Suppressor of Fused opposes Hedgehog signal transduction by impeding nuclear accumulation of the activator form of Cubitus interruptus

Members of the Hedgehog (Hh) family of secreted signaling proteins function as potent organizers in animal development. Drosophila Hh controls pattern formation during segmentation and limb development. Vertebrate Hh homologues play key roles in the morphogenesis of the neural tube, somites, axial skeleton, limbs, lung and skin (reviewed in Neumann and Cohen, 1997; Ingham, 1998; Ruiz i Altaba, 1999; McMahon, 2000). Most if not all Hh signaling is mediated through the transcriptional effectors of the Gli/Ci family (reviewed by Aza-Blanc and Kornberg, 1999). Major interest is devoted therefore to the mechanisms by which Hh signaling regulates the activity of Gli/Ci proteins.

The function of Hh is best understood in the developing Drosophila wing imaginal disc, a tissue that is formed by two cell populations, the anterior (A) and posterior (P) compartments. The selector gene engrailed (en) is expressed in P cells where it permits the synthesis of Hh while at the same time preventing a response to Hh (reviewed by Lawrence and Struhl, 1996). Hh diffuses into the A compartment and initiates a signaling cascade by binding to its receptor, the multi-transmembrane domain protein Patched (Ptc). Hh neutralizes the inhibitory hold of Ptc on the co-receptor protein Smoothened (Smo), which then signals to downstream effectors (reviewed in Alcedo and Noll, 1997; Ingham, 1998; Strigini and Cohen, 1999). Signaling results in the stabilization of the zinc finger protein Cubitus interruptus (Ci; Orenic et al., 1990) and the transcriptional activation of target genes. Several proteins, such as Cos2, PKA, Su(fu) and Slimb, prevent or downregulate signaling, but their mode of action is not well understood. Some evidence suggests that Cos2 tethers Ci to the cytoplasm (Robbins et al., 1997, Sisson et al., 1997), whereas Slimb and PKA are required for the proteolytic processing of Ci (Jiang and Struhl, 1998; Chen et al., 1998). Cleavage of Ci occurs in the absence of Hh (Aza-Blanc et al., 1997) and results in the release of a C-terminally truncated Ci (here referred to as Ci[rep]), that translocates to the nucleus and inhibits transcription of target genes such as decapentaplegic (dpp) and hh itself (Dominguez et al., 1996; Aza-Blanc et al., 1997; Méthot and Basler, 1999). Reception of Hh prevents Ci[rep] formation, and activates the transcription of several Hh target genes such as dpp and ptc (Aza-Blanc et al., 1997; Méthot and Basler, 1999). Recently, it has been shown that inhibition of Ci processing is not sufficient to activate Hh-target gene transcription. An uncleavable form of Ci (Ci^U) still requires Hh input to induce ptc expression in wing imaginal discs (Méthot and Basler, 1999), and the proteosome inhibitor lactacystin does not potentiate the response to Hh in tissue culture cells (Chen C.-H. et al., 1999). These results suggest that a Hh-dependent event other than prevention of cleavage permits full-length Ci to function as a potent transcriptional activator (Ci[act]). Several studies identified phosphorylation of Ci by protein kinase A (PKA) as a key event in mediating cleavage (Chen et al., 1998; Chen Y. et al., 1999; Price and Kalderon, 1999; Wang and Holmgren, 1999; Wang et al., 1999). Elegant genetic experiments have indicated that the serine-threonine kinase Fu plays a role in the conversion of Ci into a transcriptional activator (Ohlmeyer and Kalderon, 1998).

Our previous studies focussed on the demonstration that the formation of both Ci[act] and Ci[rep] is tightly controlled by Hh signaling (Méthot and Basler, 1999). Here we examined Hh signal transduction components for their role in the conversion of full-length Ci into either Ci[act] or Ci[rep]. We found that the formation of Ci[rep] requires PKA, Cos2 and Fu, while the generation of Ci[act] proceeds through the neutralization of PKA and Cos2 activity. Although Ci can bypass PKA and gain constitutive Ci[act] activity by mutations in its PKA phosphorylation sites, we have observed that this activity can be further stimulated by Hh signaling. Both Su(fu) and Fu alter the transcriptional output of this mutant form of Ci, by regulating its nuclear-cytoplasmic localization. The accumulation of full-length Ci in the nucleus is Hh-dependent and is blocked by excess Su(fu). We propose that Fu kinase stimulates the Hh pathway, not by promoting the formation of Ci[act], but rather by facilitating its entry into the nucleus.

Ectopic expression of Ci, Ci mutants, mycSu(fu), mycFu and mycCos2 in wing imaginal discs was achieved with the UAS-Gal4 system (Brand and Perrimon, 1993). UAS-Ci and UAS-Ci^PKA (in this paper referred to as Ci^PKA1), where Ci codons for amino acids 837, 838 and 839 were mutated to encode alanine residues, have been described in Méthot and Basler (1999). The cDNA for Ci^PKA4 is a gift from C. Dahmann and encodes a mutant protein in which the following serine residues are replaced by alanines: S838, S855, S856, S891, S892 and S962. The cDNA for CiGFP was made by fusing an EGFP (Clontech) cDNA in frame to the 3′ end of Ci cDNA. GFPCi75 DNA was obtained by fusing EGFP DNA to the 5′ end of Ci cDNA that had been truncated at the unique EcoRV site. All constructs that were obtained following a PCR step or blunt ending of restriction sites were sequenced to verify integrity. DNA templates used for in vitro translations were obtained from Ci cDNA coding for a triple HA tag at its 5′ end (Méthot and Basler, 1999), and truncated at its unique EcoRV site, to generate pBS-Ci⁴⁴⁰. The in vitro translated product derived from this construct consists of a triple HA tag followed in frame by Ci amino acids 5 to 440. Deletions (Δ212-268 and Δ268-346) were generated by PCR mutagenesis and introduced into pBS-Ci⁴⁴⁰. Fused and Su(fu) cDNAs were amplified from Drosophila embryonic cDNA and cloned in frame 5′ to a myc tag. The resulting vectors [pBS-mycFu and pBS-mycSu(fu)] were transferred as Asp718-XbaI fragments into pUAST. A construct coding for GST-mycSu(fu) was derived from pBS-mycSu(fu), where an EcoRI-blunted Asp718 fragment was transferred into pGEX3X (Pharmacia) digested with BamHI, blunted and redigested with EcoRI. Cos2 cDNA (gift of K. Ho and M. Scott) was cloned in frame 3′ to a myc-tag, and transferred to pUAST.

Reporter genes used in this study were dpp^P10638 (Zecca et al., 1995), hh^P30 (Lee et al., 1992), ptc^(10.8L)A (Chen and Struhl, 1996). Gal4 drivers used were C765 (Nellen et al., 1996), act5c>CD2>Gal4 (Pignoni and Zipursky, 1997) and apterous-Gal4 (Calleja et al., 1996).

pGEX-mycSu(fu) was transformed into E. coli BL21. Bacteria were grown to an optical density of 0.5 to 0.8 before induction with 1 mM isopropyl-1-thio-β-D-galactopiranosid (IPTG). Cells were grown for 3 hours at 30°C before harvesting by centrifugation. The pellet was resuspended in lysis buffer (phosphate-buffered saline, 1 mM EDTA, 1 mM DTT and 1 mM phenylmethylsulfonyl fluoride). Cells were lysed on ice by five sonication cycles. Cellular debris was removed by centrifugation. GST-fusion proteins were purified on glutathione-agarose (Pharmacia). Ci deletion mutants (as specified in the figure legend) were labeled with [³⁵S]methionine using the T7 TNT in vitro translation system (Promega). Translation reactions (20 μl) were diluted to 300 μl in binding buffer (20 mM Hepes pH 7.4, 200 mM KCl, 1 mM DTT, 1 mM MgCl₂, 1% NP40, 10% glycerol) and incubated for 1 hour at room temperature with 1 μg of GST-fusion protein. Glutathione-agarose beads were added, and the mixture was incubated for an additional 15 minutes on a rotator. Beads were washed five times in binding buffer, before loading on an SDS-polyacrylamide gel. Gels were fixed, treated for fluorography and exposed overnight on Kodak XAR-5 film.

Alleles used in this study were: smo¹, smo³ (Nüsslein-Volhard et al., 1984); DCO^E95 (Lane and Kalderon, 1993); slimb¹ (Jiang and Struhl, 1998); cos2⁵ (Forbes et al., 1993); fu¹, fu^A, fu^mH63 and Su(fu)^LP (Préat et al., 1993). Clones of mutant cells were generated by Flp-mediated mitotic recombination (Xu and Rubin, 1993). Larvae were heat shocked in late first instar for 30 minutes at 35°C and heat shocked again in late 3^rd instar phase to induce marker gene expression (hsp70-GFP or hsp70-CD2). Following a 1 hour recovery period, larvae were dissected, discs were fixed and stained with the appropriate antibodies. The following antibodies were used: rat monoclonal anti-Ci (gift from R. Holmgren); mouse monoclonal anti-Ptc (gift from I. Guerrero); mouse monoclonal and rabbit polyclonal anti-GFP (Clonetech); mouse monoclonal anti-CD2 OX34 (Serotech); mouse monoclonal (Promega) and rabbit polyclonal (Cappel) anti-βGal; mouse monoclonal anti-myc 9E10; Alexa 488 and 594 secondary antibodies (Molecular Probes).

Presence of Ci[rep] was assayed genetically in vivo by its effect on hh-lacZ transcription. The final genotype of larvae were the following:

Fig. 1A: y w hsp70-flp; smo³FRT39/hsp70-CD2 FRT39; UAS-Ci/ C765 hh-lacZ

Fig. 1B: y w hsp70-flp; smo³DCO^E95FRT39/hsp70-CD2 FRT39; UAS-Ci/ C765 hh-lacZ

Fig. 1C: y w hsp70-flp; smo¹FRT42 cos2⁵/smo³FRT42 P[smo⁺, hsp70-GFP]; UAS-Ci/ C765 hh-lacZ.

Fig. 1D: y w f fu^A; smo³FRT39/hsp70-CD2 FRT39; UAS-Ci hh-lacZ/ C765 hsp70-flp

Fig. 3D: y w hsp70-flp; smo³FRT39/hsp70-CD2 FRT39; UAS-Ci^PKA4/ C765 hh-lacZ.

To determine the Hh-dependency of Ci^PKA4, larvae with the following genotype were analyzed:

Fig. 3C: y w hsp70-flp; smo³FRT39 ptc-lacZ/hsp70-CD2 FRT39; UAS-Ci^PKA4/ C765.

Other marked clones:

Fig. 2A,B: cos2⁵: y w hsp70-flp; dpp-lacZ FRT42 hsp70-CD2/ FRT42 cos25

Fig. 2C: PKA^E95: y w hsp70-flp; FRT40 DCO^E95/ 2x hsp70-πmyc FRT40 ptc-lacZ

Fig. 2D fu^A: y w f fu^AFRT19/ y w hsp70-GFP FRT19; ptc-lacZ /

+; hsp70-flp.

Clones expressing mycFu, mycSu(fu) and mycFuΔN were produced by mild heat shock (30 minutes at 34°C) of larvae containing the respective UAS construct and the act5c>CD2>Gal4 transgene (Pignoni and Zipursky, 1997). To inhibit nuclear export, discs from 3^rd instar larvae were dissected and cultured for 8 hours in C8 media (Bhanot et al., 1996) in the presence or absence of 30 nM leptomycin B.

Ci functions as a potent repressor of hh expression in the absence of Hh signal transduction (Aza-Blanc et al., 1997; Méthot and Basler, 1999). Posterior cells that ectopically express Ci transduce the Hh signal and do not form Ci[rep] due to their exposure to the Hh signal. Removal of Smo function, however, prevents such P cells from transducing Hh and results in the formation of Ci[rep] and inhibition of hh transcription (Fig. 1A). To determine which components of the Hh signal transduction cascade are required for the generation of Ci[rep], we expressed Ci in P cells and removed the function of Smo together with that of either Cos2 or PKA. In contrast to smo⁻ single mutant clones, hh transcription was not reduced in smo⁻PKA⁻ (Fig. 1B) or smo⁻cos2⁻ (Fig. 1C) double mutant clones. Hence, the activities of PKA and Cos2 are required for Ci[rep] formation.

The behavior of certain fu alleles suggests that Fu, in addition to its stimulatory function, also plays an inhibitory role in the Hh pathway (Préat et al., 1993). Two types of fu alleles exist. Molecular analysis of these two classes has shown that fu type I alleles bear mutations in the kinase domain, which are thought to reduce or abolish kinase activity, while fu type II alleles lack the C terminus of the protein (Thérond et al., 1996). We assayed Ci[rep] formation in posterior smo⁻ cells that expressed Ci and were either mutant for a fu type I (fu¹) or a type II (fu^A) allele. hh transcription was eliminated in smo³ clones of fu type I discs (data not shown), but not of fu type II discs (Fig. 1D). Thus, the catalytic function of Fu is dispensible for Ci[rep] formation, whereas the function of the C terminus is required for Ci[rep].

We found that absence of Cos2 or PKA activity prevents formation of Ci[rep]. It has been shown earlier that mere prevention of Ci cleavage does not suffice for Ci[act] formation (Méthot and Basler, 1999). To test whether, in these mutant contexts, prevention of Ci[rep] formation is linked to the generation of Ci[act], we removed PKA or Cos2 and assessed Ci[act] activity. While the repression of hh transcription is indicative for the presence of Ci[rep], the upregulation of ptc transcription serves as an indicator for the presence of Ci[act] (Méthot and Basler, 1999). We examined ptc expression in A cells mutant for cos2 or PKA and found, in both cases, ptc to be upregulated (Fig. 2B,C). These results indicate that A cells mutant for cos2 or PKA generate Ci[act], and suggest that neutralization of the activities or effects of either Cos2 or PKA is an important step for the formation of Ci[act]. Interestingly, although the C terminus of Fu is required for Ci[rep] formation, its absence did not lead to ectopic ptc-lacZ expression and Ci[act] formation (Fig. 2D).

Ci[act] is generated in cells that lack PKA activity (Li et al., 1995; Jiang and Struhl, 1995; this work). An equivalent situation can be created by mutating the PKA phosphorylation sites of Ci. We mutated one (Ci^PKA1) or four (Ci^PKA4) PKA phosphorylation sites and compared the ability of these mutants to activate ptc-lacZ in wing imaginal discs. Ubiquitous weak expression of wild-type Ci led to ptc-lacZ activation only in Hh-exposed cells (Fig. 3A). This indicates that under physiological conditions, transcriptional activity of Ci is under the control of Hh. In contrast, Ci^PKA1 activated ptc-lacZ in all cells, regardless of their exposure to Hh (Fig. 3B). Thus, Ci^PKA1 is constitutively active in vivo. Identical results were recently described by several groups, and are taken as indication for a role of PKA in Ci activation (Chen Y. et al., 1999; Price and Kalderon, 1999, Wang et al., 1999). Interestingly, closer examination of discs ubiquitously expressing Ci^PKA1 reveals that P cells express ptc-lacZ at higher levels than A cells (Figs 3B, 4I). This suggests that the transcriptional activity of Ci^PKA1 can be further enhanced by Hh signaling. We have obtained similar results with Ci^PKA4 (Figs 3C, 4C). To test whether it is the reception of the Hh signal that stimulates the basal activity of Ci^PKA, we removed the function of Smo from a subset of P cells expressing Ci^PKA4. Such cells transcribed ptc-lacZ at lower levels than their smo⁺ neighbors, levels that were similar to those found in anterior smo⁺ Ci^PKA cells (Fig. 3C). Thus Ci protein that is constitutively active due to mutated PKA phosphorylation sites is further stimulated in its transcriptional activity by the Hh signal.

It is possible that the reduction of ptc-lacZ expression in posterior smo⁻ Ci^PKA4 cells is due to formation of some Ci[rep], which could compete with the Ci[act] activity of Ci^PKA. We used the repressor assay described above to test whether Ci^PKA4 can be converted to a transcriptional repressor, and found that hh transcription in smo⁻ posterior cells was essentially unaffected (Fig. 3D). This indicates that Ci^PKA4 cannot be converted to Ci[rep] and that the reduction of ptc expression in smo⁻ Ci^PKA4 cells is not due to the presence of Ci[rep].

Next, we searched for components of the Hh pathway that could stimulate the basal activity of Ci^PKA. Fu is one of the few proteins that have a positive input on Hh signaling (Préat et al., 1990, Ohlmeyer and Kalderon, 1998; Thérond et al., 1999). We examined ptc-lacZ levels in wing imaginal discs that ubiquitously expressed Ci^PKA1, in either a wild-type or fu¹ background. ptc-lacZ expression was clearly reduced in P cells of fu¹ discs (compare Fig. 4A and B). Similar results were also obtained with Ci^PKA4 (Fig. 4I,J). Slightly elevated ptc-lacZ could still be seen near the AP compartment boundary in fu¹ discs, and may be the result of cumulative activities of endogenous Ci[act] and Ci^PKA. We conclude that Fu kinase enhances the basal activity of Ci^PKA.

fu is tightly linked to Su(fu), both genetically and biochemically (Préat, 1992; Préat et al., 1993; Monnier et al., 1998). To test whether the modulation of Ci^PKA activity involves Su(fu), we ubiquitously expressed Ci^PKA4 together with myc-tagged Su(fu) or GFP as negative control. Ci^PKA4 (in the presence of GFP) induced ptc-lacZ expression everywhere in the wing imaginal disc but at higher levels in the P compartment (Fig. 4C). Co-expression with mycSu(fu) abolished ptc-lacZ expression in the A compartment and reduced ptc-lacZ levels in P cells (Fig. 4D). Thus Su(fu) inhibits the activity of Ci^PKA4. This results is strengthened by the converse experiment, where the absence of Su(fu) (in Su(fu)^LP homozygous animals) reduced the difference in ptc-lacZ levels between A and P Ci^PKA1-expressing cells (Fig. 4E).

To determine whether Su(fu) negatively acts on Ci^PKA4 by direct protein-protein interaction, we created a mutant form of Ci with impaired Su(fu) binding. Su(fu) interacts with Ci within a region that encompasses amino acids 244-346 (Monnier et al., 1998). Indeed, an N-terminal fragment of Ci (amino acids 5-440) interacts with GST-Su(fu) (Fig. 4F, lane 5). A deletion removing amino acids 212-268 of Ci almost abrogated Su(fu) binding to an N-terminal in vitro translated product of Ci (Fig. 4F, lanes 6 and 7). Removal of amino acids 268 to 346 also reduced Su(fu) binding, but to a lesser extent (lanes 8 and 9). The Δ212-268 deletion was introduced into Ci^PKA4, to create CiΔN^PKA4. This mutant was constitutively active, with P cells expressing higher ptc-lacZ levels than A cells (Fig. 4K). The activity of CiΔN^PKA4 was slightly reduced when introduced into a strong fu background (fu^mH63; compare Fig. 4K and L), but the reduction was much less pronounced compared to that observed for Ci^PKA4 (compare Fig. 4I and J). We conclude that inhibition of Su(fu) activity by Fu kinase is an important step towards stimulating the basal activity of Ci^PKA4 (and by analogy Ci[act]).

Regulation of Ci nuclear accumulation may be an important control point in the Hh signaling cascade. It has been shown recently that human GLI1 and Ci accumulate in the nuclei of tissue culture cells treated with the nuclear export inhibitor drug leptomycin B (LMB; Kogerman et al., 1999; Chen, C.-H. et al., 1999). To test if LMB has similar effects in vivo, Ci tagged at its C terminus with GFP (CiGFP) was expressed in the dorsal compartment of wing imaginal discs using an apterous-Gal4 driver (Calleja et al., 1996). Nuclear β-galactosidase derived from the ptc-lacZ reporter was used as a marker for nuclei and CiGFP activity. Under these conditions, CiGFP induced ptc-lacZ in P compartment cells only, indicating that the protein is active and under the control of Hh signaling (Fig. 5A, inset). A higher magnification shows that GFP fluorescence was largely excluded from the nuclei of untreated anterior cells, whereas some GFP fluorescence was observed in some nuclei of posterior cells (Fig. 5A, green). Treatment of wing discs with LMB resulted in a dramatic accumulation of GFP fluorescence in the nuclei of P cells or A cells close to the Hh source. A weaker accumulation can also be observed in A cells far from the Hh source (Fig. 5B). Several conclusions can be drawn from these results. In the absence of Hh, some Ci must enter the nucleus, as accumulation in A cells is seen in the presence of LMB. It is probably exported rapidly, since no nuclear Ci is detected without LMB treatment. In the presence of Hh, more Ci enters the nucleus, but it is also rapidly exported, explaining why little nuclear accumulation is seen in untreated P cells.

Next, we studied the effect of LMB on the subcellular localization of Fu and Su(fu). dpp-lacZ was used as a marker for nuclei and the AP compartment boundary, and mycSu(fu) and mycFu were expressed in cell clones using an act5c>CD2>Gal4 transgene (Pignoni and Zipursky, 1997). Su(fu) was largely confined to the cytoplasm of untreated or LMB-treated wing imaginal discs (Fig. 5C,D). Fu was also cytoplasmic in untreated cells (Fig. 5E), but showed some accumulation in the nuclei of LMB treated cells (Fig. 5F). In contrast to Ci, the quantity of nuclear Fu did not depend on Hh exposure (data not shown). The subcellular localization of myc-tagged Cos2 was also examined. mycCos2 was mainly cytoplasmic in untreated imaginal disc cells (Fig. 5G), but partly nuclear in LMB-treated cells (Fig. 5H). As with Fu, no difference in nuclear accumulation between A and P cells could be detected. Thus, Fu and Cos2 can enter the nucleus, but the rate of entry does not change with Hh signaling.

A possible mechanism by which Fu stimulates and Su(fu) counteracts Ci[act] could be the promotion or impediment of nuclear Ci[act] accumulation, respectively. We compared the subcellular distribution of Ci^PKA in cells expressing or lacking Su(fu). Strikingly, we found that, while wild-type cells show a cytoplasmic distribution of Ci^PKA4, this protein is mostly nuclear in Su(fu) mutant salivary gland cells (Fig. 4G,H). Identical results were obtained with wild-type Ci fused N terminally to GFP (data not shown). This suggests that Su(fu) influences the nuclear localization of Ci. We further tested the effect of Su(fu) on Ci localization by overexpressing Su(fu) with a GFP-tagged form of Ci75, which has been shown to be mainly nuclear (Aza-Blanc et al., 1997; Hepker et al., 1997, Fig. 6A). Expression of mycSu(fu) reduced the amount of GFPCi75 that accumulates in the nucleus (Fig. 6B). We also performed these experiments with full-length Ci (CiGFP). As shown before, CiGFP accumulates in LMB-treated nuclei, particularly in those exposed to Hh (Fig. 6C). Co-expression of Su(fu) in LMB-treated discs reduced the amount of nuclear CiGFP, both in A and P cells (Fig. 6D). We conclude that Su(fu) downregulates the Hh pathway by preventing nuclear accumulation of Ci[act].

Ci[act] and Ci[rep] are two distinct forms of Ci whose generation is strictly under the control of Hh. The mechanism by which full-length Ci is converted into Ci[act] is unknown, as are the precise events leading to the formation of Ci[rep]. Using the Ci target genes ptc, dpp and hh as diagnostic tools for the presence of Ci[act] or Ci[rep], we demonstrate that PKA and Cos2 prevent Ci[act] formation and promote the generation of Ci[rep]. The structural integrity of Fu also contributes to Ci[rep] formation. Loss of PKA activity leads to the appearance of Ci[act], an effect that can be mimicked by mutating PKA phosphorylation sites of Ci. The key finding of our present study is the discovery that, although such mutant forms of Ci are constitutively active, their activity can be further stimulated by Hh signaling. Our results indicate that this effect is mediated by Fu and counteracted by Su(fu). We suggest that Fu, rather than being involved in Ci[act] formation per se, stimulates the Hh pathway by permitting nuclear accumulation of Ci[act]. Below we describe arguments that support this proposal.

Su(fu) is not involved in the formation of Ci[rep] or Ci[act], as no ectopic dpp or ptc expression was detected in Su(fu)^LP clones (data not shown). Indeed, Ohlmeyer and Kalderon (1998) report no change in the amount of Ci75 in Su(fu)^LP discs. Rather, Su(fu) appears to restrict the activities of Ci[act]. This is evident from the observation that Su(fu) overexpression substantially curbs the transcriptional activity of constitutively active Ci^PKA, and is suggestive of Su(fu) acting after Ci[act] formation. There are several ways by which Su(fu) could fulfill such a role. One possibility is that it impedes entry of Ci[act] into the nucleus. Alternatively, Su(fu) might promote nuclear export of Ci[act]. It is difficult to distinguish between these two possibilities. The observation that little Su(fu) accumulated in the nuclei of LMB-treated cells suggests that Su(fu) functions primarily in the cytoplasm and hence might exert a negative effect on Ci[act] by preventing its nuclear entry. It cannot be excluded, however, that a minor fraction of Su(fu) negatively affects the activity, stability or localisation of Ci[act] in the nucleus.

Fu, as the main regulator of Su(fu) activity, is also controlled by Hh. In fu¹ discs, Ci^PKA expression leads to similar levels of ptc transcription in A and P cells but, in fu⁺ discs, Ci^PKA expression causes higher ptc levels in P cells. In other words, Fu enhances Ci^PKA activity only in Hh-exposed cells. From this, it can be concluded that Fu activity is subject to Hh control.

One puzzling aspect regarding Su(fu) is that it is dispensable for viability. Animals that lack Su(fu) protein do not exhibit Hh-independent Ci[act] activity. This paradox can be partly explained by viewing Su(fu) only as a partial inhibitor of Ci[act] activity, which exerts its function subsequent to Ci[act] formation. Other elements, discussed below, ensure tight control over the generation of Ci[act].

The problem of how full-length Ci protein is converted into Ci[act] is more challenging. Fu has been implicated in this process (Ohlmeyer and Kalderon, 1998; this work), but as in the case of Su(fu), Fu kinase activity is partially dispensable in wild-type discs and entirely dispensable in animals lacking Su(fu) (Préat, 1992; Préat et al., 1993). This suggests that the Fu kinase functions only to prevent Su(fu) from negatively acting on the Hh pathway. If we accept that Su(fu) acts subsequent to Ci[act] formation, we must conclude that the same is true for the Fu kinase. In short, we propose that the activity of the Fu kinase is only required to maximize the output of an already activated form of Ci, for example by opposing cytoplasmic tethering of Ci[act] by Su(fu). The precise mechanism of how these components act is not understood. No substrate for the Fu kinase has been identified and the significance of nuclear Fu protein is unclear.

Slimb, Cos2, PKA and Fu are required for the conversion of full-length Ci into the repressor form. Basal levels of dpp are transcribed in anterior slimb¹ clones, indicating absence of Ci[rep] in these cells. No Ci[rep] could be detected by our repressor assay in cells mutant for cos2 or PKA. Moreover, we found that an intact C terminus of Fu is required to generate Ci[rep], whereas Fu kinase activity is dispensable.

slimb¹ mutant cells do not ectopically express ptc (N.M. and K.B., unpublished). Similar results were recently reported by Wang et al. (1999), who have also used a slimb null allele. Thus Slimb seems to play no role in the generation of Ci[act]. PKA and Cos2, in contrast, appear to be components dedicated to prevent Ci[act] formation, as cells lacking these proteins ectopically express ptc. This suggests that some of the events leading to Ci[act] include the neutralization of PKA and Cos2 activities. PKA directly phosphorylates Ci in vitro and mutations at the PKA phosphorylation sites of Ci have a profound effect on Ci activity (Chen et al., 1998; Chen Y. et al., 1999; Wang et al., 1999; Price and Kalderon, 1999; this study). Since no change in catalytic activity has been reported for PKA upon Hh signaling (Ohlmeyer and Kalderon, 1997), it is thought that other events, such as protection of Ci from phosphorylation by PKA, or dephosphorylation of Ci, may be involved in activation. Indeed, Chen C.-H. et al. (1999) provide pharmacological evidence for specific phosphatases that modulate the phosphorylation state of Ci.

We have shown that PKA and Cos2 prevent Ci[act] formation and that the same components are required for Ci[rep] formation. This observation closely links the two events. Cos2, Fu and Ci are found in a large cytoplasmic complex that is associated with microtubules (Sisson et al., 1997; Robbins et al., 1997). Fu derived from type II alleles, lacking the C-terminal portion, fails to locate to this complex (Robbins et al., 1997). Indeed, Ci[rep] is not generated in cells expressing only type II-mutant Fu protein. In addition, exposure to Hh releases Cos2 from microtubules (Robbins et al., 1997). This links Ci[rep] formation to complex formation. Together, these findings lead us to surmise that complex formation fulfills two roles: one is to tether Ci to microtubules, thereby preventing nuclear entry. The other is to localize Ci to the site of proteolytic processing for the formation of Ci[rep]. Hh signaling would promote the formation of Ci[act] by releasing this complex (or Ci) from microtubules, and as a consequence would prevent the cleavage of Ci. Upon release, Ci[act] would be subjected to Su(fu) action, possibly by cytoplasmic tethering. Stimulation of Fu kinase activity by Hh inhibits Su(fu) and enables nuclear accumulation of Ci[act] (Fig. 7). This model is consistent with the genetic data on Fu and Su(fu). Full-length Ci accumulates in fu^A (type II mutant) clones (N.M. and K.B., unpublished, Alves et al., 1998) since it is not cleaved. Hh target genes are ectopically expressed in fu^ASu(fu)^LP cells (Préat et al., 1993) because complex formation is compromised, Ci[act] arises without Hh input and accumulates in the nucleus in the absence of Su(fu). A challenging question to be answered is whether the Hh-dependent events are all catalyzed by a single biochemical step.

We would like to thank the following persons for generous gifts: I. Guerrero and R. Holmgren for anti-Ptc and anti-Ci antibodies, respectively, D. Kalderon for fu^mH63, Su(fu)^LP and P[fu⁺] stocks, B. Limbourg-Bouchon and C. Lamour-Isnard for fu¹ and fu^A mutants, K. Ho and M. Scott for the cos2 cDNA, C. Dahmann for the UAS-Ci^PKA4 construct and lines, C. von Mering for supplying Fig. 1A, M. Yoshida for leptomycin B, P. Zipperlen and J. Berger for sequencing. We also thank E. Hafen, M. Levine and C. Dahmann for comments on the manuscript, and members of the laboratory for helpful discussions. N.

M. was the recipient of a Human Frontier Science Programme fellowship. This work was supported by the Swiss National Science Foundation and the Kanton of Zürich.