TONDU (TDU), a novel human protein related to the product of vestigial (vg) gene of Drosophila melanogaster interacts with vertebrate TEF factors and substitutes for Vg function in wing formation

The growth and patterning of the Drosophila wing is dependent on the formation of compartments in the wing imaginal disc which are defined by the restricted expression of the selector genes engrailed and invected in the posterior compartment and apterous in the dorsal compartment (reviewed in Blair, 1997). Interaction between cells at the boundary of these compartments leads to the production of the signaling molecules Wingless (Wg) at the dorsoventral (DV) boundary and Decapentaplegic (Dpp) at the anteroposterior (AP) boundary and to the activation of the Notch (N) pathway at the DV boundary (Lecuit et al., 1996; Neumann and Cohen, 1996, 1997; Serrano and O’Farrell, 1997).

These proteins will regulate the expression of several genes amongst which vestigial (vg) and scalloped (sd) play a crucial role in wing morphogenesis. These two genes are expressed in the wing primordia with identical patterns and mutations of both genes are associated with a loss of wing tissue. vg expression is regulated by two separate enhancers: the boundary (BE) and the quadrant enhancer (QE) (Kim et al., 1996; Williams et al., 1994). The BE directs vg expression at the DV boundary whilst the QE drives vg expression in the rest of the wing pouch. These two enhancers are sequentially activated by the DV signaling pathway (N, wg) and the AP signaling pathway (dpp) and allow vg expression in the wing pouch with highest expression at the DV boundary (Kim et al., 1996, 1997).

It has been shown that ectopic vg but not sd expression, in the eye-antenna or leg imaginal discs induces wing tissue outgrowth, misexpression of wing-specific genes such as sd, drosophila serum responsive factor (dsrf), spalt (sal) and spalt related gene (salr), and its own expression (Kim et al., 1996; Halder et al., 1998; Paumard-Rigal et al., 1998; Simmonds et al., 1998). In addition, several observations demonstrate that Vg must interact directly with Sd to activate target gene expression and promote wing tissue proliferation: (i) in vitro protein interaction tests and two-hybrid experiments have shown that Vg and Sd form a dimer (Paumard-Rigal et al., 1998; Simmonds et al., 1998), (ii) trans-determination as well as gene induction associated with ectopic vg expression are suppressed in sd mutants (Paumard-Rigal et al., 1998; Simmonds et al., 1998), (iii) binding sites for Sd have been identified in the enhancers of several genes regulated by Vg and in the two vg enhancers. In S2 cells, induction of these enhancers by Vg requires the function of the Sd protein (Halder et al., 1998). It can be inferred from these studies that vg integrates positional information conferred by the AP and DV signaling pathways and activates gene expression by forming a complex with Sd. However, the role of Vg in this complex has not been formerly determined. Here we show that Vg is a transcriptional activator in yeast suggesting that the function of Vg, in the Sd/Vg complex, is to activate transcription.

The sd gene encodes a protein belonging to the TEA transcription factors family (Campbell et al., 1992; Jacquemin and Davidson, 1997). Strong or null alleles are lethal (Campbell et al., 1991). During embryonic development sd is expressed in the precursors of the larval nervous system (Campbell et al., 1992). During larval development, a specific pattern of sd expression is observed in the leg, eye-antenna and wing imaginal discs (Campbell et al., 1992) suggesting a role for sd in the development of the corresponding adult structures. In the wing imaginal disc, sd is mainly expressed in the region that will give rise to the wing blade and in the presumptive scutellum region (Campbell et al., 1992; Williams et al., 1993). This expression pattern correlates with the wing defects in sd mutants (Campbell et al., 1991).

Expression in the leg imaginal disc is not associated with any defects in these appendages at least in hypomorph sd alleles while in the antenna-eye imaginal disc, only a slightly roughening of eyes has been reported in sd mutants (Lindsley and Zimm, 1992).

In the wing imaginal disc, it has been shown that Sd interacts with Vg to activate the transcription of target genes (Halder et al., 1998). However, in other imaginal discs such as the leg or antenna-eye discs, where sd but not vg is expressed, there may be other Sd-interacting coactivators, assuming that the function of Sd in these structures is to activate the transcription of specific target genes.

To determine in which structures Sd is transcriptionally active without using ex vivo system or ectopic gene expression, which could modify cell fate, we produced transgenic flies harboring a GAL4-responsive lacZ reporter gene and expressing a modified Sd protein in which the TEA/ATTS domain has been replaced by the DNA-binding domain of the GAL4 protein. We show that, in third instar larvae, ubiquitous expression of this fusion protein activates the cognate reporter gene only in the wing pouch of the wing disc where vg is specifically expressed, but does not activate transcription in the other discs where sd is expressed. This result indicates that, at this time, sd is transcriptionally active only in Vg-expressing cells. Consequently, no functional homologues of Vg are present in other discs and the molecular function of sd in these discs, if any, is likely different from its role in wing morphogenesis. Surprisingly, we found that, in the wing disc, even if Vg and Sd are present, the activity of the Vg/Sd complex is reduced at the DV and AP boundaries in late third instar larvae. We propose that the ratio between Vg and Sd regulates the activity of Sd/Vg dimer during wing development.

It has been proposed that the mammalian TEF factors require transcriptional intermediary factors (TIFs) to activate transcription (Chaudary et al., 1994; Hwang et al., 1993; Xiao et al., 1991). The existence of such TIFs have been deduced from transfection analysis where activation of a cognate reporter is severely reduced upon TEF-1 overexpression. This dominant effect is also observed with a truncated form of TEF- 1 lacking the TEA/ATTS DNA-binding domain (Hwang et al., 1993). It has been shown that TEF-1 interacts with Vg in vitro and can substitute for Sd during Drosophila development (Deshpande et al., 1997; Simmonds et al., 1998). This functional conservation between Sd and TEF-1, and the fact that Vg is a likely TIF for Sd, suggests that a protein with homology to Vg may exist in mammalian cells. We have cloned a novel gene called TONDU (TDU) which possesses a short domain homologous to the region of the Vg protein required for interaction with Sd and TEF-1. We show that this novel human protein interacts with all the mammalian TEFs and with Sd both in yeast and in vitro and is able to functionally substitute for Vg in wing blade formation and induction of sd and vg expression in vivo in Drosophila. We conclude from this functional analysis in Drosophila that TDU could be a co- activator for the mammalian TEFs.

The vg^83b27 strain was provided by J. Bell (Williams et al., 1991) and the vg-GAL4 was provided by M. Hoffmann and S. Morimura. The sd^P(ry+,ETX4) strain is described in Anand et al. (1990) and Campbell et al. (1992), and the sd⁵⁸ strain provided by K. Irvine (Simmonds et al., 1998). The ptc-GAL4 and the UAS-lacZ strains were provided by the Bloomington Stock Center. The UAS-reaper strain is described in Van de Bor et al. (1999). The vgBE (for vg Boundary Enhancer) and vgQE (for vg Quadrant Enhancer) were provided by S. Carroll and are described in Williams et al. (1994) and Kim et al. (1996).

TONDU cDNA (GenBank Accession number: AF137387) corresponds to the I.M.A.G.E. consortium clone ID 347406 (Lennon et al., 1995) which contains the entire coding sequence of TDU.

The different plasmids and methodological procedure used in this study were provided by R. Brent (Gyuris et al., 1993; Russel et al., 1995).

For one-hybrid assays: the yeast EGY48 holding the lacZ reporter plasmid pSH18-34 was transformed with pEG202 derivatives containing the lexA fused to the vg sequence corresponding to the amino acids 7-453, 7-127, 127-273, 273-357, 357-453. The β- galactosidase activity was detected with ONPG assay after growth on SCglu-UH. The LEU2 assay was performed by testing the growth on ScGlu-UHL.

For two-hybrid assays: the same yeast strain was co-transformed with a pEG202 derivative and a derivative containing the B42 fusion. To test the interaction between TEF-1, TEF-3, TEF-4, TEF-5 and TDU, we have constructed four pEG202 derivatives containing hTEF- 1, hTEF-3, mTEF-4 and hTEF-5 sequences corresponding to the amino acids 136-426, 137-427, 146-445, 141-435, respectively, and pJG4-5 derivative containing TDU sequence corresponding to the amino acids 24-50. The β-galactosidase activity was detected with ONPG assay after growth on SCGal-UHW. The LEU2 assay was performed by testing the growth on ScGal-UHWL.

Bacteria harboring plasmid pGEX2T-GST-TONDU, pGEX2T-GST- TEF-1 or pGEX2T-SdΔTEA were induced with IPTG 1mM. The fusion protein was purified using glutathione-sepharose 4B beads (Pharmacia) as described by the manufacturer.

0.5 μg of the appropriate pXJ40-TEF vectors (described in Hwang et al., 1993) or of the appropriate pXJ40-TDU vectors were used in a TnT-coupled rabbit reticulocyte lysate system (Promega) according to the supplier’s recommendations in the presence of [³⁵S]methionine. pXJ40-TDU plasmid and the three pXJ40TDU derivatives (TDU- Δ24-50, TDU-Δ24-40, TDU-Δ41-50) were produced by PCR and verified by sequencing. All details of cloning are available upon request. Interaction test procedures were performed as described in Monnier et al. (1998).

The reporter construct pUAST-lacZ vector was obtained by inserting the lacZ gene of E. coli into the pUAST vector. The pCaSpeR-hsp- GAL4_db-TDU vector encodes for a chimeric protein formed with the amino acids 1-147 of GAL4 and the amino acids 1-256 of TDU, under the control of hsp promoter. Drosophila S2 cells were maintained in Schneider medium with 5% heat-inactivated fetal bovine serum. Transfections were performed by using DOTAP liposomal transfection reagent kit (Boehringer) according to the supplier’s recommendations. Following incubation for 24 hours at 25°C, cells were heat shocked at 37°C for 1 hour and incubated for an additional 24 hours incubation at 25°C. The β-galactosidase activity was evaluated with ONPG assay. All transfections were performed in duplicate in at least three independent experiments. Transfection efficiency ranged from 5 to 10% as assayed with pCaSpeR-hsp-GFP vector.

The pCaSpeR-hsp-GAL4_db-Sd vector encodes for a chimeric protein formed with the amino acids 1-147 of GAL4 and the amino acids 213- 440 of Sd, under the control of hsp promoter. The pUAST-TDU vector was obtained by insertion of the entire TDU cDNA in the pUAST vector.

Northern blot Human Fetal Multiple Tissue Northern (MTN^TM) BlotII was purchased from CLONTECH Laboratories inc. Hybridization was carried out according to the supplier’s recommendations with the entire cDNA of TDU as a probe.

X-gal staining was performed as described in Paumard-Rigal et al. (1998) and viewed using Nomarski optics. Immunostaining, using rabbit polyclonal anti-Vg (Williams et al., 1991), was performed according standard protocols.

Previous analysis have shown that Vg and Sd physically interact (Paumard-Rigal et al., 1998; Simmonds et al., 1998). Furthermore, expression of Sd and Vg in S2 cells synergistically activates the three enhancers of dSRF or vg genes (Halder et al., 1998). These results support a model in which the two proteins form a heterodimer that binds DNA through the TEA domain of Sd, but the role of Vg remains unclear. To assess the possibility that Vg is a transcriptional activator, we tested the ability of a fusion protein between Vg and the LexA DNA-binding domain (LexA-Vg) to activate the expression of the lacZ and LEU2 reporter genes in yeast. As shown in Fig. 1, significant activation was observed. To determine which region of the Vg protein is involved in transcriptional activation, we used four LexA-Vg derivatives. Significant activation was only observed with the N-terminal and C-terminal domains of the Vg protein (Fig. 1). No activation was observed with the domain of Vg that interacts with Sd. This result indicates that Vg contains two separate autonomous domains that are able to activate transcription in yeast.

In the wing imaginal disc, it has been shown that Sd interacts with Vg to activate the transcription of target genes (Halder et al., 1998). However, in other imaginal discs such as the leg or antenna-eye discs, where sd but not vg is expressed, there must be other Sd-interacting coactivators if Sd function is to activate the transcription of specific target genes. To assess this possibility, we developed a genetic approach to allow us to determine in which structures Sd is transcriptionally active. We produced transgenic flies expressing a modified Sd protein in which the TEA DNA-binding domain has been replaced by that of the GAL4 yeast protein, under the control of the hsp70 promoter, We used two GAL4-responsive reporter strains; UAS-lacZ and UAS-reaper. In this latter line, the expression of the cell death gene reaper is under the control of UAS sequences and will produce cell ablation in the cells where Sd is transcriptionally active.

In a first step towards determining where Sd is active, heat shock was applied at the mid third larval instar and larvae collected 24 hours later, which corresponds to late third instar. We found that, in hsp-GAL4_db-sd/UAS-lacZ larvae, lacZ expression was restricted to the wing pouch in the wing and halter discs (Fig. 2C). When the experiment was done in a vg mutant context (hsp-GAL4_db-sd; vg^BG/vg^BG; UAS-lacZ), X-gal staining in the wing and halter discs was strongly reduced or absent (Fig. 2I). No expression was detected in other imaginal discs (data not shown). With hsp-GAL4_db-sd/UAS-reaper flies, we observed, in pharate adults, a loss of wing tissue (see below for a more precise description), but no visible effect on the thorax, legs or eyes. This result indicates that, in third instar larvae, Sd activates transcription of reporter genes only in the wing discs where vg is specifically expressed. This suggests that, in other imaginal discs where sd is endogenously expressed, it is not transcriptionally active and that its function is different from that in wing morphogenesis. For example, Sd may behave as a transcriptional repressor in these discs. Furthermore, as we analyzed the ability of Sd to activate transcription in third instar larvae, we cannot exclude the possibility that Sd interacts with factors other than Vg and to activate transcription at the embryonic, early larval or pupal stages. However, the absence of clear phenotype associated with sd hypomorph mutations in cuticular structures such as eyes or legs indicates that sd is likely not required for the development of these structures, but this has to be more precisely determined by clonal analysis.

To determine in which cells the Sd/Vg dimer is active during wing imaginal disc development, we applied heat shock at different stages. When heat shock was applied at the beginning of the second larval instar, we observed a disruption of the wing margin, but an intact wing blade in the hsp-GAL4_db-sd/UAS- reaper flies (Fig. 2B). This shows that, at this time, the Sd/Vg complex is preferentially active and required at the presumptive wing margin. X-gal staining performed on hsp- GAL4_db-sd/UAS-lacZ discs supports this conclusion (Fig. 2A).

In late third instar wing discs, X-gal staining was restricted to the wing pouch, but interestingly, no or lower staining was observed at the DV and AP boundary (Fig. 2C). This indicates that although Vg and Sd are present, the complex formed is not active at the DV and AP boundaries. This was confirmed with the hsp-GAL4_db-sd/UAS-reaper flies, which show a necrosis of the wing blade, but an intact wing margin when heat shock was done at the beginning of the third instar (Fig. 2D).

With hsp-GAL4_db-sd/UAS-reaper flies heat pulsed at the end of third larval instar, we observed vein disorganization, with both ectopic or absent veins (Fig. 2G,H). In prepupal discs, X-gal staining was more intense in stripes, which likely correspond to the presumptive vein regions (Fig. 2F). This result suggests a role of the Sd/Vg complex in the determination of vein/intervein cell fate and is supported by experiment from Halder et al. (1998), which shows that the dSRF gene is a target gene of the Sd/Vg dimer. In addition, we observed that vg surexpression in wing disc induces disorganisation of vein patterning (data not shown).

Altogether, these results indicate that, at the beginning of the third larval instar, the Sd/Vg complex is active at the DV boundary, whilst its activity later decreases at the DV and AP boundary to be finally restricted to specific regions of the wing blade.

How can these results be explained in terms of a functional Sd/Vg complex? Several observations suggest that the Sd/Vg ratio is critical for the function of the Sd-Vg complex. First, it has been shown that overexpression of sd by means of the UAS/GAL4 system suppresses the wing outgrowth induced by ectopic vg expression (Halder et al., 1998). Second, their stoichiometry is crucial in cell culture for the activation of target genes (Halder et al., 1998; Simmonds et al., 1998). In fact, in our system, Vg can bind either to endogenous Sd or to the GAL4_db-Sd chimeric protein. If the lack of X-gal staining that we observed at the DV boundary reflects an imbalance between Sd and Vg, we would predict that reducing sd expression will increase the quantity of GAL4db-Sd/Vg dimer and therefore lead to X-gal staining at the DV boundary. To test this possibility, we analyzed lacZ reporter expression driven by GAL4_db-Sd in the sd⁵⁸ strain, which is a strong hypomorph sd mutant. In this context, we observed a staining at the DV boundary and in the presumptive hinge region (Fig. 2E -a) at the third instar. This experiment confirms that the stoichiometry between Sd and Vg is critical for the formation of an active complex. No staining was observed in the remaining of wing pouch suggesting that vg is not expressed in these cells in a sd⁵⁸ mutant context. To confirm this, we analyzed vgBE and vgQE activity and found that vgBE but not vgQE was activated in a sd⁵⁸ context (Fig. 2E-b,E-c). Although it has never been described before, we found that less Vg is produced at the DV and AP boundaries (inset in Fig. 5D) in late third instar larvae. As vg is itself regulated by the Vg/Sd heterodimer, this decrease in Vg expression could be due to the lack of activity of this complex at the DV boundary.

The above results show that Vg is a coactivator for Sd in Drosophila (this work and Halder et al., 1998; Simmonds et al., 1998). The observation that the mammalian TEF factors are able to physically and functionally interact with Vg led us search for mammalian proteins with homology with Vg (Deshpande et al., 1997; Simmonds et al., 1998). Sequence comparison analysis with dbEST cluster database identified a putative human protein Q99990 that shows sequence homology with Vg. This protein was deduced from EST sequences and corresponds to a unique gene included in PAC 196E23 that was found to localize on chromosome Xq26.1-27.2. Interestingly the similarity between the two proteins is restricted to the Vg domain that interacts with Sd or TEF-1 suggesting that this human protein could interact with different TEF proteins. By using I.M.A.G.E. consortium cloneID 347406 (Lennon et al., 1995), we cloned a 1147 bp cDNA comprising the complete open reading of this gene, designated TONDU (TDU) encoding a 258 amino acid protein (Fig. 3A). In northern blot experiments, this cDNA reveals a 1.35 kb transcript which is expressed in fetal lung and kidney, but not in brain and liver (Fig. 3B), indicating that the isolated cDNA is almost complete and that TDU is not ubiquitously expressed.

To investigate possible interactions between TDU and the different TEFs, we used the yeast two-hybrid system. As shown in Fig. 4A, co-expression of TDU-B42 with the different TEF-LexA chimeras (TEF-1, 3, 4, and 5) in yeast leads to the activation of the two reporter genes lacZ and LEU2 (Fig. 4A). Similar results were obtained when we used a fusion protein including the complete sequence of TDU (data not shown) or only the Vg homology region. These results indicate that the Vg domain involved in the interaction with Sd is conserved in TDU and mediates interaction with the mammalian TEFs.

To determine if TDU-TEF-1 complex formation that we observed in yeast results from a direct interaction, a GST-TDU fusion protein was expressed in E. coli and immobilised on glutathione sepharose beads. In vitro ³⁵S-labeled TEF-1 was produced in a coupled transcription/translation reaction system and incubated with the beads. As shown in Fig. 4B, ³⁵S-labeled TEF-1 bound specifically to the GST-TDU beads, while no binding to GST beads was observed. In addition, when TDU was produced in vitro and incubated with GST-Su(fu) beads, no [³⁵S]TDU was recovered (data not shown). These results demonstrate that TEF-1 directly and specifically interacts with TDU.

To determine which domain of TEF-1 is involved in the interaction with TDU, three TEF-1 C-terminal deletion mutants were used (Fig. 4B and Hwang et al., 1993). As shown in Fig. 4B, ³⁵S-labeled TEF-1 (1-329) and (1-402) bound to GST-TDU, whereas no binding of TEF-1 (1-205) was observed. This indicates that a domain of TEF-1 within or overlapping with amino acids 205–329 is required for interactions with TDU. Comparison between Sd and the mammalian TEF factors shows that the region required for interaction with TDU is well conserved (64.6% identity and 89% similarity at the amino acid level, Fig. 4C).

To confirm that Vg homology region is required for interaction with TEF-1 and Sd, we tested three different TDU mutants, one with an 27 amino-acid deletion that removes all the Vg homology region (TDU-Δ24-50), one with an 17 amino-acid deletion that removes the N-terminal homology domain (TDU-Δ24-40) and one with an 10 amino-acid deletion that remove the C-terminal homology domain (TDU-Δ41-50), for their ability to interact with TEF-1. The full-length [³⁵S]TDU protein and the three [³⁵S]TDU deleted proteins were produced in a coupled transcription/translation reaction system and incubated with GST-TEF1 beads. As shown in Fig. 4D, deletion of the Vg homology region (TDU-Δ24-50) abolishes the binding of TDU to TEF-1. However, a lower but significant binding was observed when Vg homology region was partially deleted (TDU-Δ24-40, TDU-Δ41-50). Similar results were obtained when we used GST-Sd beads (Fig. 4D). These findings confirm that the Vg homology domain of TDU is required for interaction with TEF-1 and that both the C- terminal and the N-terminal domains of this conserved region contribute to this interaction.

If TDU interacts with TEF to form an active transcriptional factor, we would predict that TDU behaves as a transcriptional activator. Transfection of Drosophila S2 cells with a vector expressing a fusion protein between TDU and the DNA- binding domain of the GAL4 yeast protein (pCaSpeR-hsp- GAL4_db-TDU) resulted in a strong activation of expression from a GAL4-responsive pUAST-lacZ reporter (Fig. 4E). These results confirm that TDU interacts with TEF to form a transcriptional factor and that the Vg homology region of TDU is required for this interaction.

Altogether, these results indicate that this region of TEF/Sd and the conserved sequence of TDU/Vg constitute two evolutionary conserved protein-protein interaction domains. In addition, dimerization observed between Vg and TEF-1 (Simmonds et al., 1998) as well as with the other TEF factors (data not shown) and between TDU and Sd/TEF factors (Fig. 4A-C) further supports the conservation of these interaction interfaces during evolution.

The evolutionary conservation of the Sd/TEF-Vg/TDU interaction suggests that TDU may functionally substitute for Vg. To test whether TDU can substitute for Vg in Drosophila, we produced transgenic flies expressing TDU by means of the UAS/GAL4 system (Brand and Perrimon, 1993).

We first examined whether TDU expression can substitute for Vg in wing development. To do so, we used the vg^83b27 allele, which is associated with an almost complete deletion of intron 2 and which removes the boundary enhancer and behaves as a null allele during wing development (Williams et al., 1991, 1994). We chose to express TDU by using a driver strain in which GAL4 expression is under the control of the boundary enhancer (vg-GAL4 strain). Indeed, we observed that vg expression driven by this GAL4 driver in a vg^83b27 mutant context leads to a complete rescue of the wing phenotype (Van de Bor et al., 1999). When TDU was expressed with this driver, we observed a partial, but significant rescue of the vg^83b27 wing phenotype (Fig. 5A compared to 5B) in emergent flies. Importantly, the wings formed presented veins and wing margins, indicating that TDU not only induces wing tissue growth, but also allows correct wing patterning (inset in Fig. 5A). This result demonstrates that TDU can functionally substitute for Vg in wing development.

It has been previously shown that vg induces its own expression, by acting on boundary and quadrant enhancers, as well as sd expression (Halder et al., 1998; Paumard-Rigal et al., 1998). If TDU is able to interact with Sd to form a transcriptionally active complex, as shown for Vg, we would predict that it will be able to activate sd and vg expression. We first analyzed vg expression in vg^83b27; UAS-TDU; vg-GAL4 wing imaginal discs. Indeed, Vg is not detectable in vg^83b27 wing imaginal discs, but the vg-QE could be activated by TDU expression. As shown in Fig. 5C, we observed that TDU expression at the DV induces vg expression in the wing pouch except at the DV boundary. In fact, it has been shown that in the vg^83b27 mutant, Vg and QE reporter expression is lost, suggesting that vg expression at the DV boundary mediates vg expression in the rest of the wing pouch non-autonomously (Kim et al., 1996; Nagaraj et al., 1999). Our results show that TDU expression at the DV boundary is able to confer an identity on the cells of the DV boundary that allows them to initiate vg expression through the QE. The correct wing margin that we observed in vg^83b27; UAS-TDU; vg-GAL4 flies and the fact that TDU expression restores cut expression at the presumptive wing margin in a vg^83b27 mutant context (data not shown) are consistent the notion that TDU substitutes for Vg in the establishment of a DV boundary.

To test the effect of TDU on sd expression, we expressed TDU with the ptc-GAL4 driver in a sd-lacZ strain (sd^ETX4). Expression of the sd-lacZ reporter gene normally occurs in the wing pouch with highest activity at the DV boundary (Fig. 5F). When the lacZ expression pattern was analyzed in sd-lacZ; UAS-TDU; ptc-GAL4 discs, we observed an increase of X-gal staining at the AP boundary of the wing pouch in accordance with the ptc-GAL4 expression pattern (Fig. 5E).

Ectopic expression of vg by different GAL4 drivers leads to wing tissue outgrowth (Kim et al., 1996). In contrast, expression of TDU in other imaginal discs did not induce wing outgrowth. Nevertheless, using ptc-GAL4 or daughterless- GAL4, we observed a large increase in eye size (Fig. 5G), which can lead to the formation of additional eye tissue in the head. This result indicates that, like Vg, TDU is able to induce cell proliferation although it is not capable of directing these structures to form wing tissue. This effect is partially suppressed in sd^ETX4 males, showing that the eye enlargement induced by TDU is dependent on Sd (Fig. 5H).

Together, these results indicate that TDU associates with Sd to form a bipartite transcription factor, which is able to partially substitute for Vg function in Drosophila.

The development of appendages in Drosophila as well as limb development in vertebrates requires a complex and precise spatial and temporal regulation of gene expression. Studies of wing development in Drosophila constitute a powerful tool to better understand how these complex expression patterns are established and regulated. This latter aspect is well illustrated by studies showing that establishment of the apical ectodermal ridge (AER) of the vertebrate limb bud, like the DV boundary of the wing primordia, is dependent on Notch and Wg (reviewed in Shubin et al., 1997). It has been shown that expression of wing-specific genes such as dSRF or vg results from the activity of the Vg/Sd heterodimer, which restricts expression to the presumptive wing region, and from a combination of signaling outputs (Ci, Su(H) or Mad), which activates a specific pattern of gene expression within the wing field (Halder et al., 1998).

In this study, we have used a different approach, which can be considered as a functional test, to show that the activity of the Sd/Vg dimer is dynamically regulated during wing imaginal disc development. We show that the Sd/Vg dimer is active at the DV boundary at the end of the second larval instar, but is strongly downregulated at the DV and AP boundary in late imaginal disc development. Importantly, as these experiments were done without ectopic gene expression, they probably accurately reflect the wild-type situation for vg and sd. Our findings lead us to propose that the function of the Sd/Vg heterodimer is not solely to restrict the expression of target genes to the wing pouch, but that it also participates in setting up the pattern of gene expression within the wing pouch.

The modulation of Vg/Sd activity could also contribute to growth control during wing development. Indeed, in the wing disc, a zone of non-proliferating cells (ZNC), coinciding with the DV boundary of the wing pouch, is formed during the third instar (O’Brochta and Bryant, 1985). This ZNC is established through the coordinate action of Notch and Wg and is required for a correct specification of the wing margin (Johnston and Edgar, 1998). Several lines of evidence point to a role for vg in the control of cell proliferation; clones with a loss of vg function fail to proliferate (Simpson et al., 1981), whilst ectopic vg expression induces cell proliferation. Recently, it has been shown that cell proliferation of wing pouch cells controlled by the EGF receptor pathway is mediated by regulation of vg expression (Nagaraj et al., 1999). As, vg induces cell proliferation, a negative modulation in Vg/Sd activity at the DV boundary could be a prerequisite to the formation of the ZNC.

Our results highlight the advantage of using a transcription factor complex combining a DNA-binding molecule (Sd) and an activator molecule (Vg) for dynamically controlling the activity of several cognate enhancers, which have to be precisely and dynamically expressed during wing growth and determination. The activity of such transcription factor complexes can be modulated by changing the expression levels of either component. In the case of Sd and Vg, an increase in Sd or a reduction in Vg may lead to diminished activation as Sd alone can bind to its cognate sites and act as dominant negative inhibitor of the Sd/Vg complex. Such a situation increases the combinatorial possibilities for the fine tuning of transcriptional regulation.

Sd belongs to an evolutionary conserved family of transcriptional factors. In mammals, the functions of these different TEFs are not completely understood, but they are probably involved in the transcriptional control of cardiac, skeletal muscle and placental genes. Indeed, it has been shown that the TEFs bind to the M-CAT element found in the regulatory region of several cardiac and skeletal muscle genes (Farrance and Ordhal, 1992) and the placenta-specific chorionic somatomamotrophin gene (hCS) (Jacquemin and Davidson, 1997; Jiang and Eberhardt, 1994). Furthermore, disruption of the TEF-1 gene (Tead1) in mouse is associated with heart defects showing that TEF-1 is likely involved in the regulation of specific cardiac genes (Chen et al., 1994).

Like Sd, TEF function may depend on interaction with coactivators that are expressed in a specific pattern to activate transcription. Indeed transactivation by TEF-1 is cell-type dependent (Xiao et al., 1991) and could act as a transcriptional repressor in particular cell types (Jiang and Eberhardt, 1996). In addition, two factors, NEF-1 and NEF-2, negatively regulate transcriptional activation by TEF-1 in vitro, but none of these factors have been cloned (Chaudhary et al., 1994, 1995). Our results showing that TDU interacts with the TEF factors suggest that it is a good candidate coactivator. TDU is expressed in lung, kidney and placenta, and also probably in fetal heart since ESTs for TDU were originally identified in these tissues. Heart, skeletal muscle, lung, placenta and kidney are enriched in transcripts for various TEFs. It is therefore possible that TDU acts as a transcriptional coactivator for the TEFs in these tissues allowing that activation of specific target genes.

Two others factors have been identified which interact with the TEFs and positively regulate TEF-dependent transactivation of the rat cardiac α-myosin heavy chain (αMHC) and of the cardiac troponin T genes (Gupta et al., 1997; Butler and Ordahl, 1999); the bHLH protein Max and the chromatin-modifying protein PARP. For the troponin T promoter, gel-shift assays indicate that TEF-1 binds to M-CAT elements with different partners (Butler and Ordahl, 1999), one of these could be TDU.

Strong evidence that TDU is a transcriptional coactivator in vivo comes from our experiments which show that TDU is able to functionally substitute for Vg in Drosophila and is able to induce some of the effects observed with ectopic vg expression. Most importantly TDU expression in Drosophila is able to rescue a loss of Vg function in wing blade formation.

To some extent, the function of vg in wing morphogenesis is analogous to function of eyeless (ey) in eye morphogenesis. These two genes are able to trans-determine a tissue and can be considered as master genes control (Kim et al., 1996; Halder et al., 1995). In the case of eyeless, strong homology in the coding and in the flanking sequences indicates that it is orthologous to the murine gene Pax-6/Sey and to the human gene ANIRIDIA, which are also involved in eye morphogenesis. Consistent with this notion, expression of Pax- 6 in Drosophila induces like ey, formation of ectopic eye structures (Halder et al., 1995). In contrast, the homology between Vg and TDU is limited to a short domain of 26 amino acids essential for the interaction with the TEFs. Primarily because of this low similarity between Vg and TDU, we do not consider that TDU is orthologous to vg. However, our results show rescue of vg function by TDU demonstrating that, in terms of molecular function, TDU is analogous to vg. This observation together with the fact that TDU interacts with all the TEFs supports the hypothesis that TDU is a TEF coactivator. The future challenge will be to determine the precise function of TDU in vertebrates. This will require biochemical analysis of the transcriptional properties of the TDU/TEF heterodimers and analysis of the role of TDU development and organogenesis by homologous recombination.

We would like to thank A. Plessis and V. Monnier for the pEG-Ci and pJG-su(fu) plasmids (Monnier et al., 1998), S. Carroll for Vg antibodies, J. Bell for the vg^83b27 strain and D. Busson for reading the manuscript and its helpful discussion. We acknowledge R. Cossard and B. Legois for their excellent technical assistance and D. Weil for its helpful assistance. This work has been supported by the Association pour la Recherche sur le Cancer (ARC), grants ARC6936 and by a grant from Association Française des Myopathes (AFM).