Patched regulates Smoothened trafficking using lipoprotein-derived lipids

The Hedgehog (Hh) signaling pathway is a highly conserved regulator of growth and differentiation that is essential for both vertebrate and invertebrate development (Ingham and Placzek, 2006; McMahon et al., 2003). Its misregulation in adult life leads to the formation of a variety of different tumors (Dellovade et al., 2006; Pasca di Magliano and Hebrok, 2003; Ruiz i Altaba et al., 2002). Hh signals by regulating the level and activity of the transcription factor Cubitus interruptus (Ci). It does so by binding to the transmembrane receptor Patched (Ptc) and preventing Ptc from repressing signaling by the seven-transmembrane domain protein Smoothened (Smo). Smo signaling then prevents processing of the full-length Ci₁₅₅ to a repressor form (Ci₇₅), and promotes nuclear translocation and activation of Ci₁₅₅, apparently by separable mechanisms (Aza-Blanc and Kornberg, 1999; Kalderon, 2005; Lefers et al., 2001; Lum and Beachy, 2004; Nybakken and Perrimon, 2002; Smelkinson et al., 2007).

In the absence of Hh, Ptc catalytically represses Smo activity (Taipale et al., 2002) by a mechanism that correlates with changes in Smo subcellular localization. In Drosophila, Ptc decreases Smo stability and inhibits its accumulation on the basolateral membrane (Denef et al., 2000). In vertebrate cells, Ptc prevents Smo accumulation in the primary cilium where downstream components of the signal transduction pathway are located (Corbit et al., 2005; Haycraft et al., 2005; Huangfu and Anderson, 2005; Huangfu and Anderson, 2006; Rohatgi et al., 2007). Although localization of Smo to primary cilia is required for Smo-dependent target gene activation, it is not sufficient — deletion of the Smo N terminus or treatment with the Smo antagonist cyclopamine both cause Smo to traffic into cilia but inhibit its signaling activity. Other Smo antagonists appear to act by blocking ciliary localization, suggesting that Smo activity might be regulated at multiple steps by different ligands (Aanstad et al., 2009; Rohatgi et al., 2009; Wang et al., 2009).

The mechanisms by which Ptc exerts these effects are not completely understood. Ptc is one of a family of sterol-sensing domain (SSD) proteins that includes SCAP [sterol regulatory element-binding protein (SREBP) cleavage-activating protein] and NPC1 (the protein encoded by the Niemann-Pick type C-1 disease gene), all of which are structurally related to the larger RND (resistance nodulation division) family of proton-driven transmembrane transporters (Tseng et al., 1999). The nearest relative of the Ptc proteins, NPC1, promotes the mobilization of glycosphingolipids and low-density lipoprotein (LDL)-derived cholesterol from late endosomes to other cellular membranes in both Drosophila and vertebrates (Ikonen and Holtta-Vuori, 2004; Mukherjee and Maxfield, 2004). A functional Ptc SSD is essential for Smo repression (Martin et al., 2001), raising the possibility that lipid trafficking may be involved in this process. Smo repression also requires regions in the Ptc C-terminal tail, which promote Ptc internalization and turnover (Lu et al., 2006). Vertebrate Smo signaling can be artificially repressed or activated by the binding of a variety of small lipophilic compounds (Chen et al., 2002a; Chen et al., 2002b; Frank-Kamenetsky et al., 2002), supporting the possibility that related lipids may regulate Smo activity in vivo. However, whether these small molecules mimic the activity of endogenous cellular lipids is not clear. Vertebrate Smo activity can be increased by oxysterols, however these do not appear to bind to Smo directly (Corcoran and Scott, 2006; Dwyer et al., 2007). Furthermore, whether Ptc actually regulates lipid trafficking remains to be established.

Hh associates with the Drosophila lipoprotein particle Lipophorin (Lpp) (Panáková et al., 2005) and these particles are necessary for normal Hh signaling (Callejo et al., 2008; Panáková et al., 2005). Here, we show that specific lipids contained within Lpp are required to decrease Smo levels and Smo-dependent Ci₁₅₅ stabilization in the absence of Hh. Ptc recruits a small fraction of internalized Lpp to Ptc-positive endosomes and regulates both lipid trafficking and Smo trafficking from this compartment.

Transgenes: UAS<HcRed>dsLpp (Panáková et al., 2005), dpplacZ^BS3.0 (Blackman et al., 1991), UAS-Ptc (Martin et al., 2001), UAS-Ptc^S2GFP (Torroja et al., 2004), UASPtc^SSD (Martin et al., 2001), UAS-Ptc^Dloop2 (Briscoe et al., 2001), UAS-Ptc^1130x (Johnson et al., 2000), tubP-CFPRab5 (Marois et al., 2006), UAS<HcRed>Rab7TN (Marois et al., 2006), ap-GAL4 (Rincon-Limas et al., 1999), Adh-GAL4, Tub-GAL4 and tubP-GAL80^ts (Bloomington). Mutants: disp^S037707 (Burke et al., 1999), boca¹ (Culi and Mann, 2003), smo³ (Chen and Struhl, 1998), hh^ts2 (Strigini and Cohen, 1997), ptc^IIw (also known as Ptc¹⁶) (Strutt et al., 2001) and ptc^S2 (Martin et al., 2001).

Transgenic flies expressing double-stranded RNA (dsRNA) against LRP1 or Megalin RNAi were made using pFRiPE (Marois et al., 2006). We used a repeat unit corresponding to nucleotides 10671-11112 of CG33087-RC for LRP1 and nucleotides 11821-12342 of CG12139-RB for Megalin. UAS<HcRed>Rab9S21N was made by insertion of rab9S21N (Riederer et al., 1994) into pFRIPE (Marois et al., 2006).

LppRNAi was induced as described (Panáková et al., 2005) using Adh-GAL4 to drive fat body expression. Larvae containing Arrow, Megalin and LRP1 RNAi constructs under the control of apGAL4 were heat shocked at 37°C for 90 minutes to excise the intervening HcRed cassette, reared at 29°C and dissected after 48 hours.

boca¹ clones were generated as described (Culi and Mann, 2003). ptc^S2 and ptc^IIw clones were generated as described (Martin et al., 2001) For smo³ null clones, hsp70-flp; smo³, FRT40A/Ubi-GFPnls, FRT40A; +/TubGAL4 and hsp70-flp; smo³, FRT40A/Ubi-GFPnls, FRT40A; UAS<HcRed>dsLpp/ TubGAL4 larvae were heat shocked at 37°C for 90 minutes, reared at 25°C and dissected after 108 hours.

Ap-GAL4, tubP-GAL80^ts/UAS-Ptc animals were reared at 18°C and third instar larvae were transferred to 29°C for 15-18 hours. Ap-GAL4, tubP-GAL80^ts/UAS-Ptc; hh^ts2/hh^ts2 larvae were treated similarly.

Imaginal discs were fixed and stained as described previously (Marois et al., 2006). Antibodies were diluted as follows: anti-Ci 2A1 1:10 (Wang and Holmgren, 1999), anti-β-gal 1:100 (Promega Z378A), anti-Ptc 1:100 (DSHB, University of Iowa, Department of Biological Sciences, Iowa City, IA 52242, USA), anti-Smo 1:50 (Lum et al., 2003), anti-Megalin 1:1000, anti-LRP1 1:1000, anti-Hh 1:500 (Panakova et al., 2005) and anti-Lpp guinea pig 1:1000 (Eugster et al., 2007).

For anti-LRP1 we immunized rabbits with a glutathione-S transferase (GST) fusion to amino acids 4607-4699 of CG33087-PC (Eurogentec, Seraing, Belgium). For anti-Megalin, a peptide corresponding to amino acids 4533-4547 of CG12139-PB was conjugated to keyhole limpet hemocyanin (KLH) and used to immunize guinea pigs (Antibodies Inc., Davis, CA, USA). Rabbit anti-Ptc was generated against the second extracellular loop and used at 1:300.

Imaginal discs were dissected in Grace's insect medium and incubated for 90 minutes at 28°C with 20 nM of leptomycin B (Sigma) before fixation.

Membranes were incubated with anti-Ci 1C2 1:5 (Wang and Holmgren, 1999) and anti-actin 1:1000 (Sigma); followed by HRP-conjugated anti-rat IgM 1:2000 and anti-rabbit IgG 1:5000 (Dianova), respectively.

For confocal images, the position below the apical surface corresponds to 0.7-2.8 μm for the apical region, 1.4-2.1 μm for the sub-apical region and 2.1-4.8 μm for the basolateral (or middle) region.

All quantified immunostaining was performed on discs that were dissected, fixed, stained and imaged in parallel using the same microscope settings. To quantify Ci and Smo staining intensities, three apical sections, 0.7 μm apart, were projected using maximal intensity in ImageJ. For each image, two rectangles of 100 pixels parallel to the anteroposterior (AP) axis by 351 pixels parallel to the dorsoventral (DV) axis were selected and centered at the AP boundary in ventral and dorsal compartments. Average pixel intensity was determined as a function of distance from the AP boundary using PlotProfile and plotted using Microsoft Excel. All AP boundaries were determined according to anti-Ci or anti-Ptc co-immunostaining.

To estimate the significance of changes in staining intensities in discs of different genotypes, we measured Smo or Ci staining intensity at the same distance from the AP boundary in each disc and calculated P values using Excel. To quantify the colocalization percentage, we used the colocalization plugin of the Fiji image processing software.

Statistical significance was determined according to Costes et al. (Costes et al., 2004), where P value_Costes=1 — the fraction of randomized images giving a Pearson correlation coefficient greater than the original image.

Lpp particles were isolated from third instar larvae by first breaking them with a loose pestle and then pelleting tissues and larval carcasses at 1000 g for 10 minutes. Supernatants were spun at 33,600 rpm (142,000 g) for 3 hours in a SW40Ti rotor to remove debris and KBr was then added to a concentration of 0.33 g/ml and spun at 39,000 rpm (192,000 g) for 64 hours in a SW40Ti rotor. The top fraction was removed and desalted on a Sephadex column.

Desalted Lpp particles (see above) were labeled with an amine-reactive Alexa 546 probe (Molecular Probes), as described (Panáková et al., 2005). Ap-GAL4, tubP-GAL80^ts/UAS-Ptc larvae, reared at 29°C for 18 hours, were dissected in Grace's medium and incubated with labeled Lpp particles (at a protein concentration of 0.5 mg/ml in Grace's medium) for 10 minutes at 22°C. Ten minutes is the minimum time required to see uptake of labeled dextran by imaginal disc cells (Marois et al., 2006). The discs were then either washed immediately in PBS at 4°C and fixed (4% PFA for 20 minutes at room temperature), or washed and incubated for varying times in Grace's medium at 22°C before fixation.

Lipids were extracted from purified Lpp particles by a two-step Bligh and Dyer method (Bligh and Dyer, 1959). Total lipid concentration was measured according to Marsh and Weinstein (Marsh and Weinstein, 1966). Liposomes were prepared from dried lipids by sonication for 30 minutes into Grace's medium to produce a final total lipid concentration of 500 μg/ml. Assuming that hemolymph represents approximately half of the larval volume, this concentration should be similar to that of the lipids contributed by the Lpp fraction of the hemolymph. Wing imaginal discs from hsp70-flp/+; Adh-GAL4/+; UAS<HcRed>dsLpp/+ larvae were incubated in Grace's medium or in Grace's medium with liposomes for 2 hours at 22°C. Discs were fixed and stained as described previously (Marois et al., 2006).

Lpp particles are made in the fat body and travel through the hemolymph to imaginal discs where they bind to Hh and are required for long-range Hh trafficking and target gene activation (Panáková et al., 2005). As we investigated how loss of Lpp affected individual components of the Hh signal transduction machinery, we made the unexpected discovery that these particles also function to repress a subset of Hh signaling events when Hh is absent.

Ptc-mediated Smo repression correlates with the post-transcriptional destabilization of Smo protein and reduces Smo levels at the basolateral membrane (Denef et al., 2000). To investigate the consequences of reduced Lpp levels on Smo trafficking, we compared the accumulation of Smo protein in wild-type wing discs with LppRNAi wing discs. Although smo is transcribed uniformly throughout the wing pouch, Ptc activity reduces Smo protein levels on the basolateral membrane of most anterior compartment cells in wild-type discs. Smo accumulates to a high level only in the posterior compartment, which does not express Ptc, and in the anterior compartment near the AP boundary, where Hh inhibits Ptc activity (Fig. 1A,C) (Denef et al., 2000). Strikingly, LppRNAi strongly increases Smo protein levels on the basolateral membrane throughout the anterior compartment, but has little effect on Smo accumulation in posterior cells (Fig. 1B,C). This is surprising because LppRNAi does not expand the range of activation of any Hh target gene. Indeed, Lpp knockdown narrows the range of Hh target gene activation (Fig. 1Q,R) (Callejo et al., 2008; Panáková et al., 2005). Despite the reduced range of target gene activation in LppRNAi discs, Smo actually accumulates on the basolateral membrane as it would if Ptc activity were reduced. This disconnection between Smo trafficking and target gene activation resembles the situation in cyclopamine-treated vertebrate cells, where inactive Smo accumulates in primary cilia ligands (Aanstad et al., 2009; Rohatgi et al., 2009; Wang et al., 2009).

To investigate why target genes were not activated by the accumulation of Smo in LppRNAi discs, we first asked whether Smo was able to induce stabilization of the full-length form of the transcriptional activator Ci₁₅₅. In wild-type discs, antibodies that recognize only full-length Ci₁₅₅ show that it is stabilized by Smo signaling over a distance of 40 μm (12-15 cell diameters) from the posterior Hh-producing cells (Fig. 1D,F) (Aza-Blanc et al., 1997). When Lpp levels are reduced by RNAi, full-length Ci₁₅₅ staining is elevated, both near the AP boundary where Hh is present, and throughout the entire anterior compartment (Fig. 1E,F). Thus, increasing levels of basolateral Smo is enough to stabilize Ci₁₅₅.

To confirm that Lpp knockdown caused Ci₁₅₅ accumulation by preventing processing to Ci₇₅, we performed western blotting using imaginal disc extracts and an antibody that detects both Ci₁₅₅ and Ci₇₅. LppRNAi increases the ratio of full-length Ci₁₅₅ to that of the repressor Ci₇₅ (Fig. 1G). Thus, Lpp knockdown not only causes Smo to accumulate, but increases its capacity to stabilize Ci₁₅₅. These effects are not an indirect consequence of a failure to mobilize dietary lipids, because transfer to lipid-depleted medium does not mimic the effects of LppRNAi (see Fig. S1A-G in the supplementary material). Instead, the effects are caused by the direct action of Lpp particles on imaginal discs; addition of purified Lpp particles to explanted discs reversed both Smo accumulation and Ci₁₅₅ stabilization (see Fig. S1H-O in the supplementary material).

To confirm that ectopic Ci₁₅₅ accumulation required Smo, we induced smo mutant clones in LppRNAi animals and stained their wing imaginal discs with an antibody to Ci₁₅₅. In wild-type discs, Ci₁₅₅ is destabilized by loss of Smo only in cells near the AP boundary where Smo signaling is activated by Hh (Fig. 1H arrowhead, I). Low levels of Ci₁₅₅ are detected further away from the AP boundary, but do not require Smo since they are not reduced in smo clones (Fig. 1H arrow, I). In LppRNAi animals, the ectopically stabilized Ci₁₅₅ that is far from the AP boundary is also reduced by smo loss of function; Ci₁₅₅ is reduced in smo clones even when they lie on the anterior edge of the disc where Hh is not detectable (Fig. 1J arrowheads, K). Thus, elevated Ci₁₅₅ levels in LppRNAi discs depend on Smo.

To confirm that the increased level of Ci₁₅₅ in LppRNAi animals was Hh-independent, we asked whether LppRNAi could still stabilize Ci₁₅₅ in a dispatched (disp) mutant background. Disp is required for Hh secretion. Zygotic mutants survive to the third larval instar, but fail to secrete Hh as maternal Disp is depleted (see Fig. S2A in the supplementary material) (Burke et al., 1999). As a result, the range of Ci₁₅₅ stabilization is strongly reduced in these discs (Fig. 1L). When LppRNAi is induced in a disp mutant background, no Hh is observed in receiving tissue (see Fig. S2B in the supplementary material). Nevertheless, Ci₁₅₅ is stabilized just as efficiently as when Disp is present and Hh is secreted (Fig. 1M,N). Thus, loss of Lpp stabilizes Ci₁₅₅ independently of Hh.

In addition to stabilizing Ci₁₅₅, Smo signaling promotes the release of Ci₁₅₅ from microtubule-associated multiprotein complexes, allowing its nuclear translocation. To test whether Ci₁₅₅ that is ectopically stabilized by Lpp knockdown was able to enter the nucleus, we treated discs with leptomycin B to block nuclear export and stained them for Ci₁₅₅. LppRNAi discs accumulated ectopically stabilized Ci₁₅₅ in the nucleus within 1.5 hours after drug addition — a time frame that is consistent with that required for Ci₁₅₅ nuclear accumulation near the AP boundary in wild-type discs (Fig. 1O,P; see also Fig. S2C,D in the supplementary material). Thus, increased levels of basolateral Smo in LppRNAi discs cause Ci₁₅₅ stabilization and allow more Ci₁₅₅ to enter the nucleus, but are insufficient for target gene activation (Fig. 1Q,R) (Panáková et al., 2005). These data suggest that Ci₁₅₅ stabilization may be controlled by the amount of basolateral Smo, but that other signals are required for Smo to enable Ci₁₅₅ to act as a transcription factor. Lpp is required for normal Smo trafficking, but does not regulate Smo-dependent target gene activation.

To distinguish whether intact Lpp particles were required to reduce Smo and Ci₁₅₅ stability, or whether their lipid contents alone might suffice, we extracted lipids from purified Lpp particles, dried them and resuspended them into protein-free liposomes. The apolipophorin protein moiety is undetectable in these liposomes (see Fig. S2G in the supplementary material). We then incubated discs from LppRNAi animals with liposomes containing Lpp-derived lipids for 2 hours at a concentration that approximated that of the hemolymph. This treatment reverses the basolateral accumulation of Smo in LppRNAi discs to wild-type levels (Fig. 2A-E). Lpp lipids also reversed Ci₁₅₅ accumulation in a subset (n=27/54) of discs (see Fig. S2H-K in the supplementary material). Incubation with Lpp-derived lipids does not generally decrease levels of basolateral proteins since Arrow accumulation is unaffected in these discs (see Fig. S2L-O in the supplementary material) and similar treatment with medium alone had no effect on Smo (Fig. 2A,C); neither did liposomes containing two of the most abundant Lpp lipids, phosphatidylcholine and ergosterol (see Fig. S2P-S and Fig. S8A in the supplementary material). Thus, although Lpp lipids do not generally affect the trafficking of basolateral membrane proteins, they specifically reduce levels of basolateral Smo and consequently destabilize Ci₁₅₅.

How do Lpp lipids decrease Smo levels? One possibility is that they regulate the levels or activity of Ptc. If so, addition of these lipids to Ptc mutant cells would not be expected to reverse ectopic Smo accumulation. Alternatively, Ptc might mobilize lipids from Lpp particles, making them ‘available’ to destabilize Smo. In this case, the direct addition of lipids that have already been extracted from Lpp particles might reverse basolateral Smo accumulation in Ptc mutant cells. Lpp lipids did not reduce Smo accumulation in mutant clones that were totally missing the Ptc protein (Fig. 3A-D,I). However, interestingly, Lpp lipids did reduce Smo accumulation in tissue homozygous for ptc^S2, which harbors a point mutation in the SSD (Fig. 3E-H,J). This suggests that one function of the Ptc SSD is to make lipids within Lpp particles available to regulate Smo trafficking. However, it is clear that the Lpp lipids are insufficient to destabilize Smo when the Ptc protein is totally absent. Regulation of Smo trafficking by Ptc requires not only a functional SSD, but also an intact C-terminal tail (Lu et al., 2006). Thus, Lpp lipids may need to act in conjunction with the Ptc C terminus to destabilize Smo.

How might Ptc regulate the access of Lpp lipids to Smo? NPC1 promotes the exit of LDL-derived cholesterol and many other lipids from late endosomes (Ikonen and Holtta-Vuori, 2004; Mukherjee and Maxfield, 2004; Wojtanik and Liscum, 2003). Might Ptc act similarly? When overexpressed, Ptc causes Lpp accumulation and colocalizes with it in early endosomes (Fig. 4A-C; see also Fig. S3K-M in the supplementary material) (Callejo et al., 2008). Ptc overexpression does not obviously alter the size or number of either Rab5- or Rab7-positive endosomes, suggesting that it does not generally retard progression from early to late endosomes (see Fig. S3A-D in the supplementary material) (Rink et al., 2005). Furthermore, Lpp does not accumulate in cells overexpressing NPC1 (see Fig. S3E,F in the supplementary material). Thus, Ptc specifically recruits Lpp particles to Ptc-positive endosomes.

To investigate which regions of Ptc were required to cause Lpp accumulation, we overexpressed different mutant forms of Ptc in the dorsal compartment (for expression levels, see Fig. S3G in the supplementary material). We found that removal of the Ptc C-terminal tail, which is required for Smo repression, prevented Ptc-mediated Lpp accumulation (Fig. 4D,E). The C-terminal tail is also required for Ptc internalization and turnover (Lu et al., 2006). Another internalization-defective Ptc allele, Ptc¹⁴, also fails to cause Lpp accumulation when overexpressed (Callejo et al., 2008), suggesting that Ptc must be internalized to cause Lpp accumulation in endosomes. Although the SSD is required for Smo repression, it is not required for the accumulation of Lpp in endosomes with Ptc (Fig. 4F,G), consistent with results from Callejo et al. (Callejo et al., 2008). Thus, although Ptc SSD mutants can recruit Lpp to Ptc-positive endosomes, they cannot make use of their contents to destabilize Smo. Finally, a Ptc mutant missing sequences that are required for binding to Hh nevertheless efficiently recruits Lpp to Ptc endosomes (Fig. 4H,I). Consistent with this, expression of wild-type Ptc causes Lpp accumulation even in a hh^ts mutant background (see Fig. S3H-J in the supplementary material). Thus, although Hh is present on some Lpp particles, Ptc affects Lpp trafficking independently of Hh.

To distinguish whether the Lpp accumulation caused by Ptc resulted from increased Lpp uptake or decreased Lpp degradation, we performed uptake experiments with Alexa 546-labeled Lpp particles. Discs overexpressing Ptc in the dorsal compartment were incubated for 10 minutes at 22°C with labeled Lpp, then washed, fixed and stained. After the 10-minute incubation, labeled Lpp particles are present in endosomes throughout the wing disc and are equally abundant in the dorsal and ventral compartments (Fig. 4J-L). In the dorsal compartment, the vast majority of Lpp-Alexa 546 colocalizes with overexpressed Ptc in Rab5-positive endosomes (see Fig. S4A-F in the supplementary material). Thus, although Ptc and Lpp-Alexa 546 are incorporated rapidly into the same endosomes, Ptc overexpression does not increase the rate of Lpp-Alexa 546 uptake. By contrast, overexpression of the LDL receptor homologue LpR1-GFP does increase the rate of Lpp uptake, as expected (see Fig. S4G-J in the supplementary material), indicating that this assay is sensitive to such changes. After washing and incubation for a further 20-40 minutes in Lpp-free medium, Lpp particles begin to disappear from ventral compartment cells but are visible for up to 20 minutes in Ptc-overexpressing cells (Fig. 4M-O). By 40 minutes, Lpp is mostly degraded in both compartments (see Fig. S4K,L in the supplementary material). These data show that Ptc decreases the rate of Lpp degradation after internalization, but does not increase Lpp uptake. Thus, Ptc is unlikely to be an endocytic receptor for Lpp.

Since Ptc did not appear to promote Lpp uptake, we wondered whether any of the canonical lipoprotein receptors might be required for normal Hh signaling. We tested requirements for Megalin, the LDL receptor protein LRP1, Arrow, and the two LDL receptor homologues LpR1 and LpR2, using both mutagenesis and RNAi-mediated knockdown. None of these mutants altered Lpp accumulation in the steady state or produced Hh phenotypes similar to those caused by LppRNAi (see Figs S5-S7 in the supplementary material). These data indicate that multiple redundant mechanisms may contribute to lipoprotein uptake in the wing disc, and that none of these LDL receptor family proteins is individually required for the Lpp trafficking pathways that are relevant to Ptc-mediated Smo repression.

Although Ptc can elevate Lpp levels when overexpressed, the endogenous Ptc present in the anterior compartment is not sufficient to obviously increase Lpp accumulation there (note the Lpp levels in the wild-type ventral compartment in Fig. 4A,B). This suggests that Ptc influences the trafficking of only a small fraction of Lpp when expressed at endogenous levels. This is not surprising, considering its important nutritional functions and the large number of different receptors with the potential to contribute to Lpp trafficking in the wing disc (see Fig. S5A-D in the supplementary material). In order to detect the effect of Ptc on Lpp in the anterior compartment in wild-type discs, we exploited the fact that Lpp in these endosomes should be degraded more slowly than most of the Lpp internalized by disc cells. To specifically visualize Lpp that was degraded more slowly than average, we placed dissected discs in Grace's medium for 2 hours. In this time, most Lpp that was internalized in vivo should pass through the degradative pathway and disappear, revealing the tissue distribution of any Lpp subpopulation with a longer half-life. Immunostaining discs after a 2-hour incubation in Grace's medium reveals a stable population of Lpp that is found specifically in the anterior compartment, where Ptc is expressed (Fig. 4P-R). These data support the idea that, whereas most Lpp is internalized and degraded rapidly, Ptc redirects trafficking of a small subset of Lpp particles to a more stable endocytic compartment. Sequestration of Lpp by Ptc in these endosomes might give Ptc access to Lpp-derived lipids. We note that Lpp stability is increased uniformly throughout the anterior compartment and is not further increased near the AP boundary where Ptc is upregulated (but repressed by Hh). This may suggest that Hh interferes with Ptc-dependent Lpp trafficking.

The SSD of NPC1 is required for efflux of sterols, sphingolipids and other lipids from late endosomes. These include lipids derived from internalized lipoprotein particles (Ikonen and Holtta-Vuori, 2004; Mukherjee and Maxfield, 2004; Wojtanik and Liscum, 2003). Flies do not synthesize sterols (Clayton, 1964) and rely on Lpp for sterol delivery to discs (see Fig. S1A,C; Fig. S8A in the supplementary material). To test whether the Ptc SSD influenced the trafficking of Lpp lipids from Ptc-positive endosomes, we examined sterol distribution in cells overexpressing wild-type and mutant forms of Ptc (for expression levels, see Fig. S3G in the supplementary material). Although wild-type Ptc sequesters Lpp in endosomes, filipin staining shows that it does not perturb sterol distribution (Fig. 5A,B). By contrast, both Ptc^S2 and Ptc^SSD, which also sequester Lpp, cause sterol accumulation in Ptc positive endosomes (Fig. 5D-H). Thus, the SSD is required for the exit of sterols from this compartment. Ptc¹¹³⁰ overexpression does not cause endosomal sterol accumulation (Fig. 5C) — consistent with its inability to sequester Lpp in endosomes. These data suggest that Ptc sequesters Lpp particles in endosomes and mobilizes their lipid contents via its SSD. Since RND transporters have extremely broad substrate specificities (Piddock, 2006) and many non-sterol lipids accumulate in NPC1 mutant endosomes, it is possible that Ptc SSD activity also regulates the accumulation of other lipids.

Although it seems likely that the accumulated sterol in Ptc^SSD-positive endosomes is derived from sequestered Lpp, it is also possible that membrane sterol, previously delivered by Lpp, is also trapped by mutation of the Ptc SSD. To confirm that sterols that have been newly delivered by Lpp contribute to endosomal sterol accumulation, we fed larvae with BODIPY-cholesterol (Holtta-Vuori et al., 2008) coincident with induction of Ptc or Ptc^SSD overexpression. BODIPY-cholesterol accumulates in cells overexpressing Ptc^SSD but not wild-type Ptc (see Fig. S8B-G in the supplementary material). Thus, at least part of the accumulated sterol has been delivered to discs subsequent to the induction of Ptc expression.

Because endogenous Ptc normally sequesters only a small subset of internalized Lpp, it is unlikely to be required for endosomal sterol efflux in general — indeed, sterol does not accumulate abnormally in ptc^S2 mutant clones (not shown). However, these data indicate that endogenous Ptc probably influences the membrane lipid composition of the subset of endosomes in which it is present. How could the lipid composition of Ptc endosomes affect Smo? A clue emerged when we compared the distribution of Smo in cells overexpressing wild-type Ptc with cells overexpressing Ptc^SSD. Wild-type Ptc overexpression inactivates Smo signaling and reduces Smo levels on the basolateral membrane in both anterior and posterior compartments (Martin et al., 2001) (see Fig. S8I in the supplementary material). No Smo is detectable in apical endosomes (see Fig. S8H in the supplementary material). By striking contrast, Ptc^SSD expression causes a dramatic accumulation of Smo in apical endosomes together with Ptc^SSD and Lpp (Fig. 5I-P). In the anterior compartment, Ptc^SSD expression also elevates basolateral Smo levels (Fig. 5Q-T), consistent with activation of Smo signaling in these cells (see Fig. S8R,S in the supplementary material). This suggests that Ptc^SSD mutations specifically block Smo trafficking towards degradation — perhaps indirectly increasing Smo recycling (see Marois et al., 2006). To ask whether Ptc^SSD overexpression might generally perturb the trafficking of basolateral membrane proteins, we examined the localization of FasIII and Arrow (see Fig. S8J-O in the supplementary material). Unlike Smo, neither FasIII nor Arrow accumulates in endosomes with Ptc^SSD, nor are their levels on the basolateral membrane altered. Thus, mutation of the Ptc SSD specifically affects trafficking of both Lpp lipids and Smo without perturbing localization of other basolateral membrane proteins.

To ask whether mutation of the Ptc SSD caused endosomal Smo accumulation when expressed at endogenous levels, we examined Smo localization in ptc^S2 mutant cells. Similar, although less dramatic, colocalization of Ptc^S2, Smo and Lpp in punctate structures is also observed in ptc^S2 mutant tissue (Fig. 6A-I), consistent with the idea that the Ptc SSD is required for normal endosomal trafficking of Smo. Taken together, these observations indicate that Smo may normally traffic through Ptc endosomes, and that blocking Ptc SSD activity not only causes endosomal lipid accumulation but also alters the trafficking of Smo from this compartment.

A central feature of the Hh pathway is repression of Smo signaling by Ptc when Hh is absent. How this repression functions at a mechanistic level is not understood. Ptc represses Smo at substoichiometric levels (Taipale et al., 2002) and the two proteins do not appear to be enriched at similar subcellular locations in the steady state in vivo (Martin et al., 2001). Repression of Smo by Ptc correlates with changes in Smo subcellular localization and decreased stability (Denef et al., 2000).

Models for Ptc-mediated repression have been proposed based on its SSD and on its sequence similarity to RND family transporters. Membrane sterol levels can alter trafficking and/or the stability of other proteins with SSDs (Kuwabara and Labouesse, 2002), thus it has been proposed that lipids might control the repressive activity of Ptc. Alternatively, based on its similarity to RND transporters, it has been proposed that Ptc might regulate Smo by modulating lipid trafficking. This is an attractive idea because Smo activity can be regulated by small lipophilic molecules (Chen et al., 2002a; Chen et al., 2002b; Frank-Kamenetsky et al., 2002). Although these are plausible models, there has as yet been no evidence to suggest that Ptc alters lipid trafficking or that any lipid alters Ptc function. Furthermore, if Ptc did regulate lipid transport to repress Smo, it is not clear where contact between Smo and Ptc-mobilized lipids would occur.

We now show that one or more lipids derived from the Drosophila lipoprotein Lpp are required for normal Ptc-mediated Smo destabilization. Ptc sequesters a small fraction of internalized Lpp in the Ptc-positive endosomal compartment and regulates lipid trafficking from this compartment via its SSD. Mutation of the Ptc SSD causes at least one Lpp-derived lipid, sterol, to accumulate in Ptc endosomes. Given that RND permeases have rather broad substrate specificities (Piddock, 2006), Ptc SSD mutation may also perturb the trafficking of additional lipids from this compartment. Similar mutations in NPC1 alter endosomal trafficking of sphingolipids, as well as sterols (Ikonen and Holtta-Vuori, 2004; Mukherjee and Maxfield, 2004; Wojtanik and Liscum, 2003).

Although Ptc-dependent sterol trafficking is clearly dependent on its SSD, Ptc has been reported to reduce the accumulation of neutral lipid in a manner that does not depend on its SSD. Rather, this effect of Ptc appears to depend on its ability to sequester Lpp (Callejo et al., 2008). This may suggest that Ptc-mediated sequestration of Lpp diverts these particles away from trafficking pathways that promote neutral lipid storage.

Ptc is normally present in a small fraction of endosomes and, unlike NPC1, does not generally perturb cellular sterol trafficking when mutated. How could altered trafficking in the small subset of endosomes that contain Ptc affect Smo? We show that mutation of the Ptc SSD not only alters lipid trafficking in Ptc endosomes, but also causes Smo to accumulate in this compartment (as well as on the basolateral membrane). Similar endosomal colocalization between Ptc and Smo has been observed in vertebrate tissue culture cells (Incardona et al., 2002). This suggests that Smo normally traffics through Ptc endosomes, where it can be exposed to lipids that are mobilized by the Ptc SSD. Thus, Ptc may control the level of basolateral Smo by changing the balance of Smo degradation and recycling from Ptc endosomes. Lipid accumulation in endosomes of NPC1 mutant cells influences the activity of Rab7 (Lebrand et al., 2002), Rab9 (Ganley and Pfeffer, 2006) and Rab4 (Choudhury et al., 2004), perturbing degradation and recycling. The lipid composition of Ptc-positive endosomes may exert similar, but much more specific, effects on Smo, altering the total levels of Smo protein.

Which Lpp lipids does Ptc use to regulate Smo trafficking? Although sterols are present in Lpp and the Ptc SSD does regulate sterol trafficking from Ptc endosomes, our data do not support a role for bulk membrane sterol in Smo regulation. Membrane sterol can be reduced sixfold by dietary depletion without changing Smo levels on the basolateral membrane or other aspects of Hh signaling. Furthermore, liposome-mediated addition of ergosterol, the most abundant membrane sterol in Drosophila (Rietveld et al., 1999), does not reverse Smo accumulation caused by LppRNAi. However, our data do not rule out the possibility that signaling sterol derivatives that act at low concentrations might be responsible. We are now fractionating Lpp lipid extracts to identify this active molecule.

Loss of Lpp reproduces only a subset of the effects of Ptc mutation on Smo signaling — although it stabilizes Smo and causes Smo-dependent accumulation of full-length Ci₁₅₅, it does not allow target gene activation. Similar uncoupling of Ci₁₅₅ stability and target gene activation is seen in fused and dally mutants (Alves et al., 1998; Eugster et al., 2007; Wang and Holmgren, 1999). It seems unlikely that ‘activation’ of Ci₁₅₅ simply requires higher levels of Smo than Ci₁₅₅ stabilization; in LppRNAi discs, Smo levels reach those that result in target gene activation in wild-type discs over a broader region than in wild-type discs — nevertheless, the range of target gene activation is narrowed. Rather, these data suggest that, although Ci₁₅₅ stabilization is regulated by the level of Smo on the basolateral membrane, its activation as a transcription factor requires a separate Smo-dependent signal. Similarly, in vertebrate cells, where Ptc regulates Smo trafficking to the primary cilium, it is clear that ciliary localization is insufficient for Smo signaling to activate transcription of target genes (Aanstad et al., 2009; Rohatgi et al., 2009; Wang et al., 2009). Other G-protein-coupled receptors activate multiple signal transduction pathways in response to different ligands (Perez and Karnik, 2005). Thus, Smo trafficking and Smo activation may also be controlled by different ligands. Although Lpp lipids regulate Smo trafficking, other lipids mobilized by Ptc could have additional effects on Smo signaling activity.

The following model (Fig. 7) would be consistent with what is already known of Ptc-dependent Smo regulation and with our new observations. Ptc sequesters Lpp into Ptc-positive endosomes and regulates lipid trafficking in this compartment via its SSD. Smo that passes through Ptc endosomes can be targeted either for degradation or recycling, depending on Ptc-dependent modulation of endosomal lipid composition, and on other functions of the Ptc protein. Lipids derived from Lpp particles bias Smo trafficking towards degradation (Fig. 7A). The balance of degradation versus recycling affects total Smo levels on the basolateral membrane and its ability to stabilize Ci₁₅₅.

The association of Hh with Lpp may do more than simply promote the release of Hh from the membrane. Hh is thought to bind to the extracellular loops of Ptc (Marigo et al., 1996) — a region that is important for conferring substrate specificity in RND family transporters (Elkins and Nikaido, 2002; Mao et al., 2002). The presence of Hh on lipoproteins may block Ptc-mediated mobilization of their lipids (Fig. 7B). Alternatively, increased Ptc degradation upon Hh binding may prevent Lpp sequestration and lipid mobilization (Fig. 7C). In either case, association with Lpp particles efficiently positions Hh to interfere with, or alter, the mobilization of Lpp lipid contents by Ptc, helping to alleviate Ptc-mediated Smo repression. Thus, Hh may signal, in part, by influencing the utilization of the lipoprotein particles on which it is carried.

We gratefully acknowledge Christian Dahmann, Isabel Guerrero, Gary Struhl, Richard Mann and the Bloomington Stock Center for providing flies. We thank Robert Holmgren and the Developmental Studies Hybridoma Bank for providing antibodies. We thank Robert Bittman for providing BODIPY-cholesterol. We thank Kai Simons, Christoph Thiele, Christian Dahmann, and Elisabeth Knust and Marino Zerial for critical comments on the manuscript. We are grateful to Sven Ssykor for embryo injections. Ali Mahmoud provided essential and expert technical help throughout. Eric Marois contributed to the generation of flies expressing dsMegalin. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (EA4/2-1 and EA4/2-2) and by the Max Planck Gesellschaft.