Research ArticleDevelopment

Sequential Phosphorylation of Smoothened Transduces Graded Hedgehog Signaling

See allHide authors and affiliations

Science Signaling  05 Jul 2011:
Vol. 4, Issue 180, pp. ra43
DOI: 10.1126/scisignal.2001747

Abstract

The correct interpretation of a gradient of the morphogen Hedgehog (Hh) during development requires phosphorylation of the Hh signaling activator Smoothened (Smo); however, the molecular mechanism by which Smo transduces graded Hh signaling is not well understood. We show that regulation of the phosphorylation status of Smo by distinct phosphatases at specific phosphorylated residues creates differential thresholds of Hh signaling. Phosphorylation of Smo was initiated by adenosine 3′,5′-monophosphate (cAMP)–dependent protein kinase (PKA) and further enhanced by casein kinase I (CKI). We found that protein phosphatase 1 (PP1) directly dephosphorylated PKA-phosphorylated Smo to reduce signaling mediated by intermediate concentrations of Hh, whereas PP2A specifically dephosphorylated PKA-primed, CKI-phosphorylated Smo to restrict signaling by high concentrations of Hh. We also established a functional link between sequentially phosphorylated Smo species and graded Hh activity. Thus, we propose a sequential phosphorylation model in which precise interpretation of morphogen concentration can be achieved upon versatile phosphatase-mediated regulation of the phosphorylation status of an essential activator in developmental signaling.

Introduction

The Hedgehog (Hh) pathway, one of the few signaling cascades that operate repeatedly throughout development, coordinates cell-fate decisions, tissue patterning, and organ growth (15). Initiation of Hh signaling requires the Hh receptor Patched (Ptc) and the activator protein Smoothened (Smo). We previously demonstrated a critical role of cell surface–localized Smo in mediating Hh signaling in Drosophila melanogaster (6). Subsequent studies by other groups correlated this surface localization event with enhanced phosphorylation of Smo. Upon activation of cells by Hh (79), Smo is phosphorylated by adenosine 3′,5′-monophosphate (cAMP)–dependent protein kinase (PKA), which readies, or “primes,” Smo for phosphorylation by casein kinase I (CKI). Three consensus clusters for phosphorylation by PKA and CKI are present in the cytoplasmic tail of Smo. Mutation of serine residues to aspartic acid residues in these clusters to mimic enhanced phosphorylation of Smo results in constitutive localization of Smo at the cell surface in cultured fly cells as well as in ectopic Hh signaling in wing discs (79). These results highlight an essential role for the enhanced phosphorylation of Smo in mediating Hh signaling, which raises an important question as to how PKA- and CKI-phosphorylated Smo is regulated in response to the binding of Hh to Ptc.

One potential mechanism by which Hh might activate Smo is through the regulation of either PKA or CKI; however, Hh signaling did not seem to increase the abundance or activity of either kinase (fig. S1) (1, 1012), although a report suggests that Hh signaling reduces the concentration of intracellular cAMP in cl-8 cells, which are derived from Drosophila wing imaginal discs (13), potentially reducing the activation of PKA. Protein phosphorylation is a reversible process that is coordinated by opposing kinases and phosphatases. The inability of Hh signaling to increase the activities of PKA or CKI suggests the possibility that differential Smo phosphorylation might be achieved by protein dephosphorylation through the actions of phosphatases.

Protein phosphatases are increasingly appreciated to play specific roles in many functions, ranging from developmental signaling (14, 15), circadian rhythms (16, 17), and DNA damage repair (18, 19) to asymmetric cell division (20, 21). Although most protein dephosphorylation is mediated by a small number of protein serine or threonine phosphatases (PSTPs), the substrate specificity and selectivity, subcellular localization, and physiological regulation of these PSTPs are conferred by their obligate, highly diverse regulatory subunits (22, 23). In combination with large arrays of kinase catalytic activities, multimeric phosphatase-mediated protein dephosphorylation provides reversible and precise control of signaling cascades in response to cellular and environmental inputs that are essential for development.

Protein phosphatase 4 (PP4), a member of the PSTP family, is thought to potentially regulate the phosphorylation status of Smo; however, PP4 may act on phosphorylated residues other than those in the PKA-CKI clusters (24). Because PKA- and CKI-phosphorylated Smo plays a critical role in activating Hh signaling, an understanding of how Smo-specific phosphatases antagonize the activities of PKA and CKI will be imperative for controlling Smo-mediated Hh signaling. Here, we identified PP1 and PP2A as Smo-specific phosphatases that antagonize the activities of PKA and CKI, respectively. We describe a previously uncharacterized molecular mechanism in which two distinct Smo phosphorylation states, which are specifically regulated by distinct phosphatases, transduce graded Hh signaling. Our data reveal a model of sequential Smo phosphorylation in which the identities of phosphorylated residues in PKA-CKI clusters ensure precise interpretation of the Hh gradient.

Results

Inhibition of PSTP activity activates Smo-mediated Hh signaling

To study the role of phosphatases in regulating the dephosphorylation of Smo, we treated Hh-responsive, wing disc–derived cl-8 cells or cl-8–gsmo cells that stably express a functional gfp-smo fusion (6) with okadaic acid (OA), a specific PSTP inhibitor (25), in the absence of exogenous Hh (fig. S2, A to N). Phosphorylation of endogenous Smo and ectopic green fluorescent protein (GFP)–Smo in response to OA was enhanced compared to that of untreated cells (fig. S2, A and B); both proteins were localized to the surface of OA-treated cells (fig. S2, E and K). OA was sufficient to activate Hh signaling, as indicated by a ptc-luciferase reporter (fig. S2O), highlighting the existence of at least one OA-sensitive PSTP in the inhibition of Smo activity.

The activities of several Hh signaling components downstream of Smo, including Hh signaling transcription factor Cubitus interruptus (Ci), can be modulated through phosphorylation (14). To exclude the possibility that OA-induced activation of Smo was a consequence of feedback regulation of Hh signaling, we treated S2 cells, which do not express ci (26), with OA. We observed similar OA-dependent effects on Smo phosphorylation (fig. S3T) and localization (fig. S3, G to I) in S2 cells as were observed in cl-8 cells, indicating that phosphatase activity likely had a direct effect on Smo activity.

PP1 and PP2A differentially regulate the phosphorylation and localization of Smo

To identify phosphatases that directly dephosphorylated Smo, we performed two RNA interference (RNAi) screens with a double-stranded RNA (dsRNA) library (table S1) targeting the catalytic subunits of 28 PSTPs and 18 dual-specificity phosphatases in the Drosophila genome (27). The first screen was based on the observation that the blocking of phosphatase activity promoted the constitutive localization of GFP-Smo to the surface of cl-8 cells. Inhibition of microtubule star (mts), which encodes the catalytic subunit of PP2A, resulted in the localization of GFP-Smo to the cl-8 cell surface in the absence of exogenous Hh (Fig. 1C). The targeting, with dsRNAs, of other OA-sensitive phosphatases, including PP1, PP2B, PP4, PP5, and PP7 (25), had no effect on the localization of Smo. PP4 regulates the surface localization of Smo only in the presence of Hh (24). We failed to observe any effect of targeting PP4, which is consistent with the previous report given that our screen was conducted without added Hh (see Discussion).

Fig. 1

PP1 and Wdb-PP2A differentially regulate the phosphorylation and localization of Smo. (A to L) GFP-Smo localization in cl-8–gsmo cells transfected with the indicated dsRNAs in the absence of exogenous Hh. (A) gal80 dsRNA was used as a negative control because gal80 is not present in the fly genome, whereas (B) ptc dsRNA was a positive control, which resulted in the localization of Smo to the cell surface (arrows). dsRNAs targeting the PP2A catalytic subunit Mts (C), the scaffolding subunit CG17291 (D), or one of the regulatory B subunits Wdb (E) promoted the surface localization of Smo. dsRNAs for the other PP2A B subunits (F to H) or any of four PP1c subunits (I to L) had no effect on the localization of Smo. Scale bar, 15 μm. (M and N) Western blotting (WB) analysis of cl-8–gsmo cells transfected with dsRNAs in the absence of exogenous Hh. Slow-migrating, phosphorylated forms of GFP-Smo (GFP-pSmo) were detected with an antibody against GFP. β-Tubulin served as a loading control. (O) Western blots of cl-8–gsmo cells treated with either Hh or TC or transfected with the nipp1 plasmid in the absence of Hh. (P) Luciferase assays for cl-8–gsmo cells transfected with the ptc-luciferase plasmid. Cells were either treated with Hh for 24 hours or transfected with the indicated dsRNAs and cultured in the absence of exogenous Hh for 4 days. SDs are shown (n = 3 experiments). *P < 0.005.

The PP2A holoenzyme consists of a scaffolding A subunit, a regulatory B subunit, and a catalytic C subunit (22, 23). The fly genome encodes one A (CG17291), one C (Mts), and four B subunits, Twins (Tws), Widerborst (Wdb), Well-rounded (Wrd), and CG4733. Consistent with previous observations that removing either the A or the C subunit destabilizes PP2A and reduces its activity (17, 28, 29), we found that RNAi targeting of the A subunit induced the surface localization of GFP-Smo (Fig. 1D). Although the A and C subunits are essential for the activity of PP2A, the variable regulatory B subunits confer substrate specificity to the phosphatase. By targeting individual B subunits with the appropriate dsRNAs, we identified Wdb as the specific regulatory subunit that controlled the localization of Smo (Fig. 1E and fig. S3, P to R) and its signaling activity (Fig. 1P) in cl-8 and S2 cells.

Because only the enhanced phosphorylated forms of Smo localize to the cell surface (8), our cell-based screen might not detect phosphatases that act on moderately phosphorylated forms of Smo. Our second screen directly examined the constitutive phosphorylation of GFP-Smo in RNAi-treated cells. As expected, wdb-specific RNAi resulted in the enhanced phosphorylation of Smo compared to mock-treated cells (Fig. 1M and fig. S3U). In addition, inhibiting any of the four PP1 catalytic (PP1c) subunits moderately induced the phosphorylation of Smo in cl-8 cells (Fig. 1N); however, induction of the phosphorylation of Smo in response to PP1c-specific RNAi was not sufficient to promote the surface localization of Smo (Fig. 1, I to L). PP1c subunits are conserved, with 88% protein sequence identity and indistinguishable activities in vitro (16, 23). To address the redundancy of PP1c subunits, we inhibited PP1 activity by treating cells with tautomycetin (TC), a cell-permeable specific inhibitor of PP1 (30), or by overexpressing nipp1, which encodes an endogenous PP1c inhibitor that directly binds to PP1c (31). NIPP1 has been successfully used as a potent PP1c inhibitor in cultured cells and transgenic flies (16, 3235). Consistent with the effects of pp1-specific RNAi, GFP-Smo was moderately phosphorylated (Fig. 1O) but failed to localize to the cell surface in cl-8 cells treated with TC or in which nipp1 was overexpressed. Thus, we identified two phosphatases, PP1 and PP2A, that regulated different aspects of Smo phosphorylation (moderate and enhanced phosphorylation of Smo, respectively), which led to distinct outcomes in terms of the surface localization of Smo. These results were further supported by biochemical analyses that demonstrated that individual PP1c subunits and the PP2A Mts and Wdb subunits formed complexes with GFP-Smo in cl-8 cells (Fig. 2, A to C). Furthermore, PP1 and PP2A subunits directly interacted with the recombinant fusion proteins of glutathione S-transferase (GST) and Smo (GST-Smo) in vitro (Fig. 2, D to H), consistent with a general mechanism in which a direct interaction with its respective substrates is required for PSTP function (22).

Fig. 2

Smo forms complexes with PP1c and Wdb-PP2A. (A to C) Immunocomplexes formed in (A) cl-8–gfp, (B) cl-8–gsmo, or (C) cl-8–gsmo cells transfected with individual pp1c-V5 plasmids. GFP or GFP-Smo was immunoprecipitated (IP) with an antibody against GFP conjugated to agarose. GFP-Smo formed complexes with PP2A catalytic (Mts) and regulatory (Wdb) subunits as well as with PP1c subunits. GAPDH served as a negative control. (D and E) Three versions of recombinant GST-Smo cytoplasmic tail fusions (D), all of which contain three PKA-CKI consensus clusters (circles), were generated in bacteria (E) for in vitro direct binding assays with PP1 or PP2A. Only GST-Smo 557–1036 (asterisk) contains a putative type II PP1-binding site (F-xx-R/K-x-R/K), which is marked by a solid square at amino acid positions 597 to 602 (81). (F to H) A direct interaction between GST-Smo and PP2A or PP1c. GST-Smo bound on glutathione beads was incubated with (F) the purified PP2A A:C (Mts) dimer, (G) MBP-Wdb, or (H) recombinant PP1c. Compared with the shorter GST fusions, GST-Smo 557–1036 showed stronger binding to Mts (panel F, top). The binding between GST-Smo and the PP2A A scaffolding subunit was detectable only in a low-stringency wash condition (panel F, bottom). (G) MBP-Wdb efficiently pulled down GST-Smo 600–800 and GST-Smo 557–1036 (asterisk). (H) Only GST-Smo 557–1036, which contains the PP1-binding site, pulled down recombinant PP1c. GST served as a negative control.

Manipulation of phosphatase activity mimics Hh signaling–related adult wing phenotypes

To study the physiological consequences of differential dephosphorylation of Smo, we examined the roles of PP1 and PP2A in regulating Hh signaling in vivo. Because PP1c subunits function redundantly in flies, it is not surprising that overexpressing pp1c (any isoform) has little effect on the development of the adult wing (33). Inhibition of PP1 activity by overexpressing nipp1 produced a mild wing phenotype (fig. S4C) that is often associated with the overexpression of smo (6, 7, 36). In contrast, manipulation of the activities of Wdb-containing PP2A (Wdb-PP2A) by overexpressing wild-type or dominant-negative wdb (wdbDN; N-terminal truncated Wdb) (28) substantially disrupted patterning in the adult wing (fig. S4, D and F), which is reminiscent of stereotypical Hh signaling phenotypes (37, 38). Wing discs ectopically expressing wdb exhibited an about threefold increase in PP2A activity compared to that in wing discs overexpressing wdbDN, as quantified by a malachite green colorimetric assay for phosphatase activity (39).

Alterations in PP1 and PP2A activities have distinct consequences for Hh signaling

During patterning in the developing wing, distinct target genes are activated in response to different concentrations of Hh (24). Low-threshold Hh signaling stabilizes full-length Ci (CiFL) (Fig. 3A), which translocates to the nucleus to activate Hh target genes. The expression of one of these targets, decapentaplegic (dpp), is induced by intermediate-threshold Hh activity (Fig. 3B), whereas two other targets, ptc and collier/knot (col), respond to high-threshold Hh signaling (Fig. 3, C and D). We measured the abundance of dpp mRNA rather than that of dpp-lacZ because we found that dpp mRNA was a more sensitive indicator of intermediate-threshold Hh signaling activity than was dpp-lacZ (fig. S5).

Fig. 3

PP1 and Wdb-PP2A regulate distinct thresholds of Hh signaling. Stabilization of CiFL, induction of dpp, and induction of ptc and production of Col protein in wing discs correlate with low-, intermediate-, and high-threshold Hh signaling, respectively. Immunofluorescence was used to visualize CiFL, Col, Smo, and Ptc proteins, whereas in situ hybridization was used to detect dpp and ptc mRNA, as indicated in the panels. (A to E) Normal Hh-responsive gene expression patterns. (A) MS1096-Gal4–driven mCD8-gfp was expressed to a greater extent in the dorsal (d) as compared to the ventral (v) compartment of the wing disc. The dorsal-ventral (d/v) boundary is marked by a dashed line in panel A. (F to T) Distinct effects on Hh-responsive gene expressions in wing discs overexpressing (F to J) the PP1 inhibitor NIPP1, (K to O) WdbDN, or (P to T) wild-type Wdb driven by MS1096-Gal4. wdbDN encodes a dominant-negative mutant that inhibits the function of Wdb-PP2A (28). (U to Y) Clonal analyses of wdb function in wing discs. Ectopic wdb expression in clones (U to W, marked by mCD8-GFP; arrows) abolished Ptc protein production (W, right image), but stabilized CiFL (U, right image) and induced ectopic dpp transcription (V, right image) in anterior clones located away from the AP boundary. Both CiFL and Smo were stabilized in wdbIP loss-of-function clones (X and Y, identified by the absence of GFP; arrows). Magnified images of the areas boxed in (U to W, left panels) and (X) are shown in (U to W, right panels) and (Y), respectively. wdb-overexpressing clones were generated by the flip-out technique (54), as described in Materials and Methods.

Inhibition of PP1c activity through the overexpression of nipp1 in the dorsal compartment of wing discs stabilized CiFL and increased the extent of dpp expression (Fig. 3, F and G). Consistent with the inability of PP1 to promote the enhanced phosphorylation and surface localization of Smo in cl-8 cells, a reduction in its activity had a mild effect on the stabilization of Smo (Fig. 3J) and failed to induce the expression of high-threshold Hh targets in wing discs (Fig. 3, H and I). Another nipp1 transgene (nipp1-HA) produced the same effects in wing discs (fig. S4, G to I). These data suggest that PP1-regulated phosphorylated Smo (pSmo) species transduced low- to intermediate-threshold Hh responses (CiFL and dpp, respectively) but probably not high-threshold Hh signaling targets (ptc and Col).

A reduction in Wdb-PP2A activity through the expression of WdbDN stabilized anteriorly localized Smo and CiFL (Fig. 3, K and O) and induced both intermediate- and high-threshold Hh signaling to the anteriormost regions (Fig. 3, L to N). Conversely, increased wdb expression was sufficient to destabilize Smo, even in the posterior compartment where ci is not expressed (Fig. 3T). Furthermore, high-threshold Hh targets (ptc and Col) were inhibited along the anterior-posterior (AP) boundary (Fig. 3, R and S, and fig. S4L), indicating that Smo species dephosphorylated by Wdb-PP2A were required for high-threshold Hh signaling. The area of low- to intermediate-threshold Hh signaling was expanded in wdb-overexpressing discs at 18° and 25°C (Fig. 3, P and Q, and fig. S4K). Similarly, overexpression of ectopic wdb in clones (expressed by the strong Gal4 driver) away from the AP boundary stabilized CiFL and induced dpp expression (Fig. 3, U and V). This suggests that the contrasting responses of low- to intermediate-threshold and high-threshold Hh signaling to the increased activity of Wdb-PP2A most likely reflected the intrinsic signaling ability of pSmo. This signaling activity is subject to inhibition by Wdb-PP2A, rather than occurred as the consequence of the expanded Hh protein gradient that resulted from the reduced expression of ptc at the AP boundary (Fig. 3W) (40). This interpretation was further supported by observations in wing discs overexpressing ck1α-specific RNAi, a transgene that regulates endogenous Smo phosphorylated by CKI (7). In this genetic background, the areas of expression of CiFL and dpp in wing discs were expanded (fig. S4, M to R).

Wdb-PP2A functions at the level of Smo to regulate Hh signaling

Our findings suggested that Wdb-PP2A inhibited Hh signaling, most probably at the level of Smo. We inhibited Wdb-PP2A activity in multiple wdbDN transgenic lines that specifically disrupt endogenous Wdb function in wing discs (28). A similar N-terminal truncation in the mouse PP2A regulatory subunit results in dominant inhibition of PP2A in melanoma cells (41). Contrary to our results, Jia et al. (24) reported that wdb-specific RNAi in wing discs mildly reduces the abundance of CiFL and the expression of dpp-lacZ, but has no effect on the stability of Smo (although they did not examine the effect of wdb-specific RNAi on ptc or Col). These results led the authors to propose that Wdb-PP2A might be necessary to promote Hh signaling, potentially affecting the phosphorylation status of Ci. In our study, overexpressing wdb-specific RNAi in the same line used by Jia et al. as well as in three additional RNAi lines from the Vienna Drosophila RNAi Center (VDRC) and the Transgenic RNAi Project (TRiP) collections did not produce an effect on Hh target genes, which was probably as a result of the inefficiency of the wdb-specific RNAi in reducing the potentially large amount of wdb present in these cells (23). To address this discrepancy, we generated loss-of-function wdb clones (28) to disrupt Wdb-PP2A function in wing discs. Similar to the results from our experiments with wdbDN, both Smo and CiFL were stabilized in hypomorphic wdbIP clones (Fig. 3, X and Y). These data are consistent with the regulation of Smo activity by Wdb-PP2A and are in direct contrast to a reported role for Wdb in regulating Ci activity (24).

Whereas we uncovered an inhibitory role for Wdb-PP2A in Hh signaling, two screening studies contained peripheral examinations of PP2A in Hh signaling and concluded that the PP2A catalytic subunit Mts is required for maximal Hh signaling. Reducing the extent of expression of mts suppresses a wing phenotype caused by a dominant gain-of-function hhMrt allele (37), whereas in a first-generation, genome-wide dsRNAi screen, inhibition of either mts, wdb, or wrd (which encodes another PP2A regulatory subunit) by dsRNA abolishes Hh-induced ptc-luciferase (ptc-luc) activity in cl-8 cells in the presence of exogenous Hh (42). No validation of mts was performed in a subsequent rescreen (43) aimed at vigorously eliminating off-target effects that plagued the initial screen (wdb and wrd were not retested). Nevertheless, we studied the consequence of reduced PP2A activity on Hh signaling in wing discs with multiple RNAi lines from the VDRC and TRiP collections that targeted either the catalytic (Mts) or scaffolding A (CG17291) subunits of PP2A. Reduced PP2A activity by mts or CG17291 RNAi led to the accumulation of Smo (fig. S6, B and F) but the destabilization of CiFL (fig. S6, C and G), suggesting that PP2A may play a dual role in regulating the phosphorylation of Smo and Ci. Because ci acts downstream of smo in the Hh signaling cascade, it is not surprising that Hh signaling (that is, Ptc production) was significantly reduced (fig. S6, D and H) in the absence of PP2A activity.

We next asked which PP2A regulatory subunit(s) regulated the activities of Smo and Ci. Our biochemical analyses demonstrated that Wdb directly interacted with Smo (Fig. 2G), thus enabling PP2A to dephosphorylate Smo to control its activity. Consistent with our genetic data, neither endogenous Wdb (fig. S6I) nor overexpressed Wdb-HA (fig. S6J) formed a complex with Ci in cl-8 cells, implying that the regulation of Ci activity by PP2A might be controlled by a different regulatory subunit. Indeed, we found that Ci interacted with Tws, another PP2A regulatory subunit (fig. S6K). Furthermore, overexpressing tws (fig. S6Q), but not wdb (fig. S6M), in cl-8 cells was sufficient to promote the constitutive nuclear localization of CiFL, which is a prerequisite for Ci activation. Together, our data suggest that the dual roles of PP2A in Hh signaling by regulating Smo and Ci might be mediated through Wdb and Tws, respectively. Further study is needed to investigate whether Tws regulates localization or activity of Ci in vivo.

PP1 and PP2A dephosphorylate distinct species of pSmo

To explore the molecular mechanisms by which PP1 and PP2A regulate the phosphorylation status of Smo, we treated lysates of Hh-induced cl-8–gsmo cells with purified PP1 or PP2A (16, 17, 44). Neither phosphatase alone was able to completely dephosphorylate pSmo (Fig. 4, A and B, upper panel); however, the addition of PP1 to PP2A-treated cell lysates resulted in further dephosphorylation of pSmo (Fig. 4B, lane 6), which suggested that PP1 and PP2A might have different efficiencies or selectivities to dephosphorylate Smo. Alternatively, both phosphatases might target distinct pSmo species. Among 26 phosphorylated residues in the cytoplasmic tail of Smo, the PKA-CKI consensus sites are essential for Smo signaling (79). We generated an antibody (α-Smo-pS667) specific for Smo phosphorylated by PKA at Ser667 (fig. S7, A and B). This antibody exhibited at least fivefold more sensitivity toward PKA-phosphorylated Smo than to PKA-primed, CKI-phosphorylated Smo (fig. S7, C to F), which enabled us to distinguish PKA-phosphorylated Smo from CKI-phosphorylated Smo and to study the importance of the sequential phosphorylation of Smo to Hh signaling.

Fig. 4

PP1 and PP2A dephosphorylate distinct species of pSmo. (A and B) PP1 or PP2A alone incompletely dephosphorylated GFP-pSmo. Hh-induced cl-8–gsmo lysates were treated with increasing amounts of (A) PP1 or (B) PP2A. Western blotting analysis with an antibody against GFP (top panels) detected all forms of GFP-pSmo (box brackets), which were completely removed by CIP. The α-Smo-pS667 antibody (middle panels) specifically detected PKA-phosphorylated GFP-Smo (GFP-PKA-pSmo). PP1 efficiently dephosphorylated GFP-PKA-pSmo, whereas PP2A treatment enriched for the α-Smo-pS667 antibody–specific GFP-PKA-pSmo. GFP-PKA-pSmo species treated with PP2A were greatly reduced in abundance by additional PP1 activity (panel B, compare lane 6 with lanes 2 to 4). The Western blot analyzed with the α-Smo-pS667 antibody (A, middle panel) was overexposed to reveal PKA-pSmo in lysates containing enhanced amounts of pSmo. β-Tubulin served as the loading control. (C and D) PP1 dephosphorylates PKA-phosphorylated Smo. GST-PKA-pSmo labeled with cold ATP was dephosphorylated by PP1, but not by PP2A, as indicated by the decrease in the extent of detection of GST-PKA-pSmo with α-Smo-pS667 (C). This PP1 response was inhibited by PP1-I2, a specific PP1 inhibitor. Radiolabeled PKA–phosphorylated GST-Smo (GST-PKA-32pSmo) was dephosphorylated by PP1, but not by PP2A (D). λ-Phosphatase (λPpase) was used as a positive control in panels C to F. (E) GST-Smo (lane 1) was phosphorylated with PKA alone (lanes 2 and 3) or PKA and CKI (lanes 4 to 8) in the presence of cold ATP. Treating GST-CKI-pSmo with PP2A (lane 5), but not PP1 (lane 6), enriched for the α-Smo-pS667–reactive GST-PKA-pSmo, which was reversed by additional PP1 activity (lane 7). (F) PP2A specifically targets CKI consensus modifications. Individual PKA-CKI clusters in GST-Smo were phosphorylated by PKA in the presence of ATP followed by CKI-mediated phosphorylation in the presence of [γ-32P]ATP (lanes 1, 3, and 5), resulting in selective incorporation of [γ-32P]ATP only at CKI consensus serines within a single cluster, as schematically shown. PP2A removed CKI phosphorylation–specific 32P-phosphates from individual PKA-CKI clusters (lanes 2, 4, and 6).

Western blotting analysis of PP1-treated cell lysates with the α-Smo-pS667 antibody revealed a gradual reduction in the amount of PKA-phosphorylated Smo (Fig. 4A, middle panel). In contrast, PKA-phosphorylated Smo was enriched in PP2A-treated cell lysates (Fig. 4B, middle panel), implying that PP2A enriched the samples for epitopes that were recognized by the α-Smo-pS667 antibody. Of note, these PP2A-enriched α-Smo-pS667–specific epitopes were removed by PP1 activity (Fig. 4B, middle panel), raising the possibility that PP1 and PP2A might act on distinct pSmo species. To test this hypothesis, we performed in vitro dephosphorylation experiments with recombinant GST-Smo. GST-Smo was phosphorylated by PKA in the presence of cold adenosine triphosphate (ATP) to generate GST-PKA-pSmo that was detectable by α-Smo-pS667 (Fig. 4C, lane 1), and PP1 (lanes 2 and 3), but not PP2A (lane 4), removed these α-Smo-pS667–specific epitopes. To verify that PP1 acted specifically on PKA-phosphorylated Smo, we performed experiments in which GST-Smo was phosphorylated by PKA in the presence of [γ-32P]ATP, generating radiolabeled GST-PKA-32pSmo (Fig. 4D, lane 1). We found that PP1 (lanes 2 to 4), but not PP2A (lane 5), was sufficient to dephosphorylate GST-PKA-32pSmo, thus indicating that PKA-phosphorylated Smo was a direct substrate of PP1 and that PP2A did not act on PKA-phosphorylated Smo.

To identify which species of pSmo served as a substrate of PP2A, we incubated GST-Smo with PKA and CKI in the presence of cold ATP to generate PKA-primed, CKI-pSmo (GST-CKI-pSmo), which was not efficiently detected by the α-Smo-pS667 antibody (Fig. 4E, lane 4). However, the α-Smo-pS667 antibody regained its reactivity when GST-CKI-pSmo was treated with PP2A (lane 5), but not PP1 (lane 6), suggesting that PP2A likely removed phosphoryl groups from CKI consensus sites to enrich the sample for PKA-dependent pSmo epitopes that were detectable by α-Smo-pS667. Because α-Smo-pS667 could discriminate only between PKA- and CKI-mediated phosphorylation in the first PKA-CKI cluster (cluster 1), we examined whether this PP2A-specific action on CKI sites was also true for the other PKA-CKI clusters. We generated GST-Smo variants to enable us to study the role of PP2A on individual clusters (Fig. 4F). Individual PKA-CKI clusters were phosphorylated with PKA and cold ATP (PKA-cold) followed by phosphorylation with CKI in the presence of [γ-32P]ATP (CKI-hot) to generate GST-CKI-32pSmo (Fig. 4F). As expected, PP2A efficiently dephosphorylated GST-CKI-32pSmo at all clusters. Further study of GST-Smocluster2 also confirmed that PP1 could not act on CKI-phosphorylated Smo (fig. S8A).

Each PKA-CKI cluster consists of one PKA and two CKI consensus serines, all of which are phosphorylated upon PKA-primed phosphorylation of Smo by CK1. To determine whether PP2A specifically removed phosphates from residues in CKI consensus sites or from any serine residues at which high-density phosphates were present, we incubated GST-Smocluster2 first with PKA and [γ-32P]ATP (PKA-hot) and then with CKI and cold ATP (CKI-cold). This approach selectively radiolabeled the PKA consensus serine, whereas CKI consensus site residues were labeled with nonradiolabeled phosphates (fig. S8B, lane 1). PP1 (lane 2), but not PP2A (lane 3), removed the 32P-phosphate from the PKA consensus site serine. Together, our differential radiolabeling assays demonstrated that PP2A acted specifically on residues in the CKI consensus site, but not on random phosphorylated sites at which high-density phosphates were present.

Overexpression of phosphorylation-defective Smo inhibits distinct thresholds of Hh signaling

PP1 and PP2A act nonredundantly on PKA- and CKI-phosphorylated Smo, respectively. To explore the importance of the regulation of sequential phosphorylation of Smo to Hh signaling, we directly compared the roles of PKA- and CKI-phosphorylated Smo in transducing distinct Hh signals by taking advantage of the requirement for protein dimerization to activate Smo (45). Wild-type and glycogen synthase kinase 3 (GSK) (SmoGSK-SA), PKA (SmoPKA-SA), and CKI (SmoCKI-SA) phosphorylation–defective variants of Smo (7), produced to similar amounts in wild-type wing discs, could interfere with graded Hh signaling mediated by endogenous Smo (Fig. 5, A to D). Consistent with previous reports (6, 7, 36), overexpressing wild-type smo or smoGSK-SA in the dorsal compartment of wing discs did not affect the expression of Hh target genes (Fig. 5, E to L). Overexpressing smoCKI-SA variant only slightly expanded the expression domain of CiFL; however, we observed a notable increase in CiFL abundance along the AP boundary (Fig. 5M, arrow). This effect might have been as a result of the conversion of CiFL to labile CiA, a process that is associated with high-threshold Hh signaling (36, 46, 47). Consistent with this observation, smoCKI-SA acted in a dominant-negative manner to reduce ptc expression and Col abundance (Fig. 5, O and P). We observed similar effects of smoPKA-SA on Hh signaling (Fig. 5, Q to T), suggesting that phosphorylation by both PKA and CKI is required for high-threshold Hh signaling. However, SmoPKA-SA (Fig. 5R), but not SmoCKI-SA (which was still subject to PKA-mediated phosphorylation) (Fig. 5N), competed with endogenous PKA-phosphorylated Smo to abolish the expression of dpp, suggesting that intermediate-threshold Hh signaling might require PKA-phosphorylated Smo.

Fig. 5

Distinct effects of phosphorylation-defective Smo variants on Hh signaling. (A to D) ap-Gal4–driven mCD8-gfp in the dorsal compartment (B) did not affect Hh signaling (C and D). The dorsal-ventral boundary is marked by a dashed line in (B). (E to T) Wild-type (WT) wing discs expressing (E to H) smo (smoWT), (I to L) GSK (smoGSK-SA), (M to P) CKI (smoCKI-SA), or (Q to T) PKA (smoPKA-SA) phosphorylation–defective smo variants driven by ap-Gal4 in the dorsal compartment (box brackets). Overexpressed SmoCKI-SA (M, O, and P) and SmoPKA-SA (Q, S, and T) reduced high-threshold Hh signaling (including ptc mRNA expression and Col production, and CiFL at the AP boundary, as indicated by arrows). In contrast, these two mutants had distinct effects on intermediate-threshold Hh signaling (N and R). SmoPKA-SA substantially reduced dpp transcription (R). Overexpressed smoCKI-SA did not result in obvious expansion of the area in which dpp was expressed.

To test this hypothesis and to eliminate the possibility that the distinct roles of phosphorylation-defective Smo variants on dpp transcription were as a result of incomplete dominant-negative interference, we examined the ability of the phosphorylation variants of Smo to rescue dpp expression in smo2 loss-of-function clones (7). We found that transcription of dpp was abolished in the absence of endogenous smo (Fig. 6D, solid line). Consistent with a nonessential role of GSK-mediated phosphorylation in Hh signaling, SmoGSK-SA was sufficient to induce the expression of dpp in smo2 clones (Fig. 6I). SmoCKI-SA (Fig. 6D, dashed line), but not SmoPKA-SA (Fig. 6N), fully restored the expression of dpp, confirming that PKA-phosphorylated Smo was sufficient to activate intermediate-threshold Hh signaling. Furthermore, these results provide an explanation as to why enlargement of the pool of PKA-phosphorylated Smo, which resulted from overexpression of nipp1 (which led to the inhibition of PP1c) or wdb (which led to the increased activity of Wdb-PP2A), induced expression of ectopic dpp.

Fig. 6

Sequential phosphorylation of Smo transduces graded Hh signaling. (A to O) PKA-mediated phosphorylation of Smo (SmoCKI-SA) rescues dpp transcription in smo2 loss-of-function clones. Magnified images of the areas boxed in panels A, F, and K are shown in panels B to E, G to J, and L to O, respectively. dpp mRNA (red) was detected by in situ hybridization (D, I, and N), followed by incubation with antibodies against Myc (green) and GFP (blue). Cells in smo2 clones (C, H, and M) did not express Myc, whereas cells expressing smo phosphorylation variants (E, J, and O) did not express CFP, respectively. Cells in a smo2 clone with transgene expression (that is, Myc-negative and CFP-negative) are circled with dashed lines (B, G, and L). dpp mRNA was not expressed in cells of a smo2 clone lacking the smo transgene (that is, Myc-negative but CFP-positive, solid line in panel D). SmoCKI-SA (D) and SmoGSK-SA (I), but not SmoPKA-SA (N), rescued dpp transcription in smo2 clones. (P) Model for the transduction of graded Hh signaling by sequential phosphorylation of Smo. Shown are four different states of Smo phosphorylation in response to Hh signaling: unphosphorylated Smo, basal level of Smo phosphorylation (Basal-pSmo), moderately phosphorylated Smo (Moderate-pSmo), and enhanced phosphorylation of Smo (Enhanced-pSmo). Stabilization of the Hh signaling transcription factor CiFL, which is indicative of low-threshold Hh signaling, requires a basal amount of Smo phosphorylation. PKA- and PP1-mediated, moderately phosphorylated Smo is sufficient to induce the transcription of dpp, an intermediate-threshold Hh signaling target. Enhanced phosphorylation of Smo, mediated by CKI and Wdb-PP2A, is required to maintain the expression of high-threshold Hh targets, including ptc and Col. (Q and R) cl-8–gsmo cells were incubated with increasing amounts of Hh-conditioned medium (Hh-CM, 10 to 75% in fresh medium) for 8 hours. An antibody against GFP was used to detect Hh-induced progressive phosphorylation of Smo (top panel). The maximal extent of Smo phosphorylation was observed when cells were treated with 50 to 75% Hh-CM (lanes 4 and 5); however, the abundance of GFP-PKA-pSmo, detected by α-Smo-pS667 (middle panel), increased initially (lanes 1 to 3), reached a maximal intensity with 50% Hh-CM treatment (lane 4), and then declined sharply at 75% Hh-CM (lane 5). The amounts of Hh protein present in Hh-CM were detected with an antibody against Hh (bottom panel). Quantification of the migration and intensity of pSmo species in panel Q, as described in Materials and Methods, is shown in panel R.

Discussion

PP1 and PP2A dephosphorylate Smo to inhibit Hh signaling

The conversion of a gradient of the morphogen Hh into distinct transcriptional responses is essential for cell-fate decisions and tissue patterning during development (25). Here, we provide genetic and biochemical evidence to support a model in which sequential phosphorylation of Smo, which is established by distinct kinases and phosphatases on specific serines, transduces graded Hh signaling (Fig. 6P). A basal extent of Smo activity, regulated by as yet unknown kinases and phosphatases, was sufficient to transduce low-threshold Hh signaling. PKA and PP1 collaborated to sustain PKA-phosphorylated Smo to transduce intermediate-threshold Hh signaling, whereas CKI and PP2A facilitated high-threshold Hh signaling by maintaining PKA-primed, CKI-phosphorylated Smo.

Wdb-PP2A directly and specifically acted on CKI-pSmo to restrict high-threshold Hh signaling. Apart from PP2A, another phosphatase, PP1, specifically dephosphorylated PKA-phosphorylated Smo. This collaborative regulation between different phosphatases on the same substrate also functions in other cellular processes. For example, PP1 and PP2A dephosphorylate Par-3 to regulate cell polarity in the specification of neuroblast cell fate (20, 21). Similarly, PP2A and PP4 respond to different DNA damage signals to dephosphorylate γ-H2AX to facilitate the repair of DNA double-strand breaks (18, 19).

PP2A and PP4 may dephosphorylate other components of Hh signaling

The activity of several Hh signaling components, including Smo, Ci, and Cos2, is regulated by phosphorylation (14). For example, PKA- and CKI-mediated phosphorylation of Ci leads to its destabilization. PP2A is implicated in regulating Ci activity in flies (24, 37, 42). Our study confirmed that the catalytic PP2A subunit Mts associated with both Smo and Ci in cl-8 cells (Fig. 2B and fig. S6I). Consistent with the substrate specificity of PP2A being conferred by its obligate regulatory subunits, we found that Wdb specifically regulated the signaling potential of Smo. Another regulatory subunit, Tws, may direct PP2A activity toward Ci, which may potentially promote the translocation of CiFL to the nucleus, thereby activating Hh signaling (fig. S6Q). The use of distinct PP2A regulatory subunits in the same developmental process was also observed in transforming growth factor–β (TGF-β) signaling (14). The effect of the regulatory Bα subunit on PP2A activity activates Smad2 signaling, whereas the Bδ subunit inhibits Smad2 activity. The elaborate regulation of these two signaling systems by PP2A highlights a potential paradigm in which differential PP2A activity plays an essential role in developmental signaling. PP2A is a strong tumor suppressor (48); thus, modulation of PP2A activity provides an additional route by which development and tumorigenesis might be controlled.

Another phosphatase, PP4, may play a role in inhibiting Smo (24); however, inhibiting PP4 alone is not sufficient to promote constitutive cell surface localization of wild-type Smo, unless Hh protein is provided. The surface localization of Smo is tightly linked to PKA- and CKI-dependent enhanced phosphorylation of Smo. PP4-specific RNAi further increases the extent of constitutive surface localization of Smo mutants that mimic PKA- and CKI-mediated phosphorylation (24), which suggests that PP4 may act on sites other than those in the PKA-CKI clusters.

To delineate mechanisms whereby PP2A and PP4 might act on Smo, we systematically examined the expression of Hh signaling components as well as of Hh targets in wing discs expressing pp4 RNAi. In addition to the Smo stabilization (fig. S9C) observed by Jia et al. (24), we found that pp4 RNAi reduced the abundance of Cos2 protein (fig. S9D). This might be as a consequence of the increased Smo abundance, because smo RNAi stabilized Cos2 (fig. S9H). Alternatively, PP4 might regulate Cos2 directly, because phosphorylated Cos2 is not stable (49). To distinguish between these two possibilities, we examined the genetic relationship between smo and pp4 by monitoring the stabilization of Cos2. We found that Cos2 abundance was still reduced in wing discs containing both pp4 and smo RNAi (fig. S9L). This effect is similar to the effect of pp4 RNAi alone, thus placing pp4 downstream of smo in regulating Cos2. Consistent with this, reduced expression of pp4 compromised Ptc and Col expression at the AP boundary (fig. S9, N and P). The expanded area containing Ptc, albeit at a reduced abundance, away from the AP boundary was also observed by Jia et al. (24). Our experiments are consistent with a positive role of Cos2 in mediating maximal activation of Hh signaling in cl-8 cells (50) as well as in wing discs; Ptc and Col expression are reduced in cos2 clones abutting the AP boundary (51, 52). Our data, together with the observation of a direct interaction between Cos2 and PP4 (24), argue that PP4 might also directly affect the extent of Cos2 phosphorylation.

Does the activation of Hh signaling depend on collective phosphorylation of Smo or on regulated Smo phosphorylation occurring at specific serine residues?

Smo contains three PKA-CKI phosphorylation clusters, with one PKA and two CKI consensus serines in each cluster. A previous study (8) compared the signaling potential of phosphorylation-defective Smo by mutating PKA consensus serines to alanines in one, two, or three of the PKA-CKI clusters and concluded that at least six serines in Smo are required to fully induce the expression of ptc-lacZ, whereas only three serines are needed for the expression of dpp-lacZ. A follow-up study (45) further demonstrated that PKA- and CKI-mediated phosphorylation, which results in the generation of negatively charged residues, counteracts the positive charges conferred by nearby arginine clusters, thus enabling Smo to adopt a conformational change required to activate Hh signaling. These two studies support a model of collective Smo phosphorylation such that the identity of the phosphorylated serines in the PKA-CKI consensus clusters is probably not important; rather, the resulting negative charges collectively carried by these residues after phosphorylation are critical to determine the signaling strength of Smo.

On the basis of this model, a variant Smo (Smo-CKI) in which the CKI, but not the PKA, consensus serines are mutated to alanines, thus rendering Smo-CKI resistant to CKI-mediated phosphorylation, would be anticipated to have the same signaling potential as Smo-PKA23, a variant containing a single intact PKA-CKI cluster, because both mutants contain three serines that can be phosphorylated. In smo loss-of-function clones, Smo-PKA23 is sufficient to drive expression of dpp-lacZ (8); however, Smo-CKI fails to rescue dpp-lacZ expression (7, 8). The discrepancy between the effects of Smo-PKA23 and Smo-CKI on dpp-lacZ expression cannot be simply explained by the collective phosphorylation model. Moreover, these experiments reveal a functional distinction between different phosphorylated residues: Three PKA consensus serines in Smo-CKI may have less signaling activity than the single PKA and two CKI consensus serines in Smo-PKA23. The distinct signaling potentials of the two phosphorylation variants of Smo may be caused by different negative charge densities being carried by individual PKA-CKI clusters, or they may reflect intrinsic properties of sequential phosphorylation within each cluster.

Indeed, the hierarchy of importance among individual PKA-CKI clusters in Smo has been revealed (9). Cluster 2 [also known as region V (9)] is more prevalent than the other two clusters in activating ptc-luc reporter in smo-depleted cl-8 cells. Whether this functional distinction among PKA-CKI clusters also holds true in wing discs is unclear, because neither the hierarchical importance of individual clusters nor the relative importance of specific PKA and CKI phosphorylation events in neutralizing nearby arginine clusters has been directly studied (8, 45). Nevertheless, when the relative importance of the three serines in cluster 2 was examined, the PKA-primed, CKI consensus sites (that is, sequential phosphorylation) were essential for Hh activation in cl-8 cells (9). The differential ability of SmoCKI-SA and SmoPKA-SA variants to activate dpp transcription in wing discs uncovered in our study (Fig. 6, A to O) is consistent with results obtained in cl-8 cells (9). Both observations challenge the model of collective Smo phosphorylation by arguing that the signaling potential of individual serines between each PKA-CKI cluster, as well as within a cluster, is most probably not equal. We believe that regulated phosphorylation at specific serines may therefore contribute to graded Smo signaling.

Sequential phosphorylation of Smo is required for the transduction of graded Hh signaling

The collective phosphorylation model does not distinguish between the contributions of individual phosphorylated residues in PKA-CKI clusters. Our study of phosphorylation-defective Smo variants revealed a Smo activity gradient in which phosphorylation at the PKA consensus sites and phosphorylation at the PKA-primed, CKI consensus sites were required for intermediate- and high-threshold Hh signaling, respectively. This activity gradient of Smo was directly visualized with the α-Smo-pS667 antibody. The abundance of PKA-phosphorylated Smo species, which is uncovered in Hh-stimulated fly cells by mass spectrometric analysis (9), increased initially but then declined sharply in response to Hh (Fig. 6, Q and R). As predicted from our model, phosphorylated Smo in response to intermediate-threshold Hh signaling was sensitive to dephosphorylation by PP1 (fig. S10A) but much less so to PP2A (fig. S10B). Together, these data highlight the importance of sequential Smo phosphorylation to the transduction of graded Hh signaling. We believe that sequential phosphorylation may be required to initialize graded Smo signaling activity. In addition, collective phosphorylation between different clusters may reinforce and maximize the Smo signaling potential to ensure the appropriate Hh signaling outcome.

The presence of up to 26 serine or threonine residues in Smo that can be phosphorylated in response to Hh (9) resembles the composition of residues found in the Kv2.1 potassium channel (53). Variable calcineurin-dependent dephosphorylation of Kv2.1 at 16 phosphorylated residues generates an activity gradient for channel gating and neuronal firing. The opposing actions of kinases and phosphatases on a multisite substrate are known through mathematical modeling to efficiently generate a range of stable phosphorylated forms (54). The spectrum of such distributions can be further increased with the number of phosphorylated sites (53, 55). Two additional kinases, CK2 and G protein (heterotrimeric guanosine 5′-triphosphate–binding protein)–coupled receptor kinase 2 (GRK2), phosphorylate sites in Smo other than those targeted by PKA and CKI (46, 56). Thus, the complex composition of phosphorylated residues in the cytoplasmic tail of Smo, coupled with versatile dephosphorylation by distinct phosphatases, provides an efficient and reliable mechanism to precisely convert the concentration thresholds of Hh into a graded signaling activity.

Materials and Methods

Cell culture and phosphatase RNAi screens

Hh-responsive, Drosophila wing imaginal disc–derived cl-8 cells were cultured at 25°C as described previously (6, 57). The Effectene Transfection Reagent (Qiagen) was used for all transfections. Clonal populations of cl-8 cells stably expressing gfp or gfp-smo (gfp was inserted after the smo signal sequence) and an inducible hhN (a gift of S. Cohen, Institute of Molecular and Cell Biology, Singapore) (57) were generated and maintained with puromycin (100 μg/ml, Sigma). The UAS-gfp-smo construct that was used to generate stable cl-8 cell lines is fully functional because GFP-Smo rescues smo-null embryonic lethality and is regulated in a manner similar to that of endogenous Smo in wing discs and salivary glands (6). Hh-conditioned medium (Hh-CM) was produced from cl-8–gsmo cells in the presence of 0.5 mM CuSO4. OA (3 to 5 nM, Sigma) (17, 25) and TC (1 to 4 μM, Tocris) (16, 25, 30) were used to inhibit PSTPs and PP1, respectively, in cl-8 cells. To construct a dsRNA library targeting all predicted PSTPs and dual-specificity phosphatases (27), regions of targeted genes selected from the GenomeRNAi database (http://www.dkfz.de/signaling/e-rnai3) were amplified from genomic DNA or total RNA with T7-primed oligonucleotides (table S1). The resulting PCR products were transcribed with a T7 MEGAscript kit (Ambion), annealed, and purified as described previously (58). For cell-based RNAi screens, 5 × 104 cl-8–gsmo cells were transfected with dsRNA (0.4 μg) and were cultured without exogenous Hh for 4 days before examination under a Leica DM1L fluorescence microscope for surface-localized GFP-Smo. For Western blotting–based RNAi screens, cl-8–gsmo cells transfected with dsRNA (0.8 μg) were lysed in NP-40 buffer [1% NP-40, 150 mM NaCl, 50 mM tris-HCl (pH 8), and protease inhibitors] supplemented with phosphatase inhibitors (25 mM NaF and 400 μM Na3VO4). All forms of GFP-Smo were detected by Western blotting analysis with an antibody against GFP. The efficiency of mRNA knockdown was between 27 and 75% in cl-8 cells, as determined by semiquantitative reverse transcription–polymerase chain reaction (RT-PCR) assay. To eliminate off-target effects, we used different dsRNAs generated from distinct regions of the targeted genes. Drosophila embryonic S2 cells, which do not express ci (26), were cultured at 25°C as described previously (57). The S2 cells were transiently transfected with the pUAST-gfp-smo plasmid with Effectene. For knockdown of target genes, 6 × 105 S2 cells seeded in a 24-well plate were first soaked in serum-free medium (125 μl) containing dsRNA (6 μg). One hour later, 375 μl of complete medium containing 10% serum was added. S2 cells were cultured with dsRNA for 4 days and then split into new wells followed by the same regimen of dsRNA treatment for another 4 days.

Fly genetics

71B-Gal4, ap-Gal4, dpp-Gal4, MS1096-Gal4, dpp-lacZ, Act5C>yw>Gal4, UAS-gfp, UAS-mCD8-gfp, UAS-gfp-smo, UAS-hhN (active form of Hh), UAS-nipp1, and UAS-nipp1-HA (an endogenous PP1 inhibitor) were described previously (68, 16, 33, 36, 59). UAS-smo RNAi (#9542), UAS-pp4 RNAi (#25317), UAS-wdb RNAi (#27470 and #101406), and UAS-ckIα RNAi (CG2028, #13664) were obtained from the VDRC (60). UAS-mts RNAi (#27723), UAS-PP2A A RNAi (#29384), and two additional UAS-wdb RNAis (#28939 and #27319) were from the TRiP (61). UAS-smoGSK-SA (a PKA and CKI phosphorylation–competent, but GSK phosphorylation–defective Smo), UAS-smoPKA-SA (a PKA phosphorylation–defective Smo), UAS-smoCKI-SA (a PKA phosphorylation–competent, but CKI phosphorylation–defective Smo), UAS-wdb, UAS-wdbDN (an N-terminal truncated Wdb functioning as a dominant-negative Wdb), and FRT82B, wdbIP/TM3 flies were provided by D. Kalderon (Columbia University) (7) and S. Eaton (Max Planck Institute of Molecular Cell Biology and Genetics, Germany) (28). A hemagglutinin (HA) tag was fused to the C terminus of wild-type wdb, which was subcloned into pUAST. The resulting pUAST-wdb-HA was injected into yw embryos, and transgenic Drosophila lines were established as described previously (6). In some experiments (Fig. 3, F to J and U to W), the nipp1 or wdb transgene was recombined with UAS-P35 (62) to minimize cell lethality in wing discs. All fly crosses were maintained on standard yeast–cornmeal molasses media at 25°C, except for those described here. UAS-gfp or UAS-hhN flies were crossed with 71B-Gal4 at 29°C for analyzing kinase expression and activity (fig. S1). UAS-nipp1 and UAS-ckIα RNAi flies were crossed with MS1096-Gal4 at 18°C for detecting Hh-responsive gene expression (Fig. 3, F to J, and fig. S4, M to R). The UAS-nipp1-HA transgene gave similar results at both 18° and 25°C (fig. S4, G to I). UAS-wdb were crossed with MS1096-Gal4 at either 18°C (Fig. 3, P to T) or 25°C (fig. S4, J to L) to examine its effects on Hh signaling. UAS-wdbDN flies were crossed with MS1096-Gal4 at 18°C for visualizing Smo in wing discs by fluorescence immunohistochemistry (Fig. 3O). UAS-smo RNAi and UAS-pp4 RNAi flies were crossed with ap-Gal4 at 18°C to analyze the genetic interaction between smo and pp4 (fig. S9). UAS-PP2A A RNAi and UAS-mts RNAi flies were crossed with MS1096-Gal4 at 18°C to analyze the dual effects of PP2A on Hh signaling (fig. S6, A to H). wdb-overexpressing (“flip-out”) clones (Fig. 3, U to W) were generated by heat-shocking second instar larvae of hs-Flp; UAS-P35/+; Act5C>yw>Gal4, UAS-gfp/UAS-wdb at 37°C for 1 hour, which were then maintained at 18°C. Loss-of-function wdbIP (Fig. 3, X and Y) and smo2 clones (Fig. 6, A to O) were generated by flippase (FLP)/FLP recognition target (FRT)–mediated mitotic recombination (63) and maintained at 25°C. Second instar progenies from the cross of hs-Flp;; FRT82B, Ubi-gfp and FRT82B, wdbIP were heat-shocked at 37°C for 1 hour to generate wdbIP clones. For rescue experiments in smo2 loss-of-function clones, flies with the genotype of hs-Flp, MS1096-Gal4; smo2, FRT40A, UAS>cfp>smoGSK-SA-HA (or smoPKA-SA-HA, or smoCKI-SA-HA)/CyO, provided by D. Kalderon (7), were crossed with NM31E, FRT40A/CyO flies. In these smo transgenes, the cfp coding sequence flanked by two FRTs (indicated by “>”) was localized between the UAS sequence and the smo mutant coding sequence with a C-terminal HA tag. The Smo mutants could be expressed by MS1096-Gal4 upon heat shock–induced flippase cleavage at two FRT sites to remove cfp. Second instar progenies from these crosses were heat-shocked at 37°C for 30 min. Third instar larvae were heat-shocked again at 37°C for 30 min to activate myc expression and were allowed to recover for 1 hour before dissection. smo2 loss-of-function clones could be identified by the absence of Myc staining. Cells ectopically expressing phosphorylation-defective smo mutants did not contain cyan fluorescent protein (CFP). Note that smo2 clones (Myc-negative) and clones ectopically expressing UAS-smo phosphorylation variants (CFP-negative) did not always overlap (for example, clones circled with a solid line in Fig. 6, B to E).

Production of pSmo-specific antibody

A rabbit polyclonal antibody against pSmo (termed α-Smo-pS667) was generated (Bio-Synthesis) with a phosphorylated peptide (HGPRKNpSLDSEISVSVRHVGPRKNpSLDSEISVSVRHV) corresponding to PKA-CKI cluster 1 (amino acid residues 661 to 679) of fly Smo. This peptide included a phosphorylated PKA consensus serine (pSer667) (fig. S7A). The resulting antiserum was preabsorbed with unphosphorylated peptide and then affinity-purified with the phosphorylated peptide. Purified α-Smo-pS667 antibody preferentially detected PKA-phosphorylated Smo (fig. S7, B to F). To examine the specificity of the α-Smo-pS667 antibody, we tested it against the unphosphorylated peptide HGPRKNSLDSEISVSVRHVGPRKNSLDSEISVSVRHV, which was untreated, phosphorylated by PKA, or phosphorylated by PKA and CKI. Serial dilutions of these differentially phosphorylated peptides (25, 5, or 1 ng in a total volume of 1 μl) were absorbed on nitrocellulose membrane (Pharmacia) and dot blotting was then performed as described previously (64) with purified α-Smo-pS667 antibody.

In situ hybridization and immunofluorescence staining

The coding regions of ptc or dpp were used to generate antisense RNA probes for in situ hybridization (6). When in situ hybridization was combined with immunofluorescence staining, we used an autofluorescent alkaline phosphatase substrate (Vector) to visualize mRNA in the rhodamine channel. Endogenous Smo was visualized in cl-8 cells with the 20C6 antibody (50), and the signal was further amplified with biotin and streptavidin conjugated with fluorescein (Invitrogen). For localizing stably or transiently expressed gfp-smo, cl-8 cells or S2 cells grown on coverslips were fixed with 4% paraformaldehyde before direct visualization. The coverslips seeded with S2 cells were precoated with poly-l-lysine (Sigma). To visualize transiently transfected S2 cells expressing gfp-smo, we used 20C6 to stain fixed, but not permeabilized, cells. Under this condition, 20C6 in the absence of biotin-streptavidin amplification was able to detect only ectopic GFP-Smo, but not endogenous Smo, on the surface of S2 cells (fig. S3C). In some experiments, Alexa Fluor 633–conjugated phalloidin (at a 1:80 dilution; Invitrogen) and DAPI (4′,6-diamidino-2-phenylindole; Vector) were used to counterstain the cell cortex and nucleus, respectively. To examine the nuclear localization of Ci in cl-8 cells, we cotransfected wdb-HA or tws-HA cells with ci-gfp. Two days later, we visualized HA-tagged Wdb or Tws with an antibody against HA (at a 1:400 dilution; Y-11; Santa Cruz). To analyze the amount of a given protein produced in cultured cells or wing discs, we used mouse antibody against Smo [1:10; 20C6; Developmental Studies Hybridoma Bank (DSHB)], rat antibody against Ci (1:10; 2A1; a gift of R. Holmgren, Northwestern University) (65), mouse antibody against Ptc (1:200; ApaI; DSHB) (66), mouse antibody against Col (1:100; a gift of A. Vincent, Université de Toulouse 3, France) (67), rabbit antibody against Cos2 (1:2000, a gift of M. Scott and K. Suyama, Stanford University) (68), rabbit antibody against HA (1:200; Y-11; Santa Cruz), mouse antibody against Myc (1:2000; 9B11; Cell Signaling), Alexa Fluor 647–conjugated mouse antibody against Myc (1:2; Cell Signaling), rabbit antibody against GFP (1:1,000; A11122; Invitrogen), rabbit antibody against β-galactosidase (1:4000; Cappel), and mouse antibody against lamin (1:400; ADL67.10; DSHB) (69), as well as Alexa Fluor–conjugated secondary antibodies (1:400; Invitrogen). Fluorescence images were acquired with a Zeiss Axio Imager2 equipped with an ApoTome.

Western blotting analysis and immunoprecipitations

Cells or wing discs were lysed in NP-40 buffer supplemented with phosphatase inhibitors. Western blotting and immunoprecipitation analyses were performed with standard protocols. ImageJ software was used to measure the individual band intensities and migration distances on blots shown in Fig. 6Q and fig. S7E. To visualize progressive phosphorylation of Smo in response to increased Hh concentration, we incubated 3 × 105 cl-8–gsmo cells grown in 24-well plate with increasing amounts of Hh-CM (for example, 10% Hh-CM constitutes one part of Hh-CM and nine parts of fresh medium) for 8 hours. We used an antibody against GFP to detect both the intensity and the extent of Hh-induced progressive Smo phosphorylation. The abundance of PKA-phosphorylated GFP-Smo (GFP-PKA-pSmo) was detected with the α-Smo-pS667 antibody. Specifically, the relative migration index (shown in blue in Fig. 6R) was calculated as the migration distance between GFP-pSmo (box bracket in upper panel of Fig. 6Q) and unphosphorylated GFP-Smo in each treatment divided by the maximal migration distance shown in 75% Hh-CM treatment (Fig. 6Q, upper panel, lane 5). The amounts of GFP-pSmo (Fig. 6Q, upper panel) and GFP-PKA-pSmo (middle panel), shown in green and red, respectively, were calculated on the basis of the mean pixel intensity of each band shown in the blot. Migration distances and pixel intensities were plotted with ImageJ software. The following antibodies were used for Western blotting analysis: rabbit antibody against Smo (1:2000; a gift of P. Thérond, Université de Nice Sophia-Antipolis, France) (70), rabbit antibody against PKA-phosphorylated Smo (α-Smo-pS667; 1:1000), rabbit antibody against Hh (1:4000; NHhI; a gift of T. Tabata, University of Tokyo, Japan) (71), rat antibody against full-length Ci (1:10; 2A1), mouse antibody against the PP1c subunit (1:200; Santa Cruz), antibodies against PP2A subunits [rabbit antibody against A (1:200; Cell Signaling), mouse antibody against C/Mts (1:6000; Upstate), and guinea pig antibody against Wdb (1:3000; a gift of A. Sehgal, University of Pennsylvania) (17)], antibodies against PKA subunits [rabbit antibody against the C subunit (1:1000; Santa Cruz), rabbit antibody against the RI subunit (1:200; Santa Cruz) (72), and rabbit antibody against the RII subunit (1:10,000; a gift of D. Kalderon) (73)], rabbit antibody against CKIα (1:1000; Cell Signaling), rabbit antibody against GFP (1:1000; Chemicon), mouse antibody against GST (1:1000; Santa Cruz), rabbit antibody against HA (1:1000; Y-11; Santa Cruz), mouse antibody against V5 (1:2000; Invitrogen), rabbit antibody against glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (1:1000; Santa Cruz), and mouse antibody against β-tubulin (1:6000; Covance). Agarose conjugated to antibody against GFP (Vector) was used to immunoprecipitate GFP-Smo immunocomplexes from cl-8–gsmo cells (to analyze for an interaction with PP2A) or from cl-8–gsmo cells transiently transfected with plasmids encoding V5-tagged individual PP1c subunits (16). Agarose conjugated with antibody against GFP was also used to immunoprecipitate Ci-GFP immunocomplexes from cl-8 cells transiently transfected with pUAST-ci-gfp and wdb-HA or tws-HA for detecting the association between Ci and PP2A. All Western blotting data presented in the figures are representative of experiments that were performed at least three times.

In vitro binding assays for detecting an interaction between Smo and PP1 or PP2A

For in vitro binding assays, we constructed three versions of GST-Smo fusions, which contained amino acid residues 652 to 759, residues 600 to 800, or residues 557 to 1036 (the complete cytoplasmic tail of Smo). One hundred nanograms each of purified PP2A A:C dimer (Millipore) or of the MBP-Wdb fusion protein (a gift of A. Sehgal) (17) bound on amylose resin (NEB) or recombinant PP1c subunit (NEB) was allowed to bind to GST-Smo fusions (100 ng) at room temperature for 30 min in binding buffer [20 mM tris-HCl (pH 7.4), 10 mM MgCl2, 14 mM β-mercaptoethanol, and bovine serum albumin (1 mg/ml)] (74). To examine the interaction between Smo and Wdb, Smo and Mts, or Mts and PP1c, we washed GST pull-down complexes with a high-stringency wash protocol of seven washes alternating between 1 M and 150 mM NaCl in washing buffer [50 mM tris-HCl (pH 7.5) and 1% NP-40]. The interaction between Smo and the PP2A-A subunit was detected after a low-stringency wash protocol of five washes with 150 mM NaCl.

In vitro dephosphorylation assays

To dephosphorylate GFP-Smo in cl-8–gsmo cells that was phosphorylated to either an enhanced (Fig. 4, A and B) or an intermediate state (fig. S10), we heated 2 μg of Hh-stimulated lysates at 65°C for 15 min to inactivate endogenous kinases and phosphatases, and then incubated the samples at 30°C for 2 hours with calf intestinal phosphatase (CIP, 10 U, NEB), λ-phosphatase (400 U, NEB), or with increasing amounts of PP1 (1 to 4 U, NEB) (16) or PP2A (0.01 to 0.05 U, Millipore) (17). In some experiments, cell lysates were first treated with PP2A at 30°C for 1 hour followed by inhibition of PP2A at 65°C for 15 min and then treatment with PP1 at 30°C for 1 hour (Fig. 4B, lane 6). For the in vitro dephosphorylation assays shown in Fig. 4, C to E, 1 μg of wild-type GST-Smo fusion proteins (residues 557 to 1036) bound to glutathione beads (GE Healthcare) was phosphorylated at 30°C for 90 min with PKA (2500 U, NEB) alone or together with CKI (1000 units, NEB) (75) in the presence of 1 mM cold ATP (Sigma) or 3.33 μM [γ-32P]ATP (Perkin-Elmer). For the in vitro dephosphorylation assays shown in Fig. 4F and fig. S8, GST-Smo variants were constructed as follows. GST-Smo (residues 557 to 1036) underwent site-directed mutagenesis to convert serines to alanines at two of the three PKA consensus sites (amino acid positions 667, 687, and 740), resulting in the generation of three GST-Smo mutants each containing only one wild-type PKA consensus serine (GST-Smocluster3, GST-Smocluster2, and GST-Smocluster1). For selective phosphorylation of GST-Smo mutants to yield either GST-PKA-32pSmo (fig. S8B) or GST-CKI-32pSmo (Fig. 4F and fig. S8A), GST-Smo fusion proteins bound to glutathione beads were phosphorylated by PKA in the presence of cold ATP or [γ-32P]ATP for 90 min followed by three washes with phosphate-buffered saline (PBS) containing 0.1% Triton and then heat inactivation at 65°C for 20 min. Beads were washed a further three times followed by phosphorylation by CKI overnight at 4°C in the presence of cold ATP or hot [γ-32P]ATP (75). For in vitro dephosphorylation (Fig. 4, C to F, and fig. S8), wild-type or mutant GST-Smo (residues 557 to 1036) fusion proteins were phosphorylated with PKA, CKI, or both in the presence of either cold ATP or [γ-32P]ATP. Phosphorylated GST-Smo (40 ng) bound on glutathione beads was dephosphorylated at 30°C for 2 hours by PP1 (0.25 to 1.2 U), PP2A (0.4 to 0.8 μg; a gift of G. Moorhead, University of Calgary, Canada) (44), or both. PP1-I2 (0.5 μg, NEB) was used to inhibit PP1 (16), and OA (50 nM) was used to inhibit PP2A (17).

Luciferase, kinase, and PP2A activity assays

cl-8 cells were seeded in 24-well plates and transfected with 0.2 μg of ptcfirefly luciferase reporter (58) and 0.05 ng of CMVrenilla luciferase plasmid in the absence of Hh-CM. For OA treatment, the transfected cells were incubated with different amounts of OA or Hh-CM for 24 hours, except for cells that were treated with OA (10 nM) for 16 hours to minimize cell death. For dsRNA treatment, cells were cotransfected with 0.4 μg of dsRNA with the ptc-luciferase reporter and renilla plasmids. After 4 days, luciferase activities were measured with a dual-luciferase reporter assay (Promega). The ptc-luciferase activity was normalized to account for transfection efficiency with Renilla luciferase activity. For preparation of protein samples for kinase activity measurement, third instar wing imaginal discs of the genotypes UAS-gfp/+; 71B-Gal4/+ and 71B-Gal4/UAS-hhN grown at 29°C were collected in cold PBS and homogenized in NP-40 buffer. Protein concentration was measured with the BCA (bicinchoninic acid) protein assay (Pierce). The PKA activity from 30 μg of freshly prepared wing disc extracts or from 0.2 U of purified PKA was measured in duplicate with a PepTag assay (Promega) for nonradioactive detection of PKA activity (76). This assay uses a fluorescent PKA substrate (PepTag-LRRASLG) that changes the net charge of the peptide upon PKA phosphorylation, thus enabling the phosphorylated peptide to migrate to the positive electrode of an agarose gel apparatus, whereas the nonphosphorylated peptide migrates to the negative electrode. Briefly, each sample was added into the mixture of 5 μl of PKA reaction buffer, 2 μl of PepTag peptide (at a concentration of 0.4 μg/μl), and up to 5 μM cAMP (Tocris) (77, 78) in a 25-μl final volume followed by incubation at room temperature for 30 min. The reactions were resolved on a 0.8% agarose gel in 50 mM tris-HCl (pH 8.0). Densitometric analysis of the bands was performed with ImageJ software. The PKA activity in wing disc extracts was determined relative to that of purified PKA. Experiments were performed thrice. The CKI activity from 1 μg of freshly prepared wing disc extracts or of 50 U of purified CKI was measured in triplicate by incubating with 10 μg of CKI substrate peptide (KRRRALSpVASLPGL, NEB) in reaction buffer [50 mM tris-HCl (pH 7.5), 10 mM MgCl2, 5 mM dithiothreitol (DTT), 0.9 mM ATP, and 3.33 μM [γ-32P]ATP] (79, 80). After incubation for 15 min at 37°C, reactions were spotted onto P81 phosphocellulose discs (Whatman), dried, and washed four times in 0.5% phosphoric acid solution, twice with absolute ethanol, and once with acetone. Phosphorylation was quantified with a Packard 2100 TR liquid scintillation analyzer. Experiments were performed thrice. The CKI-specific inhibitor CKI-7 (Sigma) (79) was included at a final concentration of 3 mM, where indicated. The normalized PP2A activity of 10 μg of protein lysate extracted from the wing discs of the genotypes MS1096-Gal4/+;; UAS-wdb/+ and MS1096-Gal4/+;; UAS-wdbDN/+ (grown at 25°C) was determined by measuring phosphate release with a nonradioactive, malachite green–based phosphatase assay system (39), according to the manufacturer’s instructions (Promega). A synthetic pThr peptide, RRA(pT)VA, was used as the specific substrate of PP2A. Experiments were performed thrice.

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/4/180/ra43/DC1

Fig. S1. Hh signaling does not affect PKA or CKI abundance or activity.

Fig. S2. Phosphatase activity regulates Smo signaling in cl-8 cells.

Fig. S3. Phosphatase activity regulates the phosphorylation and surface localization of Smo in S2 cells.

Fig. S4. Alteration of Hh signaling activity by manipulating the activities of Smo-specific kinases or phosphatases during Drosophila wing development.

Fig. S5. Expression domains of dpp mRNA and a dpp-lacZ enhancer trap in a wild-type wing disc.

Fig. S6. Functional interaction between Ci and Tws-PP2A.

Fig. S7. Characterization of the phosphospecific antibody α-Smo-pS667.

Fig. S8. PP2A specifically dephosphorylates CKI-phosphorylated Smo.

Fig. S9. The role of PP4 in regulating Hh signaling.

Fig. S10. Distinct abilities of PP1 and PP2A to dephosphorylate moderately phosphorylated Smo.

Table S1. T7-primed primers used to construct the PSTP dsRNA library.

References and Notes

  1. Acknowledgments: We are grateful to S. Cohen, S. Eaton, R. Holmgren, D. Kalderon, G. Moorhead, M. Scott, A. Sehgal, K. Suyama, T. Tabata, P. Thérond, A. Vincent, the Bloomington Stock Center, the DSHB, the TRiP, and the VDRC for fly stocks, antibodies, PP2A, and other reagents. We thank A. Arif and L. Zhao for technical assistance, K. Nybakken for sharing unpublished results, and T. Blankenship, J. McDonald, A. Plessis, and M. A. Price for helpful discussions. Funding: This work was supported by a March of Dimes Basil O’Connor Starter Scholar Research Award (5-FY07-41), a Scott Hamilton CARES Initiative Award, an American Cancer Society Institutional Award (IRG-91-022-12), and an NIH/National Institute of General Medical Sciences grant (R01GM085175) to A.J.Z.; American Heart Association Postdoctoral Fellowship Awards (0825591D and 10POST4110011) to Y.S. and J.Z.; and an NIH/National Institute of Child Health and Human Development Postdoctoral Training grant (T32HD007104) to J.K.O. Author contributions: A.J.Z., Y.S., J.K.O., and J.Z. designed the experiments; Y.S., J.K.O., J.Z., A.P.M., and A.M.S. performed the experiments; and A.J.Z., Y.S., J.K.O., and J.Z. analyzed the data and wrote the paper. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Signaling

Navigate This Article