The role of buttonhead and Sp1 in the development of the ventral imaginal discs of Drosophila

One principal question in developmental biology is how the body of multicellular organisms is genetically subdivided. In Drosophilathere is a great deal of information about the subdivision along the anteroposterior axis (AP) body axis: the successive activities of maternal and segmentation genes (reviewed by Lawrence and Morata, 1994) generate a number of equivalent body units that form the parasegmental trunk. The various Hox genes become active in specific positions of the parasegmental trunk and confer different identities, thus generating the morphological diversity in the AP axis.

Within the domains defined by the activity of the Hox genes there are other genes that discriminate the identity of more restricted parts. For example, pannier (pnr) appears to be responsible for distinguishing between medial and lateral regions of all thoracic and abdominal segments(Calleja et al., 2000; Herranz and Morata, 2001). The gene apterous (ap) distinguishes between the dorsal and ventral regions of the wing appendages(Díaz-Benjumea and Cohen,1993), and Distal-less (Dll) specifies the growth and identity of ventral appendages(Cohen and Jurgens, 1989; Gorfinkiel et al., 1997).

A major morphological distinction in the embryonic, larval and adult body is that between dorsal and ventral regions. There is evidence that the dorsal and ventral adult structures of the thorax share a common lineage in early development (Steiner, 1976; Wieschaus and Gehring, 1976; Lawrence and Morata, 1977),but by late embryogenesis the dorsal and ventral primordia have different lineages. It is not clear whether this restriction corresponds to a genuine compartment segregation or whether it is the result of the physical separation of the two primordia, which can be observed during embryogenesis(Goto and Hayashi, 1997; Kubota et al., 2000). Nonetheless there is a clear difference between dorsal and ventral patterns,which is also reflected in the activity of different developmental genes. There are identity conferring genes which are restricted to either the dorsal or the ventral regions, such as pnr, Dll or vestigial(vg) (reviewed by Mann and Morata, 2000).

We have developed a method (Calleja et al., 1996) which allows the visualisation of gene expression patterns in living flies. It is a modification of the Gal4/UAS procedure of Brand and Perrimon (Brand and Perrimon,1993): mobilisation of the pGalw element yields a collection of insertion lines, which are tested using an UAS-y construct. Individual adult flies showing dark (y⁺) patches of interest were used to establish new Gal4 lines. We have used this method for a systematic screen of genes showing restricted expression in adult flies. Among those, the line MD808 was found to confer y⁺ expression in ventral adult derivatives, proboscis, antennae, legs and genitalia, suggesting the gene driving its expression may have a general function in the formation of these structures.

We report a functional characterisation of the gene driving MD808 expression. It turned out to be buttonhead (btd), which encodes a zinc-finger transcription factor(Wimmer et al., 1993) and is known to be required for the formation of the intercalary, antennal and mandibular head segments (Cohen and Jürgens, 1990). A related adjacent gene, termed Sp1(previously known as D-Sp1), has been previously characterised(Wimmer et al., 1996), the function of which is partially redundant with that of btd(Schock et al., 1999). We find that in addition to their role in head development btd and Sp1 are involved in the development of the ventral imaginal discs. Their expression is under the control of the Wg and Dpp signals and is also regulated by other factors such as the bithorax complex genes. Once activated,they induce the function of genes such as Dll and headcase(hdc), which are involved in the specification of adult ventral structures. Our results also show that the Btd product is able to trigger the entire process necessary for leg and antennal development, including the activation of the wg and dpp signalling genes.

Using inverse PCR(http://www.fruitfly.org/about/methods/inverse.pcr.html),we cloned and sequenced the flanking sequence 3′ of the pGalw element(Brand and Perrimon, 1993). The insert is located at 55,584 bp within the scaffold AE 003448 (FlyBase), 753 bp upstream of btd transcription start. Sp1 is located immediately adjacent 60 kb downstream of btd.

The btd^XG81 is a strong allele of btd with the same phenotype as the deletion of the gene(Cohen and Jurgens, 1990). The Df(1)C52 is a deletion of the 8E4-9C region,deficient for btd, Sp1 and about 60 other genes(Cohen and Jurgens, 1990)(FlyBase). The Dp(1;Y)lz⁺ covers the 8B-9A region (Schock et al.,1999) and also rescues the btd and Sp1 larval phenotypes, indicating the genes responsible for the latter are within the 8E4-9A interval. This interval contains about 44 genes (FlyBase). The Ubx¹ allele has been described by Kerridge and Morata(Kerridge and Morata, 1982)and Dll^SA1 by Cohen and Jürgens(Cohen and Jürgens,1989). The reporter genes used were dpp-lacZ (Blackman,1991), wg-lacZ (Ingham,1991), Dll-lacZ (Dll⁰¹⁰⁹²)(Spradling et al., 1999), esg-lacZ (Whiteley et al.,1992) and hdc-lacZ(Weaver and White, 1995)

Heat shocks were given to larvae of different stages of genotype btd^XG81 f^36a FRT18A/y w FRT18A; hsFLP/+. Mitotic recombination in the first chromosome would produce clones of mutant btd^XG81 cells marked with f^36a and twin btd⁺ control clones marked yellow.

The GAL4-inducible constructs for RNA interference were made as follows:for btd, a 400 bp fragment was amplified by PCR using a 5′-gaaggatccgccgccaccgccgccgct-3′ upper primer and a 5′-cggggtaccgtaactggctgttcccgcacc-3′ lower primer. For Sp1, a 813 bp fragment was amplified using a 5′-gccggatcctggctggatatggg-3′ upper primer and a 5′-gccggtaccggccccgcccgtctgtggg-3′ lower primer. Each PCR product was independently cloned as a BamHI-KpnI fragment in the pHIBS vector (Nagel et al.,2002), to make the pHIBS-btd or pHIBS-Sp1constructs. The BamHI-SacI fragments from pHIBS-btdor pHIBS-Sp1 were subcloned in the Bluescript vector, generating the BS-btd and BS-Sp1, respectively. SalI and KpnI fragments (containing the intron and the gene fragment) from pHIBS-btd and pHIBS-Sp1 constructs were cloned in the opposite direction in the BS-btd and BS-Sp1 vector, thus forming the final RNAi constructs BS-btd RNAi and BS-Sp1RNAi. The RNAi constructs were cloned in the pUAS KpnI site and injected in y w¹¹¹⁸ embryos by standard procedures. They are referred to as UAS-btdi and UAS-Sp1i respectively.

We used larvae of genotype btd-Gal4>UAS-btdi and btd-Gal4>UAS-Sp1i to measure the transcripts levels in comparison to btd-Gal4 larvae used as control.

To isolate RNA for RT-PCR, anterior halves of larvae were lysed in Trizol(Invitrogen), and extracted RNA was dissolved in water. Synthesis of first strand cDNA was primed by oligo (dT) and random hexamers. RT-PCR was performed using published primers to amplify the ribosomal protein 49 (RP49) mRNA(Foley et al., 1993), which serves as internal control, and two pair of primers outside the region cloned in the UAS RNAi to amplify btd and Sp1 mRNA.

ey-Gal4 was provided by Walter Gehring and directs expression identical to the eyeless gene. The nub-Gal4, omb-Gal4 and Ubx-Gal4 lines were isolated in our laboratory and drive expression essentially in the domains defined by the genes in which the pGalw element is inserted. ptc-Gal4 is described elsewhere(Wilder and Perrimon, 1995). The MD743 line directs expression in the hinge and the pleural region of the wing (M.C., unpublished). The 444-Gal4 line gives an overall uniform embryonic expression (Azpiazu and Morata,1998). The mata-Gal4VP16 Vp15 lines (a gift from Daniel St Johnston) yield maternal Gal4 expression. The UAS-btd stock(Schock et al., 1999) was provided by Herbert Jackle.

To induce marked clones of cells ectopically expressing btd we used three different chromosomes as convenient: y w hs FLP122 f^36a; abx<f⁺<Gal4-lac-Z(de Celis and Bray, 1997), hsFLP act<CD2<Gal4 (Pignoni and Zipursky, 1997) and y w hsFLP122;act<y⁺<Gal4 UAS-GFP(Ito et al., 1997). They were crossed to UAS-btd flies and the F1 larvae were given heat shocks at different stages.

Embryos were stained using standard procedures for confocal microscopy(Gonzalez-Crespo et al.,1998); secondary antibodies were coupled to Red-X and FITC fluorochroms (Jackson Inmunoresearch) and were analysed under a laser-scan BioRad microscope. For double fluorescent in situ/antibody label we followed the method as described by (Knirr et al.,1999). For the in situ label we used a btd RNA probe reported previously (Wimmer et al.,1996) and provided by Herbert Jackle. As probe for Sp1 we used an RNA transcribed from a cloned 2 kb DNA fragment obtained by PCR using primers from the 5′ and 3′ UTR.

Imaginal discs were dissected in PBS and fixed with 4% paraformaldehyde in PBS for 20 minutes at room temperature. They were blocked in PBS, 1%BSA, 0.3%Triton for 1 hour, incubated with the primary antibody over night at 4°C,washed four times in blocking buffer, and incubated with the appropriate fluorescent secondary antibody for 1 hour at room temperature in the dark. They were then washed and mounted in Vectashield (Vector Laboratories).

The primary antibodies used were: rabbit and mouse anti-β-gal (Capel and Promega), mouse anti-CD2 (Serotec), rabbit anti-vestigial(Williams et al., 1991), mouse anti-dachshund (Mardon et al.,1994), guinea pig anti-homothorax (a gift of Natalia Azpiazu),rabbit anti-Distal-less (a gift of Grace Panganiban), mouse anti-Distal-less(Duncan et al., 1998), mouse anti-engrailed (a gift of Peter Lawrence) and mouse anti-wingless (Hybridoma Center).

Adult flies were by the standard methods for microscopic inspection. Soft parts were digested with 10% KOH, washed with ethanol and mounted in Euparal. Embryos were collected overnight and aged an additional 12 hours. First instar larvae were dechorionated in commercial bleach for 3 minutes and the vitelline membrane removed using heptano-methanol 1:1. Then, after washing with methanol and 0.1% Triton X-100, larvae were mounted in Hoyer's lactic acid 1:1 and allowed to clear at 65°C for at least 24 hours.

In our search for genes with restricted expression in the adult cuticle,the MD808 Gal4 line was found to direct expression in the ventral derivatives of the adult body; proboscis, antennae, legs and genitalia. In the abdomen and analia we could not discern a clear expression. We also noticed that the insertion was located in the first chromosome and associated with a lethal mutation. The mutant larvae showed a head phenotype resembling that described for mutants at the btd gene: loss of antennal organ and the ventral arms of the cephalopharyngeal skeleton(Cohen and Jürgens,1990), and complementation analysis indicated that the chromosome carrying the insert contained a mutation at btd. The expression pattern found in MD808/UAS-lacZ embryos was also similar to that reported for btd (see below), suggesting that the Gal4 insertion was located at this gene. In addition, the imaginal expression of MD808 and of btd was largely coincident.

Further to the genetic analysis and the expression data we cloned DNA fragments at the insertion site to map the position of the P-element on the genome (see Material and methods). It is located 753 bp 5′ of the btd gene (FlyBase). We note that the related gene Sp1, the Drosophila homologue of the human SP1(Wimmer et al., 1993; Wimmer et al., 1996) is immediately adjacent (FlyBase). It is likely that btd and Sp1 have originated by a tandem duplication of a primordial btd-like gene.

The expression patterns of btd and of Sp1 during embryogenesis have already been described(Wimmer et al., 1993; Wimmer et al., 1996). In early embryos btd is expressed in the head region, but by the extended germ band stage the expression domain has expanded to the ventral region of cephalic, thoracic and abdominal segments. During germ band retraction most of the abdominal and thoracic expression is lost, except in derivatives of the peripheral nervous system and the primordia of the imaginal discs. Sp1 is not expressed in early embryos, but from stage 11 onwards it shows the same pattern as btd(Wimmer et al., 1996).

We paid special attention to the btd/Sp1 expression domain in the thoracic imaginal discs primordia, as it may suggest a novel function related to imaginal development. Double labelling with Dll and btd probes(Fig. 1A-C) indicates that btd precedes Dll expression, but by stage 12 the two genes are co-expressed in a group of thoracic cells. However, the Dll domain is smaller and is included within the btd/Sp1 domain: there are cells expressing btd that do not show Dll activity, although all the cells expressing Dll express btd(Fig. 1B,C).

The ventral disc primordia include not only cells expressing Dllbut also other cells containing expression of escargot (esg)and hdc, markers of the diploid cells that form the imaginal primordia (Whiteley et al.,1992; Hayashi et al.,1993; Weaver and White,1995). In late embryonic stages, esg is expressed in a ring domain surrounding the Dll-expressing cells(Gonzalez-Crespo et al., 1998; Goto and Hayashi, 1999)(reviewed by Morata, 2001) and hdc is expressed in a similar pattern (C.E., G.R., M.C. and G.M.,unpublished; Fig. 1E,G). We have carried out double label experiments with btd, hdc and esg probes and found (Fig. 1D-G) that the expression of the two latter genes overlaps with that of btd (and with Sp1, not shown) in the thoracic disc primordia.

The overlap of the btd and of esg domains indicates that btd is also expressed in the hth domain, which is coincident with that of esg (Goto and Hayashi,1999). As the hth/esg domain marks the precursor cells of the proximal region of the adult leg (reviewed by Morata, 2001) the embryonic expression data indicate that btd and Sp1 are active in the entire primordia of the ventral adult structures, including the distal and the proximal parts.

The different expression patterns in third instar leg and antennal imaginal discs are illustrated in Fig. 2. In the mature antennal disc, btd expression is restricted mostly to the region corresponding to the second antennal segment,where it co-localises with both Dll and hth. In the leg disc btd also overlaps in part with Dll and with hth(Fig. 2A-C). The latter result is significant, for the expression of Dll and hth define two major genetic domains, which are kept apart by antagonistic interactions(Gonzalez-Crespo et al., 1998; Abu-Shaar and Mann, 1998). The fact that btd is expressed in the two domains suggests that its regulation and function is independent from the interactions between the two domains. This observation is consistent with the results obtained in embryos(Fig. 1) and suggests that the btd domain includes the precursors of the whole ventral thoracic structures regions from the beginning of development.

The observation that the original btd/Sp1 domain includes the Dll, hdc and esg domains suggested that btd and Sp1 might be necessary for the activation of these genes and hence for the formation of the primordia. Therefore, we studied whether the loss of btd and Sp1 function affect Dll, esg and hdc expression. We also checked for the effect on the formation of Keilin's organs (KO), larval structures which share the same lineage of the leg discs and known to require Dll activity (Cohen and Jürgens, 1989; Cohen,1990).

Larvae mutant for the strong btd^XG81 allele are defective in the formation of KOs (Fig. 3); in 45% of the larvae the KO are absent and in the rest the KO are defective. The incomplete effect of the btd^XG81mutation in suppressing KOs suggested that its function might be redundant and shared by another gene, the obvious candidate being Sp1. As no individual Sp1 mutation is available, we tested the contribution of Sp1 to KO formation by examining first instar larvae hemizygous for Df(1)C52(Schock et al., 1999), which lacks both btd and Sp1 (see Materials and methods). Df(1)C52 larvae show btd mutant phenotype in the head as reported previously (Cohen and Jürgens, 1990), but in addition we find that no KO are present (Fig. 3). Both the btd phenotype and the lack of KOs of Df(1)C52 larvae are covered by the duplication Dp(1;Y) lz⁺, indicating that the responsible genes are located in the interval 8E-9A, which contains btd and Sp1 and about 40 other genes (FlyBase). These results suggest that both btd and Sp1 play a role in the formation of the disc primordia and that their functions are partially redundant, but we cannot rule out the possibility that some other gene(s)located in the 8E-9C interval may also be involved.

We tried to test the specific contribution of btd and Sp1to the mutant phenotype reducing their transcript levels by means of RNA interference. UAS-btdi and UAS-Sp1i genes were synthesised as described in Materials and methods and were found to reduce significantly the transcripts levels of btd and Sp1 (see Fig. 6). Various Gal4 lines(Ubx-Gal4, ptc-Gal4, 444-Gal4, mata-Gal4) were used to test the effect of the two constructs either separately or together. However, in contrast with the results obtained for adult patterns (see below), we failed to see a clear phenotypic effect. We suspect that although the RNAi experiments significantly reduce btd and Sp1 activity, the remaining function is sufficient for the formation of ventral larval patterns.

The loss of KOs in Df(1)C52 larvae activity already suggested an alteration in Dll function, known to be essential for the formation of KOs (Cohen and Jürgens, 1989). Dll expression in btd^XG81 and in Df(1)C52 embryos is shown in Fig. 4: in btd^XG81 embryos there is a strong reduction in Dll activity in the thoracic primordia when compared with the wild-type, although the majority still show some label(Fig. 4B). In Df(1)C52 embryos there is no Dll activity in the same primordia (Fig. 4C), although in a few cases we observe some hints of expression,which might be due to the maternal component of Sp1(Schock et al., 1999). In the two mutant genotypes, there is normal Dll expression in the labial and maxillary segments, where it does not depend on btd/Sp1.

The lack of btd and Sp1 not only suppresses Dll,but it also affects the expression of hdc and esg in the imaginal discs primordia. In Df(1)C52 there is no hdc activity in the ventral discs(Fig. 4E,F). There was some expression in the dorsal discs, especially in the prothoracic one, but the groups of cells corresponding to the wing and haltere discs could not be discerned clearly. The expression of esg is also affected(Fig. 4H,I); it is abolished in the leg discs, but not in the wing and haltere discs. Other markers of dorsal discs such as vestigial (not shown) are also unaffected. As hdc and esg are considered as markers of the imaginal primordia (Hayashi et al.,1993; Whiteley et al.,1992; Weaver and White,1995), their lack in the ventral region of Df(1)C52 embryos suggests that ventral discs are not formed in the absence of btd/Sp1.

The results above demonstrate that btd and Sp1 are necessary for normal Dll, hdc and esg expression. Moreover, btd expression precedes that of Dll, esg and hdc,which suggests it might act as an early activator of these genes. We have tested this possibility by forcing btd activity using the Gal4/UAS method (Brand and Perrimon,1993) and examining the effect on Dll and hdcexpression. In Ubx-Gal4>UAS-btd embryos, btd activity extends from the second thoracic segment to the seventh abdominal one. In this region, the presence of the Btd product induces ectopic Dll activity in many places, although not in all of the zones where btd is expressed (Fig. 4D). Those embryos also show gain of hdc expression. In ptc-Gal4/UAS-btd embryos there is a general gain of hdcactivity in the ventral region (Fig. 4G). In the case of Dll, we checked whether there is a mutual requirement for Dll and btd/Sp1 genes by looking at btd and Sp1 expression in Dll mutant embryos. The result was that their expression is normal, supporting the idea that btd and Sp1 act upstream of Dll.

The activator role of btd (and presumably of Sp1) on Dll suggests that btd and Sp1 may play a role in mediating some of the factors known to be involved in Dll regulation during embryogenesis. Previous work has identified some of these regulators,e.g. wingless (wg), decapentaplegic (dpp)and the bithorax complex genes (Cohen,1990; Vachon et al.,1992; Goto and Hayashi,1997) and we therefore tested whether their role is mediated by an effect on btd and Sp1.

The secreted molecule Wg is necessary for the early activation of Dll in the cephalic and thoracic primordia(Cohen, 1990; Goto and Hayashi, 1997). We find that in the absence of wg function btd/Sp1expression in the disc primordia is abolished, although it remains in the central nervous system (Fig. 5B). In addition, early Dll expression is repressed by dpp; in embryos deficient for dpp function Dllexpression is expanded (Goto and Hayashi,1997). We observe a similar situation with early btdexpression, which extends to the dorsal side in dpp mutant embryos(not shown). The BX-C genes act as repressors of Dll in the abdominal region (Cohen, 1990; Vachon et al., 1992). In Ultrabithorax (Ubx) mutations the first abdominal segment is transformed into a thoracic one and correspondingly there is an additional domain of Dll. We find that Ubx mutant embryos also exhibit an additional btd/Sp1 domain (Fig. 5C) in the first abdominal segment. These results suggest that btd mediates the regulation of Dll exerted by wgand Ubx.

There are two major points emerging from the preceding results. The first is the delimitation of btd/Sp1 activity to a particular group of cells in the mid-ventral region of the embryonic thorax. This is achieved by the regulatory activities of wg, dpp, the BX-C genes and perhaps of other factors. The second point is that once activated, btd/Sp1induces the activity of genes like Dll and hdc, which are necessary for ventral disc development.

As shown above, btd and Sp1 are necessary for the initial formation of the disc primordia, but are also expressed in mature ventral imaginal discs (Fig. 2),suggesting that they may be required during imaginal disc development. To test the imaginal function we first induced a large number of marked btd^XG81 clones (see Materials and methods) during larval period in all body structures, including legs and antennae. All these clones developed and differentiated normally according to their position, and behaved like the twin btd⁺ control clones. The specific function of Sp1 could not be tested because there is no individual mutation at this gene. We tried to recover clones homozygous for the Df(1)C52, in which both btd and Sp1 are deleted, but as expected (they are deficient for over 60 genes included in the interval 8E4-9C), these clones did not survive.

The negative result obtained with btd^XG81 clones was not conclusive because there was the possibility that the loss of btdfunction could be rescued by that of Sp1, which is expressed in the same cells. Besides, there is evidence that btd and Sp1 have partially redundant roles (Schock et al.,1999). We then tested the imaginal requirements for btdand Sp1 by RNA interference using UAS-btdi and UAS-Sp1i constructs (see Materials and methods). As shown in Fig. 6, these are able to reduce strongly btd and Sp1 transcripts levels. As a rule the btd-Gal4 driver was used to check on the morphological effects on legs and antennae. This driver was chosen because it directs expression specifically in the btd/Sp1 domain. Moreover, it is mutant for btd,thus diminishing the amount of wild-type function in the combinations. The general result is that the inactivation of either btd or Sp1alone (btd-Gal4>btdi or btd-Gal4>UAS-Sp1i) fails to affect legs and antennae; at most some reduction of the tarsus was observed in some btd-Gal4>UAS-Sp1i legs. However, when the two genes are inactivated, as in btd-Gal4>UAS-btdi UAS-Sp1i flies grown at 29°C, these are unable to hatch and the antennae and legs are affected in all cases, though to a different extent. The morphological effects are illustrated in Fig. 6A-C. There is no change of identity, but the growth is severely impaired. In the antennae the stronger effect is in the second segment, which is very reduced or absent. The first ant third segments are also affected in size, but the arista is normal. In the legs there is a strong reduction in size, which appears to affect all segments, which often fuse. We also observed similar but weaker effects with the Dll-Gal4 line (not shown).

The pattern abnormalities observed in the double RNAi experiment reflects a requirement during imaginal disc development. Taking advantage that the phenotype of btd-Gal4>UAS-btdi UAS-Sp1i flies is temperature dependent (at 29°C there is strong effect, but at 17°C there is practically no alteration) we performed temperature shifts to ascertain whether btd and Sp1 are required during the larval period. Shifting the temperature from 17 to 29°C during the second or during the third larval period produced morphological alterations in the adult flies that were very similar to those of flies grown entirely at 29°C, indicating a requirement during the late phases of imaginal disc development.

The effects of RNA interference fit very well with the expression pattern in the antennal and leg discs (Fig. 2), and indicate that btd and Sp1 have a imaginal function necessary for the growth of antennae and legs. This function appears to be redundant.

The developmental potential of the Btd protein was tested by inducing ectopic btd activity in different adult regions. In some experiments,we used Gal4 lines to direct activity in specific body regions whereas in others induced flip out clones of btd-expressing cells all over the body (see Materials and methods). The principal result is that ectopic btd activity is able to transform dorsal cephalic and thoracic adult structures into the corresponding ventral ones according to the segment.

The effect in the eye was studied with the eyeless-Gal4 line,which confers expression in the eye cells, and also inducing marked clones of btd-expressing cells. In ey-Gal4/UAS-btd flies the eye disappears and is replaced by an additional antenna which often fuses with the normal one (Fig. 7A). The same transformation was observed in clones of btd-expressing cells: clones located in the eye differentiate antennal structures(Fig. 7B).

In the thoracic segments we used the nub-Gal4, MD743 and omb-Gal4 lines (see Materials and methods). The nub-Gal4 line confers high expression levels in the wing pouch and also in the corresponding region in the haltere, reflecting accurately the domain of activity of the nubbin gene (Ng et al., 1995). The result of expressing btd in these regions is illustrated in Fig. 7C,D: the wings and halteres differentiate leg tissue. The leg tissue in these flies is not arranged in segments as normal legs are, but it is possible to recognise pattern elements corresponding to most leg segments;distal markers such as claws are found as well as tarsal and tibia-like bristles. The amount of leg tissue produced in each of these appendages is greater than the corresponding individual legs, a phenomenon for which we do not have a clear explanation. In favourable cases it is possible to find morphological markers that indicate the segmental identity of the leg tissue;in the wing segment the leg tissue often contains the `edge' bristle that indicates second leg identity.

We note that btd is able to induce a complete set of leg structures even if its activity is induced in different regions of the wing disc. For example, the MD743 and the omb-Gal4 lines drive expression in different and non-overlapping regions, the hinge/pleura and the center of the wing respectively. Yet MD743/UAS-btd and omb-Gal4/UAS-btd flies develop similar sets of leg structures in the wing disc derivatives, suggesting that btd has a capacity to induce leg development independently of the local context.

The transformation of wings and halteres to leg was also observed in the flip out clones expressing btd. Both in the haltere and the wing btd-expressing clones differentiate leg tissue. These clones normally appear as either vesicles of tissue segregated from the wing or the haltere,or as protruding tissue (Fig 7E,F). They tend to form leg patterns resembling the normal ones,which include the formation of various leg segments, suggesting that the positional mechanism to form the proximodistal leg axis is being activated(see below). Moreover, these clones develop nearly complete leg patterns independently of the wing region where they localise. In the example shown in Fig. 7E,F, the clone is located in the posterior compartment but it forms a leg pattern that includes the`edge' bristle, which corresponds to the inventory structures of the anterior leg compartment.

The transformation of the dorsal into ventral structures described in the preceding experiments can be visualised before differentiation takes place by the activation of specific ventral genes, accompanied by the repression of dorsal ones.

btd-expressing clones in the eye gain expression of Dlland hth (Fig. 8), the two selector-like genes that contribute to antennal identity(Casares and Mann, 1998; Dong et al., 2002). They also lose eyeless expression (not shown). These clones adopt a round shape, indicative of having acquired different cell affinities, probably owing to the change of identity. In the wing disc btd-expressing clones lose expression of vestigial (vg)(Fig. 8D,H), a gene that confers wing identity (Kim et al.,1996). In parallel these clones acquire hth, dachshund(dac) and Dll activity(Fig. 8E-K). The gain of hth expression is of interest, as it occurs in clones located in the center of the wing where there is no local hth activity. This suggests that btd-expressing clones tend to recapitulate the development of the entire leg disc, including the proximal region where hth is expressed (Rieckhof et al., 1997). This possibility was further explored by generating large clones in the first larval period. Their size facilitates the study of the expression of leg disc marker genes. These clones often show a spatially discriminated expression of hth, dacand Dll just like the normal leg disc; some are shown in Fig. 9. The activity of hth is always restricted to the periphery of the clone(Fig. 9A,B), whereas dac and Dll extend inside. Although the expression of dac and of Dll partially overlap, they occupy different parts of the clones (Fig. 9C-F). In some favourable cases(Fig. 9E,F), their domains define three different regions within the clone: a region containing only dac expression, a medial one containing dac and Dlland yet another only containing Dll activity. This arrangement resembles closely the normal distribution of these genes products along the PD axis of the leg disc (Lecuit and Cohen,1997).

The same observation was made using Gal4 lines. For example, in MD743/UAS-btd flies only the pleural region is transformed towards leg and this transformation can be visualised by a local outgrowth in the wing disc (not shown). Although the size of the leg outgrowth is variable, dac and Dll are differentially expressed in all discs examined. Their domains are overlapping but not co-extensive.

The differential deployment of leg markers within the groups of cells expressing btd suggested that the Btd product triggers the signalling mechanism, involving the Hh, Dpp and Wg pathways, that normally patterns the leg disc (Basler and Struhl,1994; Lecuit and Cohen,1997). We have checked this by examining the expression of engrailed (en), wg and dpp in the btd-expressing clones.

The activity of en marks the distinction between anterior (A) and posterior (P) compartment, which is fixed during embryogenesis(Lawrence and Morata, 1977; Vincent and O'Farrell, 1992). Thus, all the btd-expressing clones, which are initiated during the larval period, are originally either of A (no cell expresses en) or P(all cells express en) provenance. We find that many of these clones exhibit en activity only in part of the clone(Fig. 10A-D). Thus, depending on their original assignment they either gain or lose en activity. Some are located within the P compartment and therefore lose en,whereas others are in the A compartment and gain en. The formation of this sharp border of en expression within the clones is expected to originate a zone of hedgehog (hh) activity, which in turn induces local activation of wg and dpp(Basler and Struhl, 1994). Indeed, we find that wg and dpp are activated in the btd-expressing clones. This is illustrated in Fig. 10E-H, showing that either gene is activated in part of btd-expressing clones. It is very likely that the appearance of the Wg and Dpp signals what induces the spatially discriminate expression of dac and Dll in the clones.

These results suggest that btd has the potential to induce the formation of the ventral imaginal discs, or at least the antennal and leg discs. The effect is not restricted to the appendage part, it also extends to the proximal region, defined by the activity of genes such as hth or esg (which is expressed in the same cells as hth).

The btd gene performs several roles in development. Previous work has shown that it acts in combination with orthodenticle and empty spiracles to specify the development of mandibular, intercalary and antennal segments (Cohen and Jürgens, 1990; Wimmer et al., 1993). The related Sp1 gene, located immediately adjacent to btd (FlyBase) performs partially redundant functions(Wimmer et al., 1996; Schock et al., 1999) in the peripheral nervous system and the head segments. Both genes are expressed in the primordia of the thoracic discs(Wimmer et al., 1996) (this report), but the functional significance of this expression domain was not known.

Our work demonstrates a novel and also redundant function of btdand Sp1: they are responsible for the formation of the ventral imaginal discs by activating the genetic network necessary for their development. Furthermore, we show that the Btd protein retains the capacity of inducing the entire ventral genetic network during the larval period. We propose that the activation of btd/Sp1 is the crucial event in the determination of the ventral structures of the adult fly.

Our argument is based on the finding that btd and Sp1appear to mediate all events connected with the formation of the ventral discs. We center the discussion in the leg disc, but there is evidence that antennal primordium also requires btd(Cohen and Jürgens,1990). Moreover, we have also observed that the genital primordium is lacking in Df(1)C52 embryos (C.E. and G.M.,unpublished), suggesting that this disc is also under the same control. Most of our experiments concern the function of btd but given the expression and functional similarities between the two genes(Wimmer et al., 1996; Schock et al., 1999) (this report), we assume that Sp1 encodes the same or very similar role. Therefore, we will refer to btd/Sp1 as a single function.

One crucial feature is that btd is an upstream activator of Dll and hdc, which are considered developmental markers of disc primordia (Cohen, 1993; Weaver and White, 1995): (1) btd expression precedes that of Dll and of hdc; (2)the btd expression domain includes those of Dll and hdc; (3)in btd mutants, Dll and hdc activity is much reduced, and completely absent in Df(1)C52 embryos;(4) Ectopic btd function induces ectopic activation of Dlland hdc.

The role of btd in embryogenesis can be illustrated in the light of the models of Dll regulation(Cohen et al., 1993; Kubota et al., 2000). Dll is activated by wg and its expression modulated by the EGF spitz and by dpp, whereas it is repressed in the abdominal segments by the BX-C genes(Vachon et al., 1992). Our experiments suggest that Dll regulation is mediated by btd:in wg mutants there is no btd expression and hence neither Dll nor hdc activity. In dpp mutant embryos, btd expands to the dorsal region resembling the effect on Dll (Goto and Hayashi,1997). In Ubx^- embryos there is an additional group of cells in the first abdominal segment expressing btd; the same cells that also express Dll in those embryos. The interpretation of the role of btd mediating Dll regulation by Ubxis complicated by previous observations(Vachon et al., 1992) showing direct repression of Dll by the Ubx protein. It is possible that Ubx regulates Dll both directly and by controlling btd activity.

We propose that the localisation of btd/Sp1 activity to a group of ventral cells is a major event in the specification of adult structures. btd and Sp1 are activated in response to spatial cues from Wg, Dpp, EGF and BX-C, and in turn their function induces the activity of the genes necessary for ventral imaginal development.

This hypothesis is strongly supported by the results obtained inducing ectopic btd activity in the dorsal discs, as just the presence of the Btd product alone is sufficient to bring about ventral disc development. In the wing and the haltere discs, Btd induces a transformation into leg, whereas in the eye it induces antennal development. This indicates that it specifies ventral disc identity jointly with other factors, e.g. the Hox genes, possibly through the activation of subsidiary genes such as Dll, known to contribute to ventral appendage identity in combination with Hox genes(Gorfinkiel et al., 1997).

The requirement for btd and Sp1 activity appears to be restricted only to the ventral discs, even during the early phases of the thoracic disc primordia. In this context it is worth considering the observation that in Df(1)C52 embryos there is esg expression in the wing and haltere disc primordia, even though it is absent in the leg discs. Thus, the wing and haltere discs are formed in the absence of btd and Sp1. Because in these embryos there is a almost complete lack of Dll expression, this observation raises the question of the origin of the dorsal thoracic discs, which are currently considered to derive from the original ventral primordium, formed by cells expressing Dll (see Cohen,1993; Kubota et al.,2000). Although some of the original group of ventral cells may contribute to the dorsal disc primordia, our data suggest that there may be cells recruited to form the dorsal discs that do not originate in the initial ventral primordium. Accordingly, it is worth considering that in the absence of Dll activity the leg and wing discs are formed(Cohen et al., 1993), although the leg only differentiates proximal disc derivatives. Thus, the activity of Dll cannot be considered a reliable marker for imaginal discs.

Our RNA interference experiments also indicate that both btd and Sp1 are required for the growth of the antennal and leg discs. When the two gene functions are reduced simultaneously, leg segments fuse and there is an overall reduction in the size of antennae and legs. The reduction of growth affects the proximal and distal regions of the appendage(Fig. 6), and assigns a role to the expression observed in the imaginal discs. The two genes are able to perform this function on their own, for the inactivation of only one is not sufficient to impair growth. This conclusion is also supported by the observation that mutant btd clones do not have any effect, as they still possess Sp1 activity, which is sufficient for normal development. At this point we do not know the mechanism by which btd/Sp1 may affect growth.

One particularly significant result about the mode of action of btd comes from the analysis of the ectopic leg patterns observed in the wing and halteres. The cells expressing btd tend to recapitulate the complete development of leg and antennal discs. For example, the whole genetic network necessary to make a leg appears to be activated. btdinduces the activity of hth, dac and Dll, the domains of which account for the entire disc (Lecuit and Cohen, 1997; Abu-Shaar and Mann, 1998). Furthermore, hth, dac and Dll are activated in a spatially discriminated manner: in the clone shown in Fig. 9A,B,hth is expressed only in the peripheral region, resembling the normal expression in the leg disc; in the clone in Fig. 9E,F the discriminate expressions of dac and Dlldefine three distinct regions. The formation of the dac and Dll domains is dependent on signalling from Wg and Dpp, although they require different signal thresholds (Lecuit and Cohen,1997), but the hth domain is independent from Wg and Dpp(Abu-Shaar and Mann, 1998).

The generation of distinct hth, dac and Dll domains within the clones suggested that btd-expressing cells in the wing and haltere generate their own signalling process. Indeed, we find that within these clones there is local activation of en, the transcription factor that initiates Hh/Wg/Dpp signalling in imaginal discs (reviewed by Lawrence and Struhl, 1996). btd-expressing clones also acquire wg and dppactivity in subsets of cells (Fig. 10). It is probably in the boundary of en-expressing with non expressing cells where the Wg and Dpp signals are generated de novo;subsequently, their diffusion initiates the same patterning mechanism which operates during normal leg development. The result of this process is that the hth, dac and Dll genes are expressed in different domains contributing to form leg patterns containing DV and PD axes. One question for which do not have a clear answer is how the initial asymmetry is generated, so that a few cells within the group gain (or lose) en activity. We have observed that the cells expressing en within the clones are(Fig. 10A-D) those closer to the posterior compartment cells. It is conceivable that there might be an external signal, perhaps mediated by Hh, which triggers the initial asymmetry.

The ability of cells expressing btd to build their own patterning mechanism is also indicated by the observation that inducing btdactivity in different parts of the wing disc results in the production of similar sets of leg pattern elements. For example, in MD743/UAS-btdand omb-Gal4/UAS-btd flies, btd is induced in different,non-overlapping wing regions, and yet all leg pattern elements are produced in both genotypes. Thus, the effect of btd is by and large independent of the position where it is induced, e.g. it does not depend on local positional signals.

A relevant issue is whether the ability of the Btd product to induce the formation of the full array of ventral structures has a functional significance in normal development. We believe that this property may be a faithful reflection of the original btd/Sp1 function: the activation of the developmental program to build the ventral adult patterns. We envisage btd/Sp1 function as follows. During the embryonic period, the conjunction of several regulatory factors (Wg, Dpp, EGF, Hox genes) allows activation of btd/Sp1 in a group of cells in each thoracic segment(we assume a similar process takes place in the head). These cells become the precursors of the ventral imaginal discs and will eventually form the ventral thorax of the adult – which includes the trunk (the hth domain) and appendage (the Dll domain) regions. The activity of btd/Sp1 is instrumental in segregating these ventral discs precursors from those of the larval epidermis and determining their imaginal fate. It is involved in specifying their segment identity (in collaboration with the Hox genes) and in establishing their pattern and growth. To achieve the latter role btd/Sp1 induces the production of the growth signals Wg and Dpp,probably in response to localised activation of en and subsequent signalling by hedgehog (hh).

A problem with this model is that in normal development all the genes involved, hth, en, hh, wg and dpp, are expressed in embryos prior to btd/Sp1. Why should a new round of activation be necessary?Although we cannot offer a totally satisfactory answer, we note that clones of btd-expressing cells in wing or haltere lose their memory of en expression. Those that originated in the A compartment had no previous en expression, but gained it in some cells. Conversely, all cells in P compartment clones contained en activity but some lose it. The best explanation for this unexpected behaviour is that btdprovokes a `naïve' cell state in which the previous commitment for en activity is lost. Later, en activity is re-established. We believe this phenomenon may reflect the process that occurs in normal development. The initial btd/Sp1 domain may not inherit the previous developmental commitments and has to build a new developmental program. It is worth considering that the btd/Sp1function appears to determine ventral imaginal fate as different from larval fate. This is a major developmental decision, which may require de novo establishment of the genetic system responsible for pattern and growth. A key aspect of this would be the localised activation of en in part of the btd/Sp1 domain. We speculate that there might be a short-range signal, perhaps Hh, emanating from nearby en-expressing embryonic cells, that acts as an inducer in the btd/Sp1 primordium. There is evidence that Hh can induce en activity(Guillen et al., 1995).

We thank Ernesto Sanchez-Herrero for comments on the manuscript, and Hector Herranz, Magali Suzanne and Francisco A. Martín for discussions. Jose L Gomez-Skarmeta and Elisa de la Calle were of great help with the RT-PCR. We also thank Rosa Gonzalez for the preparation of the adult cuticles and Angélica Cantarero for technical help. G.R. was a recipient of an EMBO long-term fellowship. The work has been supported by grants from the Dirección General de Investigación Científica y Técnica.