Dose-dependent transduction of Hedgehog relies on phosphorylation-based feedback between the G-protein-coupled receptor Smoothened and the kinase Fused

The Hedgehog (HH) signaling pathway is crucial for the growth and patterning of many metazoans (Lee et al., 2016). Its deregulation leads to developmental disorders and cancers (Pak and Segal, 2016). In many cases, HH acts as a morphogen; it spreads as a gradient across a field of receiving cells to promote dose-dependent discrete transcriptional responses (Vervoort, 2000).

HH signaling requires Smoothened (SMO), a member of the G-protein-coupled receptor family (Ayers and Therond, 2010). SMO is regulated by the transmembrane protein Patched (PTC; Nakano et al., 1989), which acts as both an SMO antagonist and an HH co-receptor. Studies in flies show that in the absence of signal, PTC promotes SMO internalization and targeting to the lysosome for degradation (Fan et al., 2013; Li et al., 2012; Xia et al., 2012). These effects are blocked by HH, leading to accumulation of a hyperphosphorylated form of SMO at the plasma membrane. An important contribution to SMO hyperphosphorylation is the sequential action of the cAMP-dependent protein kinase (PKA) and Casein kinase 1 (CKI) on three clusters of serines (S) and threonines (T) located in its C-terminal cytoplasmic tail (cyto-tail; Chen and Jiang, 2013). These phosphorylations are necessary and sufficient for SMO localization to the plasma membrane and activation. They induce SMO clustering and a conformational switch in its cyto-tail. The extent of phosphorylation of the PKA/CKI clusters depends on HH levels and controls in a graded manner SMO clustering, the degree of conformational change in its cyto-tail, its accumulation at the cell surface and its signaling activity (Fan et al., 2012; Su et al., 2011; Zhao et al., 2007). Moreover, CKIγ/Gilgamesh (Gish), G-protein-coupled receptor kinase 2 (GPRK2/GRK2), Casein kinase II and atypical protein kinase C increase the ability of SMO to transduce high levels of HH by phosphorylating residues present in the membrane-proximal and central regions of the SMO cyto-tail (Jia et al., 2010; Jiang et al., 2014; Li et al., 2016; Maier et al., 2014). In vertebrates, SMO phosphorylation has been reported to depend on GRK2 and CKIγ (Chen et al., 2004). However, the role of GRPK2/GRK2 has recently been called into question by a study in zebrafish (Zhao et al., 2016).

The wing imaginal disc has been instrumental in study of the morphogenetic effects of HH. In this epithelium, HH is produced in the posterior (P) compartment and spreads towards the anterior (A) region. The dose of HH controls the fate of Cubitus interruptus (CI), a zinc finger transcription factor of the GLI family (Aza and Kornberg, 1999; Wang and Holmgren, 2000). In the absence of HH, CI undergoes partial degradation that leads to a truncated repressor form (CI-R). Toward the A/P boundary, the progressive suppression of CI processing by increased doses of HH leads to the gradual decrease of CI-R associated with a progressive accumulation of full-length CI (CI-FL). Thus, in the presence of low levels of HH, CI processing is blocked, which promotes decapentaplegic (dpp) expression. In the presence of intermediate levels of HH, the levels of CI-FL are increased, leading to ptc and collier (col; knot) expression. Finally, in the A cells that abut the P compartment HH producing cells, CI-FL is turned into a hyperactive state (CI-A), which enables engrailed (en) expression.

How is the graded phosphorylation of SMO transduced to the intracellular components of the pathway that regulate CI fate? In the fly wing imaginal disc, responses to high levels of HH rely on the protein kinase Fused (FU) and the kinesin Costal2 (COS2; Alves et al., 1998; Ho et al., 2005; Ohlmeyer and Kalderon, 1998). Both proteins are part of a SMO-bound large cytoplasmic complex that controls CI-FL processing, nuclear entry and activation (Jia et al., 2003; Monnier et al., 1998, 2002; Ogden et al., 2003). The relationships between SMO, FU and COS2 are very complex. SMO directly recruits FU to the plasma membrane and promotes its clustering and its trans-autophosphorylation, with each event being sufficient to promote FU activation (Claret et al., 2007; Shi et al., 2011). COS2 and FU control each other's levels and activity (Ruel et al., 2007; Zhou and Kalderon, 2010, 2011) and also feed back on SMO (Claret et al., 2007; Liu et al., 2007; Ranieri et al., 2014). Thus, in the absence of HH, COS2 binding to SMO inhibits its phosphorylation by PKA/CKI and its accumulation at the cell surface, whereas in response to HH, FU activity is essential for the accumulation of SMO at the plasma membrane and full phosphorylation. It has been proposed that positive feedback by FU on SMO could be attributable to FU opposing the negative effect of COS2. However, FU could control the fate of SMO by directly promoting its phosphorylation on new sites.

Here, we report the identification of four new clusters of S/T in the most C-terminal part of the SMO cyto-tail, near to where FU binds. These sites are required for the full hyperphosphorylation of SMO in response to HH. Their phosphorylation depends on SMO binding to FU and on FU kinase activity, arguing that FU might directly phosphorylate SMO. By preventing or mimicking their phosphorylation, we show that they enhance the effects of the phosphorylation of SMO by the PKA/CKI, resulting in high levels of HH signaling. Finally, phosphorylation of these sites is important for FU to phosphorylate its targets COS2 and SU(FU) and to promote high CI-FL activation. We propose that in response to high levels of HH signal, SMO and FU engage in a complex positive feedback loop, in which they mutually enhance each other's activation. This multikinase phosphorylation cascade would result in an amplification of HH signaling that reaches a threshold, promoting ‘high HH’-specific responses.

PKA and CKI act on three clusters of S/T (Fig. 1A). Blocking the phosphorylation of these residues [by replacing the PKA sites by alanines (A), SMO^PKA-SA] did not totally prevent SMO phosphorylation in response to HH (Fig. 1B). Four C-terminal (Ct) blocks (GI-GIV) of S or T residues located between amino acid (aa) 916 and 1036 are conserved among Drosophila species (Fig. 1A). To assay their phosphorylation, we transiently expressed wild-type SMO (SMO^wt) or SMO^Ct-SA (all the Ct-sites replaced by A) – both HA tagged – in HH responsive Cl8 cells (Fig. 1B). In the presence of HH, SMO^Ct-SA-HA phosphorylation levels were significantly reduced compared with that of SMO^wt-HA. Mutation of both the PKA/CKI and the Ct-sites (SMO^{PKA-SA Ct-SA}-HA) almost totally suppressed SMO phosphorylation.

We conclude that, in these conditions, the Ct and PKA/CKI sites are the major contributors to the HH-induced SMO hyperphosphorylation.

Downregulation of FU kinase activity reduces the hyperphosphorylation of SMO induced by HH (Liu et al., 2007), and FU is able to induce SMO phosphorylation (Claret et al., 2007). This latter effect depends on FU kinase activity and is enhanced with GAP-gFU, a constitutively active form of FU. GAP-gFU also induces the phosphorylation of SMO^PKA-SA-HA (Fig. 1C; Fig. S1A), showing that its target sites are outside the known PKA/CKI targets. The Ct clusters are at or near the FU binding region (59 most C-terminal aa) and, as shown in Fig. 1C,D, GAP-gFU could not induce SMO^Ct-SA-HA (or SMO^{PKA-SA Ct-SA}) phosphorylation, indicating that the Ct-sites are targeted by activated FU. Note that mutation of each Ct group separately impacted the SMO phosphorylation induced by GAP-cFU to various degrees (Fig. S1B), indicating that they all contribute to it.

Moreover, GAP-gFU^DANA, a kinase-dead form of GAP-gFU, which acts in a dominant negative manner (Claret et al., 2007), reduces the HH-induced phosphorylation of SMO^PKA-SD, but not of SMO^{PKA-SD Ct-SA} (Fig. 1E). Finally, induction of SMO^wt phosphorylation by GAP-gFU was suppressed with a form of SMO lacking its FU binding region (SMO^Δ978-HA; Fig. S1C).

These data argue that activated FU can promote the phosphorylation of the Ct-sites involved in the HH-induced hyperphosphorylation of SMO and show that this effect is FU kinase activity and interaction dependent.

We next studied whether the phosphorylation of the PKA/CKI and Ct clusters influenced each other. As shown above, phosphorylation of the PKA sites is not required for that of the Ct-sites induced by GAP-gFU. We also tested SMO^PKA-SD-HA, in which the PKA and CKI phosphorylation sites have been replaced by aspartic acids (D) to mimic their phosphorylation. SMO^PKA-SD-HA was constitutively phosphorylated in the absence of HH (Fig. 1F; Fig. S1A), probably attributable to the phosphorylation of the Ct clusters as it disappeared with SMO^{PKA-SD Ct-SA}-HA. By contrast, SMO^Ct-SD-HA (Ct-sites replaced by D to mimick their phosphorylation) displayed no sign of phosphorylation in the absence of HH but was further phosphorylated in the presence of HH, probably on its PKA sites as this was not the case for SMO^{PKA-SA Ct-SD}-HA (Fig. S1D,E).

Finally, we analyzed the effect of Ct-sites on SMO phosphorylation in vivo. The smo transgenes (smo^wt, smo^Ct-SA and smo^Ct-SD) were inserted at the same locus to ensure comparable expression levels and expressed throughout the wing primodium. As seen in Fig. 1G, blocking the phosphorylation of the Ct-sites (smo^Ct-SA) reduced the levels of SMO phosphorylation, whereas mimicking it (smo^Ct-SD) led to their increase.

In summary, the phosphorylation of the Ct-sites of SMO is neither necessary for the phosphorylation of the PKA/CKI sites in response to HH nor sufficient to promote it in the absence of HH. By contrast, phosphorylation of the PKA/CKI sites favors the phosphorylation of the Ct-sites, but is not required for their phosphorylation by activated GAP-FU, suggesting that it acts by activating FU.

Our data indicate that the phosphorylation of the Ct-sites upregulates SMO levels (Fig. 1; Fig. S1). To quantify this, we fused SMO N-terminal to an enzymatic self-labeling tag called SNAP that enables direct detection and quantification of the tagged protein after electrophoresis (Tirat et al., 2006). After checking the functionality of SNAP-SMO in vivo (Fig. S2), we analyzed the effects of HH and the Ct-sites on its levels in Cl8 cells. As shown in Fig. 2A-C (and Fig. S3), HH promoted a more than twofold increase in SNAP-SMO levels. SNAP-SMO^Ct-SA basal levels were not affected, but its HH-induced accumulation was reduced by ∼30%. Conversely, SNAP-SMO^Ct-SD) basal levels were significantly upregulated (by threefold), an effect strongly increased (sevenfold compared with SNAP-SMO^wt) in response to HH. Moreover, compared with SNAP-SMO^PKA-SD, SNAP-SMO^{PKA-SD Ct-SA} levels were downregulated by about twofold both in the presence and in the absence of HH, whereas SNAP-SMO^{PKA-SD Ct-SD} was constitutively accumulated at very high levels, reaching almost fourfold those of SNAP-SMO^PKA-SD.

We therefore conclude that the phosphorylation of the Ct-sites strongly enhances the stabilizing effects of the phosphorylation of PKA/CKI sites induced by HH.

We then examined whether the subcellular localization of SMO was regulated by the phosphorylation of its Ct-sites. The cell surface and intracellular pools of SNAP-SMO were differentially marked by sequentially using membrane impermeable and permeable SNAP substrates carrying different fluorophores (Fig. 2D-F; Fig. S3). As expected, HH stimulation led to accumulation of SNAP-SMO^wt at the plasma membrane. This initial experiment also revealed two important features. First, all the hyperphosphorylated forms of SMO are at the cell surface and, reciprocally, the intracellular pool of SMO lacks extensive phosphorylation. Second, HH stimulation mainly induced an increase in the levels of SNAP-SMO^wt at the cell surface, whereas intracellular levels were almost unaffected. By contrast with SNAP-SMO^wt, SNAP-SMO^Ct-SA stayed intracellular even in the presence of HH. Moreover, in the absence of HH, SNAP-SMO^Ct-SD displayed a weak but significant increased (twofold) localization to the plasma membrane, which was strongly enhanced by HH (15-fold). As expected, SNAP-SMO^PKA-SA remained intracellular in the presence of HH, whereas SNAP-SMO^PKA-SD was mostly at the plasma membrane even in the absence of HH. This latter effect was suppressed in SNAP-SMO^{PKA-SD Ct-SA} but was further increased in SNAP-SMO^{PKA-SD Ct-SD}.

We conclude that phosphorylation of the Ct-sites is required for the accumulation of high levels of SMO at the cell surface promoted by the phosphorylation of the PKA/CKI clusters in response to HH.

SNAP labeling was then used to analyze the effects of Ct-site phosphorylation on the subcellular distribution of SMO in vivo. Wing imaginal discs expressing SNAP-tagged forms of SMO in the dorsal compartment were labeled with a membrane impermeable SNAP ligand. The ventral compartment served as an internal negative control. Labeled SNAP-SMO thus corresponds to molecules that were – at least transiently – present at the plasma membrane during the labeling time. In agreement with the Cl8 data, labeled SNAP-SMO^wt was at higher levels both at the plasma membrane and in intracellular vesicles in HH sending posterior cells and in A/P receiving cells (Fig. 3A). By contrast, SNAP-SMO^Ct-SA was less labeled and mostly present in intracellular vesicular structures, whereas SNAP-SMO^Ct-SD was highly labeled, with an enrichment at the plasma membrane of the HH sending and receiving cells (Fig. 3B,C). We also observed distinctions in labeled SNAP-SMO^wt distribution along the apicobasal axis of the wing epithelium (Fig. 3D): labeled SNAP-SMO^wt was present at the apical side (visualized with aPKC staining) of the P cells and of the A cells that receive HH and also in the most basal region of the P cells and of the A cells abutting A/P in the region where the highest levels of HH are encountered. This basal distribution was lost with labeled SNAP-SMO^Ct-SA and increased with labeled SNAP-SMO^Ct-SD, indicating that it is dependent on the phosphorylation of the Ct-sites (Fig. 3E,F). Note that these effects were also seen by immunofluorescence with untagged SMO variants (Fig. S4).

In summary, the phosphorylation of the Ct-sites regulates the subcellular localization of SMO in response to HH by two means: (i) it increases its accumulation at the plasma membrane; and (ii) it promotes its localization in the basal region of the polarized epithelial cells of the wing imaginal disc.

We then assayed whether the phosphorylation of the Ct-sites regulates SMO activity. SMO^wt, SMO^Ct-SA and SMO^Ct-SD were expressed in the D compartment, and HH signaling was monitored by following CI-FL and PTC accumulation. Despite some variation, clear signatures specific to the different forms of SMO were observed (Fig. 3G-K). SMO^wt overexpression led to a mild upregulation of HH signaling as indicated by: (1) an upregulation of CI-FL levels at the A/P boundary (in >50% of discs) and its ectopic accumulation in the A part of the wing discs (in >65% of discs; Fig. 3G1,J); and (2) an increase in PTC levels at the A/P border (in >60% of discs) and its occasional ectopic expression towards the anterior (20% of discs; Fig. 3G2,K). By contrast, SMO^Ct-SA tended to decrease HH signaling (Fig. 3H,J,K). It never led to ectopic PTC and led to ectopic CI-FL accumulation about four times less often than SMO^wt. Moreover, it downregulated PTC levels and/or CI-FL in almost 30% of discs, two events rarely seen with SMO^wt expression. Conversely, expression of SMO^Ct-SD led to a stronger activation of HH signaling than SMO^wt, with an enhanced accumulation of CI-FL and PTC in the whole A compartment in 80 and 72% of the discs, respectively (Fig. 3I-K).

Together, these results indicate that the phosphorylation of the Ct-sites of SMO upregulates its activity.

For better assessment of the roles of the phosphorylation of Ct-sites, we analyzed the impact of mutating them on SMO^PKA-SD activity. SMO^PKA-SD is known to be active constitutively (Jia et al., 2004), and its expression in the wing pouch led to an increased expression of all known HH targets (Fig. 4; Fig. S5). This was associated with an increase in CI-FL levels throughout the A compartment, with an enlargement of the region near the A/P boundary that normally displays low levels of CI-FL (corresponding to CI-A and therefore called the ‘CI-A region’) owing to the presence of high levels of HH. All these effects were stronger with SMO^{PKA-SD Ct-SD}, with a very strong ectopic en expression and an important enlargement of the CI-A region. By contrast, SMO^{PKA-SD Ct-SA} led to CI-FL accumulation in the A region but it was unable to promote ectopic target gene expression. Moreover, en expression at the A/P boundary was suppressed along with the CI-A region, and col and ptc expression were strongly downregulated. However, the expression domains of the ‘low HH’ responsive genes dpp and iroquois (iro) were almost normal. Notably, these negative effects of SMO^{PKA-SD Ct-SA} are also seen in the absence of endogeneous SMO protein (see Fig. S6).

In conclusion, phosphorylation of the Ct clusters is required for the constitutive activity of SMO^PKA-SD and enhances its effects. By contrast, preventing Ct cluster phosphorylation in SMO^PKA-SD blocks the response to high levels of HH but has almost no effect on the target genes that respond to low HH levels.

Strikingly, the expression of SMO^{PKA-SD Ct-SA} phenocopies the effects of mutations disrupting FU kinase activity. In fu mutants altered in the kinase domain, en and col expression are strongly reduced, inactive CI-FL accumulates and the CI-A region disappears, whereas dpp is almost unaffected (Alves et al., 1998). It might thus be the case that SMO^{PKA-SD Ct-SA} traps or freezes endogenous FU in an inactive state. This prompted us to examine the consequences of SMO^{PKA-SD Ct-SA} expression on the three proteins whose phosphorylation depends on FU activity and are likely targets of FU: COS2, FU itself and its antagonist, SU(FU). Myc-tagged versions of these proteins were transiently coexpressed with SMO^wt-HA in Cl8 cells (Fig. 5A). All three displayed low levels of phosphorylation (visualized by a shift in their electrophoretic migration) in the presence of SMO^wt and HH. Consistent with their signaling activity, SMO^Ct-SA-HA and SMO^Ct-SD-HA led to a slight (but reproducibly) reduced and increased phosphorylation, respectively. As expected, SMO^PKA-SD also upregulated their phosphorylation in the absence of HH. By contrast, SMO^{PKA-SD Ct-SA} expression prevented (COS2 and FU) or reduced [SU(FU)] phosphorylation, both in the absence and in the presence of HH. This indicates that SMO^{PKA-SD Ct-SA} negatively affects the kinase activity of endogenous FU.

To test this possibility further, we examined whether the negative effects of SMO^{PKA-SD Ct-SA} were rescued by the expression of GAP-CFP-FU (GAP-cFU; Fig. 5B-E; Fig. S7). As previously reported, GAP-cFU expression led to an activation of the pathway, with ectopic ptc, col and dpp expression and ectopic accumulation of CI-FL. When GAP-cFU was coexpressed with SMO^{PKA-SD Ct-SA}, all HH targets were ectopically expressed. Surprisingly, these effects were stronger than those of GAP-cFU or SMO^{PKA-SD Ct-SA} alone as indicated by high levels of ectopic en expression throughout the entire A compartment. Notably, CI-FL levels were strongly reduced, presumably because of the generation of the CI-A form throughout the A compartment.

These data show that the expression of an active form of FU can totally suppress the negative effects of SMO^{PKA-SD Ct-SA} on the ‘high HH’ targets and that, reciprocally, its effects on all HH targets are increased by SMO^{PKA-SD Ct-SA}.

We identified new phosphorylations of the fly SMO cyto-tail and showed that they are promoted by the protein kinase FU. This phosphorylation is part of a multisite, multikinase phosphorylation cascade, which controls both SMO accumulation at the plasma membrane and its distribution along the apicobasal axis. These phosphorylation events participate in amplification feedback loops, which promote high levels of SMO-FU activation that allow responses to high doses of the HH signal.

Several kinases are involved in SMO cyto-tail hyperphosphorylation, including PKA/CKI, CKII, aPKC, GPRK2/GRK2 and CK2γ/Gish. Here, we identify four new phosphorylation clusters in its most C-terminal part, near the FU binding site, that are required for the full phosphorylation of SMO in response to HH. FU might act on SMO phosphorylation directly or indirectly through the recruitment, increased expression (via CI activation) or activation of another kinase. Although definite proof of a direct effect is lacking in the absence of an in vitro FU kinase assay, several lines of evidence support that possibility, as follows: (i) blocking endogenous FU activity by a dominant negative form reduces the HH induced phosphorylation of SMO; (ii) expressing a form of FU that is not anchored to the plasma membrane also promotes SMO phosphorylation (although less efficiently than GAP-FU); (iii) both GAP-gFU-kinase activity and its interaction with SMO are required for GAP-gFU to promote the phosphorylation of Ct-sites; and (iv) the induction of SMO phosphorylation by FU or GAP-FU on SMO is also seen with S2 cells that fail to express CI. Note also that FU might act on only a few of these sites, which would prime the phosphorylation of adjacent sites by another kinase, as shown for its own autophosphorylation (Zhou and Kalderon, 2011).

Our data also suggest that sequential phosphorylation of SMO initiated at the PKA/CKI sites leads to the phosphorylation of the Ct-sites. This could be attributable to the conformational changes induced by the PKA/CKI phosphorylation, which could increase the accessibility of the Ct-sites. Alternatively, the phosphorylation of the PKA/CKI sites could act indirectly by promoting FU activation. We favor this latter possibility for the following reasons: (i) FU can be directly activated by SMO (Claret et al., 2007; Zhou and Kalderon, 2011); and (ii) GAP-gFU can promote the phosphorylation of SMO^wt in the absence of phosphorylation of the PKA/CKI sites of SMO (Claret et al., 2007; and present study).

SMO phosphorylation by PKA/CKI is necessary and sufficient to promote an increase in SMO levels and its localization at the plasma membrane, but the temporal, spatial and functional connections between these events remain elusive. Several scenarios can be considered. For instance, SMO phosphorylation could block SMO endocytosis once it has reached the cell surface or it could upregulate the targeting of newly synthesized SMO to the plasma membrane. Our data, in agreement with (Kupinski et al., 2013), strongly support the first hypothesis, as follows: (i) the increase of SMO induced by HH, by the phosphorylation of its PKA/CKI sites, or both, is almost entirely attributable to an increase of its cell surface pool, with only very modest effects on intracellular SMO levels; and (ii) the phosphorylated SMO molecules are entirely located at the cell surface and, reciprocally, all of the cell surface fraction is phosphorylated.

Moreover, the phosphorylation of the Ct-sites seems to act by increasing the effects of the PKA/CKI sites on the stabilization of SMO at the plasma membrane. Mimicking the phosphorylation of the Ct-sites leads to a strong increase of the fraction of SMO present at the plasma membrane, and this effect is most visible either in the presence of HH – when the PKA/CKI sites are phosphorylated – or when their phosphorylation is mimicked.

Phosphorylation of the PKA/CKI sites increases SMO activity by antagonizing a region called the SMO auto-inhibitory domain (SAID), which shuts down SMO activity in the absence of HH, owing to its intermolecular electrostatic interaction with acidic aa of the most C-terminal region of SMO (Zhao et al., 2007). As a result, the SMO cyto-tail dimerizes. This effect is increased by GPRK2/GRK2 phosphorylation (Maier et al., 2012). By contrast, FU acts on sites located in or near the acidic region, and their phosphorylation further increases the negative charges in this region, which suggests that they might activate SMO by a new (so far, not understood) mechanism. Nonetheless, the effects of phosphorylation of Ct-sites strengthen the tight correlation between SMO signaling activity and its accumulation at the cell surface, and they further support the notion that the concentration or number of SMO molecules present at the cell surface is crucial for HH signaling. Finally, the phosphorylation of the Ct-sites of SMO appears also to be involved in the apicobasal localization of SMO, leading to increased basal localization in the HH receiving cells. Although the significance of this differential localization awaits further experimentation, it is interesting to consider it in the light of the double HH gradient that has been described: apical for the long range, ‘low HH’ responses and basal for the short range, ‘high HH’ responses (for review, see Guerrero and Kornberg, 2014).

Both phosphorylation of SMO Ct-sites and FU kinase activity are necessary for high levels of HH signaling and for the increase in SMO levels and cell surface localization induced by HH (Claret et al., 2007; Liu et al., 2007). So far, these effects have been attributed to the ability of FU to phosphorylate COS2 and SU(FU). Our results provide a new output for FU as they show that it is likely to upregulate the levels of SMO directly at the cell surface and its signaling activity by phosphorylation. It is noteworthy that the FU-dependent phosphorylation of COS2 has been recently shown not to be essential for its activity as SMO and COS2 interact (Zadorozny et al., 2015); it is thus possible that the effect of FU on COS2 activity could in fact occur mainly via SMO. Our work also reveals that the FU downstream activity [i.e. the phosphorylation of SU(FU) and COS2] requires the phosphorylation of SMO Ct-sites. It indicates that SMO activates FU in a two-step mechanism: (i) an initial activation of FU by recruitment to the plasma membrane and that promotes its action on SMO; and (ii) a second step that requires the phosphorylation of SMO Ct-sites, which is necessary for FU effects on its cytosolic targets and the activation of CI-FL.

These results provide evidence of the interplay between SMO and FU which, by increasing SMO phosphorylation, leads to high SMO/FU activation. We propose a model whereby SMO/FU activation is based on the progressive multisite phosphorylation of SMO and on a series of activation loops that take place between SMO and FU in the presence of high levels of HH (Fig. 6). This amplifying loop finally leads to a high accumulation of the hyperactivated SMO/FU couple at the cell surface, reaching a threshold of activation sufficient to promote high levels of CI activation and the ‘high HH’ specific responses. Interestingly, multisite phosphorylation can lead to ultrasensitivity under some circumstances and to bistability when combined with a positive feedback loop (for review, see Ferrell and Ha, 2014a,b; Lander, 2007). In this context, the crosstalk between SMO and FU might provide both robustness and switch-like behavior to the HH pathway that could account in part for its morphogenetic effects. Although SMO has been conserved from flies to mammals, significant differences have appeared between human and fly SMO and its regulation (Ingham et al., 2011). This has probably led to some changes in the molecular mechanism, but the underlying principles of the regulation might well have been conserved.

For transgenesis, wild-type and mutants smo cDNAs were introduced into the same landing site (9738) on 3R (at 99F8) by BestGene, using the PhiC31 integration system. For MARCM clones, the larvae were heat-shocked 3-4 days after egg laying for 2 hours at 37°C. Further details are given in the supplementary Materials and Methods.

Discs were dissected and treated as described by Claret et al. (2007) with minor modifications (see supplementary Materials and Methods). The primary antibodies used were: mouse anti-PTC, 1:1000 [Apa 1, from Developmental Studies Hybridoma Bank (DHSB)] (Martin et al., 2001); mouse anti-SMO, 1:100 (20C6, from DSHB) (Lum et al., 2003); mouse anti-COL, 1:100 (a gift from M. Crozatier, Center for Integrative Biology, Toulouse, France) (Crozatier et al., 2003); rat anti-full length CI, 1:5 [2A1 from R. Holmgren, Northwestern University, Evanston, IL, USA (Motzny and Holmgren, 1995)]; and rabbit polyclonal anti-β-galactosidase 1/100 (ABR, Affinity BioReagents).

All expression vectors were constructed by the Gateway recombination method (Invitrogen) and checked by sequencing the mutated regions. Details of the constructs used here are presented in the supplementary Materials and Methods.

Cl-8 cells were cultured as described previously (Claret et al., 2007). Transient transfections were then carried out using Transit Insect Reagent (Mirus) and analyzed 48 h post transfection. Note that to visualize SMO protein, the samples were not heated. The samples were run on 10% Anderson gels. See supplementary Materials and Methods for more details.

Forty-eight hours after transfection, cells were washed in PBS then either labeled for 20 min in complete cell medium with 0.75 µM SNAP TMR-star and 0.75 µM CLIP TMR-star (NEB) for total SMO accumulation or labeled for 10 min with 0.75 µM SNAP-Surface 782 in complete cell medium, before adding 0.75 µM SNAP TMR-star and 0.75 µM CLIP TMR-star for 10 min for surface and intracellular staining. In both cases, cells were then washed twice and lysed in 1% NP-40, 150 mM NaCl, 50 mM Tris, 1 mM dithiothreitol with complete EDTA-free antiprotease mix (Roche) and Phosphatase Inhibitor Cocktail Set II (Calbiochem). The lysate was centrifuged (12,000 g) for 10 min at 4°C and mixed with Laemmli sample buffer (Bio-Rad) and 0.1 M dithiothreitol. Fifteen microliters of each sample was loaded on precast 4-20% Mini-Protean TGX gels (Bio-Rad) and run as for western blots (see supplementary Materials and Methods). Gels were scanned either on a Typhoon Trio imager (GE Healthcare; excitation 532 nm, emission 580 nm, PMT 700 V) for total accumulation or sequentially (for surface and intracellular staining) on a Lycor Odyssey at 800 nm before Typhoon Trio. The merged image was obtained by overlapping the two images. Three independent experiments (with all the constructs together) were performed. All gels were run and scanned together. Images were analyzed and quantified using ImageJ software. One tailed bivariate Wilcoxon rank tests were run using R software and R Commander (R Foundation for Statistical Computing).

Third instar larva were dissected in S2 cell medium. They were then incubated for 30 min with the nonliposoluble Alexa 546 fluorescent substrate at a dilution of 1/600 in S2 medium to allow fixation of the substrate and then incubated for 10 min at 25°C. The larvae were fixed 20 min with PFA, followed by three 10 min washes (PBS with 0.3% Triton) before immunostaining or direct imaging.

We thank M. Crozatier, J. Jiang, D. Kalderon, T. Kornberg, S. Ogden, D. Robbins, P. Thérond, A. Zhu and their laboratories for flies and reagents. We are grateful to F. Schweisguth, L. Ruel, G. D'Angelo, P. Therond, A. Brigui and our colleagues from the Institut Jacques Monod for insightful discussions and to T. Herbillon for technical help. Antibodies from the DSHB were developed under the auspices of the NICHD and maintained by the University of Iowa. Drosophila embryo injections were carried out by BestGene Inc. We acknowledge the ImagoSeine core facility of Institut Jacques Monod, member of France-BioImaging (ANR-10-INBS-04) and certified IBiSA.