Mutants of cubitus interruptus that are independent of PKA regulation are independent of hedgehog signaling

hedgehog (hh) and wingless (wg) are two segment polarity genes whose initial expression is triggered by pair-rule gene products and maintained by segment polarity gene products. The initial wg and hh expression is activated by the concerted function of zygotic gene products in late cellular blastoderm (DiNardo et al., 1988; Lee et al., 1992; Tabata et al., 1992). Segment polarity gene products, including wg and hh themselves, maintain wg and hh gene expression and a mutually dependent regulatory relationship between wg and hh develops during germband extension (Martinez Arias et al., 1988; DiNardo et al., 1988).

The hh signal transduction cascade is one of the best studied signaling pathways involved in embryogenesis and disc development. The current model proposes that HH is secreted from the posterior compartment of each segment, binds its postulated receptor, Patched (PTC) (Johnson et al., 1995; Marigo et al., 1996), and relieves the PTC inhibition of Smoothened (SMO) activity (Alcedo et al., 1996; Heuvel and Ingham, 1996). The smo pathway then activates the transcription factor CI resulting in the activation of hh target gene expression. Genetic studies have shown that PKA and costal-2 (cos-2) negatively regulate this pathway (Alcedo et al., 1996; Heuvel and Ingham, 1996; Jiang and Struhl, 1995; Lepage et al., 1995; Li et al., 1995; Ohlmeyer and Kalderon, 1997; Pan and Rubin, 1995) and that fu has a positive effect (Forbes et al., 1993). Subsequent studies suggest that hh signaling stimulates the fu kinase activity that causes the maturation of full-length CI into a transient transcriptional activator (Ohlmeyer and Kalderon, 1998). This work also suggests that the role of Su(fu) is to antagonize this process. wg is one of the hh target genes. Upon secretion from the anterior compartment, WG acts posteriorly to maintain hh expression; thus hh and wg form a positive feedback loop around the anterior-posterior (A/P) boundary of the segment, reinforcing each other’s expression and orchestrating the patterning events after gastrulation (DiNardo et al., 1988).

Genetic and biochemical experiments have shown that CI is the transcription factor responsible for transducing the hh signal into the nucleus (Alexandre et al., 1996; Dominguez et al., 1996; Ohlen et al., 1997). A perplexing phenomenon associated with CI as a transcription factor is that the activated, presumably full-length, form has never been detected in the nucleus in vivo, not even in the nuclei of cells located at the A/P boundary, where CI exerts its transactivation function, or in mutants (e.g.: ptc, cos2, PKA mutants) that exhibit elevated levels of CI protein and CI target-gene expression (Motzny and Holmgren, 1995; Slusarski et al., 1995). One accepted explanation for this phenomenon is that cells require undetectable or transient amounts of activated CI in the nucleus to activate target gene expression. CI exists as two forms; a 155 kDa transcriptionally active form, and a 75 kDa N-terminal proteolytic fragment, which contains the zinc finger DNA-binding domain and is a repressor of hh target gene expression (Aza-Blanc et al., 1997). Both forms are detected in a large cytosolic complex that includes COS-2, FU and additional unidentified proteins (Aza-Blanc et al., 1997; Robbins et al., 1997; Sisson et al., 1997). In the absence of a hh signal, CI is proteolysed and the N-terminal fragment of CI can be detected in the nucleus. When cells receive a HH signal, the CI complex disassociates from the microtubules, CI is not proteolysed and the activator, presumably full-length, form of CI may then activate hh target gene expression.

Recently, we have demonstrated in Kc cells that PKA exerts its negative effect on CI activity by promoting the proteolysis of the CI protein (Chen et al., 1998). In this paper, we demonstrate that a CI protein that is mutant for consensus PKA phosphorylation sites can bypass the hh regulation of wg gene expression. Furthermore, we show that the mutation of any one of the three serine-containing consensus PKA sites suppresses the proteolytic processing of CI and that PKA directly phosphorylates these sites in vitro. Our data suggest that the hh signaling cascade and PKA regulation of gene expression converge on the CI phosphorylation status in vivo.

pPac-PKA, pPac-PKI, pPac-HACI(wt), pPac-HACI(m1-4), pPac-luciferase, ADHCAT/GLI6BS and pPac-CI PKA mutants have been described previously (Chen et al., 1998). pPac-HACI(m1), pPac-HACI(m2), pPac-HACI(m3) and pPac-HACI(m4) were made by inserting hemagglutanin (HA) tag into the MluI site at the 5th amino acid in corresponding CI PKA mutant that had Ser-838, Ser-856, Ser 892 and Thr-1006 mutated to alanine, respectively, using site-directed mutagenesis kits from Promega. pUAST-HACI(wt) and pUAST-HACI(m1-4) were made by inserting the BamHI-NotI fragment containing HACI(wt) or HACI(m1-4) into the pUAST vector (Brand and Perrimon, 1993) that had been digested with BglII and NotI. The GST-CI 441-1065 (wt) and GST-CI 441-1056 (m1-4) constructs used to generate the tryptic phosphopeptide maps were made by inserting a filled, EcoRV-SpeI fragment of either CI(wt) or CI(m1-4) into the pGEX-KG vector (Pharmacia) that had been digested with EcoRI and filled. The GST/CI junctions were sequenced to determine that the coding sequences were in frame.

Kc cells were maintained, transfected and assayed for luciferase and CAT activities as described (Chen et al., 1998). 15 μg of total DNA (100 ng of pPac-luciferase, 5 μg of ADHCAT/GLI6BS reporter gene, 2 μg of pPac-CI construct, 4 μg of pPac-PKA or PKI, plus 4 μg of pPac, or various amounts of pPac-CI construct as indicated in the figures) were transfected. Cl-8 cells (kindly provided by T. Kornberg, University of California, San Francisco) were maintained as described by van Leeuwen et al. (1994). S2 cells that stably express N-terminal HH protein (N-hh) were provided by J. M. Bishop (University of California, San Francisco) and were maintained as described (Therond et al., 1996). Fly extract was prepared as described by Currie et al. (1988). For transfections, approximately 2×10⁵ cells were seeded in 24-well dishes. A total of 3 μg of DNA (20 ng of pPac-luciferase, 1 μg of ADHCAT/GLI6BS reporter gene, different amounts of pPac-CI construct, plus 2 μg of pPac) were transfected using calcium phosphate transfection kits from GIBCO/BRL. Both Kc and Cl-8 cells were allowed to recover for 24 hours in N-hh or control medium before they were assayed for luciferase and CAT activities. Protocols for immunoprecipitations and western blots are described by Chen et al. (1998). Each dose-response curve was repeated at least two times and was done in triplicate.

Recombinant GST-CI 441-1065 (wt) and GST-CI 441-1065 (m1-4) proteins (approximately 1 μg) were phosphorylated in vitro with the recombinant catalytic subunit of PKA in the presence of [γ-³²P]ATP. The reaction mixtures containing 250 μM of [γ-³²P]ATP (5000 cpm/pmole), 50 mM MOPS [pH 6.8], 50 mM NaCl, 2mM MgCl₂, 1 mM DTT, 1.0 nM PKA, were incubated for 45 minutes at 30°C. Phosphorylated GST-CI 441-1065 proteins were separated by 6% SDS-PAGE, and identified by autoradiography of the dried gel. For two-dimensional phosphopeptide mapping, the phospho-GST-CI bands were cut from the gel and two-dimensional tryptic phosphopeptide maps were performed on the extracted proteins as described (Boyle et al., 1991).

The prd-GAL4 strain RG1 (Yoffe et al., 1995) and the UAS-R* strain (Ohlmeyer and Kalderon, 1997) were kindly provided by Dan Kalderon (Columbia University). The hedgehog allele hh^IJ was obtained from the Umea Stock Center. The hh^IJ allele is a null or near amorphic allele of hh that has been used to study the activity of exogenously expressed CI in other studies (Ingham and Hidalgo, 1993). We chose this allele in order to directly compare our results with those previously reported. The hh^IJprd-GAL4 recombinant chromosome was generated using standard genetic techniques. The balancer chromosomes TM2, TM3 and MKRS that are used in these studies can be found in Lindsley and Zimm (1992). The wild-type strain used in these studies is Canton-S. Flies were cultured on standard Drosophila cornmeal-yeast-source medium in 8 oz plastic bottles or 28×95 mm plastic shell vials. All crosses were reared at 25°C.

The HACI(wt) and HACI(m1-4) pUAST constructs were injected at a concentration of 400 mg/ml into y w embryos as described (Spradling, 1986) and using 100 mg/ml pπΔ2-3 (kindly provided by J Posakony) as a helper plasmid. At least four independent transformants were generated for each HACI construct and tested for viability and expression with the prd-GAL4 line.

UAS-HACI(m1-4) flies were crossed to prd-GAL4/MKRS flies and UAS-HACI(wt or m1-4); hh^IJ/TM2 flies were crossed to hh^IJprd-GAL4/MKRS flies. The matings were allowed to proceed for 24-48 hours. Likewise, control hh^IJ/TM2 and hh^IJprd-GAL4/MKRS flies were allowed to mate for 24-48 hours. The eggs were then collected after a 16-24 hour period and aged for an additional 48 hours. The unhatched embryos were collected, dechorionated in 50% bleach, rinsed in phosphate-buffered saline (PBS) and the vitelline membranes removed in 1:1 heptane: methanol. They were washed in 100% ethanol and mounted as described (Lewis, 1978). The cuticles were visualized by dark-field optics and photographed with Tech Pan 50 film.

For the viable Canton-S embryos and those from the cross between UAS-HACI(wt) flies and prd-GAL4/MKRS flies, the flies were allowed to lay eggs for 3 hours. The eggs were collected and aged for 24 hours when they were processed as described.

The embryos were collected after a 3 hour egg lay period, then subjected to heat shock at 29°C for 5 hours to induce the expression of GAL4 activator. The immunohistochemistry of whole-mount embryos was carried out as described by Rose et al. (1997) and the in situ hybridization protocols were those of Tautz and Pfeile (1989) and Fleming et al. (1990). The double staining was performed as described above by first performing HA staining followed by in situ hybridization. The anti-HA antibody (Boehringer Mannheim) was first diluted 1:10 in 10% horse serum/PBS and preincubated with Canton-S embryos overnight at 4°C. For HA staining only, a final concentration of 1:500 was used. For in situ coupled immunostaining, a final concentration of 1:200 was incubated with embryos at room temperature for 1 hour.

The wg mRNA probe was kindly provided by M. Forte (Vollum Institute) and was used at 1:5000 dilution. The embryos were visualized by Nomarski optics and photographed with either Ektachrome 64 or Tech Pan 50 film.

We have shown previously that PKA regulates CI activity and promotes its proteolysis in tissue culture, thus providing a mechanism for the negative regulation of the hh signal transduction pathway by PKA (Chen et al., 1998). To investigate the relative importance of the four potential PKA phosphorylation sites in CI-mediated transactivation and CI processing, we mutated the serine of each site singly to an alanine and transfected the mutant constructs into Kc cells to examine the transcriptional activity and proteolytic processing of the mutant proteins. As shown in Fig. 1A, mutating any one of the first three consensus PKA phosphorylation sites increases CI transcriptional activity. The fold increase over wild-type CI activity from m1, m2 and m3 is comparable to that of the CI(m1-4) protein that is mutant for all four sites. The fourth PKA site is not important because m4 activity is equivalent to wild-type CI activity and the m4 CI protein is proteolysed to the same degree as the wild-type CI protein (Fig. 1B). When the PKA catalytic subunit is co-transfected with the CI PKA mutants, m1, m2 and m3, the same positive two-fold effect is observed as that seen for the CI(m1-4) protein (Chen et al., 1998). Like the wild-type CI, the transcriptional activity of the CI m4 mutant decreases in the presence of exogenous PKA. When PKI, a PKA inhibitor, is co-transfected with the different CI mutant constructs, we do not observe any further increase in CI activity, indicating that a mutation of any one of the first three PKA phosphorylation sites has the same effect as mutating all the sites. However, inhibition of PKA activity by PKI increases m4 activity, which is expected if the fourth site does not contribute significantly to the regulation of CI activity.

We then determined the effect of mutating single PKA sites on the proteolytic processing of the CI protein. We expected that the enhanced transcriptional activity of the mutant CI constructs was due to the inability of these proteins to be proteolysed to the 75 kDa repressor. As shown in Fig. 1B, mutating of any one of the first three PKA sites abolishes the proteolysis of CI, suggesting that the phosphorylation of all three PKA sites is required for the processing of CI. Consistent with our transfection data, mutation of the last PKA site does not affect CI proteolysis and m4 processing is regulated in the same way as wild-type CI, i.e. PKA promotes and PKI inhibits m4 proteolysis.

We have shown that PKA negatively regulates the activity of CI by promoting its proteolysis to the 75 kDa repressor form. We wanted to determine whether hh signaling along the A/P boundary of the segments regulates the activity of CI by regulating its phosphorylation state. Reports by Ohlmeyer and Kalderon (1997) suggest that hh signaling has a positive effect on wg expression that is independent of CI protein modifications. In this case, we would expect that the activity of the CI(m1-4) mutant protein would be augmented in the presence of hh signaling. If the primary function of hh signaling were to modulate CI phosphorylation, then the activity of the CI(m1-4) mutant protein would be independent of a HH signal. Experiments were performed in Cl-8 cells, which are derived from Drosophila disc cells and can respond to a HH signal (Aza-Blanc et al., 1997; Therond et al., 1996).

We transfected Cl-8 cells with CI(m1-4) and CI(wt) constructs and treated them with N-hh-conditioned media to see whether HH could affect the activity of CI that is mutant for the putative PKA phosphorylation sites. As shown in Fig. 2A, transfection of CI(wt) increases the expression of a CI-responsive reporter gene in a dose-dependent manner when grown in the presence of control S2 cell-conditioned medium. Growth in N-hh-conditioned S2 medium causes a further increase in CI transcriptional activity, indicating that CI(wt) activity is increased by HH activity in Cl-8 cells. When the CI(m1-4) construct is introduced into Cl-8 cells, the dose-dependent increase in CI activity is larger than that of CI(wt).

Furthermore, when cells transfected with CI(m1-4) are treated with N-hh-conditioned medium, we do not observe any additional increase in transcriptional activity, suggesting that the HH-mediated increase in CI activity may be, at least in part, through dephosphorylation of CI. Endogenous CI does not contribute to these activities as the reporter is not expressed significantly either in the presence or absence of HH (0 points in Fig. 2A,B)

Because the Kc cell line displays PKA-dependent proteolytic mechanisms, we determined whether the hh signaling cascade was intact in this cell line as well. As shown in Fig. 2C, N-hh-conditioned medium increases the activity of CI(wt) in Kc cells. Furthermore, the N-hh-conditioned medium inhibits the proteolysis of CI(wt) (Fig. 2C insert) suggesting that the Kc cell line can respond to the HH signal and recapitulate the HH regulation of CI transcriptional activity. When Kc cells are transfected with increasing amounts of CI(m1-4), we observe an increase in CI transcriptional activity that is independent of HH. Furthermore, this mutant protein is not proteolysed in the presence of N-hh-conditioned medium (Fig. 2D insert). Thus the regulation of exogenously expressed CI by the hh signaling pathway is qualitatively the same in both the Kc and Cl-8 cell lines.

While our experiments demonstrate that PKA cannot regulate a CI protein that is mutant for three of the consensus PKA sites (Chen et al., 1998), we did not determine that PKA directly phosphorylated these sites. To test that PKA directly targets these sites in vivo, it is necessary to follow the phosphorylation of CI throughout development and in mutant backgrounds with antibodies directed against the phosphorylated form of CI. In the absence of these reagents, however, we can determine whether PKA phosphorylates these sites in vitro. We fused the CI fragments of CI(wt) and CI(m1-4) that contain the four PKA sites (aa441-1065) to GST, and generated two-dimensional tryptic phosphopeptide maps of the expressed fusion proteins. The two maps are presented in Fig. 3. There are at least 13 phosphopeptides that are labeled by PKA in the wild-type CI peptide. In vitro, PKA can recognize RxS/T, the subset RRxS/T, RxxS/T and RKxxS/T. The phosphorylation of S is preferred 40:6 over T (reviewed in Pearson and Kemp, 1991) and in vivo, the RRxS site is preferred 2:1 over the others (reviewed in Kennelly and Krebs, 1991). We originally chose to mutate the four consensus RRxS/T sites in CI because they would probably be the preferred phosphorylation sites in vivo. Scanning the CI fragment for all possible consensus PKA sites, we found that all of the phosphopeptides can be accounted for by the number of PKA consensus sites in the fusion protein. Three of the strong spots and two weaker spots that are present in the wild-type fragment are missing in the mutant fragment, demonstrating that PKA can specifically and directly phosphorylate the four RRxS/T consensus PKA sites in vitro. The two weak spots are difficult to distinguish and may represent only one spot or incomplete digestion of a single peptide. GST alone was not phosphorylated (data not shown).

To confirm our cell culture observations and to study the relationship between the CI proteins mutant for consensus PKA sites and hh signaling in vivo, we made transgenic flies that have UAS-HACI [hemagglutanin(HA) tagged CI (wt or m1-4)] incorporated into the genome and determined whether these transgenes could affect wg expression. Four different UAS-HACI(wt) and four different UAS-CI(m1-4) fly lines were crossed with paired-GAL4/MKRS flies, and the resulting UAS-HACI(wt or m1-4)/+; paired-GAL4/+ embryos ectopically express CI(wt) or CI(m1-4) to the same extent in the paired (prd) domain as assessed by HA antibody staining. Because all of the ci transformants have an HA tag at the N terminus of the protein, we are able to distinguish the exogenous from endogenous CI expression by staining the embryos with an anti-HA antibody. While all of the UAS-HACI(wt)/+; prd-GAL4/+ embryos survive to adulthood, only 10% of the expected UAS-HACI(m1-4)/+; prd-GAL4/+ embryos survive past the first instar. As shown in Fig. 4A-D, exogenous HACI is expressed in the broad, alternating segmental stripes that span the engrailed (en) and wg expression domains and define the prd domains. In cell culture, CI(m1-4) has a 10-fold increase in transcriptional activity over that of CI(wt). To determine whether the CI(m1-4) protein can increase hh target gene expression in vivo, we examined wg mRNA expression in the UAS-HACI(wt)/+; prd-GAL4/+ and UAS-HACI(m1-4)/+; prd-GAL4/+ embryos. Overexpression of CI(wt) in the alternate prd stripes does not cause any significant change in wg expression in the majority of the embryos examined (Fig. 4C). 93% (200/216) of the embryos double-stained with an HA antibody and wg mRNA do not show any increase in wg expression. In contrast, the ectopic expression of HACI(m1-4) in the prd stripes results in a dramatic anterior expansion of wg expression (Fig. 4D). 97% (133/137) of the embryos double-stained with HA antibody and wg mRNA shows an increase in the wg expression domains. Although the cells both anterior and posterior to the A/P boundary express HACI(m1-4) protein, wg expression only expands anteriorly, consistent with the previous finding that only anterior compartment cells are competent to express wg (Ingham, 1993). The expansion in wg expression caused by the ectopic activity of CI(m1-4) is very obvious (Fig. 4D) when compared with the neighboring segments that express only endogenous CI.

We expected that the cuticular phenotype of the UAS-HACI(wt)/+; prd-GAL4/+ embryos would be wild type because the wg expression in these animals is wild type. All of the embryos from the mating of UAS-HACI(wt) flies and prd-GAL4/MKRS flies have a wild-type phenotype, and we could not determine the exact genotype of the embryos (Fig. 4E). Because the UAS-HACI(m1-4)/+; prd-GAL4/+ embryos are virtually lethal, cuticles of unhatched embryos from the cross of UAS-CI(m1-4) flies to prd-GAL4/MKRS flies were assessed for their phenotype. The UAS-HACI(m1-4)/+; prd-GAL4/+ embryos have cuticular defects consistent with the overexpression of wg. In alternating segments, the posterior rows of denticles contain patches of naked cuticle and the anteriormost rows of the denticle belts are missing denticles (Fig. 4F).

Our cell culture experiments showed that, in Cl-8 cells, the transactivation of the CI reporter plasmid by the CI(m1-4) protein is completely independent of a HH signal. Using the UAS-HACI transgenic flies, we asked whether a change in the CI consensus PKA phosphorylation sites is all that is required for HH to signal in vivo. To test this hypothesis, we crossed UAS-HACI (wt or m1-4); hh^IJ/TM2 flies with prd-GAL4 hh^IJ/MKRS flies and assessed the ability of the CI transgenes to suppress the mutant hh phenotype in the prd domains of expression.

As a control, hh^IJ/TM2 flies were mated with prd-GAL4 hh^IJ/MKRS flies and the progeny characterized with respect to wg expression and cuticular phenotype. As expected, 25% of the embryos are defective in maintaining wg expression during germband extension (Fig. 5C). The remaining 75% of the embryos had a wild-type wg expression pattern. Because prd-GAL4 hh^IJ/hh^IJ embryos are lethal, cuticle preparations were made of the unhatched embryos from the control mating. These embryos lack naked cuticle and develop a lawn of randomly arrayed denticles (Fig. 5D). The presence of prd on the hh^IJ chromosome did not affect the hh^IJ phenotype.

In the population of embryos from the cross between UAS-HACI(wt); hh^IJ/TM2 flies and hh^IJprd-GAL4/MKRS flies, approximately 25% of the animals did not express wg in germband-extended animals. About 75% had the wild-type wg expression pattern and <5% had alternating broad stripes of wg expression. This result is consistent with the double-labeling experiments presented above. Thus, exogenous expression of CI(wt) driven by prd-GAL4 does not restore wg expression in hh mutant embryos and the primary phenotype of the UAS-HACI(wt)/+; hh^IJ/hh^IJprd-GAL4 animals is the same as hh^IJ/hh^IJprd-GAL4 animals (Fig. 5E). Because prd-GAL4 hh^IJ/hh^IJ embryos are lethal, cuticle preparations were made of the unhatched embryos from this mating. As shown in Fig. 5F, exogenously expressed CI(wt) does not rescue the hh mutant phenotype.

On the contrary, exogenous expression of CI(m1-4) in the prd domain restores wg expression in that domain (Fig. 5G) and does rescue the hh mutant phenotype (Fig. 5H). In the population of embryos from the cross between UAS-HACI(m1-4); hh^IJ/TM2 flies and hh^IJprd-GAL4/MKRS flies, 50% of the embryos had a wild-type wg expression pattern (UAS-HACI(m1-4)/+; TM2/MKRS and UAS-HACI(m1-4)/+; hh^IJ/MKRS animals), 25% had alternating wild-type and broad bands of wg expression in germband-extended embryos (UAS-HACI(m1-4); hh^IJprd-GAL4/TM2 animals) and 25% had only broad alternating bands of wg expression in germband-extended embryos (UAS-HACI(m1-4); hh^IJ/hh^IJprd-GAL4 animals). We never observed embryos without wg expression in this population. From the double-labeling experiments, we know that the broad alternating stripes only occur in the presence of the UAS-HACI and prd-GAL4 chromosomes. In the population of embryos from the cross between UAS-HACI(m1-4); hh^IJ/TM2 flies and hh^IJprd-GAL4/MKRS flies, 50% of the embryos live and 50% are lethal either because they are hh^IJprd-GAL4/hh^IJ or they are heterozygous for the UAS-HACI(m1-4) and hh^IJprd-GAL4 chromosomes. One half of the lethal embryos have the phenotype characteristic of UAS-HACI(m1-4)/+; hh^IJprd-GAL4/+ transheterozygotes. The other half are UAS-HACI(m1-4); hh^IJ/hh^IJprd-GAL4 embryos and have a suppressed hh^IJ phenotype; they develop naked cuticle and the bordering denticles exhibit some segmental organization. This phenotype is never observed in the control hh^IJ/hh^IJprd-GAL4 lethal embryos and only rarely (<5%) in the UAS-HACI(wt); hh^IJ/prd-GAL4 hh^IJ lethal embryos. The ability of CI(m1-4) to maintain wg expression in the absence of HH signaling suggests that hh-dependent target gene expression may be mediated through dephosphorylation of CI.

It has been shown that PKA has a dual effect on the hh signaling pathway in vivo (Ohlmeyer and Kalderon, 1997) and has both a positive and negative effect on CI activity in cell culture (Chen et al., 1998). We have demonstrated that CI is the target for the negative regulation of PKA; however, the target of the positive PKA effect is unknown. To determine whether PKA could modulate the CI(m1-4)-mediated stimulation of wg expression in vivo, we crossed homozygous UAS-HACI (wt or m1-4); UAS-R* flies to prd-GAL4/MKRS flies and generated embryos that express HACI (wt or m1-4) and the mutant PKA regulatory subunit, R*, in the prd domains. The UAS-R* transgene encodes a mutant form of the PKA regulatory subunit that inhibits PKA activity in imaginal discs and embryos by binding to the catalytic PKA subunit constitutively (Li et al., 1995). We expected that, if the small, positive regulation of CI by PKA was required in addition to the PKA-negative effect, the enhanced wg expression seen in the UAS-HACI(m1-4)/+; prd-GAL4/+ embryos might be suppressed by the dominant negative PKA regulatory subunit. As shown in Fig. 6, expression of the dominant PKA mutant in the prd domain causes mild wg overexpression in only 6.3% of the embryos (30/479 embryos counted), assuming that 50% of the embryos express the GAL4 activator (Fig. 6A). These results are in agreement with those reported by Ohlmeyer and Kalderon (1997). When CI(wt) is ectopically expressed in the prd domains, 40% of the embryos (288/728 embryos) overexpress the wg message (Fig. 6B). 50% of the UAS-HACI(m1-4)/+; UAS-R*/prd-GAL4 embryos (341/674 embryos) overexpress wg (Fig. 6C). Therefore, the inhibition of PKA activity sensitizes the system so that exogenously expressed CI(wt) is able to induce ectopic wg expression, while CI(wt) cannot when it is overexpressed in the prd-GAL4 genetic background. We could not detect a decrease in wg expression in UASHACI(m1-4)/+; UAS-R*/prd-GAL4 embryos compared to UAS-HACI(m1-4)/+; prd-GAL4/+ embryos.

As shown in Fig. 1, the PKA sites at Ser-838, Ser-856 and Ser-892 in CI must remain intact for the regulation of CI proteolysis and transcriptional activity. Mutating any of the sites to alanine inhibits the proteolysis of CI leading to a 10-fold increase in CI activity. Furthermore, two-dimensional tryptic peptide maps show that PKA can specifically phosphorylate these sites in vitro; with the excess kinase present in the in vitro assays, these sites are not phosphorylated when the serines are mutated to alanines. These results support the hypothesis that the RRxS/T PKA sites are directly phosphorylated by PKA in vivo as well. In addition, all three sites in CI must be phosphorylated for the processing of CI to a repressor of hh target genes. These sites must also be intact for HH regulation. In Cl-8 or Kc cells, the activity of CI(m1-4) is unaffected by the addition of a HH signal while the wild-type CI activity is increased 6- to 10-fold when the cells are stimulated by HH. CI appears to be much more active in Cl-8 than Kc cells. One explanation for this observation is that the hh signaling cascade is more active in the disc-derived Cl-8 than embryo-derived Kc cells. In our model, the PKA phosphorylation of CI would ensure the processing of CI to the repressor form of the protein in cells that do not receive a HH signal. When HH reaches the cell surface, we hypothesize that it affects the ability of CI to be phosphorylated. Because the genetic data suggests that hh does not regulate PKA directly (Jiang and Struhl, 1995; Li et al., 1995; Ohlmeyer and Kalderon, 1997), it may be that hh affects the phosphorylation state of CI by activating a phosphatase, or through changing the accessibility of CI to a phosphatase. In support of this idea is the observation that the phosphatase inhibitor, okadaic acid, stimulates CI proteolysis, even in the presence of a HH signal (Y. C., unpublished data). HH signaling stimulates fu kinase activity to transform full-length CI to a transcriptional activator (Ohlmeyer and Kalderon, 1998). It may also be that fu activity renders full-length CI inaccessible to PKA phosphorylation.

We observed that the mutation of the Thr-1006 PKA site, RRQT, does not have any effect on CI function. This is expected from the work of Prorok and Lawrence (1989) who showed that peptides with threonines at the phosphorylation position are poor substrates for PKA activity.

We have also examined the activity of CI(wt) and CI(m1-4) in the developing embryo and find that CI(m1-4) can activate hh target gene expression even in the absence of hh signaling. When overexpressed in the prd domain, CI(wt) is neither sufficient to increase wg expression nor to suppress the hh mutant phenotype. This result differs from that presented by Alexandre et al. (1996), who demonstrated that overexpression of wild-type CI in the engrailed (en) domain increases the wg message and, in the hairy (h) domain, can suppress the mutant hh phenotype. This discrepancy could be due to differences among the en-GAL4, h-GAL4 and prd-GAL4 lines used in these experiments, although we do not observe any increase in wg expression when CI(wt) is expressed in the h domain by the h-GAL4 (Brand and Perrimon, 1993) insertion (Chen and Smolik, unpublished observation). When expressed in the prd domain, CI(wt) can overexpress wg in a small percentage (7%) of the embryos. However, both the percentage and the degree of wg overexpression are significantly smaller and weaker when compared to embryos that overexpress CI(m1-4).

Ectopic expression of CI(m1-4) in the prd domain causes the expression of wg to expand into the anterior portion of the prd domain. This region includes cells that receive a HH signal as well as those that do not. By analogy to what occurs in discs (Aza-Blanc et al., 1997), it is hypothesized that, in the anterior cells that do not receive a HH signal, endogenous CI is proteolysed to a repressor of wg expression. CI(m1-4) is able to activate wg expression in these cells suggesting that, as in cell culture, it is not a substrate for proteolysis.

The phenotype that results from the CI(m1-4) overexpression is a loss of denticles in the anterior and posterior rows of the ventral denticle belts. It is known that the formation of naked cuticle absolutely requires wg activity (Bejsovec and Wieschaus, 1993). Thus the elevated levels of wg in cells that normally express wg at lower levels can account for the posterior row cells variably assuming a naked-cuticle cell fate. However, the loss of the anterior row is harder to explain. EN-expressing cells that abut the A/P boundary form naked cuticle and the row of cells just posterior to this also express EN and produce the anterior row denticles (Bejsovec and Wieschaus, 1993). These denticles absolutely require EN activity and are dependent on wg as well. Perhaps the increase in wg along the A/P boundary affects the levels or activity of EN so that the anterior row cells take on a more anterior fate. While the CI(m1-4) embryos have a phenotype consistent with the overexpression of wg, they do not exhibit the loss of entire denticle belts that is seen when wg is overexpressed in the h domain (Wilder and Perrimon, 1995) or when ubiquitously expressed by a heat-shock (HS) promoter (Noordermeer et al., 1992). One likely explanation for this result is that the overexpression of wg that is mediated by CI is not as extreme as the overexpression of wg by the HS promoter or by GAL4. An alternate hypothesis is that the constitutive activity of CI(m1-4) causes the overexpression of more than one target gene and the relative dosages of these gene products modulate the activity of high levels of wg.

As shown in Fig. 5, CI(m1-4) can rescue the hh mutant phenotype when expressed in the prd domain. Together with the cell culture results, this observation suggests that a CI protein that is independent of the negative regulation of PKA, can bypass the requirement for a HH signal. The simplest explanation for the differential effect of CI(wt) and CI(m1-4) on wg expression is that CI(wt) is still subject to hh regulation and proteolytic processing within the prd domain, while CI(m1-4) is no longer a target for proteolysis and becomes constitutively active even in cells that do not receive a HH signal.

The recent work of Methot and Basler (1999) suggests that this interpretation is too simplistic. They demonstrated that the suppression of CI proteolysis may be necessary but not sufficient to by-pass the HH regulation of CI activity. In wing discs, a mutant CI that cannot be proteolyzed is unable to activate hh target-gene expression in the absence of a HH signal. These mutant transgenes contain the consensus RRxS/T PKA sites. Taken together with these results, our data suggest that the PKA phosphorylation of these sites is important for (1) regulating the proteolysis of CI and (2) for modulating the HH regulation of CI activity. While the dephosphorylation of one PKA site is sufficient to protect CI from proteolysis, the dephosphorylation of more than one site may be required to by-pass the HH signal.

Cell culture studies demonstrated that, in addition to the negative regulation of CI, PKA could activate gene expression in a CI-dependent manner (Chen et al., 1998). Genetic studies have suggested that PKA can activate hh target genes in a ci-dependent but CI-level-independent fashion (Ohlmeyer and Kalderon, 1997). We wished to determine whether we could detect the CI-dependent PKA activation of wg expression in vivo. We reasoned that if PKA activated the full-length activator form of CI then CI(m1-4) activity could be suppressed through the inhibition of PKA in developing embryos. Although the activation in cell culture was insignificant (2-fold), it was possible that it represented a discernible activity in vivo. We expressed CI(wt) and CI(m1-4) together with a dominant negative PKA regulatory subunit (R*) in the prd domain. We expected that the increase of wg in the UAS-HACI(m1-4)/+; UAS-R*/prd-GAL4 embryos would be suppressed, but this was not the case. The wg expression pattern in these animals is indistinguishable from that found in the UAS-HACI(m1-4)/+; prd-GAL4/+ embryos.

In the absence of hh and PKA signaling, wg expression is undetectable although the full-length CI levels are increased (Ohlmeyer and Kalderon, 1997). This result suggests that wg expression can be uncoupled from the increase in full-length CI levels. We find that exogenously expressed CI(wt) can activate wg expression in the absence of hh and PKA signaling. A possible explanation for this difference is that, in our experiments, the higher levels of exogenous CI(wt) activated in response to the inhibition of PKA activity is sufficient to activate wg but the endogenous levels of wild-type CI protein are not. It is of interest that the dose of exogenous CI(wt) expressed by prd-Gal4 is not sufficient to activate wg expression unless the levels of PKA activity are decreased. In this case, the increased dosage of CI that is not phosphorylated by PKA may be sufficient to overcome the antagonistic effects of Su(fu) and allow a threshold amount of activated CI to accumulate.

This work was supported by a grant from the National Institutes of Health (DK44239) and fellowships from the Swiss National Science Foundation and the Société Académique Vaudoise (to J.-R. C.)