mutations both in ‘vein-promoting’ genes (e.g. veinlet, vein double mutants; Diaz-Benjumea and GarciaBellido, 1990) and blistered/DSRF. In veinlet, vein mutant flies, wings lack veins and consist mostly of intervein cells (Fig. 6C,D). In contrast, vein-like cells cover the entire wing blade in the rare hatching adults homozygous for the blistered 2R19 allele (Fig. 6E,F). In flies mutant for veinlet, vein and blistered 2R19, the size of the wing blade cells, their color and their adhesive properties are vein-like and correspond to those seen in flies mutant for blistered/DSRF only (compare Fig. 6G,H with E,F; see also legend). Thus, loss-of-function mutations in blistered/DSRF are epistatic to loss-of-function mutations in ‘vein-promoting’ genes with respect to epithelial wing cell morphology. This result argues strongly against the hypothesis that the sole function of DSRF/blistered is to suppress the activity of ‘vein-promoting’ genes in intervein territories. Taken together, the observations that blistered/DSRF mutations cause a haploinsufficient ectopic wing venation phenotype, whereas the complete lack of blistered/DSRF function prevents the formation of intervein cells in different genetic contexts, suggest that DSRF plays a dual role during wing development. Two fully active copies of the blistered/DSRF gene are required to ensure that the formation of wing veins is limited to vein territories; in addition, DSRF protein is essential for the proper terminal differentiation of intervein cells.

Extensive transplantation studies using imaginal disc fragments have shown that much of the information needed to pattern the adult wing appendage has been elaborated by the end of the third larval instar (Bryant, 1975). The availability of molecular probes has recently allowed these future structural territories to be visualized. In particular, the veinlet gene is exclusively expressed in the wing vein primordia in the developing wing pouch and is essential for the formation of distal vein structures (Sturtevant et al., 1993). Here we have shown that DSRF protein is expressed in the intervein territory of third instar wing discs and that DSRF function is strictly required for the differentiation of all intervein cells. To our knowledge, blistered/DSRF is the earliest known marker for the entire intervein territory of the developing wing disc, and the only gene known to be required in all intervein cells for the selection of their fate.

We do not know how the intervein-specific expression pattern of DSRF is set up in the third instar imaginal disc. It is possible that the DSRF gene is turned on throughout the wing pouch and hinge and that the activity of ‘vein-promoting’ genes suppresses DSRF expression in the vein territories. However, DSRF expression is only slightly affected in homozygous veinlet, vein third instar discs (data not shown), although such discs develop into veinless wings (DiazBenjumea and Garcia-Bellido, 1990). Conversely, veinlet expression is not altered in third instar wing imaginal discs of bs11 mutant larvae, and expands only later during prepupal and pupal stages (Fristrom et al., 1994). These findings suggest that the expression patterns of both veinand intervein-specific genes are set up for the most part independently, possibly by differential interpretation of similar positional information generated by earlier acting genes. Preliminary experiments indeed indicate that DSRF expression in distinct intervein regions is controlled by separate cis-acting DNA elements (Ute Nussbaumer, J.G., J.M. and M.A., unpublished results). The analysis of gene expression patterns in mutant backgrounds (see Sturtevant and Bier, 1995), as well as the detailed dissection of the cis-acting transcriptional control regions of veinand intervein-specific genes, should provide more insight into the mechanisms that control these spatially restricted expression domains.

Our genetic and molecular analyses demonstrate that blistered mutations affect the structure and the function of the DSRF transcription unit. A detailed phenotypic and developmental analysis of blistered function in wing development has recently been reported (Fristrom et al., 1994). Using light and electron microscopy, blistered defects were first seen in the prepupal period as a failure of apposition of dorsal and ventral wing epithelia. Correspondingly, during definitive vein/intervein differentiation in the pupal period, the extent of dorso/ventral reapposition was found to be reduced, and intervein regions that failed to become apposed showed differentiated properties of vein cells (Fristrom et al., 1994). Our finding that blistered mutations affect the function of a MADS-box transcription factor suggests that differentiation of intervein cells requires transcriptional control, and that DSRF regulates certain target genes that suppress vein identity and promote intervein identity. A number of genes have been found to be expressed more strongly or exclusively in intervein cells during prepupal and/or pupal stages. These genes include inflated (if; codes for an integrin α chain) and myospheroid (mys; codes for an integrin β chain). Both integrin subunits are required for the tight apposition of dorsal and ventral surfaces, and mutations in mys and if result in adult wings with blisters (Brower and Jaffe, 1989; Wilcox et al., 1989; Zusman et al., 1990), a phenotype often associated with hypomorphic blistered mutations or homozygous mutant blistered clones. In addition, blistered and integrin mutations interact genetically (Fristrom et al., 1994; Wessendorf et al., 1992), and mys expression appears to be greatly reduced in bs11 (D. Fristrom, personal communication). These results suggest that DSRF might directly control integrin expression at the transcriptional level during prepupal and/or pupal stages. In contrast to mutations in DSRF, intervein development appears normal in mys and if, suggesting that DSRF must also regulate other genes. One characteristic feature of intervein cells is the formation of a ‘transalar cytoskeleton’ and characteristic basal junctions (Fristrom et al., 1993; Tucker et al., 1986). One or several components of these structures might be under the control of DSRF. Interestingly, potential DSRF target sequences have been identified in the promoter region of the actin 5C gene (Bond-Matthews and Davidson, 1988), which is expressed during wing development (Petersen et al., 1985). The actin 5C gene encodes a cytoplasmic form of actin and might be involved in the formation of the transalar cytoskeleton. The identification of direct target genes of DSRF will certainly provide additional insight into the control of cell differentiation in the wing epithelium.

The biochemical properties of human SRF and its DNA binding site SRE (Serum Response Element) have been analyzed extensively. The SRE has been shown to be a nuclear target of multiple signal transduction pathways and to be activated via various extracellular stimuli (Treisman, 1994). At the SRE, SRF is known to interact with a subset of Ets domain proteins, the activity of which is regulated via a phosphorylation cascade involving the ERK MAP-kinase. In addition, SRF can also be activated independently of its interaction with Ets proteins, as part of a novel Rho-mediated signaling pathway (Hill et al., 1995; Whitmarsh et al., 1995). The ERK-MAP kinase-mediated EGF receptor pathway is implicated in the formation of veins and in the control of cell size (DiazBenjumea and Hafen, 1994; Sturtevant et al., 1993). It is possible that the control of the size of intervein cells by the EGF pathway is mediated via DSRF. It is also possible that other, Rho-mediated signaling pathways positively induce intervein differentiation by triggering DSRF activity. DSRF function is required during a developmental period when the dorsal and ventral wing disc cells undergo a conversion from a folded single cell layer to a flat bilayered wing. During this process, each of four steps (apposition, adhesion, expansion and separation; see Fristrom et al., 1993) occurs twice, and cell-cell communication during these developmental events is well documented (Garcia-Bellido et al., 1994; Garcia-Bellido and de Celis, 1992). Consistent with the biochemical studies in vertebrate systems, the activity of DSRF, or of proteins that interact with it, could be modulated by these cell-cell interactions. In this context, it is worth mentioning that the blistered 2R19 allele, which encodes a protein containing the entire MADS-box but lacking C-terminal sequences of the MADS domain (i.e. αII), results in a haploinsufficiency and a complete failure to produce intervein tissue. Based on biochemical and structural evidence, such a protein might still be able to bind DNA targets but would fail to form ternary complexes with accessory factors (Mueller and Nordheim, 1991).

To test the hypothesis that DSRF function is modulated by distinct signal transduction pathways during wing morphogenesis in vivo, it will be crucial to identify target genes of DSRF and monitor their behavior in transgenic flies expressing dominant negative or constitutively activated forms of various signaling components. In parallel, the haploinsufficient phenotype of DSRF mutations should provide a simple and sensitive screening procedure for enhancers and suppressors of DSRF activity. The characterization of the genes corresponding to these modifiers should reveal more about the regulation of DSRF activity in vivo.

We are indebted to Dianne Fristrom for providing the blistered alleles and for stimulating discussions. We thank Dan Eberl, Linda Hall and Elisabeth Raff for the generous gift of the fly strains containing deletions in the Plexate area. We would like to thank Konrad Basler and Fernandez Diaz-Benjumea for the fly strains. We are grateful to Felix Loosli for initial in situ detection of DSRF mRNA in wing discs. Special thanks go to Erika Marquardt-Wenger for help in editing the manuscript. Ernst Hafen, Dan Eberl and Dianne Fristrom are acknowledged for comments on the manuscript. J. M. was funded by a HFSP long term fellowship and a short term fellowship from the Sandoz Foundation. M. A. is supported by the Swiss National Science Foundation and M. A. and J. G. by the Kantons Basel.