A Broadly Conserved G-Protein-Coupled Receptor Kinase Phosphorylation Mechanism Controls Drosophila Smoothened Activity

Hedgehog (Hh) signaling is essential for normal growth, patterning, and homeostasis of many tissues in diverse organisms, and is misregulated in a variety of diseases including cancer. Cytoplasmic Hedgehog signaling is activated by multisite phosphorylation of the seven-pass transmembrane protein Smoothened (Smo) in its cytoplasmic C-terminus. Aside from a short membrane-proximal stretch, the sequence of the C-terminus is highly divergent in different phyla, and the evidence suggests that the precise mechanism of Smo activation and transduction of the signal to downstream effectors also differs. To clarify the conserved role of G-protein-coupled receptor kinases (GRKs) in Smo regulation, we mapped four clusters of phosphorylation sites in the membrane-proximal C-terminus of Drosophila Smo that are phosphorylated by Gprk2, one of the two fly GRKs. Phosphorylation at these sites enhances Smo dimerization and increases but is not essential for Smo activity. Three of these clusters overlap with regulatory phosphorylation sites in mouse Smo and are highly conserved throughout the bilaterian lineages, suggesting that they serve a common function. Consistent with this, we find that a C-terminally truncated form of Drosophila Smo consisting of just the highly conserved core, including Gprk2 regulatory sites, can recruit the downstream effector Costal-2 and activate target gene expression, in a Gprk2-dependent manner. These results indicate that GRK phosphorylation in the membrane proximal C-terminus is an evolutionarily ancient mechanism of Smo regulation, and point to a higher degree of similarity in the regulation and signaling mechanisms of bilaterian Smo proteins than has previously been recognized.


Introduction
The Smoothened (Smo) family of seven-pass transmembrane proteins initiate cytoplasmic Hedgehog (Hh) signaling. Smo proteins are activated by multisite phosphorylation in the cytoplasmic C-terminal tail, which counteracts the electrostatic effects of adjacent clusters of positively charged residues thought to maintain the protein in an inactive conformation. Upon phosphorylation, Smo undergoes a conformational change, dimerizes, and accumulates at the plasma membrane (Drosophila) or primary cilium (mammals), and activates downstream signaling, leading to stabilization of Ci/Gli family transcription factors and Hh target gene expression [1][2][3]. Smo is phosphorylated in a graded manner, with higher levels of Hh ligand inducing more extensive phosphorylation and thereby driving expression of higher-threshold target genes [2,4].
Although the extracellular N-terminus and seven-transmembrane regions of Smo are highly conserved, only the first 100 amino acids of the cytoplasmic C-terminus is broadly conserved. The more distal C-terminus, in many cases hundreds of amino acids long, is completely divergent in different phyla. Because they target different portions of the C-terminus, the phosphorylation mechanisms regulating Smo differ fundamentally between invertebrates and vertebrates. In Drosophila, phosphorylation at three clusters of Protein kinase A (PKA) and Casein kinase I (CKI) sites located in the Smo autoinhibitory domain (SAID) is necessary and sufficient for activation [5][6][7]. However, neither the SAID nor the critical PKA sites in Drosophila Smo are conserved in vertebrate Smo proteins; instead G-protein-coupled receptor kinase (GRK) 2 and CKI are the principal kinases that activate vertebrate Smo proteins by phosphorylating them at a different set of sites [2].
The means by which Smo engages the downstream signaling machinery through its C-terminus also appears to have diverged. In Drosophila and vertebrates, Smo binds to the downstream effector Cos2 and its orthologue Kif7, respectively [8][9][10]. In addition to a negative role in Hh signaling, Cos2 and Kif7 are required for full activation of the pathway [11][12][13], likely in the case of Cos2 because it helps to recruit the kinase Fused to Smo and to activate it [14]. Previous studies localized two separate binding sites for Cos2 in the Drosophila Smo C-terminus [8,9], neither of which is conserved in vertebrates. Given that Drosophila and vertebrate Smo signal through similar effectors [3], this divergence has been puzzling [15].
GRKs are also implicated in Smo activation in Drosophila. Gprk2, one of the two Drosophila GRKs, is required for Smo to drive high-threshold Hh target gene expression [16][17][18]. Two pairs of Gprk2 phosphorylation sites (called GPS1 and GPS2) have been mapped to the non-conserved portion of the Smo Cterminus, with phosphorylation at GPS1 suggested to contribute to the charge mechanism that overcomes inhibition by the SAID [18]. Dimerization of the kinase itself was also suggested to help promote Smo dimerization and activation in a catalytic-activityindependent manner [18]. On the other hand, loss of gprk2 causes a reduction in cyclic AMP (cAMP) levels, affecting PKAdependent Smo activation. Target gene expression can be largely rescued in gprk2 mutants by increasing cAMP levels, suggesting that this indirect effect of Gprk2 on Smo plays an important role in Hh pathway function [19].
To further explore the effects of direct Gprk2 phosphorylation on Smo activity, we mapped four new clusters of Gprk2dependent phosphorylation sites in the cytoplasmic C-terminus. We find that mutation of these sites to non-phosphorylatable residues reduces Smo dimerization and its ability to promote Hh target gene expression. Phosphomimetic mutations bypass the requirement for Gprk2. Importantly, the phosphosites we mapped overlap with CKI/GRK sites mapped in the mouse Smo Cterminus, and are remarkably well-conserved in Smo orthologues throughout the bilaterian lineages. We demonstrate that the evolutionarily conserved core of Smo is a functional, GRKregulated protein that is sufficient to activate downstream signaling, suggesting that all bilaterian Smo proteins share a common regulatory and signaling mechanism.

Results
We previously reported that Gprk2 affects phosphorylation of Smo at sites other than GPS1 and GPS2 [19]. Even when using a PKA-and CKI-phosphomimetic form of Smo, Smo SD [5] to circumvent the effects of gprk2 depletion on cAMP levels, there was a substantial reduction of Smo phosphorylation (detected by increased mobility in SDS-PAGE) upon dsRNA-mediated depletion of Gprk2 that could not be accounted for by the GPS sites ( Figure S1A). Previous analysis suggested that mutation of the GPS sites to non-phosphorylatable Ala residues (in Smo SD.GPSA12 ) had only a small effect on Smo signaling activity [18]. We found that Smo SD and Smo SD.GPSA12 had similar ability to drive expression of the highest-threshold Hh target gene, anterior engrailed (en), in vivo ( Figure S1B and C). These results suggest that the GPS sites account for at most a small fraction of functionally important Gprk2 phosphorylation sites in Smo. In order to localize the functionally important sites, we tested the ability of Gprk2 to phosphorylate a series of C-terminally truncated, GFP-tagged forms of Smo in cells ( Figure 1A). Smo truncated at amino acid 663 (Smo core -GFP), which retains the highly conserved core of the protein but lacks more than 75% of the cytoplasmic tail including all three clustered PKA/CKI sites, still shifted in response to gprk2 depletion ( Figure 1B, lanes 1 and  2). Further truncation to amino acid 603 (Smo D603 -GFP) eliminated this response ( Figure 1B, lanes 5 and 6). The intervening 60 amino acids include 15 Ser/Thr residues, nine of which were previously identified as phosphosites by mass spectrometry [6], grouped into three clusters ( Figure 1A). Mutation of all three Ser/Thr clusters in Smo core to non-phosphorylatable Ala residues (Smo core.c1-3A -GFP) eliminated Gprk2dependent phosphorylation ( Figure 1B, lanes 3 and 4). When fractionated in SDS-PAGE gels containing Phos-Tag, which specifically retards the migration of phosphorylated proteins [20], Smo core -GFP migrated as a high molecular weight smear extending from ,100 kDa to the top of the gel ( Figure 1C, lanes 1 and 5; quantified in Figure 1D). After Gprk2 depletion, Smo core -GFP ran as a more discrete, faster-migrating band ( Figure 1C Figure 1D), whereas Smo core.c1-3A -GFP co-migrated with Smo core -GFP from Gprk2depleted cells ( Figure 1C, lanes 6 and 7). These results suggest that Gprk2 phosphorylates residues in each of the three clusters. However, mutation of these sites in full-length Smo SD reduced but did not eliminate Gprk2-dependent phosphorylation ( Figure S2), indicating that additional sites exist.
To look at Smo phosphorylation more directly, we used a labelfree quantitative liquid chromatography/tandem mass spectrometry (LC-MS/MS) approach to compare the relative levels of phosphorylated:non-phosphorylated forms of individual Smo phosphopeptides in control and gprk2 dsRNA-treated cells (see Materials and Methods) [21]. Consistent with the results from the Smo truncations, we detected phosphorylation at all four Ser/Thr residues in cluster 1 (Ser 604 , Thr 606 , Thr 610 , Thr 612 ), and at three of five sites (Ser 658 , Ser 659 , Ser 660 ) in cluster 3 (Table 1; more detailed quantification included in Table S1). For each phosphospecies we detected, phosphorylation was markedly lower (generally ,10-fold or more) following gprk2 depletion. We were unable to obtain peptide coverage in the region of cluster 2. However, we note that others have shown four of the six Ser/Thr residues in this cluster (Ser 626 , Ser 627 , Thr 629 , Ser 633 ) are phosphorylated in response to Hh [6].
Phosphorylation at a fourth cluster of three Ser/Thr residues (Ser 675 , Ser 679 , Ser 682 ) located in the SAID ( Figure 1A) was also much lower in Gprk2-depleted cells ( Table 1). Mutation of these sites to Ala in the Smo SD.c1-3A background resulted in a protein

Author Summary
Hedgehog proteins are critical regulators of embryonic tissue growth and organization in species ranging from flies to humans. Binding of the secreted Hh protein to its receptor at the surface of cells triggers an intracellular signal that is initiated by Smoothened (Smo). Upon exposure of cells to Hh, Smo becomes active and signals through a series of downstream proteins to regulate gene expression. Although Smo proteins in flies and mammals are similar, the critical regions involved in activation and signal initiation differ between the two, implying that different mechanisms have evolved in different organisms. Using the fruit fly as a model organism, we identified regions in Smo that are phosphorylated by a protein kinase called Gprk2 to enhance Smo activity. These phosphorylation sites overlap with previously identified sites in mouse Smo and are conserved in Smo proteins in many animals. Phosphorylation at these sites regulates the recruitment of Costal2 to Smo, a critical step in signal initiation, through a region of the protein that is also highly conserved. Our results indicate that Gprk2 phosphorylation represents an evolutionarily ancient and conserved mechanism for regulating Smo activity, and suggest that Smo regulation and signaling are more similar between different species than previously thought.
(Smo SD.c1-4A -GFP) that no longer shifted in response to gprk2 depletion ( Figure 1E) confirming that the four mapped phosphorylation clusters account for all or most Gprk2 phosphorylation. In contrast, we consistently observed comparatively small changes in phosphorylation at GPS1 (Ser 741 and Thr 742 ) in gprk2-depleted cells (Table 1). This suggests that another kinase is responsible for a substantial fraction of the phosphorylation at these sites in vivo.
As a validation of the LC-MS/MS results, we generated phosphospecific antisera to cluster 1 sites. In Western blots, both anti-Smo-pSer 604 and anti-Smo-pThr 610 /pThr 612 antisera recognized wild-type Smo-GFP in extracts from Hh-stimulated cells ( Figure 1F and G, lane 1), as well as Smo SD -GFP ( Figure 1F and G, lane 4). The signals were abolished when the cluster 1 sites were mutated to Ala ( Figure 1F and G, lanes 3 and 6), indicating that the antisera specifically recognize the appropriate sites. gprk2 depletion strongly reduced reactivity of both antisera with Smo ( Figure 1F and G, lanes 2 and 5), confirming that the cluster 1 sites are bona fide Gprk2 phosphorylation sites.

Multisite phosphorylation by Gprk2 at clusters 1 and 2 promotes Smo activity
To assess their functional importance, we tested the effects of mutating the Gprk2 phosphosites using ptc-luciferase (ptc-luc) transcriptional reporter assays [22] in S2-R+ cells. In control experiments, gprk2 depletion using a mix of dsRNAs targeting the gprk2 59-and 39-untranslated regions (UTRs) significantly reduced both Smo phosphorylation and ptc-luc reporter activity in Smo SD -GFP transfected cells ( Figure S3A and B). Both effects could be rescued by expressing a wild-type gprk2 transgene lacking the UTRs, but not catalytically inactive mutants ( Figure S3A and B). gprk2 depletion had a similar effect on Hh-dependent signaling by  Figure S4A). We conclude that phosphorylation at clusters 1 and 2, and to a lesser extent at cluster 3, is required for full Smo activation.
In tests of Smo SD -GFP transgenes bearing mutations in only a subset of sites within cluster 1 or 2, we found that no subset impaired activity as much as mutating all sites within a cluster ( Figure S5A). We conclude that phosphorylation at many sites rather than a critical activating residue is important for Smo activity. The partial effect on activity when only some sites are mutated suggests that Gprk2 phosphorylation affects Smo activity in a graded manner.
Next, we tested if mimicking the charge effect of Gprk2 phosphorylation at Smo clusters 1 and 2 could compensate for the lack of phosphorylation in Gprk2-depleted cells. In control cells, the Asp-substituted phosphomimetic Smo SD.c12D -GFP mutant was ,25% less active than Smo SD -GFP ( Figure 2B). This may be because substituting Asp introduces less negative charge than does Ser/Thr phosphorylation under physiological conditions [23]. Importantly, the activity of Smo SD.c12D -GFP was unaffected by Gprk2 depletion ( Figure 2B). This insensitivity to Gprk2 required Asp substitution in both clusters, as each single cluster mutant (Smo SD.c1D and Smo SD.c2D ) showed only a partial resistance to depletion of the kinase ( Figure 2B). Thus mimicking phosphorylation of Smo SD by Gprk2 at clusters 1 and 2 circumvents the requirement for the kinase itself.
To confirm that the effects on Smo SD activity we observed reflect the normal situation, we mutated the Gprk2 phosphosites in a wild-type Smo background. To minimize any complication arising from endogenous Smo activity, we depleted it by treating the cells with smo 39-UTR dsRNA. Smo depletion was efficient, reducing Hh-stimulated ptc-luc reporter activity by 86% ( Figure 2C). As expected, re-expressing a wild-type smo-gfp transgene made insensitive to the dsRNA by removing the 39-UTR (Smo WT -GFP) restored Hh responsiveness ( Figure 2C). Mutation of all four Gprk2 phosphorylation clusters to Ala (Smo c1-4A -GFP) strongly impaired this rescue ( Figure 2C). Analysis of single cluster mutants confirmed that this was mainly attributable to mutation of clusters 1 and 2 ( Figure S5B). In cell surface biotinylation assays, we did not detect any effect of Gprk2 phosphosite mutations on the ability of Smo to traffic to the plasma membrane in response to Hh ( Figure S4B). We conclude that wild-type Smo shows a similar dependence on Gprk2 phosphorylation for full activity as Smo SD . Unlike PKA and CKI phosphorylation [5], Gprk2 phosphorylation was not sufficient to constitutively activate Smo, as the activity of Smo c12D -GFP was as dependent on Hh as Smo WT -GFP ( Figure 2D). It was also not sufficient to enable Hh-dependent activation of Smo that has all three PKA sites in the SAID mutated to Ala (in Smo SA.c12D ) ( Figure 2D). These results suggest that Gprk2 acts downstream of PKA to enhance Smo activity in Hh-responding cells, consistent with the idea that GRKs preferentially phosphorylate GPCRs in their activated state [24].

Direct phosphorylation by Gprk2 promotes Smo dimerization
Phosphorylation by PKA and CKI triggers dimerization of Smo C-terminal tails, which promotes high-level signaling activity [1]. To see if direct phosphorylation by Gprk2 affects Smo dimerization, we adapted previously-described biosensors [1] for use in bioluminescence resonance energy transfer (BRET) experiments to measure Smo SD dimerization in intact cells. In control experiments, co-expression of Smo SD molecules C-terminally-tagged with GFP10 (Smo SD -GFP10) and Renilla luciferase II (Smo SD -Luc) yielded a net BRET signal between the two which responded to Gprk2. Gprk2 overexpression, which increased Smo SD -GFP phosphorylation ( Figure S3B, lanes 1 and 3), also significantly increased net BRET ( Figure 2E). Conversely, gprk2 depletion significantly reduced net BRET ( Figure 2E), consistent with a previous report [18]. Re-expressing Gprk2 restored the BRET signal in gprk2-depleted cells ( Figure 2E), confirming the specificity of the effect of Gprk2 on Smo SD dimerization. Importantly, the cluster 1 and 2 phosphomimetic mutations eliminated the effect of gprk2 depletion on Smo SD dimerization ( Figure 2F). Thus, as in the ptc-luc reporter assay, mimicking phosphorylation of Smo SD at clusters 1 and 2 circumvented the requirement for Gprk2. Taken together, the results of our functional studies suggest that Gprk2 directly enhances dimerization of active Smo by phosphorylating it at clusters 1 and 2, driving it into or stabilizing its most active state.
Loss of direct Gprk2 phosphorylation impairs high threshold Hh target gene expression gprk2 mutant animals display several defects, including ectopic accumulation of Smo in Hh-responding cells and strong impairment or loss of intermediate/high but not low threshold target gene expression in the developing wing imaginal disc. To determine to what extent loss of direct phosphorylation of Smo by Gprk2 contributes to these defects, we expressed the wild-type and mutant forms of the protein during wing development. All transgenes were recombined into the same site in the genome using the WC31-based integration system [25] to ensure equal mRNA expression. For analysis of mutations in the Smo WT background, we co-expressed a smo 39-UTR dsRNA transgene to deplete endogenous Smo. Expression of this dsRNA (together with GFP as a negative control) throughout the developing wing disc using the nub-GAL4 driver strongly suppressed Hh signaling, leading to loss of the central region of the adult wing patterned by Hh ( Figure 3A   and E). Smo c1-4A -GFP had less activity, with no sign of gain-offunction phenotypes ( Figure 3D). However, it rescued development of the central region of the wing, restoring growth to about 84% of the normal size ( Figure 3D and E), suggesting that loss of direct Gprk2 phosphorylation has relatively mild effects on Smo activity in vivo.
Analysis of Hh target gene expression in wing discs led us to a similar conclusion. Expression of smo 39-UTR dsRNA in the dorsal compartment of the disc using ap-GAL4 eliminated expression of the intermediate threshold target gene ptc and high threshold target anterior en, and strongly reduced expression of the low threshold target dpp ( Figure 4A and B). Reintroducing Smo WT -GFP in this background restored Hh-dependent expression of all three genes, with signs of weak ectopic dpp and ptc expression apparent in far anterior cells ( Figure 4C and D). In contrast, Smo c1-4A -GFP rescued low (dpp) but not high (en) threshold target gene expression ( Figure 4E and F), matching observations in gprk2 mutants [16][17][18]. Hh-dependent expression of ptc was partially rescued ( Figure 4E). However, Ptc levels in Smo c1-4A -GFP-expressing dorsal cells were higher than is typically seen in gprk2 mutants, consistent with the observation that reduced levels of cAMP and PKA also contribute to the impairment of Hh signaling activity in the absence of Gprk2 [19]. The Gprk2 phosphosite mutations had a similar partial effect on the activity of Smo SD . Hh-independent ectopic expression of anterior en [5] was substantially lower in Smo SD.c1-4A -GFPexpressing discs than in those expressing Smo SD -GFP, but was still readily detectable ( Figure 4G and H). As in S2 cells, mimicking Gprk2 phosphorylation restored the ability of Smo SD to drive strong ectopic en expression in gprk2 mutant discs ( Figure 4I and J). Taken together, our in vivo analyses support the conclusion that direct phosphorylation by Gprk2 is not essential for Smo activity, but enhances it to its highest level.
Interestingly  [16,18]. We conclude that Gprk2 phosphorylation is required for the downregulation of Smo seen in cells where the Hh signaling pathway is active.  CKI sites, and binding sites for the kinesin-like protein Cos2 [8,9] are only conserved among arthropods. The absence of PKA sites in vertebrate Smo orthologues was recently explained with the identification of GRK2 and CKI as the activating kinases for these proteins. GRK2 and CKI phosphorylate twelve sites in the cytoplasmic tail of mSmo, inducing a change to an active conformation, possibly by neutralizing a nearby stretch of basic amino acids as in flies [1,2]. Most of these sites are conserved in vertebrate Smo orthologues, suggesting that PKA and CKI together replace the function of PKA in activating vertebrate Smo through a mechanism that is similar to, but molecularly distinct from, that in flies [2].
In broader sequence alignments that include Smo orthologues from each of the three main bilaterian clades, we noted that the Gprk2 phosphorylation sites we mapped in Drosophila Smo overlap with the mSmo phosphorylation sites and show a remarkable degree of conservation in other species ( Figure 5A). In total, eight of the sites in Gprk2 clusters 1-3 correspond to phosphorylation sites in mSmo that are nearly universally conserved throughout the bilaterians, either as Ser/Thr residues or as Asp/Glu residues (consistent with the evolution of phosphosites from functionally similar negatively charged residues [23]). The first two clusters (PS0 and PS1) in mSmo, which contain most of the conserved sites, were functionally the most important [2], as are the corresponding cluster 1 and 2 sites in Drosophila. Thus both at the level of sequence and function, the Gprk2/GRK2/CKI phosphorylation sites appear to represent an evolutionarily ancient and conserved mechanism of Smo regulation.
If they share a common regulatory mechanism, we hypothesized that the bilaterian Smo orthologues may also have retained the activity of the ancestral form of Smo that was being regulated, i.e. they may share a conserved molecular signaling mechanism.
To address this, we tested the effects of expressing just the highly conserved portion of Smo (Smo core -amino acids 1-663, truncated just after the broadly conserved third Gprk2 phosphorylation cluster; Figure 5A) on Hh target gene expression. Consistent with our hypothesis, we found that Smo core was capable of activating ptc-luc reporter expression in endogenous Smo-depleted cells, to about 25% the level of Smo SD ( Figure 5B). This constitutive activity of Smo core was strongly reduced or abolished by gprk2 depletion or mutation of the Gprk2 phosphorylation sites (Smo core.c1-3A ), respectively ( Figure 5C), indicating that Smo core -GFP activity is regulated by Gprk2 phosphorylation. In endogenous Smo-depleted wing discs, Smo core -GFP rescued Hh-dependent expression of dpp, ptc, and even en (though en levels remained lower than wild-type) ( Figure 5D and E), as well as overall wing development ( Figure 3F). It also drove ectopic expression of dpp and ptc ( Figure 5D), producing anterior Hh gain-of-function phenotypes in adult wings that were even stronger than those observed with Smo WT -GFP ( Figure 3F). Consistent with the results of ptc-luc reporter assays, Smo core.c1-3A -GFP displayed no ability to rescue target gene expression in endogenous Smo-depleted discs ( Figure 5F and G). Similarly, when expressed in gprk2 mutant discs, the ability of Smo core -GFP to activate ptc expression was lost, and it appeared to inhibit the residual ptc expression resulting from endogenous Smo activity ( Figure 5H). Thus, as in the S2 cell assays, signaling in vivo was dependent on phosphorylation by Gprk2. We conclude that the highly conserved core of Smo is a GRK-regulated protein that contains sequences sufficient for activating downstream signaling.
Interestingly, deleting the C-terminus altered the pattern of Smo accumulation in discs in a manner that was inversely correlated with its activity. Normally, Smo levels are low in the far A compartment, where Smo is inhibited by the activity of Ptc, and high in the P compartment and first few rows of A Hh-responding cells, where Smo activity is high because Ptc is either not expressed (P compartment) or strongly inhibited by Hh (e.g. Figure 4K). In contrast, Smo core -GFP levels were highest in far A cells where Smo core activity is lowest, and low throughout the Hh-responsive A zone and P compartment, where Smo core activity is expected to be highest ( Figure 5I). This pattern is reminiscent of the regulation of many GPCRs, which undergo GRK phosphorylation-dependent internalization and, in many cases degradation, after being activated [27]. Removing Gprk2 or mutating its phosphorylation sites eliminated the downregulation of Smo core in cells where the Hh pathway is strongly activated ( Figure 5J and K), indicating that the effect is indeed the result of direct phosphorylation. Thus Gprk2 controls both the accumulation and activity of Smo core , as GRK2 does ciliary accumulation and activity of mSmo [2].
In Drosophila and vertebrates, Smo binds to Cos2 and its orthologue Kif7, respectively [8][9][10], and these proteins are required for full activation of the pathway [11][12][13]. To see if Smo core recruits Cos2, we adapted previously-described FRET biosensors [14] to measure Cos2-Smo core interaction by BRET. Co-expression of Smo core C-terminally-tagged with GFP10 (Smo core -GFP10) and Cos2 C-terminally tagged with Renilla luciferase II (Cos2-Luc) yielded a net BRET signal that was responsive to Gprk2. Gprk2 overexpression increased Smo core -Cos2 BRET more than 2.2-fold, whereas gprk2 depletion reduced it by 60% ( Figure 6A). The effects were specific, as re-expressing Gprk2 in gprk2-depleted cells fully rescued Smo core -Cos2 BRET ( Figure 6A  Previous studies localized separate binding sites for Cos2 in the Smo C-terminus between amino acids 651-686 and 818-1035 [8,9], most of which is missing in Smo core . To see if the Cos2 recruitment activity we observed was due to binding between amino acid 651 and the end of Smo core at amino acid 663, we made a further truncation to amino acid 651 (see Figure 5A). Smo D651 interacted with Cos2 in the BRET assay even more efficiently than Smo core , in endogenous Smo-depleted cells ( Figure 6B), indicating that recruitment was not to the previously mapped sites. Further truncation to amino acid 625 (Smo D625 ), just N-proximal to Gprk2 phosphorylation cluster 2 ( Figure 5A), eliminated Cos2 recruitment ( Figure 6B). Importantly, the ability of these truncated Smo proteins to recruit Cos2 correlated with their activity in ptc-luc reporter assays ( Figure 6C). We confirmed that all proteins were expressed at similar levels ( Figure S6). These results suggest that the region between 625 and 651 contains a novel Cos2 binding site that positively transduces signals downstream of Smo. Unlike previously mapped sites, this one is located in a region that is broadly conserved among bilaterian species.

Discussion
Previous studies of GRK phosphorylation of Drosophila and mouse Smo proteins identified sites that were not conserved between the two [2,18], consistent with the high degree of sequence divergence in much of the cytoplasmic C-terminus. In fact, a large body of data concerning kinase regulation of Smo proteins and their engagement of the downstream signaling apparatus have supported the view that, although they ultimately do similar things, fly and vertebrate proteins must do so through distinct molecular mechanisms [3,28]. In this study, we characterized the effects of direct Gprk2 phosphorylation on Smo in Drosophila by mapping four new clusters of Gprk2 phosphorylation sites, two of which are situated in the highly-conserved portion of the cytoplasmic C-terminus. We find that direct phosphorylation by Gprk2 is not strictly required for Smo activation, but acts to enhance Smo dimerization and signaling activity in the Hh response. Furthermore, we demonstrate that the highly conserved core of Smo contains sequences sufficient to recruit Cos2 and activate downstream signaling, in a GRK phosphorylationdependent manner. These results suggest that GRK phosphorylation in the membrane proximal C-terminus is an evolutionarily ancient mechanism of Smo regulation, and point to a higher degree of functional similarity at a molecular level among bilaterian Smo orthologues than was previously recognized.

Multisite phosphorylation of Smo by Gprk2
GRKs tend to phosphorylate multiple Ser/Thr residues within short stretches of amino acids in their GPCR substrates [29][30][31][32]. We identified 18 Ser/Thr residues in four such clusters in the Smo C-terminus, mutation of which abolished Gprk2-dependent phosphorylation. Of these, we confirmed by LC-MS/MS analysis that 10 sites (Ser 604 , Thr 606 , Thr 610 , Thr 612 , Ser 658 , Ser 659 , Ser 660 , Ser 675 , Ser 679 , Ser 682 ) are phosphorylated by Gprk2, with Ser 604 , Thr 610 , and Thr 612 being further validated using phosphospecific antisera. Two of the sites in cluster 3 (Thr 551 and Thr 555 ) are likely not Gprk2 targets, as we did not detect phosphorylation at either site in control cells. We did not obtain peptide coverage in the region containing the remaining six sites (Ser 626 , Ser 627 , Thr 629 , Ser 633 , Ser 634 , and Ser 635 ), which make up cluster 2; however, phosphorylation at four of these sites has been observed by others [6]. Our results suggest that Gprk2 does phosphorylate at least some of these residues, since phosphomimetic mutations in cluster 2 are required to fully rescue Smo activity in gprk2-depleted cells. In mSmo, CK1 phosphorylates the sites corresponding to cluster 1 whereas GRK2 phosphorylates cluster 2 [2]. The role of CKI does not seem to be conserved, as CKI depletion had no appreciable effect on phosphorylation of the sites we mapped in Drosophila Smo ( Figure S7). We observed Gprk2-dependent changes in phosphorylation at the GPS1 sites by LC-MS/MS but they were relatively small, suggesting that another kinase also phosphorylates these sites. In total, then, Gprk2 appears to be the principle kinase responsible for phosphorylating between 11 and 16 sites in the Drosophila Smo C-terminus.

Direct phosphorylation by Gprk2 enhances Smo activity and turnover
Gprk2 regulates Smo stability and activity, with both effects mediated at least partly through direct phosphorylation. Several observations indicate that Gprk2 directly enhances Smo activity in Hh-responding cells. The extent of the conformational shift that Smo undergoes upon activation is lower in gprk2-depleted cells, indicative of a lower activity state [18]. We confirmed previous observations that its ability to dimerize is also partly compromised [18]. The result is less robust activation of target gene expression. The reduced activity of Smo mutants with Ala substitutions at the Gprk2 phosphorylation sites in ptc-luc reporter assays confirms that Gprk2 enhances Smo activity by phosphorylating it. In particular, phosphorylation at clusters 1 and 2 seems to be critical, as Ala substitutions at these clusters caused the strongest impairment of target gene activation whereas phosphomimetic substitutions at both rendered Smo resistant to the effects of gprk2 depletion, as assessed by dimerization and target gene activation. The latter observations strongly argue against Gprk2 having a catalytic activity-independent function in regulating Smo, as has been suggested [18]. The effects of mutating the Gprk2 phosphorylation sites in Smo were more subtle in vivo than in ptc-luc reporter assays in S2 cells. We speculate that this is due to the artificial nature of the ptc-luc reporter assay itself. Although Smo SD expression or Hh treatment can yield 50-fold or more activation of the reporter, proteins that induce a ,10-fold increase in these assays (Smo SD.c1-4A , Smo core ) are capable of activating most target gene expression in vivo. Activity above this level may simply be non-physiological.
Although Gprk2 phosphorylation contributes to activating Smo, it is neither necessary nor sufficient. Smo activation appears to be a two-step process, with phosphorylation by PKA and CKI in the SAID serving as the principal trigger ( Figure 7A and B). SAID phosphorylation has at least two effects. First, it inhibits Smo ubiquitination and its subsequent endocytosis and degradation, leading to Smo accumulation at the cell surface [33,34]. Second, it promotes Smo dimerization and a shift to a more active conformation [1]. In our analysis, mimicking PKA/CKI phosphorylation at all nine sites was sufficient for full expression of all target genes except en, which was only partially activated, independent of Gprk2. Thus full PKA/CKI phosphorylation is sufficient to strongly, but not completely, activate Smo. Full activation requires phosphorylation by Gprk2.
Mimicking Gprk2 phosphorylation alone had no effect on Smo activity in the absence of Hh, nor could it activate in the absence of phosphorylation by PKA/CKI. This could be because access to the Gprk2 sites is blocked without prior PKA/CKI phosphorylation. However, it seems more likely that the effect of PKA/CKI in controlling accumulation of Smo at the cell surface, where GRKs are typically localized [35], limits the influence of Gprk2 on Smo. Once Smo accumulates at the cell surface, Gprk2 appears to phosphorylate it constitutively.
PKA and CKI phosphorylation disrupts electrostatic interactions between the SAID domain and the distal C-terminus, thereby promoting a more open and active Smo conformation [1]. Gprk2 phosphorylation appears to act by a different mechanism. The functionally important Gprk2 phosphorylation sites are not located in the SAID domain, and Gprk2 phosphorylation regulates the activity of the truncated Smo core protein that lacks both the SAID domain and distal C-terminus. These observations favour a model in which Gprk2 phosphorylation more directly affects the conformation of the proximal C-terminus or seven transmembrane domain portion of Smo to enhance its activity.
Previous studies have shown that Gprk2 promotes Smo internalization and degradation in response to Hh [16][17][18]. The Smo c1-4A mutant accumulated ectopically in wild-type Hhresponding cells, as endogenous Smo does in gprk2 mutants, demonstrating that Gprk2 triggers Smo turnover by directly phosphorylating it. This is consistent with the typical role of GRKs in receptor desensitization, and supports the conclusion that Gprk2 phosphorylation limits the duration of Smo signaling [16].

Broadly conserved Smo regulatory and signaling mechanisms
The striking conservation of the first two Gprk2 phosphorylation clusters in all bilaterian Smo proteins clearly points to an ancient origin and common function. Indeed, there are some parallels between the functions of Gprk2 phosphorylation of Drosophila Smo and GRK2/CKI phosphorylation of mSmo [2]. In both cases, it is the same two membrane-proximal clusters that are most important for function. Ala substitution of these sites in either protein impairs dimerization of the C-terminal tail and target gene One important difference is that GRK phosphorylation is required and sufficient for Smo activation in mammals but not flies. In this regard, mSmo behaves more like the truncated Smo core protein ( Figure 7C and D). Our analysis indicates that Smo core contains all the sequences necessary for activating downstream signaling, although it may do so less effectively than full-length Smo. Like mSmo, Smo core signaling is strongly or completely inhibited by Ala substitution of Gprk2 clusters 1-3, or by removal of the kinase, indicating that it is strictly dependent upon phosphorylation by Gprk2. Smo core displays some constitutive activity. However, it is also regulated by Ptc. For example, in Smo core -GFP-expressing discs, cells that have higher Ptc activity (such as those in the far A) express target genes at lower levels than Hh-responding cells at the A/P boundary. Ptc overexpression in S2 cells reduces Smo core -driven ptc-luc reporter expression (not shown). Because Ptc downregulates full-length Smo through a mechanism involving ubiquitination of the SAID [36], the absence of this domain in Smo core could explain its accumulation in far A cells. How Ptc regulates Smo core activity is unclear, but could be related to its proposed function in regulating the levels of a Smo agonist or antagonist [37].
Signaling downstream of both Drosophila and mammalian Smo proteins involves Cos2/Kif7. Despite lacking previously mapped Cos2 interaction domains, Smo core is capable of recruiting Cos2, and the ability of successive C-terminally truncated forms of Smo core to do so correlates with their ability to stimulate target gene expression. This truncation approach allowed us to identify a region required for Smo-Cos2 interaction between amino acids 625-651. Gprk2 phosphorylation cluster 2 falls within this region, and the ability of Smo core to recruit Cos2 is strongly influenced by Gprk2 phosphorylation. Phosphorylation may influence Smo core conformation in a way that favours Cos2 interaction. Alternatively, Cos2 may interact with this region preferentially in a phosphorylated state, as b-Arrestins do with GRK-phosphorylated GPCRs [38]. As this region falls in the portion of the Smo C-terminus that is broadly conserved, it could represent a mechanism of Smo regulation and signaling that is common to all bilaterian species, something that has previously been lacking.

Evolutionary origin of Smo proteins
Our analysis of Smo core provides some insights into the potential evolutionary origin of Smo. Smo core is a highly conserved, minimal functional form of Smo, and we speculate that it may closely resemble the ancient form of Smo in the common bilaterian ancestor. It displays a mode of regulation that is typical of GPCR desensitization, suggesting that the ancient form of Smo may have behaved more like a classical GPCR. The evidence suggests that different mechanisms have evolved in different lineages for restricting the activity of Smo core , in the form of C-terminal negative regulatory domains. In vertebrates, a stretch of positively charged residues in the C-terminus serves to keep the core in an inactive conformation, and phosphorylation primarily at the first two GRK/CKI clusters overcomes this effect. This same phosphorylation mechanism has been retained in Drosophila and other arthropods. However, through evolutionary time, this group appears to have acquired a PKA/ CKI-regulated autoinhibitory domain that has come to dominate Smo activity. In contrast to vertebrates, the ancestral GRK mechanism has been relegated to a modulatory role in flies, where it is required to achieve the highest level of Smo signaling. The principal role of driving Smo into an open conformation and activating downstream signaling has been assumed by PKA/CKI phosphorylation of the SAID domain. One consequence of SAID phosphorylation is recruitment of Cos2 and Fu to binding sites in the nonconserved distal C-terminus, leading to Cos2-dependent Fu dimerization and activation [14,39,40]. Fu dimerization is sufficient to strongly activate Hh signaling [14]. This Fudependent mechanism, mediated via the nonconserved Cos2 and Fu binding sites, is not thought to exist in vertebrate Hh signaling, and may account for the difference in signaling strength between full-length Smo and Smo core . Further analysis of Smo core will be necessary for a full understanding of how Drosophila Smo connects to the downstream signaling apparatus, and should provide insights into a signaling mechanism common to all Smo proteins. The resulting EcoRV-NotI fragments were cloned into Smo expression vectors. All Smo constructs were C-terminally tagged with either GFP, GFP10 or RLucII. The tags were engineered as a cassette flanked by NotI and KpnI restriction sites. Coding fragments were cloned into expression constructs for use in cell culture (pRmHa3.puro [26] containing the metallothionein promoter) and flies (pUAST-AttB [25]). UAS-smo-39UTR-dsRNA was generated by cloning a genomic PCR-generated fragment containing nucleotides 2L:281756..281981 of the smo 39-UTR. The 226 nt long fragment was cloned between the EcoRI-AvrII sites and in the opposite orientation between the NheI-XbaI sites of pWIZ (Drosophila Genomics Resource Centre). To generate catalytically inactive forms of Gprk2, point mutations changing Lys 338/339 RMet (Gprk2 kd1 [18]) or Asp 453 RAsn (Gprk2 kd2 ) were generated by PCR mutagenesis and cloned downstream of a Myc-epitope tag in pRmHa3.puro. A C-terminal luciferase-tagged Cos2 expression construct used in BRET assays was engineered by flanking the Cos2 coding sequence at the 59 and 39 end with an EcoRI and a NotI site, respectively. The resulting EcoRI-NotI fragment was cloned into pRmHa3.puro. The RLucII cassette described above was cloned downstream of the Cos2 sequence at the NotI site. Fu coding sequence was cloned downstream of a Myc-epitope tag in pRmHa3.puro. Sequences of all constructs were verified. For experiments involving Hh treatment, cells were co-transfected with pRmHa3.puro/Hh N [26], which encodes an active Nterminal fragment of Drosophila Hh. For ptc-luc reporter assays a mixture of the following constructs was used: pRmHa3/Ci (a gift from S. Cohen, University of Copenhagen, Denmark), pGL.basic/ptcD136-luc [22], and pRL/CMV (Promega).
Anti-pSer 604 and anti-pThr 610 /pThr 612 phosphospecific antisera were generated by GenScript. Rabbits were immunized with phosphorylated peptides (KGRL{pS}ITLYNTHC or CSI-TLYN{pT}H{pT}DPVGL), and antibody was isolated from serum by modified peptide affinity column purification and unmodified peptide cross-adsorption. Anti-Smo antibody was raised in guinea pigs against the same His-tagged fragment of the Smo C-terminus (amino acids 560-1036) as previously used [26]. Other antibodies were: rabbit a-GFP (Torrey Pines Scientific); rabbit a-b-galactosidase (Santa Cruz Biotechnology); mouse a-Ptc (ApaI; developed by I. Guerrero) and mouse a-En (4D9; developed by C. Goodman) were obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at the Department of Biology, University of Iowa.  [16]. For processing of adult wings, flies were collected in 50% ethanol/50% glycerol. After rinsing with water, wings were transferred into a drop of Faure's solution on glass slides and cover-slipped. For imaginal disc analyses, wandering third instar larval wing discs were dissected in phosphate-buffered saline (PBS) and kept on ice for a maximum of 20 min before fixation in PBS/ 0.2% Tween (PBT) containing 4% parafomaldehyde for 20 min. Discs were washed three times in PBT, followed by incubation for 30 min in PBT with 0.1% BSA (BBT). Primary antibodies were diluted in BBT, added to the discs and incubated over night at 4uC. After four washes with PBT, the discs were incubated with fluorescently-labeled secondary antibodies (Invitrogen and Jackson ImmunoResearch Laboratories) diluted in BBT, for 2 hours at room temperature. After four to five more washes with PBT, discs were mounted on slides in mounting medium (10% PBS, 90% glycerol, 0.2% n-propyl gallate), cover-slipped, and imaged using a Zeiss LSM700 confocal microscope. Most experiments were performed using S2-R+ cells grown in Drosophila Schneider's medium (Lonza) supplemented with 10% fetal bovine serum (Gibco) and 50 U/ml penicillin and streptomycin (Gibco). Exceptionally, experiments for Figures 1B and C were performed using Drosophila S2 cells adapted to growth in serum-free medium (EX-CELL 420, Sigma), which show more pronounced Smo phosphoshifts due to a higher basal level of phosphorylation [19]. Cells were cultured at 25uC unless otherwise indicated. dsRNA was prepared by in vitro transcription using templates PCR-amplified from genomic DNA (nt 281756..281981 of genomic scaffold 2L for smo 39-UTR; nt 372-870 of the gprk2 coding sequence; nt 27259182..27259345 of genomic scaffold 3R for gprk2 59-UTR; and nt 27282732..27283011 of genomic scaffold 3R for the 39-UTR of gprk2). b-Gal dsRNA was used as a control. Forward and reverse primers included T7 (59-TAATACGACT-CACTATAGGGAGA-39) and T3 (59-AATTAACCCTCAC-TAAAGGGAGA-39) promoter sequences, respectively. Top and bottom strand RNAs were generated using MEGAscript T7 and T3 in vitro transcription kits, mixed in equal amounts, and heated to 95uC followed by slow cooling to room temperature to anneal.

Genotypes
For biochemical analysis, ,1610 6 cells were typically plated on day 1 in 24 well plates in 0.5 ml of complete Schneider's medium and each well was transfected with 100-250 ng of the indicated pRmHa.puro expression constructs using X-tremeGENE HP transfection reagent (Roche) according to manufacturer's instructions. On day 2, the cells of each well were split into 2 new wells of a 24 well plate and treated with 5 mg of the indicated dsRNA. On day 3 to 4, a second dose of dsRNA was applied and transgene expression was induced by addition of CuSO 4 to a final concentration of 0.5 mM. Cells were harvested and processed on day 7.
For ptc-luc reporter assays, S2-R+ cells were transfected in 24 well plates on day 1 of the experiment as described above. 100 ng pRmHa/Ci, 75 ng pGL.basic/ptcD136-luc [22], 75 ng pRL/CMV, and 100 ng of each additional expression plasmid (Smo/Gprk2 variant; Hh N , as indicated) were typically used. Total DNA amounts in the transfection mix were normalized using empty pRmHa.puro vector. On day 2 the cells were split into 4 wells of a 96 well plate and each well was treated with 0.5-1 mg dsRNA. On day 3 or 4 transgene expression was induced by addition of CuSO 4 and a second dose of dsRNA was administered. Cells were processed on day 7 and luciferase activity measured using the Dual Luciferase Reporter system (Promega) according to manufacturer's instructions. Assays were performed at least two times in quadruplicate, and the data was pooled. Statistical significance was assessed using two-tailed Student's t-tests.
For bioluminescence resonance energy transfer (BRET) experiments measuring Smo dimerization S2-R+ cells were transfected with 100 ng of the Smo SD -RLucII variant, 300 ng of the Smo-GFP10 variant and 100 ng of myc Gprk2 (if applicable) per well of a 24-well plate. For BRET assays monitoring recruitment of Cos2 75 ng of Cos2-RLucII, 300 ng of the indicated Smo-GFP10 variant, and 75 ng myc Fu plasmids were transfected. Cells were replated in 4 wells of a white-walled 96-well plate and subjected to dsRNA treatments and transgene induction as described above. BRET measurements were performed on day 7 as previously described [19]. Assays were performed at least two times in quadruplicate, and the data was pooled.
Immunoprecipitations, SDS-PAGE, and immunoblotting  Figure 1C, Phos-tag acrylamide (Wako Pure Chemicals Industries, Ltd.) was added to a final concentration of 7.5 mM to improve resolution of phosphoproteins [20]. Fractionated proteins were transferred to nitrocellulose membranes using a wet transfer apparatus and immunoblotted according to standard methods. Quantitation of signal intensity was performed using the Gels.Plot Lanes function of ImageJ 1.42q. Plots were normalized to equal total signal intensity (area under the curve), to correct for differences in loading.

Immunoaffinity purification of Smo
S2-R+ cells were plated in 1 to 3 wells per condition of a 6-well plate and transfected with 2.5 mg/well of pRmHa3.puro/Smo SD -GFP as above. A day later, medium was replaced and 20 mg/well control (b-gal) or gprk2 dsRNA was added to the cells. After three days of growth, the cells were harvested and replated in a 10-cm plate, along with 100 mg/plate of the appropriate dsRNA. Smo SD expression was induced by addition of 0.5 mM CuSO 4 . 2-3 d later, cells were washed with ice-cold PBS and lysed in 3 ml RIPA buffer for 15 min on ice. Lysates were cleared by microcentrifugation for 15 min at 12,0006 g and 4uC. Smo SD -GFP was immunoprecipitated using anti-GFP mAb agarose for 2 h at 4uC with rotation. Beads were washed 4 times with ice-cold RIPA buffer before addition of 16 SDS-PAGE sample buffer and heating at 75uC for 6 min. Samples were frozen at 280uC and typically 2 to 3 such preps were pooled for subsequent analysis. Pooled samples were fractionated by SDS-PAGE on 4-15% polyacrylamide gradient gels (BioRad), stained using Colloidal Blue (Life Technologies) according to manufacturer's protocol, and the band corresponding to Smo was excised from the gel.

Protein digestion
Gel pieces were washed with water for 5 min and destained twice with the destaining buffer (50 mM ammonium bicarbonate, acetonitrile) for 15 min. An extra wash of 5 min was performed after destaining with a buffer of ammonium bicarbonate (50 mM). Gel pieces were then dehydrated with acetonitrile. Proteins were reduced by adding the reduction buffer (10 mM DTT, 100 mM ammonium bicarbonate) for 30 min at 40uC, and then alkylated by adding the alkylation buffer (55 mM iodoacetamide, 100 mM ammonium bicarbonate) for 20 min at 40uC. Gel pieces were dehydrated and washed at 40uC by adding ACN for 5 min before discarding all the reagents. Gel pieces were dried for 5 min at 40uC and then re-hydrated at 4uC for 40 min with enzyme solution. Tryptic digestion was performed with a 6 ng/ml solution of sequencing grade trypsin from Promega in 25 mM ammonium bicarbonate buffer, incubated at 58uC for 1 h and stopped with 15 ml of 1% formic acid/2% acetonitrile. Chymotryptic digestion was performed with a 40 ng/ml solution (Roche) in 100 mM Tris HCl-25, mM CaCl2, pH 8 buffer, incubated at 25uC for 4 h and stopped with 15 ml of 1% formic acid/2% acetonitrile. Supernatant was transferred into a 96-well plate and peptide extraction was performed with two 30-min extraction steps at room temperature using the extraction buffer (1% formic acid/50% ACN). All peptide extracts were pooled into the 96-well plate and then completely dried in vacuum centrifuge. The plate was sealed and stored at 220uC until LC-MS/MS analysis. Protein digestion with Asp-N was performed in solution on tryptic digests. Samples were re-solubilized in a 50 mM ammonium bicarbonate buffer and 1 ng of Asp-N was added to each sample. Samples were incubated at 37uC for 3 h.

Liquid chromatography/tandem mass spectrometry (LC-MS/MS) analysis
Prior to LC-MS/MS, peptide extracts were re-solubilized under agitation for 15 min in 11 ml of 0.2% formic acid and then centrifuged at 2000 rpm for 1 min. The LC column was a C18 reversed-phase column packed with a high-pressure packing cell. A 75 mm i.d. Self-Pack PicoFrit fused silica capillary (New Objective, Woburn, MA) of 15 cm length was packed with the C18 Jupiter 5 mm 300 Å reverse-phase material (Phenomenex, Torrence, CA). This column was installed on the Easy-nLC II system (Proxeon Biosystems, Odense, Denmark) and coupled to the LTQ Orbitrap Velos (ThermoFisher Scientific, Bremen, Germany) equipped with a Proxeon nanoelectrospray ion source. The buffers used for chromatography were 0.2% formic acid (buffer A) and 100% acetonitrile/0.2% formic acid (buffer B). During the first 12 min, 5 ml of sample were loaded on column with a flow of 600 nl/min and, subsequently, the gradient went from 2-80% buffer B in 60 min at a flow rate of 250 nL/min and then came back at 600 nL/min to 2% buffer B for 10min. LC-MS/MS data acquisition was accomplished using an eleven scan event cycle comprised of a full scan MS for scan event 1 acquired in the Orbitrap which enables high resolution/high mass accuracy analysis. The mass resolution for MS was set to 60,000 (at m/z 400) and used to trigger the ten additional MS/MS events acquired in parallel in the linear ion trap for the ten most intense ions. Mass over charge ratio range was from 360 to 2000 for MS scanning with a target value of 1,000,000 charges and from ,1/3 of parent m/z ratio to 2000 for MS/MS scanning with a target value of 10,000 charges. The data-dependent scan events used a maximum ion fill time of 100 ms and 1 microscan. Target ions already selected for MS/MS were dynamically excluded for 25 s. Nanospray and S-lens voltages were set to 0.9-1.8 kV and 50 V, respectively. Capillary temperature was set to 225uC. MS/MS conditions were: normalized collision energy, 35 V; activation q, 0.25; activation time, 10 ms.

Peptide identification and quantification
The peak list files were generated with extract_msn.exe (version January 10, 2011) using the following parameters: minimum mass set to 600 Da, maximum mass set to 6000 Da, no grouping of MS/MS spectra, precursor charge set to auto, and minimum number of fragment ions set to 10. MS/MS spectra were queried against the Smo SD sequence using Mascot 2.3 (Matrix Science). The mass tolerances for precursor and fragment ions were set to 10 ppm and 0.6 Da, respectively. Search parameters allowed for up to two missed enzyme cleavages. Oxidation of methionine and phosphorylation of serine, threonine and tyrosine were allowed as variable modifications while carbamidomethyl was set as a fixed modification. Matches for phosphopeptides were validated manually. In a few cases (twice phosphorylated species of cluster 1 peptide W.AKRKDFEDKGRLSITLY.N in Chymotrypisin digest, once and twice phosphorylated species of the cluster 3 peptide R.MALTGAATGNSSSHGPR.K in trypsin+AspN digests), the phosphopeptides were not confirmed by MS2, but were detected in full scan with mass accuracies of less than 2 ppm, and eluted with very similar retention times to other phosphospecies of the same peptide. Peptides were quantitated by manual integration of precursor ion LC spectra using Qual Browser (Xcalibur from Thermo Scientific) [21,41]. For each phosphopeptide identified, the relative level of phosphorylation in each sample was calculated as the ratio of the amount of phosphorylated: non-phosphorylated forms of the peptide.