A large complex containing Patched and Smoothened initiates Hedgehog signaling in Drosophila

Hedgehog (Hh) is a secreted protein that patterns and specifies cell fate in several different tissues. In Drosophila, response to Hh is mediated by Cubitus interruptus (Ci), a transcription factor with either activator or repressor functions (Dominguez et al., 1996; Aza-Blanc et al., 1997; Methot and Basler, 1999). The processing of Ci into activator or repressor is achieved through the Hh cytoplasmic complex that includes the kinesin-like protein, Costal-2 (Cos2), the serine-threonine kinase Fused (Fu), and Suppressor of Fused [Su(fu)] (Robbins et al., 1997; Sisson et al., 1997; Stegman et al., 2000). Secreted Hh binds to its receptor Patched (Ptc). As a result, Fu and Cos2 are hyperphosphorylated (Therond et al., 1996; Robbins et al., 1997; Nybakken et al., 2002), the cytoplasmic complex dissociates from microtubules, and full-length Ci, Ci155, translocates into the nucleus as a transcriptional activator (Ohlmeyer and Kalderon, 1998; Chen et al., 1999; Wang and Holmgren, 1999). In the absence of Hh, Ci is proteolyzed to its repressor form, Ci75. This proteolysis involves phosphorylation of Ci by PKA, GSK3 and CKI and the activity of the F-box protein, Slimb (Aza-Blanc et al., 1997; Robbins et al., 1997; Jiang and Struhl, 1998; Chen et al., 1999; Methot and Basler, 2000).

The fate of Ci is controlled by Ptc, a twelve-pass transmembrane protein, and Smoothened (Smo), a seven-pass transmembrane protein. In the absence of Hh, Ptc suppresses Smo and this triggers the proteolysis of Ci. When present, Hh relieves the Ptc-mediated suppression of Smo, leading to phosphorylation and stabilization of Smo, activation of Ci, and degradation of Ptc and the Hh ligand (Aza-Blanc et al., 1997; Denef et al., 2000; Alcedo et al., 2000; Zhang et al., 2004). The Smo C-terminal tail has been shown to directly bind to Cos2 (Ogden et al., 2003), and upon Hh inactivation of Ptc, activates signaling through Cos2 and the associated Hh cytoplasmic components (Hooper, 2003; Lum et al., 2003; Jia et al., 2003; Ogden et al., 2003).

Sex lethal (Sxl) functions as the master switch in sex determination, controlling somatic sexual development and differentiation in Drosophila. It is activated in females but is inactive in males. These two modes of expression are maintained throughout the life cycle (Sanchez and Nöthiger, 1983; Cline, 1984). Sxl promotes female differentiation by regulating transformer (Boggs et al., 1987; McKeown et al., 1987) and the dosage compensation process (reviewed by Lucchesi et al., 2005).

Previously, we showed that Sxl enhances the Hh signal and proposed that this was the mechanism by which Sxl generates the larger female body size (Horabin, 2005). In the wing disc anterior compartment, Sxl responds to the presence of Hh in a Ptc-dependent manner but Smo activity is not required (Horabin et al., 2003). Here we show that Sxl is part of the cytoplasmic Hh complex that is tethered to the Smo carboxyl tail. We examined whether Ptc is also a member of the Hh signaling complex, and found that Ptc and Smo can be co-immunoprecipitated from embryonic extracts. This association could be enhanced or disrupted, depending on the activation state of Hh signaling. Immunoprecipitation and colocalization studies suggest a large signaling complex at the plasma membrane that includes both membrane proteins and the cytoplasmic components. This complex is endocytosed in a dynamin-dependent manner. Our data, together with that of others, suggest an altered model for passage of the Hh components through the endocytic pathway. Colocalization of Ptc, Smo, Ci and Sxl in salivary gland cells when signaling is on versus off suggests that, in females, Sxl may decrease the proteolytic cleavage of Ci by altering its endocytic cycling with the Hh membrane components, thereby enhancing the Hh signal.

As Sxl co-immunoprecipitates with the cytoplasmic Hh components Cos2, Fu and Ci, and studies show that Smo physically interacts with the Cos2-Fu complex via its C-terminal tail (Jia et al., 2003; Lum et al., 2003; Ogden et al., 2003; Ruel et al., 2003), we investigated whether Sxl would also co-immunoprecipitate Smo. Wild-type embryonic extracts show that Sxl immunoprecipitated Smo protein (∼5.2%) and the converse is also true; Smo immunoprecipitated Sxl (∼2.7%; Fig. 1A). By comparison, full-length Ci (Ci-155) immunoprecipitated ∼4.7% of Sxl and Smo; Cos2 ∼5.1% of Sxl and ∼3.5% of Smo. These immunoprecipitation (IP) values are in the range of or are slightly more efficient than those seen by others (Ogden et al., 2003) [see Fig. 5 of Jia et al. (Jia et al., 2003)] and indicate that Sxl is present in all the known Hh cytoplasmic complexes described to date.

Although Sxl resembles Ci in its association with the Hh signaling components, removing Smo activity has no effect on Sxl nuclear entry in the wing disc. It is Ptc which promotes nuclear entry of Sxl in a Hh-dependent manner (Horabin et al., 2003). This dependence on Ptc suggested that an association between Sxl and Ptc might exist. To investigate this possibility, we determined whether Ptc was also present with the Hh signaling components in immunoprecipitates from wild-type Drosophila embryo extracts. Sxl immunoprecipitated a significant amount of Ptc (∼10.7%, Fig. 1A) and inversely Ptc immunoprecipitated ∼3.1% of Sxl. As Sxl is in a complex with the other Hh components, we determined whether Ptc was present in immunoprecipitates of full-length Ci, Smo and Cos2. We obtained IP values of ∼6.8, ∼2% and ∼1.6%, respectively, for Ptc. Although the cause of the differences between these proteins is not clear (there might be different complexes or the stability of each component within the complex may be different), the fact that Sxl, Ci and Smo can co-immunoprecipitate Ptc suggests that Ptc is part of the Hh signaling complex. As a negative control, we probed immunoprecipitates with antibodies for proteins with similarities to those in the Hh signaling pathway (Fig. 1D). Bicuadal-D (Bic-D), a kinesin-like protein, Discs-large (Dlg), a member of the guanylate kinase family, Frizzled (Fz), the Wnt receptor a seven-pass transmembrane protein homologous to Smo, and evenskipped, an active repressor of transcription (data not shown) were tested. The immunoprecipitates were positive for the Hh component, but not for any of the controls.

To additionally show that Ptc is a part of the Hedgehog signaling complex, we fractionated cytoplasmic extracts from wild-type embryos (Fig. 1B). Three distinct populations can be described. In population A we found that Sxl cofractionates with Ptc, Cos2, and Ci (trace amounts of Fu were also detected). This population was relatively large (>700 kDa as noted by the size standards), suggesting the presence of additional proteins or from trimerization of Ptc (Lu et al., 2006). Population B, contained Ci in the first few fractions, and had Ptc, Smo, Fu, Cos2 and Su(Fu) [Su(Fu) not shown] along with Sxl. Population B is also relatively large (centered around 669 kDa) and suggests that the Hh signaling complex can contain both Sxl and Ci, or only Sxl, as in population A. In population C, which had little Sxl, Ptc is still present along with the other Hh signaling components. These complexes were smaller than 440 kDa, so they are unlikely to have several of the Hh components together, and in the fractions of the smallest size, probably exist as monomers.

Populations A, B and C were tested for co-IP of Sxl and Ptc with Cos2, Ci, Sxl or Smo for which they were positive (Fig. 1C), supporting the contention that Ptc and other Hh signaling members exist in a complex. Analysis by native gel electrophoresis also supported this conclusion; at least two extremely large complexes that contain the various Hh signaling proteins, including Ptc and Smo, were detected (see supplementary material Fig. S1A,B). These large complexes entered the stacking gel but failed to enter the resolving gel. The negative controls Eve (see supplementary material Fig. S1B) and Dlg (not shown) were in the resolving gel alone. Fz was detected in the stacking as well as the resolving gel (not shown) but as demonstrated above it did not co-immunoprecipitate with the Hh components, suggesting that it exists in a large complex of its own.

We next examined whether the activation state of the pathway affected the association of Ptc with the other Hh components. The pathway was activated using the Ptc^1130X variant, which has the last 156 amino acids of the cytoplasmic tail deleted. Ptc^1130X is a dominant-negative allele and leads to Ci activation independently of the Hh morphogen, inducing targets which require high Hh levels for activation (Johnson et al., 2000). Embryos expressing this and the variants that follow were generated using the UAS-GAL4 system; homozygotes of the ubiquitous daughterless-GAL4 (da-GAL4) driver were crossed to a line homozygous for the UAS-Ptc construct. Western blots showed that the prevailing Ptc is the variant form, which for Ptc^1130X migrated a little faster than the endogenous protein.

Immunoprecipitates of Sxl, Ci, Cos2 and Smo were probed with antibodies for Sxl, Ptc and Smo and each IP yielded ∼5-6% of the tested proteins (Fig. 2A). The exception was Smo which immunoprecipitated ∼9% of Sxl. Compared with wild-type embryos, the amount of Ptc immunoprecipitated increased regardless of the IP protein, and the amount of Sxl in complexes that contain Smo also increased from ∼3% to ∼9%. These data demonstrated that the activity state of the pathway alters the association of Ptc with the other Hh components. Removal of the last 156 amino acids appears to not only hinder the ability of Ptc to inhibit Smo, it also `locked' Ptc into the complex.

Intermediate activation of the pathway had a different effect. Ptc^D584N has an aspartic acid changed to asparagine at position 584 in the sterol sensing domain (SSD), is also a dominant negative and activates the pathway, although not as strongly as Ptc^1130X (Johnson et al., 2002). The amount of Ptc immunoprecipitated (∼1%) by Sxl and Ci appeared to decrease compared with the wild type (Fig. 2B), Smo also immunoprecipitated slightly less Ptc whereas Cos2 immunoprecipitated ∼4% of Ptc, elevated from the wild type (∼2%). The amounts of Sxl and Smo that were associated with Ci did not change substantially and remained close to ∼5%.

These data suggest that complexes that contain full-length Ci with Sxl or Smo remain unaltered when the pathway is activated or `on'. However, Smo was able to more effectively immunoprecipitate Sxl than in the wild-type condition (from ∼3% to ∼5%), suggesting that the association of Sxl with Smo within the complex may be enhanced while the association of Ptc is disrupted.

Although Ptc^1130X and Ptc^D584N were overexpressed, each displayed different interactions with the other components, suggesting that overexpression per se does not determine association. Both activate the Hh pathway; however, with respect to protein turnover, Ptc^D584N more closely resembles wild-type protein bound to ligand (Martin et al., 2001; Strutt et al., 2001; Lu et al., 2006). This suggests that the association of Ptc within the complex is disrupted when the pathway is turned on.

For the `off' state, wild-type Ptc or Ptc^ΔLoop2 was expressed in embryos. Overexpression of wild-type Ptc switches the system off as there is not sufficient endogenous Hh. Ptc^ΔLoop2 also suppresses pathway activation; it is unable to bind Hh as it lacks the extracellular loop between transmembrane segments seven and eight necessary for Hh binding. Embryos overexpressing wild-type Ptc showed full-length Ci, Cos2, Smo as well as Sxl co-immunoprecipitated with significant amounts of Ptc (10% or greater, Fig. 2C). The Hh components also immunoprecipitated significant amounts of Sxl (∼8% or greater), suggesting that Sxl remains a part of the complex when the pathway is off. Association of Ptc with Smo and the cytoplasmic Hh components appeared to be enhanced when the pathway is off. Compared with the wild type, all the components tested showed substantial increases in the amounts of Ptc that co-immunoprecipitated with them. Similar results were obtained when the pathway was off through Ptc^ΔLoop2 expression (Fig. 2D).

The data taken together, suggest a large complex that consists of Ptc, Smo and the cytoplasmic Hh components (and Sxl in females). The similarity in the IP profile of wild-type embryos to that of embryos expressing Ptc^D584N rather than either variant that turns the system off, also suggests that wild-type embryos reflect primarily an activated state.

As embryonic extracts do not distinguish between males and females, it could be argued that Sxl accounts for the IP of Ptc with the other Hh components. Wild-type male and female flies showed that females do indeed give a greater yield of Ptc in their Ci and Cos2 immunoprecipitates (Fig. 3G), however, males also showed Ptc, indicating that its presence is not sex specific.

If the Hh cytoplasmic complex directly interacts with Ptc, it should be able to do so in the absence of Smo. To demonstrate that Ci can associate with the plasma membrane when Ptc is the lone Hh membrane component, we removed Smo in male salivary gland cells using the Smo^D16 deletion allele. As the levels of Ci are greatly reduced under these conditions, a mutation in PKA was introduced to increase the levels of Ci. Clones of cells mutant for Smo^D16 and PKA showed that 15% of the Ptc and 14% of the Ci colocalize at the plasma membrane and the vesicular network (Fig. 3A-F). Although this was a little lower than we observed in the wild-type condition (see below) and might suggest a more stable complex exists when both Hh membrane proteins are present, the colocalization is consistent with the idea that the Hh cytoplasmic complex is associated with Ptc.

As Sxl depends on Ptc to respond to Hh in wing discs, we reasoned that Sxl might associate directly with Ptc. This prompted us to determine whether overexpression of Ptc would affect the localization of Sxl (Fig. 4A-F), because, unlike most of the other Hh components that reside in the cytoplasm, Sxl in embryos is primarily nuclear (Fig. 4A').

Overexpression of Ptc^1130X, PtcC (deletion of the N-terminus up to the seventh transmembrane region) and Ptc¹³ titrated Sxl out of the nucleus (Fig. 4D',E'; Ptc¹³ not shown). Wild-type Ptc and Ptc^ΔLoop2 gave intermediate effects: Sxl was detected in both cytoplasm and nucleus (Fig. 4B',F'), whereas Ptc^D584N did not significantly alter the nuclear localization of Sxl (Fig. 4C'). Both dominant negatives, Ptc^1130X and Ptc^D584N, result in the activation of Ci; however, only Ptc^1130X sequesters Sxl in embryos and facilitates its nuclear import in wing discs (our unpublished results), correlating these two properties of Ptc.

The two Ptc dominant negatives also suggest that pathway activation is not responsible for the observed change in Sxl localization. To confirm this, the state of Hh signaling was directly assayed by staining embryos for full-length Ci and Wingless (Wg; Fig. 4G-L). As predicted, the levels of Ci increase when Ptc^1130X, Ptc^D584N or Smo is expressed, and decrease when Ptc⁺ or Ptc^ΔLoop2 is expressed. The Wg signal was also altered in these backgrounds, reflecting its dependence on Hh signaling. That transcriptional activation by Hh was not involved in changing the subcellular localization of Sxl in embryos, is also supported by the observation that overexpression of Hh itself had little effect, as did the overexpression of Smo (data not shown). Since Ptc variants that both activate or inhibit Hh signaling can sequester Sxl, and the sequestration does not show segmental repeats, it is more likely that the effect is caused directly by the overexpression of the Ptc proteins.

As PtcC can titrate Sxl out of the nucleus, the carboxyl half of Ptc must provide the docking site. The 156 C-terminal residues do not appear to be involved, however, as their removal in Ptc^1130X did not severely compromise sequestering Sxl out of the nucleus. Between PtcC and Ptc^1130X, the cytoplasmic regions remaining are the loop between transmembrane segments 10 and 11 and the last ∼27 amino acids before the 1130X deletion point.

Depending on the signaling condition, Ptc and Smo appeared to be in the same complex. To determine whether they colocalize in vivo, we examined their distribution with full-length Ci in the large salivary gland cells (Fig. 5A-L); Ci serves as the marker of the cytoplasmic Hh complex. The state of Hh signaling in wild-type salivary glands is normally off (Zhu et al., 2003). To activate it, Hh was expressed using a salivary gland GAL4 driver, sgs3 (Fig. 5M-X).

Optical sections were taken near the plasma membrane as well as deeper within the cell to include a cross section of the nucleus. Sections near the plasma membrane sample vesicles more recently budded from the plasma membrane, whereas those deeper in the cell sample later stages of the endocytic cycle. Colocalization of the two Hh membrane proteins with Ci was compared between males and females, in the presence and absence of Hh (Fig. 5, Fig. 7A1).

Several observations stand out. First, for both Ptc and Smo the plasma membrane as well as the underlying vesicular network, appeared as sites of colocalization with Ci. Second, in the absence of Hh the degree of colocalization of the two Hh membrane proteins with Ci was generally a little lower in females than males, both near the plasma membrane and deeper within the cell. Third, in the presence of Hh, these trends were quite dramatically altered. Females showed more Ci colocalizing with Ptc at the plasma membrane than males, presumably reflecting a heightened response to signaling (Ptc is the receptor). They also showed a more dramatic decrease of Ci colocalized with Ptc deeper within the cells than males. This enhanced separation of Ci from Ptc in females was mirrored by an increased level of colocalization of Ci with Smo within the cell. Combined, this suggests that in the presence of Hh female cells might recruit more Ci to receive the signal at the plasma membrane, but as Ptc traffics towards degradation deeper within the cells, females segregate more of their Ci away from Ptc. Changes were also seen in the colocalization of Sxl with Ptc, Smo and Ci in the presence of Hh (Fig. 7A2); more of the Sxl and Ptc signal overlapped at the plasma membrane, and much of the Sxl appeared to dissociate from these Hh components within the cell.

Although this colocalization approach is limited – the images are not a perfect sampling of specific subcellular compartments – it did show Ci colocalization with Ptc, besides the expected colocalization of Ci with Smo. Additionally, the differences between the sexes are reliable as they are generated with similarly treated samples.

The wild-type data support the idea that Ptc, Smo and Ci are together at the plasma membrane. Since Dynamin [Shibire (Shi)] is required for the pinching of clathrin-coated pits from the plasma membrane (van der Bliek and Meyerowitz, 1991), reducing Shi activity should inhibit the first step of endocytosis and `freeze' the Hh components there. Consistent with this prediction, expressing the Shi^K44A dominant negative essentially doubled the colocalization of Ci with Ptc in sections near the plasma membrane in both sexes [from 20% to 41% in males, 15% to 33% in females (Fig. 6G-L, Fig. 7B1)]. Ci and Ptc colocalization is also increased in sections deeper within the cell. The colocalization of Ci and Smo near the plasma membrane did not change as dramatically, and was affected even less within the cell (Fig. 6A-F, Fig. 7B1,C1). Sxl showed the same trends as Ci in its changes in colocalization with Ptc and Smo, particularly at the plasma membrane (Fig. 7B1,B2).

Besides affecting the proteins at the plasma membrane, inhibiting the first step of endocytosis decreased the sexual dimorphism in the Ci and Ptc colocalization, and eliminated it for Ci and Smo. This suggests that the sexual difference is primarily a function of endocytic cycling, occurring after the events at the plasma membrane.

Blocking endocytosis at the first step appears to increase the colocalization of Ci, and Sxl with Smo and Ptc. To confirm this, IPs using extracts from embryos expressing the dominant negative Shi^K44A were performed. Relative to the wild type, the dominant negative Shi increased the amount of Ptc complexed with some of the Hh components. Ci and Sxl immunoprecipitated ∼8.4% and ∼8.7% of the Ptc, which is similar to the wild type. However, Smo and Cos2 immunoprecipitated ∼4.9 and 4% of the Ptc, an increase from ∼2% for both of them (Fig. 8B). The amount of Sxl immunoprecipitated by Smo and Ci was also elevated. This is in keeping with the idea that blocking the first step of endocytosis traps Ptc and the entire Hh signaling complex at the plasma membrane and that Ptc segregates from Smo and Cos2 as it progresses through the endocytic pathway.

From clathrin-coated pits, budding vesicles fuse with endosomes in a process that is mediated by the small Rab5 GTPase. Inhibiting the endocytic pathway at this stage using a dominant negative Rab5^S43N variant (Stenmark et al., 1994; Entchev et al., 2000), showed the amount of Ptc immunoprecipitated by Sxl is considerably less (∼3.9%, Fig. 8B) than for the Shi dominant negative. The amount of Ptc in the Smo and Cos2 IP appeared to elevate a little (from ∼4 and 4.9% to ∼7.2 and 7.3%, respectively), while Ci levels appeared to drop (from ∼8.4 to 5.2%). The co-IP of Sxl with Ci and Smo also decreased relative to the Shi dominant negative condition. This suggests that the association of both Sxl and Ci with Ptc is weakened, while Ptc continues to progress with Smo and Cos2 through the endocytic pathway. Sxl also appeared to reduce its association with Smo and Ci but its association with Cos2 was not significantly altered.

Rab7 is a GTPase essential for transporting endocytic cargo from the early to late endosomes and lysosomes (Vitelli et al., 1997). Embryos expressing a dominant gain-of-function Rab7 (Rab7^Q67L) (Entchev et al., 2000) also showed lower Sxl and Ptc association (∼2.9%, similar to the Rab5 dominant negative and down from ∼8.7% in the Shi dominant negative). However, the degree of association between Ptc with Ci, Smo or Cos2 was not substantially different from the Shi dominant negative condition (Fig. 8B). The Rab7 gain of function increased the Ci and Ptc association compared with the Rab5 dominant negative, as well as the interaction of Sxl with Smo and Ci (Fig. 8A). Consistent with this observed inhibition in sorting of Hh components, overexpression of wild-type Rab7, which increases overall Rab7 activity, also impairs the motility of endosomes (Lebrand et al., 2002).

In summary, these data suggest that Sxl and Ci begin to segregate from Ptc relatively early, prior to the stage requiring Rab5. Sxl also decreases its association with Ci and Smo. The colocalization analyses (Fig. 7A2) also suggest that within the cell, Sxl has low levels of colocalization with Ptc, Smo and Ci when signaling is on: the state embryos more strongly reflect. Overactive Rab7 reverses the segregation of Sxl from Ci and Smo, but not from Ptc suggesting that Sxl and Ptc separate earlier in endocytosis. Ci appears to separate from Ptc a little later than Sxl, whereas segregation of Smo from Ptc must occur beyond the stage regulated by Rab5. Overactive Rab7 prevents Smo and Ptc segregation, however, as they have opposite endocytic fates they would have to separate prior to sorting to the lysosome (Denef et al., 2000; Incardona et al., 2002; Zhu et al., 2003).

The sexually dimorphic colocalization of Ci with Ptc and Smo presumably reflects the effects of Sxl. Sxl appears to dissociate from some of the Hh components relatively early in the endocytic cycle. We therefore compared the colocalization of Ptc and Smo with Ci (and Sxl in females) in males and females expressing a Rab5 dominant negative and a Rab7 gain-of-function variant.

Inhibiting the endocytic cycle at the Rab5 stage showed a drop in the level of Ci and Ptc that colocalized at the plasma membrane in both sexes compared with the Shi dominant negative (41% to 29% males, 33% to 22% in females; Fig. 6R', Fig. 7B1). These values are still higher than the wild type, and since the Shi dominant negative presumably reflects the state of the components as they begin endocytosis at the plasma membrane, suggest that sorting of the proteins is still in progress. Colocalization of Ci with Smo showed a smaller decrease relative to Shi, in both sexes (33% to 27% in males, 33% to 29% in females, Fig. 6O', Fig. 7B2). Within the cells, these trends were upheld for Ptc, but not for Smo, which remained almost unchanged. These data suggest that when the pathway is off (the ground state of salivary glands), Ptc and full-length Ci segregate from each other by the Rab5 stage; but the segregation of Ci and Smo does not change considerably.

Progression from early to late endosomes involves displacement of Rab5 by Rab7 (Rink et al., 2005). The dominant gain-of-function Rab7 (Rab7^Q67L) (Entchev et al., 2000) might be predicted to accelerate Rab7 function and compromise sorting of components in Rab5 positive vesicles. We found that the colocalization values resembled the Shi dominant negative and also the Rab5 dominant negative condition (Fig. 7B1-2,C1-2, images not shown), but not the wild type suggesting a reversal in sorting events. In general, the changes in Smo and Ci colocalization were smaller than for Ptc with Ci. What stands out most clearly is that females showed an elevation over males of Ci and Ptc colocalization at the plasma membrane, and Ci and Smo colocalization within the cell. These differences indicate that the sorting effects on the Hh proteins caused by the Rab7 gain of function do not occur similarly in males and females.

With respect to Sxl, altering endocytosis with the Rab5 dominant negative and Rab7 gain-of-function variants, generally increased the amounts of Sxl that colocalized with all three Hh proteins relative to the wild type near the plasma membrane. Sxl did show a drop in its colocalization levels with both Ptc and Ci with the Rab5 dominant negative, suggesting a separation of Sxl from these two components around the Rab5 stage. The Rab7 gain of function appeared to reverse this dissociation. Within the cell, the colocalization of Sxl with Ptc was enhanced by both Rab5 and Rab7 variants, particularly the Rab7 gain-of-function variant; the association of Sxl with Smo and Ci was generally also slightly elevated.

These dynamics suggest a complex picture of components in different associations, depending on their stage within the endocytic cycle. They suggest that (1) in the wild-type condition, the overall colocalization of Ptc and Smo with Ci is low, and perturbing the early stages of endocytosis tends to increase their (and Sxl in females) colocalization levels; (2) females generally show lower protein colocalization levels than males and perturbing the early stages of endocytosis tends to decrease the difference; (3) males and females show variances in their response to sorting perturbations, particularly by the Rab7 gain-of-function variant, where females show greater fluctuations; (4) changes in Ptc colocalization with Ci tend to be greater than for Ci with Smo, with Smo showing greater changes near the plasma membrane than within the cell.

We proposed that Sxl enhances the Hh signal to generate the larger female body size (Horabin, 2005). Sxl is part of the Hh cytoplasmic signaling complex (Horabin et al., 2003) as well as the complex attached to the Smo tail, suggesting that it is an integral signaling member. To elucidate how Sxl enhances the Hh signal, we attempted to quantify the associations of the various components under different signaling conditions. We found that Sxl influenced the interactions between the components during endocytosis and, more importantly, that Ptc is also a member of the Hh signaling complex.

The IP data showed Ptc with full-length Ci, Sxl and the other Hh signaling proteins in wild-type embryos; complexes containing both Ptc and Smo suggest that this association becomes stronger when signaling is off. The complex containing Ptc is not sex specific, although female flies give greater yields than males.

The idea that Drosophila Ptc and Smo are in a common complex runs counter to prevailing thought. Experiments in mammalian cells suggested that Ptc might associate with Smo (Stone et al., 1996; Murone et al., 1999), but subsequent experiments in S2 cells questioned whether these observations applied to the Drosophila proteins (Johnson et al., 2000). Ruel et al. (Ruel et al., 2003) also reported that Ptc did not immunoprecipitate with Cos2 in cl-8 cells. This report is the first using Drosophila embryonic extracts, and we also find a weak association of Ptc with both Cos2 and Smo. However, this result changes depending on the state of Hh signaling.

The addition of Ptc to the Hh signaling complex was further corroborated by size fractionation experiments, which show Ptc in large complexes that also contained other known Hh components. It is not clear what proteins account for the very large complexes of population A in Fig. 1B, which contain Ptc, Cos2, Sxl, Ci and trace amounts of Fu. Ptc can trimerize (Lu et al., 2006), and in vivo data suggest Ptc functions as a trimer (Casali and Struhl, 2004) requiring its co-receptors Ihog and Boi (Yao et al., 2006). Additionally, some of the other Hh components besides Ptc may be present as multimers, or the complex may also contain unknown proteins.

Colocalization analyses also showed Ptc and Smo together with Ci at the plasma membrane and in the network of the endocytic pathway. When endocytosis is blocked in embryos expressing a dominant negative Shi variant, an enhancement in co-IP of Hh complex proteins with Ptc is observed, suggesting a large complex with all known signaling components at the plasma membrane. Images also show an increase in the amounts of Ptc and Smo that colocalize with Ci at the plasma membrane, supporting this view.

Finally, the Ptc tail must play a key role in influencing the interactions within the complex. Removal of the last 156 amino acids not only hinders Ptc^1130X from inhibiting Smo, it also stabilizes most of the interactions within the complex. It also appears to affect the endocytic cycling and stability of Ptc. As opposed to Ptc^1130X being proteolyzed, which is the normal fate for Ptc when Hh signaling is on, high levels of the dominant negative protein are detected at the plasma membrane (Zhu et al., 2003; Lu et al., 2006).

The yields of Ptc with the Hh cytoplasmic components was greater in females than males, suggesting Sxl stabilizes Ptc within the complex. Sxl may bind directly to Ptc, as overexpression of Ptc can titrate Sxl out of the nucleus.

How does Sxl enhance the Hh signal particularly as it stabilizes the association of Ptc, a negative component in signaling? We find that Sxl influences the segregation of the Hh components during endocytosis. Sex-specific differences in the colocalization of Ci with both Ptc and Smo are mostly eliminated by blocking the first step of endocytosis (Fig. 7B1-2), suggesting that Sxl impacts signaling events at or immediately after the plasma membrane. Additionally, sorting responses when endocytosis is perturbed using a Rab5 dominant negative or Rab7 gain-of-function variant, showed differences between males and females. Protein sorting must still be in progress during these early steps in endocytosis as perturbing them increases the levels of colocalization of the proteins relative to the wild type (including Sxl).

When signaling is off, females show lower amounts of full-length Ci colocalizing with both Ptc and Smo. When signaling is on however, females show more Ci and Ptc colocalizing near the plasma membrane, and a concomitant increase in the colocalization of Sxl with Ptc (Fig. 7A2). As Ptc is the Hh receptor, this might reflect enhanced signaling or a delay in endocytosis, or both, relative to males. A delay at the plasma membrane would not compromise signaling, but could influence the interactions and subsequent segregation of components within the cell. The Hh complex is capable of signaling at peak levels when endocytosis is blocked by the loss of shi function (Han et al., 2004; Torroja et al., 2004). This indicates that although the Hh components are trapped at the plasma membrane, Hh alters the interactions within the complex and Smo is no longer inhibited by Ptc. In keeping with the idea that Hh modifies interactions within the complex, alterations that affect Ptc activity also affect whether it immunoprecipitates strongly or weakly with the other components.

In the presence of Hh, females also show higher levels than males of Ci and Smo colocalization within cells, which should favor Ci activation and signaling. Conversely, the association of Ci with Ptc within cells is lower in females. The latter would implicate a decrease in proteolysis of full-length Ci because Ptc is proteolyzed when Hh is present. This suggests that with both membrane proteins, females favor production of full-length Ci more than males, sparing more of it from proteolysis as well as activating more. Consistent with this proposal, the amount of full-length Ci is higher in females and Hh signaling is enhanced (Horabin, 2005).

These effects during Hh signaling would appear to mostly occur near the plasma membrane, as the levels of colocalization of Sxl with the Hh components are relatively low within the cell. Indeed, IP experiments suggested Sxl begins to dissociate from the Hh cytoplasmic complex relatively early, before the Rab5 stage (Fig. 8A). The association of Sxl with Ptc appeared to be disrupted even earlier, because unlike the cytoplasmic complex components, the Rab7 gain-of-function variant does not reverse its dissociation from Ptc (Fig. 8B).

Observations by others taken with the data described here, suggest a model for Hh signaling (Fig. 9). In the absence of Hh Ptc, Smo and the Hh cytoplasmic components are endocytosed. All the components, except Ptc, are degraded while Ci is proteolyzed to its repressor form. Most of the Ptc recycles (cyloheximide slowly decreases Ptc levels in the absence of Hh so low levels of Ptc are likely to also turn over) back to the plasma membrane (Denef et al., 2000; Incardona et al., 2002; Zhu et al., 2003), permitting Ptc to repeat the cycle and regulate Smo levels in a `catalytic' manner (Taipale et al., 2002).

In the presence of Hh the initial events are similar. Ptc is now sorted for degradation while Smo is activated, and full-length Ci is generated. Membrane and microtubule association of Cos2 and Fu decreases (Stegman et al., 2004) favoring their early release from Smo and vesicles. Activated Smo recycles to the plasma membrane where it can activate more Ci. Activated Smo destabilizes Cos2 and Fu (Lum et al., 2003; Ruel et al., 2003), suggesting that once Ci is activated, Cos2 and Fu are degraded. The fate of Ci (and associated Hh cytoplasmic components) thus depends on whether Ptc or Smo returns to the cell surface by progression through the endocytic cycle.

Ptc has been described as cycling to and from the membrane independently of Hh (Strutt et al., 2001). Cells that are not exposed to Hh maintain Ptc at the cell surface and in unidentified internal stores, while Smo colocalizes with late endosomes and lysosomal markers. Vertebrate cells transfected with the Hh membrane proteins show that in the presence of Shh (the vertebrate Hh homolog), Ptc and Smo segregate in the late endosomal compartment. Smo recycles back to the membrane while Ptc is fated for lysosomal degradation (Incardona et al., 2002). Endocytosis has thus been proposed to segregate Ptc from Smo, relieving Smo of its repression while inducing the degradation of the Hh ligand (reviewed by Piddini and Vincent, 2003; Torroja et al., 2005).

Hh signaling elevates Smo levels and induces its phosphorylation at the plasma membrane, while lowering the levels of Ptc (Denef et al., 2000; Zhu et al., 2003; Jia et al., 2004; Zhang et al., 2004). The oncogenic Smo proteins, which signal constitutively, do not cycle with Ptc and remain at the plasma membrane (Incardona et al., 2002; Zhu et al., 2003). These observations are accommodated by the above model.

The homology of Ptc to bacterial transporter proteins taken with the findings that small molecules such as cyclopamine can inhibit mammalian Smo, and Ptc acts catalytically in its inhibition of Smo, led to the suggestion that a small molecule transported into the cell by Ptc serves to inhibit Smo (Taipale et al., 2002). This obviated the need for a close interaction. [Note that Casali and Struhl (Casali and Struhl, 2004) suggest that in vivo, the degree of the `catalytic effect' of Ptc on Smo may be smaller than proposed.] Our model does not preclude Ptc from inhibiting Smo through a small molecule. Indeed, their proposed proximity would enhance the efficacy, elevating the local concentration. Additionally, within small endocytic vesicles, Ptc would more rapidly affect Smo as it would require fewer molecules to alter the concentration.

The association of Ptc with Smo need not be direct. Cells incapable of giving a Hh response show very high colocalization of the two human proteins, but the two proteins do not co-immunoprecipitate (Incardona et al., 2002). A direct interaction between the two membrane components may not exist or it may be unstable. Alternatively, the data of Incardona et al. (Incardona et al., 2002) may suggest a requirement for (some of) the cytoplasmic components to bridge the two Hh membrane proteins. Future analysis will determine which is the case.

Flies were raised at 25°C; OreR was the wild-type control. Immunoprecipitations and embryo analysis used the following strains: UAS-ptc1130X, UAS-ptcC, UAS-ptc⁺ (Johnson et al., 2000), UAS-ptcD584N (Johnson et al., 2002), UAS-ptcΔLoop2 (Page, 2002), UAS-shi^DN (Moline et al., 1999), UAS-rab5^S43N, UAS-rab7^Q67L (Entchev et al., 2000), UAS-hh (Azpiazu et al., 1996), UAS-ptc13 (Strutt et al., 2001). Transgenes were expressed using the UAS-Gal4 system (Brand and Perrimon, 1993). da-Gal4 (Wodarz et al., 1995), or the salivary gland driver sgs3-gal4 were used for expression. Description of genes can be found in FlyBase (http://www.flybase.org/).

To generate clones Smo^D16, DCO^H2, FRT40/CyO were mated to y, w, hs70-FLP flies and first instars heat shocked for 30 minutes. Salivary glands from third instars were used for immunostaining.

Except for the wild type which used 50 μl, immunoprecipitations used 75 μl of 0-12 hour embryo extracts. Western blots, immunoprecipitations and whole mount stains were done as described (Vied and Horabin, 2001) varying only the lysis and wash buffers: lysis buffer 100 mM Tris-HCl pH 8.0, 300 mM NaCl, 1% NP40, 2 mM EGTA plus protease inhibitors; wash buffer 50 mM Tris-HCl pH 8.0, 300 mM NaCl, 0.5% NP40, 1 mM EGTA. Antibody dilutions were as described (Horabin et al., 2003) except for anti-Sxl (1:600; western blot), anti-Ptc (1:400; from P. Ingham, Johns Hopkins University, Baltimore, MD), anti-Smo (immunoprecipitation 1:3, from P. Beachy, University of Sheffield, Sheffield, UK; western blot 1:100, from J. Hooper, University of Colorado Health Science Center, Boulder, CO), anti-Cos2 (immunoprecipitation 1:3, from P. Beachy; western blot 1:250, from D. Robbins, Dartmouth Medical School, Hanover, NH) and Fu rabbit polyclonal serum (1:100). The anti-Fu serum was made as described (Robbins et al., 1997). Quantification of immunoprecipitates used IP lab from Scanalytics, taking the average of two or more experiments, except for the Ptc variants where amounts are for samples shown. Whole mount stains used anti-Ci at 1:2.5, anti-Ptc 1:300, anti-Smo 1:100, anti-Sxl 1:250 and anti-Wg at 1:10.

Images were obtained on a Leica TCS_NT or Zeiss LSM 510 Laser Scanning confocal microscope. For colocalization analysis, salivary gland cell images were captured with a 63× objective lens and a zoom of 1-2X. Plasma membrane sections capture the region near the membrane; within cell sections included cross sections of the nucleus. Colocalization percentages used IP Lab colocalization software (Scanalytics). For each image, the threshold value to eliminate background first used a visual guide. The threshold exclusion value was then raised or lowered to determine whether the percentage of signal pixels was significantly altered. The cut off was set as the value that did not significantly alter the percentage of signal pixels included when it was increased (indicating inclusion of significant signal), but did change significantly if it was lowered by two or more increments (indicating inclusion of noise as signal). For a given antibody most images had relatively similar threshold values. The overlap percentage for two probes was generated by the software. Two or more different images were averaged for each genotype and section type.

Fractionations were done as described (Gay et al., 1988).

We are grateful to D. Page, G. Struhl, I. Guerrero, J. Jiang, M. Gonzales-Gaitan, K. Ho and M. Scott, for fly stocks. Thanks also to R. Holmgren, J. Hooper, D. Robbins, P. Ingham for antibodies, R. Huijbregts and I. Chesnokov for help with the fractionation experiment, and G. Marques and J. Engler for input on the manuscript. The anti-Smo and Cos2 antibodies developed by P. Beachy, anti-Wg developed by S. M. Cohen, were from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242. We also thank Albert Tousson from the UAB imaging facility. This work was supported by a grant from NIH to J.I.H. S.L.W. was supported by a fellowship from the Comprehensive Minority Faculty Student Development Program at UAB.