Nuclear import of Cubitus interruptus is regulated by Hedgehog via a mechanism distinct from Ci stabilization and Ci activation

hedgehog (hh) and its vertebrate homologues, most importantly Sonic hedgehog (Shh), play critical roles in development. In vertebrates, members of the hh family function in various developmental processes ranging from neural tube patterning to hair follicle formation (Echelard et al., 1993; St-Jacques et al., 1998). In Drosophila, hh is required for embryonic segmentation, for CNS development, and for the patterning of the primordia of adult appendages, the imaginal discs.

The wing imaginal disc is an ideal system for studying the Hh signaling pathway. Cells in a wing disc are subdivided into two adjacent populations that do not intermingle, the anterior and posterior “compartments” (Garcia-Bellido et al., 1973; Lawrence and Morata, 1976). hh is exclusively expressed in posterior compartment cells, and the Hh protein traverses the compartment boundary to regulate gene expression in the anterior compartment (Lee et al., 1992; Tabata et al., 1992; Tabata and Kornberg, 1994). hh has at least two major roles in the anteroposterior (AP) patterning of the wing: a long-range, indirect activity (Zecca et al., 1995) and a short-range, direct activity (Muller et al., 1997; Strigini and Cohen, 1997). The long-range activity of hh is mediated by one of its target genes, decapentaplegic (dpp), a member of the transforming growth factor β (TGFβ) family (reviewed by Gelbart, 1989). The short-range activity is independent of dpp and is essential for correct AP patterning of the medial section of the wing.

The Hh signal transduction pathway is initiated when Hh binds to its receptor Patched (Ptc), which both transduces and sequesters the Hh signal (Chen and Struhl, 1996). Ptc forms a complex with Smoothened (Smo), a serpentine transmembrane protein with features of G-protein-coupled receptors (Alcedo et al., 1996; van den Heuval and Ingham, 1996), and represses Smo activity in the absence of Hh binding (Chen and Struhl, 1998). Upon Hh binding, the repression of Smo is relieved by an as yet unknown mechanism. Through the action of a number of downstream molecules including the kinesin-like protein, Costal-2 (Cos2) (Robbins et al., 1997; Sisson et al., 1997), the serine/threonine kinase, Fused (Fu) (Sanchez-Herrero et al., 1996; Alves et al., 1998), protein kinase A (PKA) (Jiang and Struhl, 1995; Johnson et al., 1995; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995) and the PEST protein, Suppressor of fused [Su(fu)] (Ohlmeyer and Kalderon, 1998), the Hh signal modulates the activity of the zinc-finger transcription factor, Cubitus interruptus (Ci) (Alexandre et al., 1996; Dominguez et al., 1996; Hepker et al., 1997), thereby regulating the expression of its target genes.

Ci is the Drosophila homologue of the vertebrate Gli proteins (Orenic et al., 1990) and the only identified transcription factor in the Hh signaling pathway. Given the critical role that Ci plays in the expression of Hh target genes, the mechanisms by which the Hh signal regulates Ci have been the subject of intense study, and significant progress has been made in the past several years. In the wing disc, ci is expressed throughout the anterior compartment (Motzny and Holmgren, 1995). The full-length Ci protein is proteolytically cleaved to generate a 75 kDa repressor form (Ci-75) in cells away from the compartment boundary, where the Hh signal is absent or of very low concentration (Aza-Blanc et al., 1997). The F-box/WD40-repeat protein, Slimb, a relative of yeast Cdc4p, which targets cell-cycle regulators for ubiquitination and subsequent proteasome degradation, is required for Ci cleavage (Jiang and Struhl, 1998). PKA and Cos2 are also required (Jiang and Struhl, 1998; Wang and Holmgren, 1999), and Ci becomes resistant to cleavage when putative PKA sites are mutated (Chen et al., 1998, 1999b; Price and Kalderon, 1999). The Hh signal stabilizes Ci by inhibiting cleavage. Hh also up-regulates the activity of Ci through a mechanism distinct from Ci stabilization (Methot and Basler, 1999, Wang et al., 1999; Wang and Holmgren, 1999), an event referred to as “Ci activation”. Finally, the amount of nuclear Ci appears to be tightly controlled through a cytoplasmic/nuclear shuttling process, and it has been proposed that Hh regulates the nuclear import of Ci (Chen et al., 1999a), which contains a functional nuclear localization signal (NLS) but is primarily cytoplasmic (Wang and Holmgren, 1999).

While Ci stabilization can be monitored by staining and by western blotting, the other two aspects of Ci regulation have been hard to follow directly, and many questions remain to be answered. In spite of the suggestion that Hh accelerates the nuclear import of Ci, Ci remains predominantly cytoplasmic in cells receiving high levels of Hh in vivo (Motzny and Holmgren, 1995), as well as in tissue culture cells subjected to Hh-N treatment (except when the cells are transfected to over-express ci) (Chen et al., 1999a). Similarly, while nuclear export has been implicated in the regulation of Ci subcellular distribution, inhibition of nuclear export in cultured cells, with or without Hh-N treatment, does not result in dramatic changes in Ci localization (Chen et al., 1999a). Here we address several questions regarding the regulation of Ci. To what extent is Ci nuclear import regulated by Hh, and what are the mechanisms that coordinate this process? What is the significance of nuclear export? Moreover, what is the relationship between Ci nucleasr import, stabilization, and activation? Although Ci stabilization alone does not suffice for Ci activation (Methot and Basler, 1999; Wang and Holmgren, 1999), could the mechanism of Ci activation be accounted for by accelerated nuclear import of stabilized Ci?

In this report we studied the mechanism of Ci nuclear import in vivo using a novel assay in which imaginal discs are cultured as whole-mount organs for short periods of time. Our results show that Hh signaling dramatically increases the rate of Ci nuclear import and that rapid nuclear export has a major role in maintaining the normal pattern of Ci subcellular distribution. The mechanism regulating Ci nuclear import is distinct and genetically separable from that regulating Ci stabilization and activation. Analysis of the roles of cos2 and pka in Ci nuclear import and Ci activation reveals dual functions for cos2 and pka in Ci regulation.

(For convenience, in this report “Ci” will refer to the full-length form of Ci, Ci-155, unless otherwise specified.)

cos2¹FRT42, DC0^H2myc FRT40A and fu^mH63, smo^D16DC0^H2FRT40A, smo^IIG26FRT40A, UAS-ci(m1-3), ptc^IN and ptc^G20 were kindly provided by T. Orenic, D. Kalderon, G. Struhl, S. Blair, R. Whittle and S, Smolik, respectively. cos2¹ is an embryonic lethal, loss-of-function allele. DC0^H2 is a null allele for the catalytic domain of protein kinase A. smo^D16 and smo^IIG26 are loss-of-function and hypomorphic alleles, respectively. ptc^IN is a null and ptc^G20 is a hypomorphic allele. ci(m1-3) encodes a form of Ci in which the PKA sites at Ser-838, Ser-856 and Ser-892 have been mutated (Chen et al., 2000). fu⁹⁴ is a strong hypomorphic allele.

ptc hypomorphic discs are of the genotype ptc^IN/ptc^G20.

Somatic recombinant clones were generated by means of the FLP FRT technique (Xu and Rubin, 1993). Larvae were heat-shocked for 1 hour at 37°C, 48-72 hours after egg laying. Genotypes of the larvae were as follows:

smo clones: y hsflp; smo^IIG26FRT40A/myc FRT40A pka clones: y hsflp; DC0^H2myc FRT40A/FRT40A

pka clones, 4bslacZ: y hsflp; DC0^H2myc FRT40A/FRT40A; 4bslacZ/+

pka smo clones: y hsflp; smo^D16DC0^H2FRT40A/myc FRT40A cos2 clones: y hsflp; cos2¹FRT42/myc FRT42

cos2 clones, 4bslacZ: y hsflp; cos2¹FRT42/myc FRT42; 4bslacZ/+

Imaginal discs were examined by deconvoluting microscopy. Antibodies used were as follows: anti-Ci, rat monoclonal 2A1 (1:2); anti-β-galactosidase, rabbit polyclonal (1:2000); anti-Ptc, mouse monoclonal ascites (gift from I. Guerrero) (1:1000); and all fluorescent secondary antibodies (1:200, Jackson ImmunoResearch Labs).

Wing imaginal discs from mid-third instar larvae were incubated at 25°C in cl-8 cell medium for up to 3 hours prior to staining. For treatment with leptomycin B (LMB), discs were cultured in medium containing 50 ng/ml LMB (experimental discs) or an equal volume of solvent (ethanol) (control discs). For treatment with MG132, discs were cultured in medium containing 10 μM MG132 (experimental discs) or an equal volume of solvent (DMSO) (control discs).

In Drosophila wing imaginal discs, ci is expressed throughout the anterior compartment. In cells away from the boundary, the level of full-length Ci is low due to proteolytic cleavage. In cells along the AP compartment boundary, Hh signaling inhibits Ci cleavage, creating a stripe of elevated Ci. The Hh target gene dpp is expressed in a stripe roughly corresponding to the high-level Ci stripe, and ptc is expressed at high levels in the 2-3 cells immediately adjacent to the AP boundary (Fig. 1A). Ci contains a functional NLS but is primarily cytoplasmic throughout its expression domain (Fig. 1B). We examined the subcellular distribution of Ci in discs treated with leptomycin B (LMB), a potent inhibitor of protein nuclear export (Fukuda et al., 1997). Within 1 hour, LMB-treated discs exhibit nuclear accumulation of Ci in a stripe of cells along the AP boundary (Fig. 1C), suggesting that within these cells, Ci is normally rapidly shuttled between the nucleus and the cytoplasm. Ci in cells away from the boundary remains cytoplasmic, and there is no visible change in the pattern of Ci subcellular distribution in these cells even after 3 hours of treatment. In LMB-treated discs, the region of nuclear Ci correlates with that of Ci stabilization. To determine whether the absence of nuclear Ci in cells away from the compartment boundary is caused by cleavage, discs were simultaneously treated with LMB and MG132, an inhibitor of proteasome activity. In discs co-treated with LMB and MG132, Ci is stabilized throughout the anterior compartment but only accumulates in the nuclei along the compartment boundary (Fig. 1D-F). Taken together, these results show that the rate of Ci nuclear import is significantly different between cells along the boundary and those away from the boundary and furthermore, this difference is not a simple reflection of the difference in Ci protein levels. Although Ci is uniformly cytoplasmic in untreated discs, its cytoplasmic distribution is maintained by different mechanisms in different regions of the disc. On the one hand, dynamic import takes place in cells along the boundary but is masked by rapid nuclear export. On the other hand, away from the boundary Ci either does not translocate to the nucleus or translocates at a very low rate.

To test whether the nuclear import of Ci is regulated by Hh, we analyzed mutations that either mimic or block Hh signaling. In LMB-treated ptc hypomorphic wings, Ci is nuclear throughout the anterior compartment (Fig. 1G). Conversely, smo loss-of-function clones, which cannot transduce the Hh signal (Alcedo et al., 1996; Chen and Struhl, 1996; van den Heuval and Ingham, 1996), do not accumulate nuclear Ci upon LMB treatment (Fig. 1H). Therefore, the rapid nuclear import of Ci is dependent on Hh signaling.

Previous studies have shown that pka has a negative role in the expression of hh/ci target genes. Mutations in pka are associated with disc overgrowth and duplication, signs of up-regulated Ci activity. The level of full-length Ci is elevated in pka loss-of-function clones and these clones ectopically express ptc and dpp (Jiang and Struhl, 1995; Johnson et al., 1995; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995). Subsequent work demonstrated that PKA activity is required for the cleavage of Ci (Chen et al., 1998; Jiang and Struhl, 1998; Chen et al., 1999b; Price and Kalderon, 1999). We studied the role of pka in regulating Ci nuclear import by treating discs carrying pka clones with LMB. Without LMB treatment, Ci is primarily cytoplasmic in all clones (Johnson et al., 1995) as well as in all wild-type cells. After LMB treatment, Ci is nuclear within pka clones along the boundary and cytoplasmic within those away from the boundary (Fig. 2A-C), suggesting that in pka mutant cells, the regulation of Ci nuclear import is unaltered. To further demonstrate that Ci nuclear import and stabilization are genetically separated in pka clones, pka smo double mutant clones were generated along the compartment boundary. Unlike in pka single mutant clones at similar locations, Ci failed to accumulate in nuclei in the double mutant clones (compare Fig. 2D with 2C). Thus, pka clones exhibit cell-autonomous Ci stabilization but Hh-regulated Ci nuclear import.

Another molecule that negatively regulates Ci stability and hh target gene expression is cos2 (Sisson et al., 1997; Wang and Holmgren, 1999). To directly test the role of cos2 in Ci nuclear import, discs carrying cos2 clones were treated with LMB. All clones exhibit cell-autonomous nuclear accumulation of Ci regardless of the clone’s distance from the AP boundary (Fig. 2E-G). cos2 encodes a kinesin-like molecule that is part of a multi-protein complex including Ci (Robbins et al., 1997; Sisson et al., 1997; Monnier et al., 1998). Over-expression of cos2 blocks nuclear entry of Ci in tissue culture cells (Chen et al., 1999), and it has been proposed that Cos2 regulates the microtubule-association of the complex and consequently the ability of Ci to translocate to the nucleus (Robbins et al., 1997). Our results are consistent with this model.

Discs mutant for fu show signs of compromised Ci activity, including fusion between wing veins 3 and 4, lack of late en expression in the anterior compartment, and diffuse ptc expression at lower levels. Examination of these discs after LMB treatment reveals a lack of Ci nuclear accumulation (Fig. 2H), suggesting that fu function is required for rapid Ci nuclear import. It has been shown previously that overexpression of ci can rescue the wing vein phenotype in fu mutants (Alves et al., 1998). Consistent with this finding, the Ci nuclear import phenotype is also partially rescued in fu discs in which a UAS-ci transgene was over-expressed along the AP boundary via ptcGAL4 (Fig. 2J).

We have previously studied the role of cos2 in Hh signal transduction by examining the expression of ptc in cos2 clones. High level ptc expression is dependent on Hh signaling and is normally found only in the Boundary Zone, namely the 2-3 rows of cells immediately anterior to the compartment boundary. Although cos2 clones away from the boundary can ectopically induce ptc expression (Wang and Holmgren, 1999), a small number of such clones do not express ptc (data not shown). A striking variability in the expression of ptc was observed for cos2 clones abutting the boundary: approximately 50% of cos2 clones overlapping the Boundary Zone disrupt the wild-type high-level Ptc stripe (Fig. 3A-C). It is not known why cos2 activity is more critical along the boundary than away from the boundary. Nonetheless, this result suggests that cos2 is required for Ci to become fully active in the Boundary Zone in response to a high level of Hh signaling. A requirement for cos2 is more evident when we assayed an in vivo reporter of Ci activity, 4bslacZ, whose enhancer element contains only four Ci binding sites and four Scalloped binding sites (Hepker et al., 1999) and whose response to Ci exhibits a more stringent requirement for Hh signaling than that of ptc (Wang and Holmgren, 1999). Not only do cos2 clones away from the boundary fail to ectopically induce 4bslacZ, but clones abutting the boundary invariably disrupt the wild-type 4bslacZ stripe (Fig. 3D-F). Given the cell-autonomous stabilization and nuclear import of Ci in cos2 clones, we infer that the disruption of ptc and 4bslacZ expression in such clones is due to a lack of Ci activation.

The role of pka in Ci activation was assayed in the posterior compartment of discs, in which all cells are exposed to a high level of Hh ligand. An actin promoter was used to drive a low level of ci expression throughout the discs, as previously described (Wang and Holmgren, 1999). In discs with ubiquitous ci, high-level ptc expression is observed in the posterior compartment cells (P cells) but not in the anterior compartment cells (A cells) (Alexandre et al., 1996; Hepker et al., 1997; Methot and Basler, 1999). pka mutant P cells in this background do not express ptc (Fig. 4A-C), indicating that the loss of pka function disrupts Hh signal transduction and compromises maximal Ci activity. Consistent with this result, loss of pka function also disrupts 4bslacZ expression in the Boundary Zone (Fig. 4D-F).

PKA has the potential to phosphorylate many substrates. While the loss of ptc and 4bslacZ expression in pka clones could reflect a disruption of Ci activity, it could also be due to negative effects on other proteins that bind these enhancers. To distinguish between these two possibilities, we compared wild-type Ci and a PKA site-mutant form of Ci, Ci(m1-3), for their abilities to induce ectopic 4bslacZ in the posterior compartment. Wild-type Ci exhibits a sensitive response to Hh and, at a modest protein level, induces robust 4bslacZ expression in the posterior compartment but not in the anterior (Wang and Holmgren, 1999; Fig. 4G). In contrast, Ci(m1-3) shows no response to Hh and does not induce 4bslacZ in either compartment (Fig. 3I,J). Thus, the integrity of specific PKA sites within Ci is essential for it to respond to high levels of Hh and become fully active. Since pka is dispensable for the acceleration of Ci nuclear import in Hh-receiving cells (see Fig. 2C), the requirement of PKA phosphorylation for maximal Ci activity presumably reflects a requirement for Ci activation.

The predominantly cytoplasmic localization of Ci has always been intriguing because Ci has a role in the nucleus as a transcription factor. It has been speculated that Hh may play a role in regulating the nuclear import of Ci. However, there is no significant difference in the subcellular distribution of Ci between cells that receive Hh and those that do not, either in vivo (Motzny and Holmgren, 1995) or in vitro (Chen et al., 1999b). Quantitative studies in tissue culture cells show that the rate of Ci nuclear import is increased to a certain extent in the presence of Hh. However, this increase was not enough to cause visible shifts in Ci staining pattern, and accumulation of Ci can only be seen in cells that overexpress ci (Chen et al., 1999a). Importantly, the same study suggests that one of the factors regulating the subcellular distribution of Ci is nuclear export, which can be inhibited by treating the cells with LMB. In order to analyze the nuclear import of Ci in vivo, we subjected imaginal discs to whole-mount organ culture experiments, which allowed us to perform studies in different genetic backgrounds and under conditions that preserve the spatial context of Hh signaling. Because imaginal discs are sacs with only two layers of epidermal cells, drugs that inhibit cellular processes work as well in this context as on tissue culture cells. When protein export is blocked with LMB, the amount of Ci peptides that enter the nucleus can be monitored, and we observe a remarkable difference in cells close to and those away from the AP boundary. Ci accumulates in the nucleus only in cells close to the boundary (Fig. 1C), and this has been further shown to be Hh dependent, since mutations in smo prevent nuclear accumulation of Ci (Figs 1H, 2D). Unlike the situation in cl-8 cells, the rate of Ci nuclear import in the presence of Hh is very rapid in vivo, and the subcellular distribution of Ci shifts from predominantly cytoplasmic to predominantly nuclear within as short a time as half an hour of treatment. The rate difference of Ci nuclear import in cells receiving and not receiving Hh is also significant. The cytoplasmic localization of Ci is unchanged in cells away from the boundary after three hours of treatment (data not shown). From these results, we conclude that the rate of Ci nuclear import is dramatically increased by Hh signaling in vivo. The fact that tissue culture cells do not have a similarly strong response to Hh signaling may reflect the absence of certain essential molecules. It is worth pointing out that although we could not observe Ci nuclear accumulation in cells away from the boundary, in cultured cells Ci does translocate to the nucleus at a low rate in the absence of Hh.

Cytoplasmic tethering has been proposed as a mechanism for regulating the subcellular distribution of Ci, and Cos2 is thought to mediate the interaction of the Ci/Cos2/Fu/Su(fu) complex with the cytoskeleton (Robbins et al., 1997). In response to Hh signaling, the whole complex was found to dissociate from the cytoskeleton, though it is as yet unclear whether the whole complex or Ci alone translocates to the nucleus. In tissue culture cells, over-expression of ci leads to increased nuclear import, which can be blocked by co-expression of cos2 (Chen et al., 1999a). These results suggest that (1) a stoichiometric interaction between Cos2 and Ci is important for cytoplasmic tethering; and (2) the negative regulation provided by Cos2 is saturable. fu is normally required for the rapid Ci nuclear import along the AP boundary, but this requirement can also be circumvented by ci over-expression, suggesting that once the cytoplasmic tethering system is saturated, Hh signaling is no longer required for Ci nuclear import. A diagram of this model of Ci nuclear import regulation is given in Fig. 5A.

Although the domain of rapid Ci nuclear import coincides with that of Ci stabilization in wild-type discs (Fig. 1C), the two events appear to be regulated through different mechanisms. Two lines of evidence suggest that the nuclear import of Ci is regulated even under conditions in which Ci is fully stabilized. The difference in import rates between cells away from and those close to the AP boundary persists when proteasome activity is blocked and Ci protein level is uniform (Fig. 1D-F). Furthermore, while Ci is fully stabilized in all pka clones, rapid nuclear import takes place only in those clones that are close to the AP boundary and receive Hh (Fig. 2C,D). We consider the latter result significant because it represents the first example in which different aspects of Ci metabolism have been genetically separated. The identification of mutant backgrounds which disrupt one aspect of Ci regulation but not others will be particularly useful in elucidating the mechanisms of Hh signal transduction.

It has been shown that Hh signaling up-regulates the activity of Ci through a process distinct from Ci stabilization, which is itself a Hh-induced event (Methot and Basler, 1999; Wang et al., 1999; Wang and Holmgren, 1999). When Hh-regulated nuclear import is taken into consideration, the question arises whether nuclear import on its own or combined with Ci stabilization could account for the mechanism of Ci activation. A prerequisite to answering this question is to define Ci activation. The activated form of Ci has not been characterized and at present cannot be followed. What can be followed is the apparent activity of Ci, which is assayed by the expression of Ci targets, including endogenous targets such as ptc and reporter constructs such as 4bslacZ. ptc is expressed throughout the anterior compartment, but only the elevated expression in the Boundary Zone is Ci dependent (Methot and Basler, 1999). 4bslacZ is expressed only in the Boundary Zone in wild-type discs (Hepker et al., 1999; Wang and Holmgren, 1999). Notice that the domains of Ci nuclear import and stabilization extend into the anterior compartment as far as 8-9 cells from the AP boundary, but only a subset of these cells – those in the Boundary Zone – express ptc and 4bslacZ. This observation strongly suggests that in the Boundary Zone, a high level of Hh signaling elicits an activating event(s) that is distinct from Ci stabilization or nuclear import and is necessary for Ci to reach its maximal activity. The mosaic analysis of cos2 and pka mutants is consistent with such a model. In cos2 clones, Ci is stabilized and rapidly translocated into the nucleus, but the disruption of ptc and 4bslacZ expression along the boundary suggests that Ci activation is blocked in the mutant cells.

Similarly, pka clones along the boundary lose expression of 4bslacZ, even though these clones exhibit cell-autonomous Ci stabilization and nuclear import (Fig. 2C). Based on results described here and on those from previous studies (Methot and Basler, 1999; Wang et al., 1999), we conclude that Ci stabilization, nuclear import and activation are each regulated through a distinct mechanism and cannot substitute for each other.

Our data suggest a central role for cos2 in the regulation of Ci (Fig. 5A,B). It functions not only in Ci stability (Sisson et al., 1997; Wang and Holmgren, 1999), but also in Ci nuclear import and activation (Figs 2E-G, 3A-F). Studies in mammalian cells show that the cytoskeleton plays important roles in mediating fast and precise protein-protein interactions (Torres and Coates, 1999), and this is likely to be true in Drosophila as well. Given the close association of Cos2 with the cytoskeleton, it may control the association of Ci with various modulators and its intracellular movements, including its nuclear translocation. Two other components of the Hh signaling pathway, Fu and Su(fu), have been found in close association with Cos2 and Ci (Preat et al., 1993; Robbins et al., 1997; Sisson et al., 1997; Monnier et al., 1998). Thus, this complex could serve as a platform for regulating many aspects of Ci function (Fig. 5B).

Previous work has suggested that pka regulation of Ci is complex. In embryos, pka appears to have both negative and positive effects on Ci function (Ohlmeyer and Kalderon, 1997). The situation is equally complex in the imaginal discs. In loss-of-function pka clones, Ci cleavage is blocked and the Hh target genes ptc and dpp are ectopically expressed (Jiang and Struhl, 1995; Lepage et al., 1995; Li et al., 1995; Pan and Rubin, 1995). Reciprocal experiments, in which a constitutively active form of PKA is expressed, have given different results depending upon the level of expression. At modest levels, pka expression has no effect (Jiang and Struhl, 1995) while at very high levels, pka can block the expression of hh target genes (Wang et al., 1999). The results from modest level pka expression would suggest that pka activity is a necessary precondition for proper Hh regulation but does not itself mediate the Hh signal. The high level pka expression experiments could be interpreted to implicate pka as a component of the Hh signaling transduction cascade. One concern with these latter experiments is that PKA at very high levels could be expected to phosphorylate many targets and have pleiotropic effects.

Our results suggest that pka is required for two aspects of Ci function, its previously defined role in cleavage and a second role in activation (Fig. 5B). Experiments showing that 4bslacZ is not induced by a mutant form of Ci lacking three potential PKA phosphorylation sites strongly implicate the direct phosphorylation of Ci by PKA in the process of activation. These studies do not distinguish whether the role of pka in activation is instructive (as a component of the Hh signaling cascade) or permissive (establishing a necessary precondition for proper Hh regulation). Given that there is little evidence for Hh regulation of PKA activity, we at present favor the view that pka is permissive (Fig. 5B).

Having PKA regulate both the activation and cleavage of Ci may provide a check on inappropriate activation of Ci. It could be imagined that Ci proteins are occasionally activated away from the compartment boundary. Such inappropriate activation could have deleterious consequences on imaginal disc patterning. However, in the absence of Hh signaling, PKA phosphorylation of Ci will target any inappropriately activated Ci molecules to the proteasome.

The expression of hh target genes has always been used as a readout of Hh signaling and Ci regulation. The interpretation of the expression pattern, however, is not always straightforward. For example, pka clones away from the AP boundary ectopically express dpp and ptc, yet the results presented in this paper suggest that Ci in these clones is neither activated nor translocated to the nucleus at a high rate. Why, then, is Ci stabilization in pka clones sufficient to induce ectopic ptc? In addressing this question, several points need to be taken into consideration.

In pka clones, the protein level of Ci is elevated due to the lack of cleavage. Although the majority of Ci stays tethered in the cytoplasm in the absence of Hh signal, the cytoplasmic tethering sites for Ci could be saturated with some Ci molecules translocating to the nucleus where it can act as a transcriptional activator. In stainings, the nuclear import of these molecules would be masked by the preponderance of cytoplasmic Ci.

Secondly, the expression of a gene is affected by its enhancer structure and all the proteins that interact with the enhancer. A 400 bp element from the dpp enhancer harbors binding sites for multiple factors, and activation of this element has been shown to depend upon the synergistic action of Ci with other DNA-binding proteins (Hepker et al., 1999). The whole dpp disc enhancer is 10 kb in length and probably contains many such elements. The ptc and en enhancers are likely to be equally complex. The behavior of these enhancers is certainly not due to the action of Ci alone, and it can be imagined that mutation in pka may affect the activity of factors involved in their regulation. Our results with 4bslacZ are the most straightforward and this probably reflects the simplicity of this artificial enhancer.

Finally, it is likely that what we call “stabilized”, “activated” and “nuclear imported” Ci each consists of a variety of isoforms with distinct properties. The overall activity of these molecules is not always easy to predict. For instance, Ci is stabilized in both pka and slimb clones, but is barely active in the latter (Wang et al., 1999). A better understanding of the regulation of Ci awaits further genetic, biochemical and cell biological studies.

We thank I. Guerrero, D. Kalderon, G. Struhl, R. Whittle, S. Blair and T. Orenic for reagents; T. Orenic, H. Dierick, R. Hays, S. Li, J. Chapin and M. Lefers for critically reading the manuscript. We are particularly grateful to M. Yoshida for generously providing LMB; S. M. Smolik for sharing UAS-ci(m1-3) before publication; S. Blair for comments and G. Struhl for insightful discussions. This work is supported by NIH grant GM057450 and the Robert H. Lurie Comprehensive Cancer Center.