SUMO regulates the activity of Smoothened and Costal-2 in Drosophila Hedgehog signaling

In Hedgehog (Hh) signaling, the GPCR-family protein Smoothened (Smo) acts as a signal transducer that is regulated by phosphorylation and ubiquitination, which ultimately change the cell surface accumulation of Smo. However, it is not clear whether Smo is regulated by other post-translational modifications, such as sumoylation. Here, we demonstrate that knockdown of the small ubiquitin-related modifier (SUMO) pathway components Ubc9 (a SUMO-conjugating enzyme E2), PIAS (a SUMO-protein ligase E3), and Smt3 (the SUMO isoform in Drosophila) by RNAi prevents Smo accumulation and alters Smo activity in the wing. We further show that Hh-induced-sumoylation stabilizes Smo, whereas desumoylation by Ulp1 destabilizes Smo in a phosphorylation independent manner. Mechanistically, we discover that excessive Krz, the Drosophila β-arrestin 2, inhibits Smo sumoylation and prevents Smo accumulation through Krz regulatory domain. Krz likely facilitates the interaction between Smo and Ulp1 because knockdown of Krz by RNAi attenuates Smo-Ulp1 interaction. Finally, we provide evidence that Cos2 is also sumoylated, which counteracts its inhibitory role on Smo accumulation in the wing. Taken together, we have uncovered a novel mechanism for Smo activation by sumoylation that is regulated by Hh and Smo interacting proteins.

It has long been studied that the Hedgehog (Hh) morphogen controls development processes such as proliferation, embryonic patterning, and cell growth 1,2 . It has also been shown that malfunction of Hh signaling, e.g. mutations in the Hh pathway components, causes many human disorders, including several types of cancers [3][4][5] . One good example is that abnormal activation of Smoothened (Smo), an atypical G protein-coupled receptor (GPCR), results in basal cell carcinoma (BCC) and medulloblastoma 1,2 , therefore Smo has been an attractive therapeutic target, exemplified by the newly FDA approved drugs 6 .
Most of what is known about the Hh signaling cascade comes from studies of Drosophila, where the pathway was originally identified 7 . Hh receiving system consists of Patched (Ptc) and Smo at the plasma membrane.
Smo acts as a signal transducer whose activity is inhibited by Ptc in the absence of Hh. How Ptc inhibits Smo is not completely understood, although recent studies indicate that phospholipids act in between Ptc and Smo in Drosophila Hh signaling [8][9][10] . Binding of Hh to Ptc alleviates Ptc-mediated inhibition of Smo, allowing Smo to activate Cubitus interuptus (Ci)/Gli transcription factors and ultimately induce the expression of Hh target genes, such as decapentaplegic (dpp), patched (ptc), and engrailed (en) 11,12 . The regulation of Smo is obviously a key event in Hh signal transduction.
Among the types of protein-based modifications, phosphorylation and ubiquitination of Smo have been extensively studied. In the absence of Hh, cytosolic Smo is highly unstable because of rapid degradation through both the proteasome-and lysosome-mediated pathways, which involve ubiquitination [13][14][15] . In a dose-dependent manner, thresholds of Hh promote Smo differential phosphorylation by multiple kinases including PKA, CK1 isoforms, aPKC, CK2, and G protein-coupled receptor kinase 2 (Gprk2) [16][17][18][19] , which induces the dimerization and cell surface accumulation of Smo 1,12,20 . The stimulation of Hh promotes Smo deubiquitination by ubiquitin-specific protease 8 (USP8), which blocks Smo endocytosis and enhances Smo cell surface accumulation 14,15 . Although Smo behaves differently from a typical GPCR, it has been shown that Krz, the Drosophila non-visual arrestin 21 , downregulates Smo signaling by promoting Smo internalization and degradation in ubiquitin-and Gprk2-independent manners 15,22 . It is possible that Krz downregulates Smo activation through a mechanism in parallel with phosphorylation and ubiquitination.
The small ubiquitin-like modifier (SUMO) is post-translationally conjugated to lysine residues of nuclear proteins as well as cytosolic and plasma membrane proteins, resulting in changes in their transcriptional activity or intracellular trafficking 23,24 . Sumoylation is promoted by the SUMO-activating enzyme E1, SUMO-conjugating enzyme E2, and SUMO ligase E3, and the SUMO ligase E3 is responsible to recognize the substrate 25 . As a reversible process, SUMO-protein cleavage, or desumoylation, is carried out by SUMO protease, which is also highly regulated in cellular mechanisms such as nuclear transcription factor regulation and intracellular protein trafficking 26 . Interestingly, sumoylation was recently reported to regulate proteins involved in G-protein signaling 27 , however, it is unknown whether the activity of GPCR itself is regulated by SUMO.
To explore the possibility that SUMO regulates Hh signaling proteins, we performed a small scale genetic screen with RNA interference (RNAi) lines. This screen allowed us to determine whether inactivation of the SUMO pathway regulated Hh signaling activity in vivo. Further, we were able to examine sumoylation levels of individual Hh signaling component in S2 cells using immunoprecipitation assays. We found that Smo, Costal-2 (Cos2), Fused (Fu), and Ci are all SUMO modified. In this report, we focused on the sumoylation of Smo 29 . We found that RNAi of Smt3 also modified the Smo DN phenotype and resulted in smaller wings (Fig. 1H), suggesting that SUMO may possibly regulate Smo activity in the wing. Protein sumoylation is a dynamic process that often involves a desumoylase. In Drosophila, Ulp1 is one of the desumoylases with SUMO-specific protease activity. We found that RNAi of Ulp1 attenuated the activity of Smo DN , resulting in a partial rescue of Vein 3 and Vein 4 fusion phenotype (

Inactivation of Sumoylation inhibits Smo accumulation by decreasing Smo stability.
To further examine the roles of sumoylation in regulating Hh signaling, we carried out a ptc-luciferase (ptc-luc) reporter assay in S2 cells to monitor Hh pathway activity when using RNAi to inactivate the SUMO pathway. We found that double-stranded RNA (dsRNA) targeting Ubc9, Smt3, or PIAS significantly reduced ptc-luc activity induced by the treatment with Hh in cultured S2 cells ( Fig. 2A). RNAi of GFP did not change ptc-luc activity thus served as a control. dsRNA treatment consistently had high efficiency to knock down gene expression (Fig. 2B).
We wondered whether the inactivation of SUMO pathway could regulate endogenous Smo in the wing, given the fact that RNAi of sumoylation protein expression changed Smo DN activity (Fig. 1). We found that knockdown of Ubc9 by RNAi severely reduced Smo accumulation and attenuated dpp-lacZ expression in the wing imaginal disc, an early stage of wing development (Fig. 2D, compared to WT immunostaining shown in Fig. 2C). Similarly, RNAi of PIAS or Smt3 decreased Smo accumulation (Fig. 2E,F). We further found that the expression of a UAS-Ulp1 transgene inhibited Smo accumulation in wing disc (Fig. 2G), indicating that Ulp1 played a negative role in regulating Smo, which was consistent with the finding that RNAi of Ulp1 expression reduced the dominant negative activity of Smo DN in the wing (Fig. 1I). The severe or mild changes in Ci accumulation and ptc-lacZ expression ( Fig. 2E-G, red and green panels) may not solely reflect the changes in Smo activity because it has been shown that Ci undergoes sumoylation regulation 30,31 , and because Cos2 also undergoes sumoylation that regulates its activity (see below).
In addition to examining Smo accumulation in wing discs, we further carried out a protein stability assay to determine the levels of Smo regulated by sumoylation pathway proteins. We found that inactivation of Ubc9 RNAi, or Velo RNAi by C765-Gal4. Of note, the C765-Gal4 is weaker than the MS1096-Gal4 therefore mild phenotypes were observed when using C765-Gal4.  lower panels). This was further demonstrated by a quantification analysis (Fig. 3D). These data support the idea that sumoylation enhances the stability of Smo and promotes the accumulation of Smo in wing disc.
To determine whether Smo indeed undergoes sumoylation, we performed a sumoylation assay by transfecting S2 cells with the epitope-tagged Smo and SUMO, followed by immunoprecipitation with one epitope tag and western blot with another tag. We found that the overexpression of Ubc9 elevated the levels of Smo sumoylation, which was further increased by Hh treatment (Fig. 3E). This finding suggests that the sumoylation of To explore the mechanisms of Smo regulation by sumoylation, we examined the physical interaction between Smo and its desumoylase Ulp1. We found that Myc-Smo WT was associated with HA-Ulp1 in the immunoprecipitation assay (Fig. 4C). In addition, the treatment of Hh severely reduced Smo-Ulp1 interaction (Fig. 4C), indicating that Hh promotes Smo sumoylation by disassociating the desumoylase. We further explored the possibility of Ulp1 interaction with other forms of Smo. We found that Ulp1 physically interacted with the wild-type, phosphomimetic, and phosphorylation-deficient forms of Smo, and such interaction was attenuated by Hh stimulation (Fig. 4D) 15,22 . In this study, we carried out experiments to determine how Krz blocks the accumulation and prevents the activation of Smo. Domain functions of β -arrestin have not been well characterized, although β -arrestin interacts with many protein partners and exhibits conformational changes during cell signaling 33 . The C-terminal regulatory domain consisting of 44 amino acids is highly conserved among β -arrestins 34,35 . We found that expression of Krz ΔR lacking the regulatory domain did not prevent Smo accumulation in wing discs (Fig. 5B, compared to Fig. 5A). We also found that Krz, but not Krz ΔR , interacted with Smo in an immunoprecipitation assay (Fig. 5C) (Fig. 5E). We then examined whether Krz was associated with Ulp1 in cultured S2 cells. Interestingly, both Krz and Krz ΔR physically interacted with Ulp1 in the immunoprecipitation assay (Fig. 5F). Considering Krz ΔR not interacting with Smo (Fig. 5C) (Fig. 5D) and blocks Smo accumulation in wing disc (Fig. 5A), we wondered whether it was possible for Krz to regulate Smo phosphorylation. Using an in vitro kinase assay, we found that Krz and Krz ΔR did not inhibit Smo phosphorylation detected by a phospho-Smo antibody (Fig. 5G). Consistently, Krz and Krz ΔR did not inhibit Smo phosphorylation in cultured S2 cells (Fig. 5H). These results indicate that Krz inhibits Smo activation by specifically preventing Smo sumoylation, but not phosphorylation.

The inhibitory role of Cos2 on Smo is attenuated by sumoylation. When we examined Smo sumoy-
lation in S2 cells using the immunoprecipitation assay, we also discovered that other components in Hh signaling cascade were sumoylated. As shown in Fig. 6A, the sumoylated HA-tagged Cos2 was detected by an anti-Flag antibody in S2 cells cotransfected with HA-tagged Cos2 and Flag-tagged SUMO (Fig. 6A, lane 2, top panel). Similar to the pattern of Smo sumoylation, the Flag signal exhibited lower mobility shifts compared to the major band detected by the anti-HA antibody (Fig. 6A, lane 2, middle panel), indicating that these bands correspond to sumoylated forms of Cos2. Overexpression of Ubc9 caused an increase in the levels of Cos2 sumoylation, especially the stronger signal in the band with lowest mobility shift (Fig. 6A, lane 3, top panel). To further characterize Cos2 sumoylation, we transfected S2 cells with Myc-tagged Cos2 and treated cells with dsRNA targeting Ubc9, PIAS, or Ulp1. The ladder pattern of Flag signals was severely decreased by either Ubc9 or PIAS RNAi (Fig. 6B, lane 2 and 3, top panel), however increased by Ulp1 RNAi (Fig. 6B, lane 4, top panel). These data suggest that Cos2 undergoes sumoylation that is regulated by the same sets of proteins involved in Smo sumoylation. We further narrowed down the sumoylated region in Cos2. As shown in Fig. 6C,D, the N-terminus of Cos2 containing the microtubule-binding domain and the neck domain barely exhibited any sumoylation, whereas Cos2 C-terminus containing the coiled-coil and C-tail domains was sumoylated, indicated by the ladder pattern of sumoylation in the western blot with the immunoprecipitated Cos2.
We have previously shown that Cos2 exhibits an inhibitory role on Smo phosphorylation and accumulation 36 . We therefore questioned the possibility for Cos2 to regulate Smo sumoylation. We thus transfected S2 cells with Myc-Smo WT and HA-Cos2 WT and carried out immunoprecipitation assay to examine Smo sumoylation. We found that Cos2 blocked Smo sumoylation induced by the expression of Ubc9 (Fig. 6E), suggesting that Cos2 regulates the phosphorylation and sumoylation of Smo, adding an additional layer feedback regulation of Smo by Cos2.
To determine whether Cos2 sumoylation regulates its inhibitory activity on Smo accumulation, we turn to the Drosophila imaginal disc to examine the levels of Smo accumulation. Consistent with our previous findings 36 , the expression of Cos WT blocked Smo accumulation in posterior compartment cells (Fig. 7A). Interestingly, coexpression of Cos2 WT with Smt3 RNAi partially rescued Smo accumulation, although the accumulated Smo exhibited a punctate staining pattern (Fig. 7B). Changing the Ser182 of Cos2 to Asn (S182N) in the P-loop gave rise to a dead or a dominant-negative form 37 . Interestingly, the expression of Cos2 S182N also blocked Smo accumulation in the wing disc (Fig. 7C)     Cos2 S182N with Smt3 RNAi also partially rescued Smo puncta accumulation (Fig. 7D). These data suggest that Cos2 sumoylation may antagonize its inhibitory role to block Smo phosphorylation and sumoylation.

Discussion
The To identify the sumoylation residue(s) in Cos2, we carried out immunoprecipitation assays using a series of Cos2 truncations as previously described 36 . We found that the microtubule binding domain and the neck domain of Cos2 did not undergo any sumoylation. In contrast, the coiled-coil domain and C-tail were sumoylated (Fig. 6D). However, mutating four lysine residues (Lys715, Lys892, Lys922, Lys979) in the coiled-coil domain did not affect Cos2 sumoylation. It has been shown that Cos2 C-tail is responsible for Cos2 to inhibit Smo 36 , however Cos2 still underwent the same levels of sumoylation when Lys1083 in the C-tail was mutated. Furthermore, mutating these lysine residues did not change the ability of Cos2 to inhibit Smo sumoylation. We speculate that Cos2 is sumoylated at other lysine residues.
In this study, we found that both the wild-type Cos2 WT and dead form Cos2 S182N inhibited Smo accumulation in wing disc (Fig. 7A,C), suggesting that Cos2 inhibits Smo accumulation in a Cos2-activity-independent manner. In addition, the inactivation of Smt3, the single form of SUMO in Drosophila, counteracted the inhibitory activity of Cos2 WT and Cos2 S182N in regulating Smo accumulation in the wing (Fig. 7B,D), indicating the inhibitory activity of Cos2 is regulated by the SUMO pathway. Interestingly we found that Smo was partially recovered as many tiny puncta in the wing (Fig. 7B,D). This punctate pattern suggests that Smo is located in the intracellular compartments, and these forms of Smo are likely in their inactive states. It is possible that these forms of Smo are not sumoylated because Smt3 is inactivated by RNAi. The downregulation of Smo sumoylation by RNAi of Smt3 likely prevent Smo accumulation and activation in the wing disc, although the inhibitory activity of Cos2 is compromised. Future studies could be designed to examine the co-localization of Smo with intracellular compartments labeled by different markers, which will provide a better understanding Smo intracellular trafficking. Multiple Hh signaling components are sumoylated, resulting in difficulty analyzing the phenotypes in adult wings and wing discs. For example, Ci is sumoylated, which promotes the activity of Ci in regulating cyst stem cell (CySC) proliferation 30 (Fig. 7E). However, we are able to distinguish the roles of Smo and Cos2 sumoylation using both the wing disc phenotypes and cell cultured assays to examine the levels of protein sumoylation.
The other very critical finding in this study is that  15 . However, our data presented in this study and those from the previous study 15 (Fig. 5E). Although the RNAi of Krz was very efficient, there was still weak interaction between Smo and Ulp1 (Fig. 5E), raising the possibility that other arrestins might have similar roles. Cell culture, transfection, immunoprecipitation, and western blot. S2 cell culture and transfection using Effectene transfection reagent (Qiagen) has been previously described 16 . Forty-eight hours post-transfection, cells were harvested and treated with lysis buffer [100 mM NaCl, 50 mM Tris-HCl (pH8.0), 1.5 mM EDTA, 10% (vol/vol) glycerol, 1% (vol/vol) Nonidet P-40, and protease inhibitor tablet (Roche)]. Cell lysate was obtained by centrifugation at 12,000 rpm for 10 min. A total of 6 × 10 6 cells were harvested and lysed in 450 μ L lysate buffer. 50 μ L was saved for direct western blots, out of which 4 μ L was used for each load. The remaining 400 μ L was used for IP assay. The cell lysate was added with beads of Protein A Ultralink Resin (Thermo Scientific) after adding the proper primary antibody for 2 h. The samples were then resolved by SDS-PAGE and transferred onto PVDF membranes (Millipore) for western blot. About 16 times more of the immunoprecipitation sample was analyzed compared with the corresponding lysate. Western blot analysis was performed using the indicated antibodies and the enhanced chemiluminescence (ECL) protocol. To normalize the levels of Smo, 50 μ M MG132 (a proteasome inhibitor, Calbiochem) and 15 mM NH 4 Cl (a lysosome inhibitor, Sigma-Aldrich) were used to block Smo degradation, and samples were normalized for loading 14,16 . For Smo stability assay, Cycloheximide (Sigma) treatment was performed at a final concentration of 100 μ M for the indicated time points before harvesting S2 cells 36 . Density of the western blot was analyzed by ImageJ software. Hh treatment achieved by transfection with HhN cDNA combined with treatment with 60% of HhN-conditioned medium to achieve high level of Hh signaling activity, and RNAi achieved by adding dsRNA to cell culture in 6-well plates have been previously described 16,39 . GFP dsRNA was used as previous described 14 . Ulp1 dsRNA was synthesized against coding sequence 171-720, Velo dsRNA was against 2100-2640, CG12717 dsRNA was against 1-500, PIAS dsRNA was against 131-680, Krz dsRNA against 192-798, Smt3 and Ubc9 dsRNAs were against full coding sequence plus 3′ -UTR. Antibodies used for western blotting: mouse anti-Myc (9E10, Santa Cruz, 1:5,000), anti-HA (F7, Santa Cruz, 1:5,000), anti-Flag (M2, Sigma, 1:10,000), and anti-GFP (Millipore, 1:1,000); rabbit anti-Krz (Thermo Scientific, 1:2,000). The consistency of western blots was confirmed by three to five individual repeats. To examine the levels of gene expression, RT-PCR was carried out using S2 cells with the primers for Ubc9 (5′ -TGG CGC AAG GAT CAC-3′ ; 5′ -GCC CGC CCT CCC AGG-3′ ), PIAS (5′ -CAG CTG CCT AAT GTC ATT C-3′ ; 5′ -GAC ACC ACT GAA CCG-3′ ), Smt3 (5′ -AGA AGG GAG GTG AGA C-3′ ; 5′ -CGT TCA TCA GCT TCC TC-3′ ), and Ulp1 (5′ -CGG GAT TCC AGG CTC-3′ ; 5′ -GTC CAC ACG CCG GTA C-3′ ).

In vitro
The ptc-luc reporter assay has been described with S2 cells cultured in 6-well plates and transfected with 50 ng tub-Ci and 150 ng ptc-luc reporter constructs followed by activity analysis using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA) combined with the GLOMAX Multi Detection System (Promega) 16 . Each ptc-luc experiment was repeated three times and the error bars indicated standard deviation (S.D.) from four repeats.
Immunostaining of wing imaginal discs. Wing discs from third instar larvae with specific genotypes were dissected in PBS then fixed with 4% (vol/vol) formaldehyde in PBS for 20 min. After permeabilization with PBST [PBS supplemented with 1% (vol/vol) Triton X-100], discs were incubated with the indicated primary antibodies for 3 h and the corresponding second antibodies for 1 h, and then washed with PBT for three times, for 20 min per wash. Primary antibodies used in this study: mouse anti-SmoN (DSHB, 1:10); rabbit anti-β -Gal (Cappel, 1:1,500), anti-HA (Santa Cruz, Y-11, 1:100); rat anti-Ci (Developmental Studies Hybridoma Bank, 1:10). Secondary antibodies were from Jackson ImmunoResearch Laboratories Inc., affinity-purified for multiple labeling (1:500). Samples were mounted on slides in 80% glycerol and Fluorescence signals were acquired with the 20 x objective on an Olympus confocal microscope. The images shown represent five or more images collected for each experiment.