Activation of Smoothened in the Hedgehog pathway unexpectedly increases Gαs-dependent cAMP levels in Drosophila

Hedgehog (Hh) signaling plays a key role in the development and maintenance of animal tissues. This signaling is mediated by the atypical G protein–coupled receptor (GPCR) Smoothened (Smo). Smo activation leads to signaling through several well-characterized effectors to activate Hh target gene expression. Recent studies have implicated activation of the heterotrimeric G protein subunit Gαi and the subsequent decrease in cellular cAMP levels in promoting the Hh response in flies and mammals. Although Hh stimulation decreases cAMP levels in some insect cell lines, here using a bioluminescence resonance energy transfer (BRET)-based assay we found that this stimulation had no detectable effect in Drosophila S2-R+ cells. However, we observed an unexpected and significant Gαs-dependent increase in cAMP levels in response to strong Smo activation in Smo-transfected cells. This effect was mediated by Smo's broadly conserved core, and was specifically activated in response to phosphorylation of the Smo C-terminus by GPCR kinase 2 (Gprk2). Genetic analysis of heterotrimeric G protein function in the developing Drosophila wing revealed a positive role for cAMP in the endogenous Hh response. Specifically, we found that mutation or depletion of Gαs diminished low-threshold Hh responses in Drosophila, whereas depletion of Gαi potentiated them (in contrast to previous findings). Our analysis suggested that regulated cAMP production is important for controlling the sensitivity of cellular responses to Hh in Drosophila.

The seven-transmembrane-spanning receptor protein Smo is an atypical member of the G protein-coupled receptor (GPCR) 3 protein family that activates cytoplasmic signaling in response to secreted Hedgehog (Hh) family proteins. Insufficient or excessive Smo activity can have severe consequences for the formation and maintenance of tissues (1,2). In fact, activating mutations in Smo have been shown to drive tumor formation in several tissues (1,3). This suggests that Smo activity must be tightly regulated in cells. Consistent with this, several complex and intersecting regulatory mechanisms for finely controlling Smo activity have been identified. Many of the details are best understood in Drosophila melanogaster, where ubiquitination (4,5), sumoylation (6,7), and phosphorylation (8 -13)  PKA phosphorylation primes the SAID for phosphorylation at adjacent sites by CkI, and the combined effects are sufficient to stabilize Smo, promote its accumulation at the plasma membrane, and drive it into an active conformation (8 -10, 14). However, other kinases, including GPCR kinase 2 (Gprk2) (11), CkI␥ (12), and Fu (13), act after PKA to modulate Drosophila Smo activity by phosphorylating other sites in the SmoCT. Like the PKA phosphorylation sites, most of these sites are not conserved between flies and vertebrates. Interestingly, however, Gprk2 and a homologous mammalian kinase, GRK2 (together with casein kinase I), phosphorylate conserved sites in the proximal SmoCT in Drosophila and mice, respectively (11,15). In both organisms, phosphorylation of these sites promotes Smo signaling activity (11,15,16), indicating that GRK phosphorylation of the SmoCT is an evolutionarily conserved mechanism for Smo activation (although it may not be critical in all species) (17). However, precisely how GRK phosphorylation enhances Smo activity is not known.
Once activated, Smo signals through a series of downstream canonical effectors that includes the atypical kinesin Costal-2/ KIF7 and (in some organisms like Drosophila and zebrafish (18,19)) the protein kinase Fused, which are components of the so-called Hh Signaling Complex (HSC), as well as Suppressor of Fused (Sufu) (1,2). Canonical signaling ultimately converges on the transcription factor cubitus interruptus (Ci)/GLI. In the absence of Hh, Ci/GLI2/3 are converted through limited proteolysis to transcriptional repressors. Hh-dependent activation of canonical signaling prevents this processing, causing Ci/GLI2/3 to accumulate in full-length transcriptional activator forms that drive transcription of Hh target genes (1,2). Phosphoregulation of Ci/GLI2/3 processing is another key point of control in the Hh pathway. PKA again plays the key role, although in this case it acts as a negative rather than positive regulator. Ci/GLI2/3 phosphorylation by PKA primes the proteins for further phosphorylation by other kinases, ultimately promoting ubiquitination and subsequent partial proteolysis to the repressor forms (1,2).
The central role of PKA in Drosophila and vertebrate Hh signaling pathways has long suggested that regulation of the second messenger cAMP, the activator of PKA, could be important for pathway function. Indeed, Smo shares some structural and functional similarities with classical GPCRs (20) and there is evidence that Smo can couple to heterotrimeric G proteins, specifically to G␣ i . Activation of G␣ i by Smo suppresses cAMP production and can promote target gene expression in flies (21,22) and mammals (23,24), and is involved in GLI-independent noncanonical responses in mammals as well (25)(26)(27)(28). Recent studies have demonstrated that certain other GPCRs can influence Hh signaling by activating G␣ s or G␣ i . For example, Gpr161 is a constitutively active GPCR that signals through G␣ s to keep cAMP levels high in the absence of ligand, ensuring efficient PKA-dependent processing of GLI proteins and silencing of target gene expression (29). Conversely, the G␣ icoupled GPCR Gpr175 suppresses cAMP production and promotes Hh pathway activation (30).
The physiological importance of heterotrimeric G protein signaling in the Hh response has been controversial. Experiments using a cAMP-insensitive form of the PKA catalytic subunit initially shed doubt upon the importance of regulated PKA activity in the Hh response (31,32). However, several studies cited above have now shown clear links between G␣ proteins, cAMP, and Hh pathway activity. The link between decreased cAMP production and increased Hh target gene expression seems fairly straightforward in mammals, where the principal role of PKA is to inhibit GLI proteins (2). However, there is conflicting evidence showing that PKA can also promote Shh signaling in vertebrate cells and tissues (33,34). The situation is similarly complex in Drosophila, where PKA functions both as an activator and inhibitor of the Hh response. For example, complete loss of PKA activity causes constitutive Ci activation and ligand-independent target gene expression (31,35), but more moderate PKA inhibition or reduction of cAMP levels can actually decrease Hh-dependent target gene expression (8,36). Similarly, increased cAMP levels or PKA activity can either prevent (31,37) or enhance (8,36,38) expression of Hh target genes. The evidence suggests that control of cAMP levels within certain upper and lower limits is important for proper functioning of the pathway. Some of the conflicting evidence may result from cell-type-specific effects that could reflect differences in basal cellular cAMP levels in different systems (39).
To address some of the conflicting evidence about the implication of heterotrimeric G proteins in the Hh response and to examine how G protein signaling could be regulated at the level of Smo, we sought to characterize the effects of Hh signaling on cAMP levels and target gene expression using cell-based assays in Drosophila S2 cells. To our surprise, we did not observe changes in cAMP levels in Hh-treated cells. However, stronger activation of Smo caused a substantial G␣ s -dependent increase in cAMP levels specifically in response to Gprk2 phosphorylation of the SmoCT. This effect on cAMP production was mediated by the evolutionarily conserved core of Smo, and occurred independently of the canonical signaling effectors Fu and Cos2. Although not essential for signaling, G␣ s and G␣ i depletion experiments in vivo both suggest that heterotrimeric G proteinmediated regulation of cAMP levels is important for adjusting the sensitivity of cells to endogenous Hh ligand levels.

Smo can activate G␣ s -dependent signaling
Drosophila Smo has been shown to activate G␣ i in transfected Sf9 cells and Hh-treated Clone-8 cells (21,40). Initially to see if we could observe the same response, we used a bioluminescence resonance energy transfer (BRET) assay based on the EPAC-BRET biosensor (41) to measure changes in cAMP levels in live S2 cells. The BRET signal produced by this biosensor is inversely correlated to cAMP levels. In control experiments, stimulation of cAMP production by expression of a partially activated mutant form of G␣ s (G␣ s Q215L ) (42) caused the expected decrease in the EPAC-BRET signal (Fig. 1A). Conversely, inhibiting cAMP production with the comparable activated form of G␣ i (G␣ i Q205L ) significantly increased the EPAC-BRET signal (Fig. 1A). Thus this assay can reliably detect both increases and decreases in cellular cAMP levels in S2 cells.
We then conducted EPAC-BRET assays to test for Hh-dependent changes in cellular cAMP levels. Transfection of S2 cells with an expression plasmid for Hh N , a secreted and biologically active N-terminal fragment of Hh (43) did not yield the expected decrease in cellular cAMP levels, although it did activate target gene expression (assessed using a ptc-luciferase reporter assay) (44) (Fig. 1, B and C). Even when using a myristoylated, membrane-targeted form of the EPAC-BRET biosensor to more specifically assess changes in cAMP pools at the plasma membrane (45), we did not observe the anticipated decrease (Fig. S1).
To see if enhancing Smo activity would lead to an effect in this assay, we expressed GFP-tagged, WT and mutant forms of Smo (see Fig. 1D). ptc-luc reporter activity was similarly low in WT Smo (Smo WT )-transfected cells compared with mock transfected controls (Fig. 1E), indicating that the level of Smo overexpression was moderate. Co-expression with Hh N caused strong activation of ptc-luc reporter transcription (Fig. 1E). Unexpectedly, we observed a substantial Hh-dependent increase in cellular cAMP levels under these conditions (Fig.  1F). We obtained the same effect when we strongly activated signaling by expressing the PKA-and CkI-phosphomimetic,  (Fig. 1, G and H). Thus strong activation of the Hh pathway increased both target gene expression and cAMP levels in S2 cells.
To determine whether the increase in cAMP levels we observed is an acute response to Smo activation rather than a gradual or adaptive response to days-long pathway activation in our Hh N co-transfection experimental setup, we carried out a time course analysis on Smo WT -expressing cells treated with medium conditioned by control-or Hh N -expressing cells. Treatment with control-conditioned medium had no effect on cAMP levels at any time point tested (Fig. 1I). In short-term treatments (between 1 and 10 min), we did not observe Hh Ndependent changes in cAMP levels (results not shown). Increases in cAMP levels were observed after 30 and 60 min of exposure to Hh N , but they did not reach statistical significance (Fig. 1I). After 2 h of Hh N treatment, cAMP levels were significantly increased, and by 4 h the magnitude of increase was similar to that observed in control cells that had expressed cotransfected Hh N for 2-3 days (Fig. 1I). In Western blot analysis at each of these time points, Hh N -dependent stabilization of Smo-GFP was just detectable after 30 min of treatment, and the shifted, hyperphosphorylated active form of Smo-GFP first became readily apparent after 2 h of Hh N exposure and increased at 4 h (Fig. 1J). Thus the time course of cAMP regulation closely matched the profile of Smo activation. Although

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
the cAMP increase was similar in cells expressing Hh N for days and those treated for only 4 h, the transiently exposed cells had only a small percentage of the levels of Smo protein observed in the continuously exposed cells (Fig. 1J). This suggests that the effect does not require extremely high levels of Smo protein.
The most straightforward interpretation of these results is that strong activation of Smo activates G␣ s , either directly or indirectly (e.g. by regulating another GPCR). If this is true, we expected that double-stranded RNA (dsRNA)-mediated depletion of G␣ s would block the increase in cAMP levels. We first confirmed that treatment of G␣ s -transfected cells with dsRNA targeting the g␣s transcript efficiently blocked production of the protein ( Fig. 2A). Consistent with our expectation, the ability of Smo to increase cAMP levels was strictly dependent upon G␣ s . dsRNA-mediated depletion of G␣ s had no significant effect on basal cAMP levels in control cells (Fig. 2B). However, the Smo SD -dependent increase in cAMP levels was almost completely inhibited (Fig. 2B). We confirmed that Smo SD protein levels were unchanged by G␣ s depletion (Fig. 2C). We conclude that strong activation of Smo activates G␣ s , either directly or indirectly, to stimulate cAMP production in S2 cells.

GRK phosphorylation of SmoCT triggers G␣ s -dependent downstream signaling
Although PKA/CkI phosphorylation is the principal trigger for Smo activation, more than half of the transcriptional responseobservedinptc-lucreporterassaysdependsuponphosphorylation of Smo by Gprk2 (11). To see if Gprk2 affects heterotrimeric G protein-dependent signaling downstream of Smo, we treated S2 cells with a gprk2-specific dsRNA. The ability of Smo to activate G␣ s depended upon Gprk2, as efficient depletion of the kinase strongly impaired the ability of Smo SD to stimulate cAMP production (Fig. 3, A and B). To confirm that this is due to a direct effect of Gprk2 phosphorylation on Smo activity, and not to an indirect effect of Gprk2 on cAMP regulation (as described previously in Ref. 36), we targeted the Gprk2 phosphorylation sites in SmoCT rather than the kinase. Consistent with the depletion experiments, mutating the four clusters of Gprk2 phosphorylation sites to Ala (in Smo SDc1-4A ) prevented Smo SD from stimulating cAMP production (Fig. 3C). However, Smo SDc1-4A was well-expressed and retained substantial ability to stimulate target gene expression (ϳ20-fold over baseline levels) (Fig. 3, D and E). Although phosphorylation by PKA and CkI is sufficient to stabilize Smo and activate some signaling, we conclude that it is phosphorylation by Gprk2 that specifically triggers G␣ s -dependent downstream regulation of cAMP levels.

G␣ s -dependent signaling is mediated via the conserved core of Smo
We previously showed that a C-terminally-truncated form of Smo that retains only the broadly conserved core of the protein (amino acids 1-663; Smo core ) is constitutively active and capable of activating Hh target gene expression, although less effectively than Smo SD (Fig. 4A) (11). This suggested that Smo proteins from different species share common aspects in their signaling mechanisms (2,46). We found that Smo core possesses all the sequences necessary to promote G␣ s -dependent signaling, in a constitutive manner. In Smo core -expressing cells, cAMP levels increased to a similar extent as in Smo SD -expressing cells (Fig. 4, B and C), both throughout cells and more specifically at the membrane (Fig. S1), and in a G␣ s -dependent manner (Fig. 4, D and E). In Smo core -expressing cells, co-expression of Ptc almost completely suppressed both ptc-luc reporter activation and the increase in cAMP production ( Fig.  4, A-C). This suggests that the stimulation of cAMP production is a specific, Hh-regulatable signaling activity of Smo core .
We tested additional mutants to identify the determinants in Smo core required to activate G␣ s -dependent signaling. Unlike full-length Smo, Smo core activity is entirely dependent upon Gprk2 phosphorylation ( Fig. 4F) (11). Its ability to drive G␣ s -dependentsignalingwassimilarlyreliantuponGprk2phosphorylation, as stimulation of cAMP production by Smo core was strongly impaired by either depletion of Gprk2 (Fig. 4, G and H) or mutation of the Gprk2 phosphorylation sites (Smo core.c1-3A ) (Fig. 4, I and J). We previously showed that the C-terminus of Smo core (between amino acid 626 and 663) is required for its activity in ptc-luc reporter assays (11). This same region is also required for G␣ s activation (Fig. 4, K-L). Thus the 7-transmembrane domain region of Smo is not sufficient to activate G␣ sdependent signaling; sequences in the broadly-conserved membrane-proximal portion of the cytoplasmic tail are also required.

Activation of G␣ s -dependent signaling downstream of Smo is distinct from HSC signaling
Phosphorylation of Smo by Gprk2 affects canonical signaling through the HSC in complex ways. For example, it promotes Smo C-terminal dimerization, which itself promotes Fu activa-

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
tion (11,47). It also promotes interaction of Costal-2 with Smo core (11). We wondered to what extent the G␣ s -dependent signaling we observed is linked to activation of these more typical effectors of Smo. To assess this, we examined the effects of manipulating Fu and Cos2 activity on the ability of Smo to stimulate cAMP production. Fu is recruited to the plasma membrane and activated by Smo and Cos2, and in turn promotes high-threshold target gene expression (47,48).  (Fig. 5G), and this was associated with a strong reduction in Smo protein levels (as others have seen) (40) (Fig. 5H). Interestingly, however, CSBD had no significant effect on the ability of Smo SD to stimulate cAMP production despite the decrease in Smo SD protein levels (Fig. 5I)

G␣ s enhances the response to low levels of Hh
Having not seen any obvious role for heterotrimeric G protein signaling in the endogenous response of S2 cells to Hh Nconditioned medium, we turned to an in vivo system to see if G␣ s plays any physiological role in Hh signaling. First, we generated clones of cells homozygous for a null allele of g␣s (g␣s R60 ) in wing discs using the FLP-FRT system (52). These clones were small when generated in an otherwise WT background. Even when generating them in a Minute heterozygous background to give them a growth advantage (53), g␣s R60 clones were only about half the size of WT clones generated in parallel (Fig. 6, A-C). We did not observe changes in intermediate or highthreshold Hh responses (expression of Ptc and Engrailed, respectively) in g␣s R60 clones at the anterior/posterior (A/P) boundary (not shown). However, we did observe an effect on expression of the low threshold target gene dpp (assessed using a dpp-LacZ enhancer trap line). Normally, dpp-LacZ staining is fairly symmetrical in dorsal and ventral compartments of the wing pouch (see below) (Fig. 6D). However, in some larger g␣s R60 clones located away from the A/P boundary, dpp-LacZ

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
expression was reduced (Fig. 6, E and F). As the concentration of Hh is thought to decline with increased distance from the P compartment where it is produced, this suggests that G␣ s may play a role in the normal response to Hh where ligand levels are low.
We were concerned that this approach might underestimate the role of G␣ s because of the potential for nonautonomous rescue in mosaic discs. Wing discs cells have been demonstrated to be highly connected through gap junctions, with small dyes able to diffuse over many cell diameters at late third instar stage (54). cAMP has a lower molecular weight than these dyes, suggesting that it too may be able to diffuse in discs. Diffusion of cAMP from neighboring WT cells might limit the cell-autonomous effects in g␣s R60 clones.
To circumvent this potential complication, we used expression of a dsRNA transgene to deplete G␣ s throughout a much larger territory in the disc. Expression of G␣ s dsRNA throughout the developing wing pouch (using nubbin-GAL4) caused a failure in wing inflation (Fig. 7, A and B), a phenotype that has been linked to a failure in G␣ s activation downstream of the GPCR Rickets (55). This suggests that the dsRNA transgene does in fact deplete G␣ s .
Consistent with the clonal analysis, depletion of G␣ s selectively impaired low threshold dpp expression. We used apterous (ap)-GAL4 to drive expression of G␣ s dsRNA throughout the dorsal compartment of the wing disc, leaving the ventral compartment as an internal WT control. In control ap-GAL4/ϩ wing discs, low threshold Hh responses (such

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
as stabilization of full-length Ci (Ci 155 ) and expression of dpp) and intermediate threshold responses (expression of Ptc) appeared similar in the GAL4-expressing dorsal and WT ventral compartments (Fig. 7E). Quantification and averaging of the relative fluorescence intensities from several discs confirmed that both compartments show virtually identical Hh responses both in terms of intensity and distance from the A/P boundary (Fig. 7F). ap-GAL4-driven expression of G␣ s dsRNA had little effect on Ci stabilization or Ptc expression (Fig. 7, G  and H). The most striking effect was a strong and statistically significant reduction in dpp expression (Fig. 7, G and H). Most notably, the A-P distance over which dpp was expressed was substantially less in the G␣ s -depleted dorsal compartment, indicating that cells exposed to low levels of Hh were no longer able to mount an effective transcriptional response. This fits well with our observations of g␣s R60 mutant clones, and sug-gests that G␣ s plays a positive signaling role in situations where Hh levels are low and limiting for a response.
Consistent with this interpretation, we observed a Hh concentration-dependent requirement for G␣ s in Smo-transfected S2 cells. In a Hh N dose-response test, we found that the standard conditions we used for testing Hh induction of the pathway (100 ng of transfected Hh N plasmid) produced approximately a 60% maximal response in ptc-luc reporter assays (Fig.  8A). Under these conditions, we did not observe a significant effect of G␣ s depletion on ptc-luc reporter expression (Fig. 8B). However, at a lower level of Hh (20 ng of Hh N plasmid), which was sufficient for ϳ25% maximal response in ptc reporter assays (Fig. 8A) and for stimulating cAMP production (Fig. 8C), depletion of G␣ s decreased Hh-dependent ptc-luc reporter activation by Smo-GFP (Fig. 8D). The effect was modest (ϳ25% decrease), suggesting that signaling through G␣ s is not essential

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
for the response. Rather, this result together with our in vivo analysis suggest that G␣ s enhances responsiveness to low levels of Hh in vivo.

G␣ i depletion reveals an inhibitory role for this protein in the Hh response
Our analysis of G␣ s function appears to contradict previous studies in several systems showing that inhibition (rather than stimulation) of cAMP production, through G␣ i , promotes Hh target gene expression (21,22). In particular, g␣i mutant clones in wing discs, which appeared fairly small, failed to express dpp (22). We tested whether depletion of G␣ i over a larger territory would produce a similar effect. Expression of a dsRNA targeting G␣ i throughout the developing wing pouch (using the nubbin-GAL4 driver) impaired growth, causing the resulting adult wings to be undergrown, similar to those of g␣i hypomorphic mutants (Fig. 7C) (40). This suggests that the dsRNA did target G␣ i . Interestingly, the wings were not uniformly smaller. Instead, the size of the central region bounded by longitudinal wing veins 3 and 4 (L3 and L4) as a proportion of total wing area was significantly increased upon G␣ i depletion (Fig. 7D). This is a characteristic feature of increased Hh signaling in the wing disc. Consistent with this interpretation, G␣ i depletion throughout the dorsal compartment significantly increased the level and extent of dorsal Ci 155 stabilization and expression of dpp and Ptc (Fig. 7, I and J). Compared with control discs, cells further from the A/P boundary (and thus exposed to lower levels of ligand) in the G␣ i -depleted discs were able to activate robust target gene transcription. This suggests that G␣ i depletion enhanced sensitivity of cells to low levels of Hh, which is the opposite of the effect of G␣ s depletion. Our analyses of G␣ i and of G␣ s function in wing discs are thus consistent in revealing a positive effect of cAMP on sensitivity to Hh in vivo.

Discussion
Despite conflicting evidence about the ability and necessity of Smo itself activating G␣ signaling, heterotrimeric G proteins have increasingly been linked to Hh signaling under normal and pathological circumstances (e.g. Refs. 29 and 56). The emerging model (57), based largely on observations in mammalian cells, suggests that cAMP plays primarily an inhibitory role in the pathway. In the absence of Hh, a G␣ s -coupled GPCR (like Gpr161 in mammals) signals to maintain cAMP at high levels, thereby ensuring efficient PKA phosphorylation and processing of GLI proteins to their repressor forms. In response to Hh ligands, decreased G␣ s signaling and/or increased G␣ i signaling by Smo or certain other G␣ i -coupled GPCRs lowers cAMP levels, allowing GLI proteins to avoid PKA phosphorylation and accumulate in their full-length activator forms. However, the situation must be more complex, as cAMP/PKA cannot only inhibit but also promote Hh signaling in both flies and mammals (8,33,34,36,38), most directly in the case of Drosophila by phosphorylating and activating Smo. Consistent with this, we have observed a G␣ s -dependent increase in cellular cAMP levels downstream of Smo activation, and identified a positive role for G␣ s in Hh signaling in vivo.
We expected to see G␣ i activation and a decrease in cAMP levels in Hh-treated S2 cells, similar to what was previously observed in other insect cell lines (21,22). However, we observed no consistent effect on total cAMP levels in S2 cells exposed to Hh N . Studies in mouse cerebellar granule cell progenitors suggested that Hh signaling is controlled by a relatively small pool of cAMP that is restricted in its localization, likely near the cilium (58). Hh-dependent changes in cAMP levels may similarly be restricted to a smaller pool in S2 cells such that they do not significantly impact the total cAMP pool; in fact, using a membrane-anchored, myristoylated FRET biosensor of PKA activity in S2 cells, Li et al. (45) observed an ϳ2-fold increase in PKA activity in response to Hh. However, we saw no change in membrane cAMP levels with a comparable myristoylated cAMP biosensor. It may be that the EPAC-BRET biosensor is not sensitive enough to detect relatively small changes in cAMP levels in response to activation of endogenous signaling, where negative feedback may also limit the magnitude of the response. Alternatively, these cells may differ in some other way from the other cell lines previously tested (discussed below). Nevertheless, hyperactivation of the pathway at the level of Smo produced a substantial cAMP response. The fact that G␣ s depletion blocked the increase in cAMP levels strongly suggests that it is due to increased cAMP production rather than a decreased rate of degradation by phosphodiesterases. The relatively slow accumulation of cAMP over several hours is somewhat unusual, as GPCRs typically activate heterotrimeric G protein signaling on a time scale of seconds to minutes. However, it fits with the previously characterized prolonged time frame for Smo activation (59), which involves phosphorylation by PKA and CkI as well as trafficking to the plasma membrane.  ). B, wing imaginal disc bearing clones of g␣s R60 homozygous mutant cells generated and analyzed in parallel to WT clones in A. C, measurements of wing pouch clone areas from discs as in A and B, expressed as the percentage of the wing pouch area that is composed of clone cells (normalized to the value for WT clones). t test: ***, p Ͻ 0.001. D, disc in which no g␣s R60 homozygous mutant clones (marked by absence of GFP (green)) were generated. The disc was immunostained with an anti-␤-Gal antibody to reveal dorsal/ventral symmetry of expression of a dpp-LacZ enhancer trap (pink). E and F, discs bearing larger g␣s R60 homozygous mutant clones, stained as in D. Clone outlines are traced in yellow.

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
At the level of Smo, we identified several features required for this response. The ability of Smo to trigger G␣ s -dependent cAMP production lies within its highly conserved core. The seven-transmembrane domain is not sufficient, as the C-terminal 26 amino acids of Smo core (between amino acids 626 and 651), situated at the end of its cytoplasmic tail, are required for the effect. This same region is required for Smo core to interact with Cos2 and to promote expression of Hh target genes (11). However, we did not observe a link between HSC-dependent signaling and the cAMP response. In fact the ability of CSBD,

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
Smo by PKA and CkI is required. Crucially, however, whereas PKA/CkI phosphorylation is sufficient for activation of Hh target gene expression, the cAMP response is strictly dependent upon SmoCT phosphorylation by Gprk2. This is reminiscent of the phenomenon of phosphorylation barcoding that has been described for some GPCRs, where phosphorylation at different sets of sites in the receptor tail by different kinases dictates distinct signaling responses (61).
Given these constraints, how could Smo activity lead to signaling through G␣ s ? The most straightforward model would be that Smo directly couples to G␣ s , like a typical GPCR. If so, then we could explain the difference between our observations and others' demonstration of Smo-G␣ i coupling as a cell-type or context-dependent ability of Smo to couple to different G␣ proteins, as observed for many GPCRs (62). However, we have been unable to obtain compelling evidence for a direct interaction between Smo and G␣ s . We favor the alternative explanation that Smo regulates a GPCR that couples to G␣ s in S2 cells. Gpr161 provides a precedent for this in mammalian cells. The ability of Gpr161 to localize to the cilium where it can couple to G␣ s is determined by the activity of Smo in the cilium (63), thus providing an indirect link between Smo and G␣ s . Interestingly, there is some evidence that Smo is trafficked to the cilium in response to phosphorylation by GRK2 in mammalian cells (15) (although this has been disputed) (16). In Drosophila, Gprk2 is required for internalization of Smo from the plasma membrane in response to Hh (64). Thus it could be that the regulation of Smo-G␣ s signaling by Gprk2 that we observe in S2 cells reflects GRK-dependent trafficking of Smo to a subcellular location where it regulates a G␣ s -coupled GPCR, perhaps on endosomes (from which mammalian GPCRs have increasingly been shown to signal through heterotrimeric G proteins after GRK-dependent internalization) (65). If this is the case, the difference in G␣ signaling between S2 cells and other insect cell lines could reflect differences in the GPCRs they express.
Regardless of whether Smo directly or indirectly regulates G␣ s in S2 cells, we find that G␣ s does play a role in the endogenous Hh response. Because Smo activation leads to localized PKA activity at the plasma membrane (45), we imagined that a localized burst of cAMP production might be required to enhance Smo phosphorylation to allow high-threshold target gene expression. However, our results suggest that such a mechanism is not required for high-threshold signaling. Instead, we consistently see selective impairment of lowthreshold responses when G␣ s is depleted or mutated in S2 cells, in wing disc clones, and most clearly, throughout an entire compartment. The narrowing of the domain of dpp expression in response to g␣s-depletion suggests that the sensitivity of cells to low levels of Hh was reduced, consistent with the impairment of ptc-luc reporter activation we observed in S2 cells in response to low but not high levels of Hh. Thus, rather than boosting Smo activity toward its maximum, G␣ s -driven cAMP production may be most important for enhancing Smo activation above a signaling threshold under limiting ligand conditions. Interestingly, while this manuscript was under review, Pusapati et al. (66) reported similar findings in vertebrates. They showed that GPCR signaling through G␣ s , downstream of both Smo and GRKs, regulates the sensitivity of responding cells to Sonic Hedgehog. The parallel between these findings and ours indicates that regulation of cAMP in target cells is an evolutionarily conserved mechanism for fine-tuning Hh responsiveness.
Finally, in contrast to previously published work, we observe an inhibitory rather than stimulatory role for G␣ i in the Hh response. Specifically, G␣ i depletion led to ectopic expression of Hh target genes in cells located away from the A/P boundary, where ligand levels are lower. This difference from previous work may be due to a difference in the extent of signaling impairment in G␣ i -depleted cells versus g␣i mutant clones, similar to the way that PKA inhibition can suppress the Hh response (8), whereas complete elimination of the catalytic subunit leads to constitutive target gene expression (31). In any case, these results are consistent with our G␣ s depletion experiments and other work showing a positive role for G␣ s /cAMP in the Hh response (36,67,68). Together they clearly suggest a positive role for regulated cAMP production in controlling the sensitivity of cells to Hh.

Expression constructs
Expression plasmids encoding C-terminally-tagged GFP Smo mutants and Smo truncations under control of the metallothionein promoter (in pRmHa3.puro) were previously described (11). The expression construct for Smo ⌬Fu (lacking the C-terminal-most 52 amino acids containing a Fu-binding domain) (49) was generated by PCR amplification of a portion of the Smo coding sequence spanning the EcoRI site at codon 797-798 to the Leu-984 codon. Primers were designed to introduce a 3Ј stop codon and NotI site immediately following codon 984. The resulting EcoRI-NotI fragment was used to replace the corresponding C-terminus-encoding fragment in a pRmHa3.puro backbone containing the Smo SD coding sequence fused to C-terminal GFP (engineered as a cassette flanked by NotI and KpnI restriction sites). To create the Cos2 Smo-binding domain (CSBD) expression construct (40), sequences encoding amino acids 1001-1201 of Cos2 were PCR amplified, introducing a 5Ј EcoRI and 3Ј KpnI site. The resulting EcoRI-KpnI restriction fragment was cloned into a modified pRmHa3.puro plasmid encoding an in-frame N-terminal Myc epitope tag. For expressing Ptc, an EcoRI-KpnI fragment encoding full-length Ptc fused to a C-terminal GFP tag (obtained from Stephen Cohen) was cloned into pRmHa3.puro. The plasmids encoding the EPAC-BRET biosensor or the same protein lacking the GFP10 moiety (used as a control for background emission), as well as plasmids encoding constitutively active G␣ s Q215L and G␣ i Q205L , were prepared as described (36). pRmHa3.puro/Hh N , which encodes an active N-terminal fragment of Drosophila Hh plasmid (59), was used in the preparation of Hh N -conditioned media and for co-transfection of cells transfected with Smo expression plasmids in some experiments. An empty pRmHa3.puro vector (Ϫ) was used for transfection as a control.

BRET and ptc-luciferase reporter assays, dsRNA treatment
EPAC-BRET assays were performed essentially as described (36). Briefly, for BRET experiments not involving dsRNA treat-

GRK-phosphorylated Drosophila Smo activates G␣ s signaling
ment, S2-Rϩ cells were seeded in 24-well plates and transfected on day 1. Each well was transiently transfected with 100 -150 ng of pMT.puro/GFP10-EPAC-RLucII_T781A,F782A encoding the EPAC-BRET cAMP biosensor protein (described previously in Refs. 36 and 41) consisting of the cAMP-binding domain of the EPAC protein fused to GFP10 and Renilla luciferase II (RlucII) at the amino and carboxyl termini, respectively. cAMP binding promotes a more open conformation of the EPAC-BRET protein, causing the BRET signal to decrease (41). Cells were also transfected with pMT.puro expression constructs for Smo variants (with C-terminal GFP tag) or Hh N , depending on the experiment. On day 2, transgene expression was induced and cells were transferred to 3-4 wells of whitewalled, clear-bottomed 96-well plates. BRET measurements were performed on day 4 or 5 as described previously (36).
For experiments involving dsRNA treatment, dsRNA was prepared as described (11). dsRNAs targeted ␤-gal (11), gprk2 (11), fu (nucleotides 1043-1547 of fu-RA (FBtr0074602), g␣s (nucleotides 282-697 of G␣s-RA (FBtr0072144)), or g␣i (nucleotides 533-986 of g␣i-RA (FBtr0076934)). Cells were transfected as above on day 1. On day 2, cells were split and plated into 96-well plates and each well was treated with 0.5-2 g of dsRNA. On days 4 or 5, a second dose of dsRNA was administered and transgene expression was induced. Cells were processed for BRET measurements on day 7. BRET signals were determined by calculating the ratio of the light emitted by GFP10 (515 nm) and RlucII (410 nm). The net BRET value represents the BRET signal from which the background (represented by the BRET signal generated by cells expressing a biosensor lacking the GFP10 acceptor moiety) was subtracted. All graphs represent the composite results of at least two independent experiments, each performed in triplicate or quadruplicate measurements, with the exception of the experiment in Fig. 1I, which was performed once in triplicate. Statistical significance was determined by two-tailed Student's t tests.
ptc-luc reporter assays were performed essentially as described (11). Assays were performed at least two times each with triplicate or quadruplicate measurements and data compiled, with the exception of the experiment in Fig. 8A, which was performed once in quadruplicate. Due to interference with the normalization plasmid encoding Renilla luciferase in some dsRNA experiment setups, all data are presented as measurements of firefly luciferase reporter activity. Statistical significance was assessed using two-tailed Student's t tests.

Hh N treatment, Smo activation time course experiments, and immunoblotting
For time course experiments, Smo-expressing S2-Rϩ cells were treated with medium conditioned by either control cells or cells expressing Hh N as previously described (64). Cells were transfected with 150 ng each of pMT.puro/GFP10-EPAC-RLucII_T781A,F782A, pMT.puro/SmoWT-GFP, and (in the control cells) pMT.puro/Hh N and processed as above. On day 4, growth medium was gently removed and 0.1 ml of either control or Hh N -conditioned media was added to each well at the appropriate time. Cells treated at different intervals were processed simultaneously by adding DeepBlueC directly to the conditioned media and BRET values were determined as above. Time-dependent changes in Smo WT phosphorylation following exposure to control-or Hh N -conditioned media were determined by immunoblotting with rabbit ␣-GFP antibody (Torres Pines Scientific) as described (11). In other experiments, immunoblotting was performed with mouse monoclonal anti-myc (9E10; Santa Cruz Biotechnology), mouse monoclonal anti-␣-tubulin (12G10; obtained from the Developmental Studies Hybridoma Bank (DSHB)), guinea pig anti-Gprk2 (64), or guinea pig anti-Smo.

Drosophila crosses, immunostainings, and image analysis
For generating g␣s mutant clones, we used the g␣s R60 mutant allele (stock obtained from the Bloomington Drosophila Stock Centre). This allele contains a mutation that converts Tyr-231 to a stop codon, deleting the C-terminal ϳone-third of the protein, and should be a null allele. To make g␣s clones in a Minute background, w;dpp 10638 ,FRT42D,g␣s R60 /CyO,Kr-GAL4, UAS-GFP males were crossed at 25°C to w,hsFLP;FRT42D, Ub::GFP,M(2R)53 1 /CyO virgins. For WT clones, w;dpp 10638 , FRT42D males were used. Offspring from the two crosses were heat shocked at 37.5°C and dissected in parallel. For no clones control, the heat shock was omitted. Wing disc-bearing anterior halves were dissected from wandering third instar larvae and processed for immunofluorescence staining with rabbit anti-␤-gal antibody (Santa Cruz Biotechnology). To quantify clonal growth, the total area of the wing pouch and of the GFPnegative regions contained in the wing pouch for 10 discs were measured using the Histogram function of Photoshop. Clone size was expressed as the total GFP-negative clonal area over total wing pouch area.
For G␣ s and G␣ i depletion experiments w,UAS-Dcr;ap-GAL4,dpp 10638 /CyO or w,UAS-Dcr;nub-GAL4,dpp 10638 /CyO females were crossed at 27°C to w,UAS-GFP males (control), w;UAS-G␣s.dsRNA males (from Vienna Drosophila Resource Centre; stock number v24958), or w;UAS-G␣i.dsRNA males (69) (kindly provided by J. Knoblich). Wandering larvae were dissected and anterior halves processed for immunofluorescence with rabbit anti-␤-gal, rat anti-Ci 155 (2A1 from DSHB), and mouse anti-Ptc (Apa-1 from DSHB) and discs were imaged by confocal microscopy as described (11). To quantify fluorescence, boxes of equal size spanning the A/P boundary were drawn in the dorsal and ventral regions of each wing disc image and fluorescence values along the anterior-posterior axis were calculated using the Plot Profile function of Image J. Dorsal and ventral data were normalized by dividing each data point by the average maximum intensity value from the ventral compartment of that disc (defined as the average of the 10 highest values), effectively converting them to % maximum WT response. Data were then arranged to align the A/P boundaries and pixelby-pixel data from four discs was averaged to yield a mean intensity plot.