Differential requirements of the Fused kinase for Hedgehog signalling in the Drosophila embryo

The segmentation process during Drosophila embryogenesis leads to the formation of a set of repeated organized metameres along the anteroposterior axis. A major step in this process happens during the morphogenetic movements of the germ band, when the initial segmentation pattern is further refined to establish and stabilize the positional information of all cells within each metamere (Nüsslein-Volhard and Wieschaus, 1980; Akam, 1987; Ingham, 1988; Ingham and Nakano, 1990; Hooper and Scott, 1992). This process is believed to occur through cell-cell interactions and involves segment polarity genes.

Further genetic and molecular studies identified two critical signalling proteins; Wingless (Wg) and Hedgehog (Hh) which are secreted by adjacent rows of cells located on either side of the parasegmental boundaries (Martinez Arias et al., 1988; Baker 1987, 1988; Van den Heuvel et al. 1989; Gonzales et al., 1991; Mohler and Vani, 1992; Tabata et al., 1992; Lee et al., 1992). Cells expressing hedgehog also express engrailed (en) which is known to maintain hh expression (Tabata et al., 1992). Wg and Hh signals stabilize each others expression through short range interactions between 3 and 6 hours of embryonic development (Bejsovec and Martinez-Arias, 1991; Heemskerk et al., 1991; Heemskerk and DiNardo, 1994). These short range interactions establish a constant source of organizing molecules that maintain and reinforce the compartmental boundaries.

Besides their early role in reciprocal stabilization, Wg and Hh play additional roles in embryonic development since they can act directly as morphogens to ultimately pattern the entire compartment. Indeed it has been shown that Wg behaves as a morphogen for the specification of naked ventral cuticle (Dougan and DiNardo, 1992; Bejsovec and Wieschauss, 1993,1995; Lawrence et al., 1996). Evidence has also been presented for direct specification of dorsal cells in the embryonic ectoderm by the Hh protein in a concentration-dependent manner (Heemskerk and DiNardo, 1994; Bokor and DiNardo, 1996).

In the ventral ectoderm, Hh signals in both directions leading to the expression of the patched (ptc) gene in all neighbouring cells (Nakano et al., 1989; Hooper and Scott, 1989; Ingham et al., 1991) indicating that the Hh signal is not polarized. However two observations suggest that anterior and posterior cells might have different abilities to respond to Hh: (1) only the anterior neighbouring cells express wg in a Hh-dependent manner; (2) Hh patterns only those cells posterior to its expression domain in the dorsal ectoderm (Heemskerk and DiNardo, 1994). The molecular basis for this asymmetric process is unclear.

In anterior neighbouring cells, a pathway is known to transduce the Hh signal. Two transmembrane proteins are thought to constitute the Hh receptor: Ptc and Smoothened (Smo) (Ingham et al., 1991; Ingham and Hidalgo, 1993; Stone et al., 1996, Marigo et al., 1996; Alcedo et al., 1996, van den Heuvel and Ingham, 1996). Ptc function leads to repression of wg transcription as well as its own. By binding to Ptc, Hh antagonizes Ptc activity which results in the stabilisation of a transactivating isoform of Cubitus interruptus (Ci), a zinc finger protein which activates the Hh target genes ptc and wg (Eaton and Kornberg, 1990; Orenic et al., 1990; Motzny and Holmgren, 1995; Alexandre et al., 1996; Dominguez et al., 1996; Hepker et al., 1997; Aza-Blanc et al., 1997). The means by which Ci is activated by the Hh signal is not yet fully understood. Ci is associated with a cytoplasmic protein complex with the serine/threonine protein kinase Fused (Fu) (Préat et al. 1990) and a kinesin-like molecule, the Costal-2 (Cos2) protein (Robbins et al., 1997; Sisson et al., 1997) and interacts with the Supressor of fused (Su(fu)) protein (Monnier et al. 1998). This complex is thought to mediate Hh signalling: in the absence of Hh it is associated with microtubules, however in the presence of Hh the complex is released which may allow Ci to localise to the nucleus. It is still unknown if this scheme can be extended to all Hh receiving cells and if the protein components of the Hh pathway are equivalent in all embryonic cells. Indeed one hypothesis to explain the difference between cells expressing both wg and ptc genes and the posterior neighbouring cells expressing only ptc could be the existence of two distinct signal transduction pathways.

To test this hypothesis we focused on the Fused protein kinase and tested if it is required in all or only in a subset of Hh responding cells. Fu protein was thought to be uniformly distributed because the transcripts are ubiquitously expressed in the embryo (Thérond et al., 1993). It is believed to be involved in Hh signal transduction for the following reasons. (1) Fu is required for the proper expression of wg and en, and probably hh, during the period when the expression of each of these genes is interdependent (Limbourg Bouchon et al. 1991; van den Heuvel et al., 1993; Forbes et al., 1993). (2) Fu activity is necessary for extended anterior wg induction by the Hh signal in HS-hh embryos (Ingham, 1993). (3) Fu was shown to be phosphorylated in an Hh-dependent manner in embryos and in tissue culture cells (Thérond et al., 1996a). (4) Epistatic studies have shown that Fu inhibits the negative regulator Costal-2 in anterior cells (Préat et al., 1993). Together these results strongly suggest that Fu is involved in Hh signal transduction, at least in Wg-expressing cells, but they are not conclusive for its exact role. For example all of these results can be rationalized if Fu was required to modify Hh before secretion. Furthermore, it has been suggested that Fu could also play a role in the cells posterior to Hh-expressing cells. Indeed in fu embryos ptc transcription in both cells flanking the hh domain is not maintained (Forbes et al., 1993).

It was thus necessary to determine with greater precision in which cells Fu is required. First we re-examined the expression pattern of Fu using two polyclonal antibodies. We show that the Fu product is present in all embryonic cells with a preferential accumulation in the anterior cells that are capable of responding to Hh but that high levels of ectopic Fu do not affect segmentation. We then used the UAS-GAL4 system to express Fu in defined regions of each segment. With this approach we have been able to determine whether Fu activity is required in all Hh responsive cells. We show here that Fu product can correct the deficiency in the Hh pathway observed in fu mutants when expressed only in the anterior neighbouring cells of Hh, which are the wg-expressing cells. Such rescued embryos show a wild-type ptc expression that is symmetrical to the Hh source of production. Similarly, we show that the dorsal cuticle can be differentiated in cells posterior to Hh cells without Fu contribution. Thus the major and perhaps exclusive role of fu in embryogenesis appears to be the early transduction of the Hedgehog signal to control wg expression in anterior cells. This allows Hh expression to be maintained and consequently Hh-dependent ptc transcription on both side of Hh-expressing cells and differentiation of dorsal cuticle. Furthermore we show that Smo and Ci are necessary for ptc expression in cells posterior to Hh cells. Together these data imply that in these cells, Hh can signal independently of the Fu kinase. This result and the expression of Ci and Su(fu) during embryogenesis suggest that position specific modulators of the Hh response exist in embryos.

A 0-3 hour cDNA library constructed by Poole et al. (1985) (gift of T. B. Kornberg) was screened with a 1.4 kb BamHI-SalI genomic DNA fragment (Thérond et al., 1993). A full-length fu cDNA was recovered and subcloned into the Bluescript KS+ vector (Stratagene) before sequencing using the Sequenase system (USB). This cDNA (fuD6) includes the entire fu coding region from nucleotide position +861 to +3849 (Thérond et al., 1993).

The fu cDNA was cloned in the sense orientation into the EcoRI site of pUAST (Brand and Perrimon, 1993) and pCaSpeR-hs. These vectors, which contain the mini white gene, were coinjected with a Δ 2-3 helper plasmid into a w1118 host line (Robertson et al., 1988) under standard conditions (Spradling and Rubin, 1982).

To produce UAS-fu transgenic flies, stocks were established from 5 independent transformant lines. Two of these lines (denoted as w UAS-fu71 and w UAS-fu140 contain a transposon located on the X chromosome and were used to obtain the w fu^AUAS-fu and w fu¹UAS-fu lines by chromosomal recombination. These strains were kept with the FM7 balancer chromosomes.

Several independently Hs-fu transformant lines were analyzed. Two of them were viable insertions on autosomal chromosomes that completely suppress the Fu phenotypes of fu^A and of fu¹. Embryos were heat shocked for 30 minutes or 60 minutes and fixed 1 hour later for in situ hybridization and immunocytochemistry. Larvae were heat shocked over three consecutive days with ten times 30 minutes separated by a 30-minutes recovery period at room temperature.

Flies were kept at 20°C on standard medium as described by Gans et al. (1975). The strains fu¹, fu^A used in this study were described previously (Busson et al., 1988). Gal4 lines used were en-Gal4 (gift from A. Brand), prd-Gal4 (Brand and Perrimon, 1993) and wg-Gal4 (gift from J. Pradel) which contains the promoter of wg upstream of the Gal4 coding sequence. Ci^cell2 and Ci^DR50 alleles were described by Slursarski et al., (1995). Ci^cell2 allele encodes only for the repressive form of Ci and lacks any transactivation activity (Méthot and Basler, 1999). Ci^DR50 is a partial revertant of CiD (Slursarski et al., 1995). The two UAS-Ci strains used are described by Alexandre et al. (1996) and Dominguez et al. (1996). The smo germline clones were generated using smoIIX43 FRT40A (see van den Heuvel and Ingham, 1996). The UAS-smo construct will be described elsewhere (van den Heuvel, unpublished data).

A fusion protein was created by cloning the BamHI-Pvu2 fragment of fu cDNA E92 (Thérond et al., 1993) into M13mp18. A BamHI fragment was subcloned into pGEX-2T. The corresponding fusion protein (GST-5G) joined GST to almost the entire catalytic domain of Fu (residues 13 to 247). The 5’ junction of the fusion constructs was confirmed by sequencing with Sequenase II (USB).

Transfected E.coli DH5α cells were used to produce GST alone and GST-5G as previously described (Smith and Johnson, 1988) with the following modifications: the incubation temperature was lowered to 25°C and the concentration of IPTG (isopropylthiogalactoside) reduced to 0.05 mM to produce soluble fusion proteins. Cells were then harvested and the GST proteins isolated essentially as described previously (Smith and Johnson, 1988). The proteins were fractionated by SDS-PAGE and the gel slices corresponding to the protein of interest were cut out and electroeluted, for 12 hours, using a Schleicher & Schuell Electroeluter apparatus at 100 V at 4°C in a buffer containing 0.25 M glycine, 25 mM Tris pH 8. Part of these eluted proteins were used to inoculate rabbits for polyclonal antisera production (BABCO). The rest of the eluates were dialyzed 2 times against PBS at 4°C. The GST and GST-5G proteins were independently bound to Aminolink columns (Pierce) according to the manufacturer’s instructions. Antisera from the inoculated rabbits were diluted 1:1 in phosphate-buffered saline (PBS) and were passed through the GST-linked column three times consecutively to clear anti-GST antibodies. The antibodies were then affinity purified essentially as outlined by Harlow and Lane (1988). Another antibody was used in this study recognizing residues 419-493 of Fu as previously described (Robbins et al., 1997). The two different purified antisera gave similar results with immunodetection analysis. The antibody recognizing the Fu catalytic domain was used in all experiments described.

Dechorionated embryos were fixed in 7:3 vol. heptane:4% paraformaldehyde in PEM buffer (0.1 M Pipes pH 6.8, 1 mM MgCl₂, 1 mM EGTA) for 20-40 minutes. Alternatively, embryos were fixed for 5 minutes in 1:1 vol. heptane:37% formaldehyde as described previously (Theurkauf, 1992). Devitellinized embryos were washed three times with PBT (PBS, 0.1% Tween 20), blocked 30 minutes with PBTN (PBT + 5% donkey serum) and incubated overnight with gentle rocking at 4°C with diluted primary antibodies in PBTN as follows: Fu, purified rabbit polyclonal antisera 1:500 (for the anti-kinase domain) and 1:5000 (for the anti-hinge domain); En, mouse monoclonal (4D9) (1:1000) gift from Dr T. B. Kornberg; Ci, rat monoclonal (2A1) (1:15) gift from Dr R. Holmgren; Ci rabbit polyclonal (AbN) (1:1000) gift from Dr T. B. Kornberg. Embryos were then washed 4 times for 1 hour in PBT, blocked 30 minutes with PBTN and incubated for 2 hours at room temperature with secondary antibody. For light microscopy, preabsorbed donkey anti-rabbit or anti-mouse antibody conjugated to HRP (Jackson) were diluted at 1:200. Embryos were washed four times in PBT followed by incubation with Vectastain reagent A and B (Vector Laboratories) as specified by the manufacture. Embryos were washed four times for 15 minutes and stained using the DAB staining kit (Vector Laboratories) as specified by the manufacturer. Indirect immunofluorescence double labelling was performed in parallel using mouse and rabbit primary antibodies. The following preabsorbed secondary antibodies were used: donkey anti-mouse conjugated to Cy3 (1:1000); donkey anti-rabbit conjugated to FITC (1:1000). Images were taken on a Laser Scanning Confocal Microscope (BioRad MRC600).

For Su(fu) labelling embryos were blocked in PBTA (PBS 1×, crystallized BSA 0.1%, Triton X-100 0.1%, sodium azide 0.1%) 3 hours at 4°C. Embryos were incubated overnight at 4°C with preadsorbed anti-Su(fu) purified antibody (Monnier et al., 1998), diluted at 1:50 in PBT (PBS 1×, BSA 0.1%, Triton X-100 0.1%). The biotinylated anti-rabbit secondary antibodies (BIOSYS) were preadsorbed overnight at 4°C against fixed embryos (dilution 1:20), and then diluted 1:200 in PBT. Embryos were incubated 3 hours at room temperature. Embryos were incubated with 5 mg/ml propidium iodide for 30 minutes during secondary antibody washes and mounted in Citifluor (Citifluor Ltd).

Whole-mount in situ hybridization was performed according to Tautz and Pfeifle (1989) with some modifications provided by U. Waldorf (Biozentrum, Basel). Embryos were collected and fixed for 20 minutes at room temperature in a two-phase mixture consisting of heptane-[4% formaldehyde, phosphate-buffered saline (PBS), 50 mM EGTA]. Once devitellinized in methanol/heptane mixture, embryos were rinsed several times in methanol, then in ethanol and stored in ethanol at –20°C. After a second fixation in a PBS/0.1% Tween 20/4% formaldehyde solution, embryos were treated with proteinase K (50 mg/ml) for 5 minutes, their pretreatment being stopped with 2 mg/ml glycine. Once re-postfixed in a PBS/0.1% Tween20/4% formaldehyde buffer, embryos were hybridized overnight at 65°C. For probe labelling, we used a 2.4 kb wg cDNA cloned in Bluescript and a 5.5 kb ptc cDNA cloned in PNB40 (gift from R. Phillips). The cDNA probes were prepared using the digoxigenin DNA labelling kit from Boehringer Mannheim, following the manufacturer’s instructions. For the identification of Ci⁻/Ci⁻; wg-GAL4 UAS-Ci embryos, immunolocalisation was performed using the 2A1 Ci antibody followed by a Cy³-conjugated donkey anti-rat secondary antibody (Jackson).

The fu transcript is evenly distributed within all germ layers up to late extended germ band stage (Stage 11), but from the beginning of germ band retraction, fu transcript abundance is reduced and is no longer detected in stage 13 (Thérond et al., 1993). Polyclonal rabbit antisera directed against, respectively, the catalytic and the hinge domains of Fu (Robbins et al. 1997, Material and Methods) were used to analyze the expression of Fu during embryonic development. Prior to cellularisation, Fu is distributed uniformly within the embryo (Fig. 1A). In the cellular blastoderm and after the onset of gastrulation, Fu is in the cytoplasm of all cells (Fig. 1B,C). Immunoreactivity of Fu showed a striped pattern during germ band extension (stage 9-11) (Fig. 1D) while in fu^A mutant embryos no staining is observed (Fig. 1E). Each stripe is continuous along the dorsoventral axis in both the ectoderm and the underlying mesoderm. The stripes are no longer detectable at stage 12. The stripes seem to correspond to the anterior compartment, just anterior to parasegmental grooves. Unfortunately the weak signal provided with both the Fu sera was not sufficient to allow the precise localization of the compartments.

Fu accumulates in the anterior compartment of the wing imaginal disc (Alves et al., 1998). This pattern parrallels Fu expression during stage 9-11 of embryonic development. Fu accumulation in the anterior compartment could be due to a higher protein level. Alternatively a difference in the accessibility of the epitope to the antibodies could explain the observed pattern. Although our experiments do not definitively separate these two possibilities, they tend to argue against the latter. Indeed two polyclonal sera recognizing two different domains of the protein showed the same pattern. We conclude that Fu is present in all cells of the embryonic ectoderm but seems to accumulate in the anterior compartment.

To analyze the functional relevance of Fu modulation, we have constructed transgenic flies (Hs-fu) in which fu expression is directed by the heat shock protein hsp70 promoter. High levels of Fu could be induced at any stage of development after 60 min heat shock pulses (data not shown). Surprisingly, we observed no differences in either survival or phenotype between control and treated animals. Even multiple heat shocks at several developmental stages fail to induce alterations in the pattern of the segments, and experimental larvae developed into normal adults. No differences in wg or en expression were observed between Hs-fu embryos and wild type (data not shown).

To assess whether Fu produced by the Hs-fu transgene is functional we tested its ability to rescue phenotypes of two different fu alleles. fu mutant embryos die before hatching with a mirror-image duplication of the abdominal dentical belts (Fig. 2B). By contrast, fu⁻; Hs-fu larvae were perfectly viable with no significant segmental defects (Fig. 2F). Because of the maternal and zygotic basal level of transcription associated with the non-induced hsp70 promoter, rescue was observed even with no heat shock and a stock of fu⁻; Hs-fu could be maintained. In these embryos we observed the same Fu localisation in a striped pattern as that observed in wild-type embryos (data not shown). Thus, the Hs-fu construct produces ubiquitous functional Fu, it does however not appear to affect the normal cellular interactions occurring during embryonic development.

To identify precisely the cells in which Fu is required we used the GAL4 system to express Fu in restricted regions of each segment. To avoid interference with endogenous wild-type Fu we used the fu^A allele that produces a segment-polarity lethal embryonic phenotype. Because fu^A homozygous mutant embryos can be rescued by maternal wild-type Fu we were able to obtain viable fu^A adults that did not express functionally active endogenous Fu. We first checked that the UAS driven fu cDNA does not induce a partial rescue of the fu⁻ embryonic phenotype in the absence of GAL4 by crossing homozygous fu^AUAS-fu females with fu^A males. Only 3% (13/514) of the resulting embryos hatch and all the embryos are indistinguishable from fu^A embryos derived from fu^A females (Fig. 2B). This confirms that a UAS driven expression of Fu in the segment could be induced without detectable Fu⁺ background of the transgene.

We used a strain bearing a GAL4 coding sequence under the control of the wg promoter (kindly provided by Dr Pradel). When fu^AUAS-fu females are crossed with fu^A; wg-GAL4/+ males at 25°C, Fu is expressed in wg-GAL4 bearing embryos, with a pattern similar to that of wg at all developmental stages (Fig. 1F-G). In this cross 47% (427/909) of the embryos hatch. When examined for cuticular pattern defects, half of the embryos show a characteristic fu pattern, as expected for embryos without the wg-GAL4 chromosome. The other half of the progeny show cuticular patterns either indistinguishable from wild type (Fig. 2E) or with defects in only a small number of denticle bands (Fig. 2D) involving partial or total fusion of adjacent denticle bands. These defects could be attributed to an insufficient amount of GAL4 protein produced by the wg-GAL4 transposon preventing complete rescue. This result clearly demonstrates that the presence of Fu only in the wg-expressing cells is sufficient to rescue the segment polarity defects of fu⁻ mutant embryos leading to wild-type segmentation. Temperature shift experiments using the thermosensitivity of the GAL4 activity suggest that Fu is required early in wg cells during a critical period before 5 hours AEL: providing the wg-expressing cells with an increased amount of Fu by raising the temperature from 19°C to 25°C at 5 hours AEL does not improve hatching compared to embryos raised at 19°C, in contrast to earlier temperature shifts performed before 5 hours AEL (data not shown).

In fu⁻ embryos it has been shown that Wg protein disappears from the dorsal epidermis by stage 10 before complete extinction at stage 11 (Limbourg Bouchon et al., 1991; van den Heuvel, 1993; Fig. 3A-D). Wg function is required to maintain en expression in the adjacent cells. Cells expressing en also express hh; en maintains hh expression (Tabata et al., 1992). At a stage when expression of wg, en and hh are interdependent, the decay of Wg induces the loss of en and hh expression in posterior adjacent cells during stages 9-10 (Lee et al., 1992; Tabata et al., 1992; Ingham and Hidalgo, 1993; Bejsovec and Wieshaus, 1993). While in wild-type embryos en is expressed in full stripes in every segment (Kornberg et al., 1985), in fu and hh embryos gaps appear in the en expression stripes during stages 11-12 (Limbourg Bouchon et al., 1991, van den Heuvel 1993) (Fig. 3G,H). The phenotypic rescue of fu^AUAS-fu/+; wg-GAL4/+ embryos strongly suggests that proper cell-cell interactions between wg and hh-expressing cells have been restored. We confirmed this by looking at wg and En expression in rescued embryos.

In fu^AUAS-fu/+; wg-GAL4/+ embryos the transcription pattern of wg and the expression of En are indistinguishable from wild type (Fig. 3: compare A,B with E,F and see also K,L) at all stages, although the level of wg transcription is reduced compared to wild type (perhaps due to late induction of fu by GAL4 system). By double labelling immunofluorescence we showed that En expression is rescued in the fu^AUAS-fu/+; wg-Gal4/+ flies and is restricted to the cells posterior of the wg domain where Fu is expressed (Fig. 3K,L). From these results we conclude that the presence of Fu solely in the wg domain is necessary and sufficient to restore wild-type wg and en expression particularly at the stage when wg and en expression are co-dependent. The localized requirement for Fu is in good agreement with its role in mediating transduction of Hh secreted from en-expressing cells.

The previous result indicates that Fu is required for the maintenance of wg expression in cells anterior to the en/hh domain and therefore it is indirectly required for maintenance of hh expression in this domain. We directly tested a possible role of fu in posterior cells of the segment by generating embryos in which Fu is present only in en-expressing cells. When fu^AUAS-fu females were crossed with fu^A; en-GAL4/+ males at 25°C, the embryos did not hatch and displayed a fu phenotype indistinguishable from each other (Fig. 2C). We further looked at the fu expression pattern which presumably reflects the en transcription pattern (through enGAL4) and found that in all embryos the en stripes decay by stage 10-11 with an uneven pattern (Fig. 3I,J), as observed for the en transcription in fu⁻ embryos (Limbourg Bouchon et al., 1991). Thus, the presence of Fu in en-expressing cells does not induce any visible correction on the segmentation polarity defects of fu⁻ embryos and en transcription. Therefore Fu appears to be dispensable in the posterior compartment and its effect on en expression occurs through wg regulation in the cells anterior to the en/hh domain.

The complete rescue of fu^AUAS-fu/+; wg-GAL4/+ embryos is intriguing in view of previous results concerning ptc expression (Taylor et al., 1993). It has been shown that late ptc transcription in two narrow stripes flanking the En/Hh-expressing cells (Fig. 4A-E) can be broadened in Hs-hh induced embryos (Tabata and Kornberg, 1994). Furthermore ptc expression disappears in hh or fu mutant embryos (Forbes et al., 1993). These authors proposed that ptc transcription is dependent on the transduction of Hh by Fu in both cell stripes flanking the Hh domain. Nevertheless, since in fu⁻ mutant embryos wg expression decays, and consequently hh expression in adjacent cells would also, the decay of ptc expression in cells posterior to hh domain could be a consequence of the loss of Hh rather than absence of Fu activity.

To address this issue ptc transcription was investigated in fu^AUAS-fu/+; wg-GAL4/+ embryos and fu^A embryos. In the latter, ptc expression is initially similar to that observed in wild-type embryos, with large stripes complementary to the En stripes (data not shown). By stage 9 to 12 ptc transcription is reduced (Fig. 4K,L) as compared to the wild type, the most striking feature being the absence of ptc transcripts in the laterodorsal cells. A strong ventral area of expression remains around the developing tracheal pits. By stage 12 ptc transcription has completely disappeared. In contrast, at the same stage, the fu^AUAS-fu/+; wg-GAL4/+ embryos have a wild-type pattern of ptc transcription (Fig. 4F-J). Not only is the latero-dorsal expression restored but also subtle regulations of ptc transcription, such as modulation of transcription of the two thin stripes flanking the En stripe, are similar to that observed in wild-type embryos (compare Fig. 4D and 4J). These results show that maintenance of ptc expression in cells posterior to the en/hh domain is Fu independent.

To verify the differential requirement of Fu for ptc expression in cells anterior versus posterior to En/hh cells we crossed fu^A; UAS hh females with fu^A; en-GAL4/+ males. In half of the embryos from this cross (genotype fu^A; en-GAL4/+;UAS-hh/+), we expect that the hh transcription will persist longer in the En cells than in those of fu^A; UAS hh/+ embryos. This maintenance of hh transcription is not sufficient to restore the wg transcription in anterior cells which decreases by stage 10 in all the embryos with only a slight delay in fu^A; en-GAL4/+;UAS hh/+ embryos (data not shown), confirming the major role Fu plays in relaying the Hh signal to maintain wg expression. Nevertheless, in the double transgenic embryos the ptc expression anterior to hh-expressing cells is lost whereas the expression posterior to En/Hh-expressing cells can be observed up until stage 11-13 (Fig. 4N,O). Similar results were obtained with the fu¹ allele, demonstrating that this Hh induced ptc transcription is not related to special characteristics of the fu^A mutation. Furthermore, because wg expression decays in fu^A; en-GAL4/+;UAS-hh/+ embryos we conclude that the posterior ptc expression observed in these embryos is not dependent on long range diffusion of the Wg protein. Thus, at stage 11-12, the ptc transcription in cells posterior to the En stripe depends on the Hh expression in En cells. These experiments suggest that there is a differential requirement for Fu downstream of Hh to maintain ptc expression on each side of Hh-expressing cells.

To confirm that the Hh pathway is required in the cells posterior to the hh/en domain, we deleted the activity of Smo, a protein required in cells receiving the Hh signal as part of the receptor complex to lead to target gene transcription (Alcedo et al., 1996; van den Heuvel and Ingham, 1996). In smo⁻ germline clone mutant embryos, ptc and wg transcription are completely lost at stage 11 (Fig. 5A and unpublished data) as in hh⁻ embryos. In wg-GAL4 UAS-smo, smo⁻ germline clone mutant embryos, the anterior stripe of ptc expression is rescued (Fig. 5B) as well as wg expression (data not shown). However, the stripe of ptc expression posterior to the en/hh domain is lost at stage 11 (Fig. 5B). This result is consistent with a requirement for smo in these cells.

Using the same approach, we tested the requirement of the transcription factor Ci for the maintenance of ptc transcription on both sides of the hh/en domain. Ci is the most downstream component of the Hh cascade and in mutant embryos which are homozygous for Ci^ce, lacking the Ci activator activity (Méthot and Basler, 1999 and see below), ptc transcription is lost during germ band extension (Fig. 5C,D). In wg-GAL4 UAS-Ci embryos where Ci is overexpressed in the wg domain, the two stripes of ptc expression are present, the anterior one being the strongest (6E,G). However, in wg-GAL4 UAS-Ci, Ci^ce embryos ptc transcription is maintained anteriorly to hh/en while its expression is lost posteriorly (Fig. 5F,H). Similar results were obtained using the Ci^DR50 allele (data not shown). Taken together these results indicate that ptc expression is dependent on Hh signalling mediated by Smo and Ci, on both sides of the hh/en expression domain.

Since regulation of the Hh target genes wg and ptc is believed to depend on the activity of Ci, we compared the localization of the Ci proteins in wild-type, fu^A and fu^AUAS-fu/+; wg-Gal4/+ embryos.

In wild-type embryo, in cells that do not receive Hh, the full length 155 kDa Ci (Ci¹⁵⁵) protein is processed into a transcriptional repressor of 75 kDa, presumably through a proteasome-dependent degradation involving the Slimb protein and the PKA kinase (Aza-Blanc et al., 1997; Jiang and Struhl, 1998; Ohlmeyer and Kalderon, 1998). In cells exposed to Hh, this processing is inhibited by the Hh signalling pathway which results in the accumulation of the Ci¹⁵⁵ in cells surrounding the Hh-expressing cells during stage 10 and 11, as recognized by a monoclonal antibody specific for the Ci¹⁵⁵ isoform (Fig. 6A). Nevertheless, several factors showed that the Ci¹⁵⁵ protein is not the transcriptional activator form of Ci but that another step of activation is needed to convert Ci¹⁵⁵ into a labile activated form (Ohlmeyer and Kalderon, 1998; Méthot and Basler, 1999). By stage 11 in wild-type embryos the intensity of the Ci¹⁵⁵ stripe anterior to En cells decreases and by stage 12 only the posterior stripe can be detected (Fig. 6B). The total expression of the different isoforms of Ci, as revealed by the polyclonal antibody AbN is complementary to the En expression domain until the end of stage 11 and is then restricted to a stripe posterior to the En-expressing cells (data not shown). These regulations are posttranscriptional since the ci transcripts are still uniformly expressed in all cells except en/hh cells by stage 12 (data not shown).

In fu⁻ embryos the amount of Ci¹⁵⁵ is similar in stripes of cells anterior and posterior to the en/hh domain by stage 11 (Fig. 6C) but the anterior stripe of Ci¹⁵⁵ does not decrease in later stages. Thus at stage 12 two stripes of equal intensity are still present (Fig. 6D,G). The same is seen when all the different isoforms of Ci, as revealed by the polyclonal antibody AbN, are detected (data not shown). These data do not make it possible to distinguish between an inhibition of the slimb/PKA degradation pathway and a stabilization mechanism of the detectable Ci¹⁵⁵ isoform in the absence of the Fu protein. Nevertheless, in fu^AUAS-fu/+; wg-GAL4/+ embryos, the late Ci pattern is fully restored (Fig. 6E,F,H). Thus, as for ptc transcription, Fu is required in the cells anterior to the Hh-expressing cells for proper Ci processing but not in the posterior ones.

Since it has been shown that the Su(fu) and fu gene dosage can modulate the Ci¹⁵⁵ protein level and transcription of the Hh target genes (Préat et al., 1993; Pham et al., 1995; Alves et al., 1998; Ohlmeyer and Kalderon, 1998) we investigated the expression of Su(fu) protein in embryo. Previous results have shown that the Su(fu) gene is expressed uniformly in the embryo until the end of gastrulation (Pham et al., 1995). We found that, at least until stage 10, the Su(fu) protein is uniformly expressed in the embryo. Starting at germ band retraction the uniform labelling begins to fade and completely disappears by the end of retraction (Fig. 1G-I). So the differences of Ci¹⁵⁵ protein level anterior and posterior to the En-expressing cells cannot be attributed to different levels of Su(fu) in these regions.

In addition to its role in wg maintenance, Hh acts as a morphogen for dorsal cellular patterning posterior to its domain of expression (Heemskerk and DiNardo 1994, Bokor and Dinardo, 1996). We addressed a possible role of Fu in this process by looking at the dorsal cuticles of fu^AUAS-fu/+; wg-GAL4/+ embryos and fu^A embryos. Four different cell types (1, 2, 3 and 4) had been identified along the dorsal epidermis of the segment (Heemskerk and DiNardo, 1994 and see Fig. 7A). hh is expressed in cell type 1 and specifies cell types 1, 2, 3 whereas wg specifies the large group of cell type 4 where it is expressed in some of them (Bokor and Dinardo, 1996). In fu^A embryos, dorsal cuticle is mostly covered by a lawn of type 4 fine hairs (Fig. 7B,C). Cell type 1 are missing and the extension of naked cuticle (cell type 2) is greatly reduced. In most embryos cell type 3 are present but on rare occasions they are also missing (data not shown). In contrast, cell type 4 are unaffected by the lack of Fu. This phenotype is similar to the phenotype of dorsal cuticle of embryos in which the Hh function has been inactivated at 6-7 hours AEL (Bokor and Dinardo, 1996). It also suggests that the early Wg production independent of Hh is sufficient to engage most of the dorsal cells in the type 4 fate.

The dorsal cuticle of fu^AUAS-fu/+; wg-GAL4/+ rescued embryos are well differentiated with all the four cell types described in wild-type embryos present (Fig. 7D). The rescue is almost complete, with a smaller number of type 1 cells than in the wild-type embryos, which could be a consequence of the lower level of Wg protein in these embryos, as rescued by Fu, than in the wild type (see Fig. 3).

Thus, the late Hh patterning pathway of dorsal cell types 1, 2, 3 appears to be mainly independent of Fu presence, since the protein is present only in a small subset of cell type 4. In consequence, in these embryos, the Hh signal is maintained by the Fu-dependent wg expression in adjacent anterior cells (type 4) and this Hh activity is sufficient to pattern dorsally all the cells of the segment independently of Fu.

Our results clearly support the involvement of the Fu serine/threonine kinase in Hh signal transduction in the embryonic segment within cells producing Wg. Although Fu is present in all embryonic cells, expression of Fu only in the wg-expressing cells of the anterior compartment in a fu mutant context is sufficient to restore a wild-type transcription pattern of both wg and en and a normal cuticular pattern. In contrast its expression in the posterior compartment – the en-expressing cells – has apparently no effect, either on the transcription of wg and en or at the phenotypic level. This is in good agreement with recent data (Alves et al., 1998) showing that the adult fu mutant phenotype can be rescued by expressing Fu in the most posteriorly located cells of the anterior compartment in the wing imaginal disc. Together these data show that Fu is not necessary in the majority of cells where it is expressed and suggest that its activity could be induced in wg-expressing cells in response to Hh.

Here we show that Fu protein is evenly distributed in the embryo until stage 9 when it begins to accumulate in the anterior compartment. What is regulating Fu accumulation? Because fu mRNA distribution is uniform at least until stage 10 we hypothesized that the localisation is due to post-translational regulation. Another component of the Hh pathway, Cos2, also accumulates in the anterior compartment at this stage independently of the level of the Hh signal (Sisson et al., 1997). As for Cos2, the uniform level of Fu in the anterior compartment seems to be constitutive to anterior cells and independent of Hh signal. Indeed, this regulation is observed during the time when local signalling by Wg and Hh stabilize each other’s expression. At this stage anterior cells at the A/P border are receiving Hh and responding to it. Other anterior cells distal to the A/P border do not receive Hh but have the potential to respond to it. Thus, these results are inconsistent with Hh regulating Fu and Cos2 accumulation in the entire anterior compartment. Fu accumulation could be related to its association with Ci and Cos2 within the same protein complex (Sisson et al., 1997; Robbins et al., 1997). Nevertheless, since we show that fu expression is only required in the wg-expressing cells for proper patterning, the higher Fu protein levels in the whole of the anterior compartment do not seem to have any functional significance.

We show that increasing the levels of Fu protein by heat shock induced expression or by using the binary Gal-4 UAS system failed to induced any alteration in embryonic development. In contrast to Fu, high level expression of the transcription factor Ci is sufficient to activate transcription of Hh target genes in the absence of Hh activity (Alexandre et al., 1996; Hepker et al., 1997). Thus, it is possible that endogenous Ci concentration is limiting, probably due to a Hh-dependent proteolytic regulation (Aza Blanc et al., 1997). We conclude from this result that the endogenous concentration of Fu does not provide a major regulatory step in the Hh pathway, whereas the concentration of Ci does.

Because Fu is necessary for the Hh-induced wg expression it was believed that in the embryo, the Hh signal requires Fu to transduce its effects (Ingham, 1993). This idea was further strengthened by recent results demonstrating that Fu is phosphorylated, in vitro and in vivo, in response to the Hh signal (Thérond et al., 1996a) and is present in a protein complex involving the Ci protein that is thought to be the transcription factor mediating the Hh signal (Robbins et al., 1997; Sisson et al., 1997).

Two lines of evidence in this study suggest that Hh can signal independently of Fu in the embryo. First of all, in contrast to hh⁻ or fu⁻ embryos where late ptc transcription is lost, fu^AUAS-fu/+; wg-GAL4/+ embryos show a normal ptc transcription pattern at stage 12 even in cells posterior to the En stripe where Fu is not present. Moreover we have shown that this latter ptc expression can also be observed in fu^A; en-GAL4/+;UAS-hh/+ embryos where late hh expression is maintained. This strongly suggest that (1) the absence of both stripes of ptc expression in fu⁻ embryos is caused by the lack of maintenance of the Hh signal due to the loss of wg expression through a Fu-dependent pathway, (2) the maintenance of the Hh signal in the En-expressing cells in the rescued embryos allows the transcription of ptc in cells posterior to En cells through a Hh signalling pathway independent of Fu.

Nevertheless others interpretations are also possible. For instance a signal other than Hh originating from En or Wg cells could induce ptc transcription in neighbouring cells. It is unlikely that the diffusable Wg signal could be responsible for the late posterior ptc transcription, since in fu^A; en-GAL4/+;UAS-hh/+, the fading of Wg is only slightly delayed compared to fu^A embryos while the late posterior ptc transcription is restored. Normal ptc expression has also been observed in gooseberry mutant (gsb⁻) embryos throughout embryogenesis while wg expression disappears at stage 11 (Hidalgo and Ingham, 1990). Such a putative signal would have to depend on the level of hh transcription, according to the restoration of the late ptc transcription observed in fu^A; en-GAL4/+;UAS-hh/+ embryos. Also the fact that the Hh receptor Smo is required in a cell autonomous manner for ptc transcription in the posterior cells strongly argues that Hh is the signal directly involved in controlling posterior ptc transcription.

A simpler situation is seen in the patterning of dorsal cuticle. In this region Hh behaves as a morphogen since the differentiation of the denticles depends on Hh in a concentration-dependent manner (Heemskerk and DiNardo, 1994). In fu^A embryos the dorsal cuticle lacks cell types 1 and 2 (and occasionally type 3) as observed in embryos where the Hh product is inactivated at 6-7 hours AEL (Heemskerk and DiNardo, 1994). In fu^A, UAS-fu/+; wg-GAL4/+ embryos the wild-type denticles are restored. In particular type 1 and 2 cells can be observed in a region where Fu is not present. This strongly suggests that (1) the abnormal differentiation of dorsal cuticle in hh⁻ and fu⁻ embryos is due to the lack of maintenance of the Hh signal as discussed before, and (2) the maintenance of the Hh signal in the En cells allows the differentiation of the dorsal cuticle through an Hh signalling pathway independent of Fu.

The previous conclusions for the existence of Fu-independent Hh signaling are based on the assumption that there is a lack of functional Fu protein in fu^AUAS-fu/+ embryos. Even if these embryos show a strong fu embryonic phenotype we cannot completely exclude that a partially active Fu protein, below our detection threshold, could allow correct ptc transcription and dorsal patterning by the Hh signal. This is possible if the Fu concentration requirement in these pathways is much lower than in the anterior pathway involved in wg maintenance. However this possibility seems very unlikely. First of all, we never detect Fu protein outside the wg-expressing cells in fu^AUAS-fu/+; wg-GAL4/+ embryos. Secondly, similar results of rescue were obtained with the class I fu¹ allele (data not shown), in which the catalytic kinase domain is affected, in contrast to fu^A, a class II allele (Thérond et al., 1996b) in which only the extracatalytic domain is impaired. Thus our results seem to be independent of the nature of the fu mutations, which strongly suggests that the observed rescue is not due to a remnant Fu wild-type function. Then, in fu mutants the Hh pathway has been shown to be severely affected: the Wg protein disappears by embryonic stage 11 inducing a defect in En and Hh maintenance; late ptc transcription is absent by embryonic stage 12 and only two types of dorsal denticles are observed. Thus it seems unlikely that a putative weak Fu activity, needed for a transduction of the Hh signal outside the Wg cells, could explain the complete rescue that we observe in fu^AUAS-fu/+; wg-GAL4/+ embryos for ptc expression and dorsal patterning while absolutely no phenotypic improvement is seen in fu^AUAS-fu/+ controls.

Our results prove that Hh can signal independently of Fu in the embryo for ptc transcription and dorsal epidermis morphogenesis. This is clearly in contrast to standard models of Hh signaling in which Fused is absolutely necessary for Hedgehog transduction (Ingham, 1993).

Our data raise the possibility that Hh uses completely different pathways to signal in neighbouring anterior versus posterior cells. However, we showed clearly that Hh signal is using the same receptor component in both cell types, Smo being necessary for Hh target genes expression in all cell types. Thus, starting from Smo, two alternative hypothesis may be proposed to explain the response of the cells posterior to the Hh-expressing domain. (a) Completely new components downstream of Smo are used to control ptc expression, which seems unlikely because of the Ci requirement, or (b) the same Hh signaling components except Fu are used. In the latter case several mechanisms may explain the asymmetric Hh signaling.

(1)The target genes wg and ptc may be sensitive to different levels of the Hh signal. In this case Fu would facilitate Hh signaling and thus allow expression of genes that require a high threshold of Hh. Such a process has been described in wing imaginal discs: a high level of Hh signal is necessary for a Fu-dependent en expression while a lower level of Hh is sufficient to promote a Fu-independent dpp expression (Strigini and Cohen, 1997; Alves et al., 1998; Ohlmeyer and Kalderon, 1998). Nevertheless our results suggest that this scheme cannot account uniquely for the Fu-independent ptc transcription posterior to Hh stripe. Indeed both cells bordering the Hh source of production should receive the same amount of Hh signal and yet they show differential transcription of Hh target genes. Moreover, we showed that both wg and ptc transcriptions depend on Fu activity in the adjacent anterior cells.

(2)Alternatively, the Hh signal could be modulated by position-specific modulation of known components of the pathway. For instance it has been shown for wing disc morphogenesis that the severity of the fu phenotype in fu^-flies depends on the Su(fu) gene dosage: increasing the number of copies of Su(fu) increases the fu phenotype while a lack of Su(fu) supresses it (Pham et al., 1995). Moreover, recent results in wing imaginal discs and in embryos (Ohlmeyer and Kalderon, 1998) suggest that Fu could promote, through Hh activation, the maturation of the Ci protein into a short lived transcriptional activator. In the absence of this signal the Ci¹⁵⁵ protein would be inactivated through the formation of a complex with the Su(fu) protein. Thus, in this model, Hh transduction is modulated by the stochiometry of the Fu and Su(fu) proteins which has been shown to operate in opposite directions (Préat et al., 1993; Pham et al., 1995; Ohlmeyer and Kalderon, 1998). The existence of Fu-independent Hh signaling in some cells could then result from different Su(fu) concentrations in the embryonic segment. Our results suggest that this scheme cannot account for the Fu-independent ptc transcription posterior to the En stripe. Indeed, initially, the Su(fu) protein appears to be uniformely expressed in the embryo at least until stage 13. Secondly, by stage 12 the Ci¹⁵⁵ protein is differentially matured on either side of the En stripe: in wild-type and rescued embryos, the anterior stripe of Ci¹⁵⁵ disappears, which suggests a complete processing of the Ci protein. By contrast a high level of non processed Ci¹⁵⁵ persists in the posterior cells which suggests that the level of the putative short lived Ci transcriptional activator would be higher in the anterior cells than in posterior ones. Thus one would expect a preferential transcription of Hh target genes in anterior cells. Nevertheless in fu⁻; en-GAL4/+; UAS-hh/+ embryos, where the Hh signal level is increased compared to fu⁻ embryos, only the posterior ptc stripes are maintained by stage 12. Altogether these data suggest strongly that the Fu-independent posterior ptc transcription cannot be accounted for by an assymetric maturation of the Ci¹⁵⁵ protein into a short lived transcription activator on either side of the En cells. Nevertheless we can not exclude that yet unindentified transcriptionally active forms of Ci are differentially expressed in these cells.

(3)A third hypothesis, that we favor, is that the Hh signalling pathway may act in synergy with other spatially localized transcription activators to promote Hh targets transcription. Indeed we have shown that Ci is necessary for the transduction of the Hh signal in posterior cells independently of Fused. Moreover the full length form of Ci is observed at high levels in these cells. This form, unprocessed by Fused, could promote transcription of the ptc gene in the presence of a coactivator absent of the Wg cells. Alternatively, this form could have a constitutive activity which is repressed in the Wg cells by another localized factor.

In the dorsal cuticle the level of the Hh signal has been shown to be directly responsible for the morphogenesis of the dorsal denticles (Bokor and DiNardo, 1996). As in embryonic ventral epidermis and imaginal patterning, the antagonistic roles played by Hh and Ptc have been identified as being responsible for the patterning of cell types 1 and 2 in the dorsal epidermis of the embryo (Bokor and DiNardo, 1996). By contrast, cell type 3 which is also dependent upon Hh activity is not specified by the usual antagonism between Ptc and Hh (Bokor and DiNardo, 1996), suggesting that the Hh morphogen does not always act through the usual pathway Ptc – Smo – Fu – Ci. Our results clearly support this assumption. In particular, it is significant that cell type 3 can still be observed in a fu^- mutant, probably due to the early hh transcription. The maintenance of the Hh source in the En cells in fu^A, UAS-fu/+; wg-GAL4/+ embryos allows the differentiation of cell types 1 and 2 by a different process which requires Ptc repression but not a functional Fused protein which is absent in these cells.

In fu⁻ cells the short lived Ci transcription activator, dependent on the Hh signaling, is reduced or may be suppressed. Thus the complete restoration of the dorsal cuticular pattern in fu^A, UAS-fu/+; wg-GAL4/+ embryos cannot be explained if the short lived Ci transcription activator is the only effector of the Hh signal. Interestingly, the dorsal cuticular pattern in Ci^Ce2, UAS-Ci/+; wg-GAL4/+ embryos is similar to the dorsal pattern in hh mutant embryos (data not shown). This suggest that another form of the Ci protein which does not require a Fu maturation to become a transcriptional activator could be partly responsible for this dorsal patterning. As for ventral cells, new position-specific coactivators of Ci could be part of the Hh signaling pathway for dorsal cell specification.

Thus as more data becomes available, a very complex scheme of Hh signaling emerges. This complexity could originate from different forms of the Hh protein which could be used for signaling, and/or, as we suggest in this paper, from the existence of position-specific modulators of the Hh response. New investigations on the interplay between these two sources of variation of the Hh signal response are clearly necessary to understand the patterning of the embryo.

We thank T. Kornberg, K. Nybakken, D. Busson, A. Brand, J. Pradel for providing strains and reagents. We thank also B. Theunissen and F. Nourrit, students of the Molecular and Cellular Genetics DEA of Paris VI and Paris XI Universities, for their contribution to this work. Some of these experiments were initiated while P. P. T. was a post-doctoral fellow in J. M. Bishop’s laboratory at the University of California, San Francisco, and P. P. T. is deeply indebted to him for generous support and encouragement. P. P. T. wants also to acknowledge members of the Bishop laboratory for helpful and stimulating discussions. We thank C. Desplan for critical comments on this data. This work was supported by the ACCSV program of the Centre National de la Recherche Scientifique (CNRS), by grants from the Association pour la Recherche contre le Cancer, from Ligue Nationale contre le Cancer, from Fondation pour la Recherche Médicale and from the ATIPE program of the CNRS to P. P. T.