To gain a more comprehensive view of the Hh pathway at early stages of drug development, we developed and validated a novel High Content Screening (HCS) method based directly on Smo translocation to the PC(Wang, et al., 2012). Herein we report our findings while using the method to identify agonists of Smo ciliary accumulation. An EGFP tagged form of human Smo was introduced into Hh responsive NIH3T3 cells(Wang, et al., 2009) (Fig. S1 A) to generate a clonal cell line in which Hh-dependent accumulation of Smo∷EGFP in the PC mirrored movement of endogenous Smo(Wang, et al., 2009). An Inversin(Ivs)∷tagRFPT expression cassette provided a constitutive, independent PC marker (Watanabe, et al., 2003).

Custom algorithms were developed to perform quantitative multi-parametric image analyses(Wang, et al., 2012). Robust dose dependent responses were observed upon treatment with several known small molecule modulators of Smo: the agonist SAG and the antagonist cyclopamine (Cyc), both of which directly bind Smo, and forskolin (FKL), whose stimulatory action on protein kinase A inhibits Smo signaling (Fig. S1 A–F). Despite the fact that Cyc and SAG physically interact with Smo in a competitive fashion suggesting a common binding mechanism, and that both induce ciliary accumulation, Cyc bound Smo is inactive. Thus, accumulation within the primary cilium appears to be essential but not sufficient for downstream activation of the Hh pathway. In contrast, FKL likely induce Smo ciliary accumulation indirectly potentially by accelerating anterograde intraflagellar transport (Besschetnova, et al., 2010). A better understanding awaits a clearer picture of the cellular trafficking processes. As a demonstration of the assay’s ability to detect local changes within the PC, elongation of the PC on FKL treatment was detected as an expanded Ivs+ domain (last panel in Fig. S1 F), consistent with a recent report(Besschetnova, et al., 2010).

We conducted a screen with a library consisting of 5,672 compounds with annotated activities, including FDA approved drugs and drug candidates in preclinical or clinical development. Representative examples of plates including small-molecule control wells are shown for the analysis (Fig. S1 G). Z-prime scores(Zhang, et al., 1999) consistently >0.4 indicate a reasonable reliability of the primary screen.

Approximately 60 compounds in 15 distinct chemical classes were confirmed to induce Smo accumulation at the PC, after rigorous assessment of the dose-response curves for primary hits. As expected, these comprised both pathway agonists and antagonists. For example, LY 294002, an inhibitor of phosphatidylinositol 3-kinase (PI3K) (Vlahos, et al., 1994), induces Smo ciliary accumulation, but inhibits Hh signaling (Fig. S1 H–K). The PI3K pathway is important in a variety of cancer types and may intersect with the Hh pathway in tumorigenesis (Hambardzumyan, et al., 2008). In combination treatment, a PI3K inhibitor and a Smo antagonist delayed the onset of drug resistance in a mouse model of medulloblastoma (Buonamici, et al., 2010; Dijkgraaf, et al., 2011). PI3K action has also been linked to the regulation of Gli proteins through the Akt pathway (Riobo, et al., 2006). These data suggest that PI3K may act at multiple levels in Hh signaling.

Strikingly, the most predominant chemical class identified comprised naturally occurring and synthetic glucocorticoids (GC), several of which are widely used as anti-inflammatory agents in the clinic (Fig. 1 A–B; Fig. S1 L) (Sommer and Ray, 2008). Interestingly, a recent screen examining β-arrestin aggregation identified an overlap with a subset of these compounds, lending additional support to a GC intersection in Smo-directed Hedgehog signaling, but also raising the possibility of alternative mechanisms (Wang, et al., 2010). Structure-activity relationship (SAR) analysis suggests that fluorine at position 9, a ketal at positions 16 and 17, and protonation at position 11 significantly enhance the potency of this class of compounds in directing Smo accumulation to the PC (Fig. 1 C).

To investigate in more detail the consequences of GC-induced Smo accumulation in the PC, and to obtain mechanistic insights into GC action in the Hh pathway, we first chose one compound in clinical use, fluocinolone acetonide (FA). FA displays an EC50 of around 5 µM for accumulation of Smo in the PC; in addition, no obvious cytotoxic effects are observed in vitro at much higher doses (up to 200 µM; Fig. 1 B and D, Fig. S1 M). Localization of an inversin-based PC reporter and other PC markers including Arl13b, acetylated tubulin, and detyrosinated α-tubulin (glu-tub) were unaltered in response to FA (Fig. 1D; Fig. S1 M and N)(Caspary, et al., 2007; Schroder, et al., 2007; Watanabe, et al., 2003). Further, no change was detected in the activity of a Wnt-signaling reporter in response to FA concentrations that modify Smo distribution (Fig. S1 O). Together these data suggest that FA’s effects in this assay are specific to the Hh pathway.

The accumulation of Smo in the PC is thought to be essential for transcriptional activation of the Hh pathway (Kim, et al., 2009; Rohatgi, et al., 2007). However, we observed a marked disparity between FA-induced Smo accumulation in the PC and Hh pathway activation in transcription reporter assays. At low levels of FA that effectively promote Smo accumulation in the PC (10µM), no pathway activation was observed. Higher concentrations (>50 µM) invoked a weak transcriptional response measurable in a Gli-luciferase reporter assay (4 fold versus 25 fold for 1µM of the Smo agonist SAG in the same assay (data not shown)), and on quantitative reverse transcription–polymerase chain reaction (qRT-PCR) measurement of Hedgehog target gene expression (Ptch1 and Gli1; Fig. 2A–B). The EC50 for weak transcriptional activation (>50µM) was 10 fold higher than that of FA-induced accumulation of Smo within the PC.

The effects of FA resemble over-expression of Smo in that constitutive accumulation of wildtype Smo within the PC only results in weak pathway activation (Fig. S2 A). Ciliary accumulation of Smo sensitizes cells to subsequent Sonic hedgehog (Shh) ligand input, raising the possibility that FA-driven Smo accumulation may sensitize Hh responsive cells. Indeed, costimulation of cells with 10µM FA results in a dose-dependent enhancement of a Shh-induced transcriptional response (Fig. 2 C–D). Furthermore, this effect was measurable after prolonged withdrawal of FA; cells treated for 24 hours with FA followed by compound withdrawal prior to Shh addition showed a higher induction of pathway activity than DMSO treated controls (Fig. 2 E–F). The EC50 of a FA induced response to priming is approximately 4µM, in good agreement with the dose required for efficient accumulation of Smo in the PC (Fig. 1 B). Smo turnover in the PC is relatively slow after Shh-invoked pathway activation(Wang, et al., 2009), or compound withdrawal (Fig. S2 B–C), providing a potential explanation for a FA induced pathway priming effect. FA treatment showed no effect on Wnt pathway activity (Fig. S1 O), consistent with Hh pathway specificity.

To determine whether FA interacts with Smo, we performed a competition assay with Bodipy-Cyc. Cyc binds Smo directly (Chen, et al., 2002) and its fluorescent analog, Bodipy-Cyc, shows strong Smo-dependent fluorescence within cells over-producing Smo (identified by co-expression of a nuclear localized tagRFP-T, Fig. 2 G–H). An oncogenic mutation within the 7^th transmembrane domain (SmoM2, also named SmoA1, Fig. 2 H–I) (Chen, et al., 2002), and a recently described drug resistance mutation within the 6^th transmembrane domain (SMOD473H) significantly impair Cyc binding to Smo, suggesting that these are critical sites for chemical interaction (Yauch, et al., 2009).

FA displayed a dose-dependent competition of Bodipy-Cyc binding to wild-type Smo, similar to other small molecules that directly bind Smo (SAG, and GDC0449), or that likely interact directly with Smo based on similar competition assays (SANT-1) (Fig.2 H–I) (Chen, et al., 2002; Chen, et al., 2002; Frank-Kamenetsky, et al., 2002; Yauch, et al., 2009). In contrast, FKL induces Smo accumulation in the PC but does not compete with Bodipy-Cyc, reflecting an indirect action through its protein kinase A target (Milenkovic, et al., 2009; Wilson, et al., 2009). Weak pathway activation induced by FA was attenuated by Smo antagonists (Fig. 2 A) and depended on endogenous Smo as activation was not observed in fibroblasts lacking Smo activity (Fig.S2D). SANT-1 and GDC0449 inhibit FA promoted accumulation of Smo in the PC (Fig. S2 E – F). Collectively, these data support a direct interaction between FA and Smo.

Considering that GCs and various Hh pathway antagonists may share a common Smo target, and GCs are widely used to suppress inflammation in conjunction with cancer therapy, we next asked whether we could observe a potential GC crosstalk with Smo antagonists in cell culture assays. Hh pathway inhibition by GDC0449, Cyc and SANT-1, as measured by both Gliluciferase induction (Fig. 3A and Fig. S3) and Smo ciliary localization (Fig.3 B–C and Fig. S3), was dramatically reduced in vitro in the presence of FA. Thus, FA co-treatment leads to a drug-dependent alteration of cellular response to chemical inhibitors of Smo. This may occur through competition, or the requirement for a higher level of GDC-0449 to inhibit Hh-driven pathway activity in the presence of GC, but the outcome resembles the genetic resistance seen with a dominant active Smo mutation (SmoM2) (Fig. 3A).

We next assessed whether the observations for FA were replicated by a second clinically approved GC, Triamcinolone Acetonide (TA). TA was slightly more potent than FA in Smo ciliary translocation assay (Fig.1B). Similar to FA, TA only evoked a Gli-mediated transcriptional response at much higher doses than those that induced Smo ciliary accumulation, although the Hh pathway was activated to higher levels than measured on FA treatment (Fig.S4A). No activation was observed in Smo−/− embryonic fibroblast cells as expected (Fig.S2D). Further, at 10µM TA enhanced the response to Hh ligand (Fig.S4B), a dose that does not sufficient to induce ligand-independent pathway activity (Fig.S4A). TA also displayed a dose dependent competition with Bodipy-Cyc for binding to Smo (Fig.S4C–D). More importantly, 10µM TA induced a dose-response shift for GDC0449 mediated inhibition of Hh pathway activity, and Smo ciliary accumulation induced by ligand treatment (Fig. S4E–G). Taken together, our results indicate that these, and possibly other GCs that alter Smo localization share broadly similar biological properties but further work will be required to examine the extensive set of compounds identified in our study.

To further explore FA actions, we isolated cerebellar granule neuron precursors (CGNPs) from Ptch1+/− neonates. Proliferation of CGNP is Shh-dependent and Ptch1 heterozygosity predisposes both mice and humans to develop CGNP-derived medulloblastoma (Schuller, et al., 2008). Consistent with results on Hh pathway activation in NIH3T3 cells, only very high doses of FA (120µM) elevated the number of proliferative, phospho-histone H3 (pH3) positive GCNPs (Fig. 4 A–B). However, a lower dose of FA (10µM) markedly enhanced Shh-driven CGNP proliferation (Fig. 4 C–D). Further, co-administration of FA(10µM), with the Smo antagonist GDC0449, impaired GDC0449 inhibition of Shh-stimulated GCNP proliferation (Fig. 4 E–F).

While a large number of GCs promote Smo ciliary accumulation, secondary assays of small molecules sharing the core GC scaffold identified two inhibitory GCs: Budesonide (Bud) and Ciclesonide (Cic) (Fig. 5 A–C; Fig. S5 A – C). When compared with Smo promoting GCs, Bud and Cic are distinguished by bulky hydrophobic groups at positions 16 and 17 (Fig. 1; Fig. S1 L; Fig.5A; Fig. S5 A). In contrast to FA and TA, Bud had no pathway inducing activity, nor did Bud induce a hypersensitive response to Hh ligand (Fig. 2 A–F; Fig. 5 D), reinforcing the association of hyper-responsiveness to Smo ciliary accumulation activity. As expected from the inhibition of Smo accumulation in the PC, Bud and Cic inhibited Shh dependent activation of a Gli-reporter (Fig. 5 E and Fig. S5 D). Further, Bud attenuated Smo ciliary accumulation and pathway activation by SAG (Fig. 5 E; Fig. S5 E – F), and also suppressed Cyc induced Smo accumulation to the PC (Fig. S5 E – F). Bud treatment showed no effect on Wnt pathway activity (Fig. S5 G), consistent with a specific modulation of Hh signaling outside of its GC activity.

SmoM2 encodes a dominant active Smo variant identified in a human cancer that is resistant to inhibition by available Smo antagonists at concentrations that completely suppressed wildtype Smo activity (Fig. S5 H – K)(Xie, et al., 1998). Unexpectedly, both Bud and Cic attenuated SmoM2 ciliary localization, and downstream pathway activity, as effectively as wildtype Smo (Fig. 5 F–H; Fig. S5 L – M). Bud and Cic did not disrupt ciliary structure or ciliary trafficking: acetylated-tubulin (acet-tub), Ivs∷tagRFPT, and Arl13b∷tagRFPT within the PC were unaltered on treatment (Fig. S5 N – R).

The emergence of a drug resistant form of Smo with a D473H mutation was reported in a MB patient during treatment with GDC0449. The appearance of this mutation associated with a reemergence of the tumor (Yauch, et al., 2009). This finding has triggered a search for antagonists that effectively inhibit the activity of both wildtype and mutant forms of Smo (Dijkgraaf, et al., 2011; Kim, et al., 2010; Tao, et al., 2011). We examined Bud and GDC0449 in parallel for their inhibition of Hh induced SmoD473H activity, and the corresponding ciliary localization. Smo−/−MEF cells were transfected independently with wildtype and D473H mutant forms of Smo. Both forms rescued the cell’s response to Hh ligand (Fig.S5S). As expected, the D473H mutation conferred a dramatic resistance to GDC0449’s inhibitory action on both Hh pathway activity and Smo ciliary localization (Fig. S5T–V). In contrast, Bud showed similar efficacies in inhibiting wildtype Smo and SmoD473H activity in both assays (Fig.5I–K).

To examine the site of Bud action in the Hh pathway, we examined Hh signaling activity following removal of suppressor of Fused (suFU) activity, a Gli repressor functioning downstream of Smo. Distinct from GANT61(Lauth, et al., 2007), Bud failed to suppress ligandindependent Hh pathway activity induced by loss of suFU function (Fig. 5L). Together these data suggest that Bud may act at the level of Smo but through a different mechanism than other Smointeracting antagonists including SANT-1, Cyc, and GDC0449, and also distinct from FA and SAG. Consistent with a unique inhibitory action, Bud failed to compete with Bodipy-Cyc even at levels well above the inhibitory maximum (100 µM; Fig. 5 M–N). Further, whereas FA competed with GDC0449 to suppress effective pathway inhibition (Fig. 3), Bud enhanced GDC0449's activity to block Smo accumulation at the PC and Hh pathway inhibition (Fig. 6).

NIH/3T3 cells were maintained in DMEM containing 10%(v/v) calf serum, penicillin, streptomycin, and L-glutamine. HEK293, L, cos7 , and suFU−/− mouse embryonic fibroblast cells were maintained in DMEM containing 10%(v/v) fetal bovine serum, penicillin, streptomycin, and L-glutamine. Smo∷EGFP and Ivs∷ tagRFPT were cloned into pBabe retroviral constructs. Smo∷EGFP/Ivs∷tagRFPT stable cell lines was generated through viral infecting NIH/3T3 cells according to the procedure described previously(Wang, et al., 2009). A ShhLightII cell line (ATCC) was used for Gli-luciferase reporter assays. This line contains a stably integrated Gliresponsive firefly luciferase reporter and a constitutive Renilla luciferase expression construct (Taipale, et al., 2000). A subclone of this cell line was created expressing a stably integrated SmoM2 expression construct. Shh conditioned medium was collected from cos7 cells transfected with an expression construct encoding the amino terminal 19kDa signaling peptide of Shh and used at 13.7(±3.0)nM unless stated otherwise. Control conditioned medium was collected from cos7 cells transfected with an empty plasmid. Wnt3a conditioned medium was collected from an L cell line stably expressing a Wnt3a expression construct. Control conditioned medium was collected from wild-type L cells. All conditioned medium were diluted 1:10 prior to assay.

Chemical libraries screening utilized the Library of Pharmacologically Active Compounds (LOPAC, Sigma-Aldrich), the Spectrum Collection (Microsource Discovery Systems), and the Prestwick Chemical Library (Prestwick Chemical), along with a custom collection of additional biologically annotated chemistries absent from the above pre-plated reference collections. Glucocorticoids, cyclopamine, forskolin, mouse monoclonal anti-acetylated tubulin antibody for follow-up studies were purchased from Sigma. SAG was purchased from Axxora Platform. SANT-1 was obtained from Tocris Biosciences. GDC0449 was purchased from Selleck Chemicals. BODIPY-cyclopamine was purchased from Toronto Research Chemicals. All small molecule stock solutions were prepared by dissolving in DMSO at 1 or 10 mM and stored at −20°C. Mouse recombinant ShhN purified protein (IIShhN) was a gift from Dr. Pepinsky (Biogen Inc). Rabbit polyclonal anti-detyrosinated α-tubulin (Glu-tub) was from Chemicon, Mouse monoclonal anti-Arl13b antibody was from Antibody Incorporated. Secondary antibodies were from Life Technologies. Transfection was performed using Fugene6 or Fugene HD (Roche).

Cells were cultured and treated in 384-well imaging plate precoated with poly-D-Lysine (Greiner Bio-one) , fixed with 4% paraformaldehyde (Electron Microscopy Sciences) , and stained with Hoechst(Life Technologies). Immunofluorescence staining was conducted with standard procedures when necessary. Images were collected using Opera High Content Screening System (Perkin Elmer). ActivityBase (IDBS), Pipeline Pilot (Accelrys), Excel (Microsoft), and Prism (GraphPad) were used for high content screening data management and analysis. Acapella 2.0 software (Evotec Technologies/PerkinElmer) was used to perform multiparametric image quantification. All the comparative images were scanned with identical microscopic setting and analyzed with the same input parameters.

ShhLightII cells and SmoM2/LightII cells were cultured and treated in 96 well assay plates (Corning) and incubated with Duo-Glo luciferase substrates (Promega) to sequentially measure firefly and renilla luciferase activity. Smo, or GFP, expression plasmids were co-transfected into 3T3 cells together with a Gli-responsive firefly reporter and a TK-renilla luciferase reporter contruct to monitor effects of Smo overexpression. Co-transfection of the two reporter constructs was conducted in assays measuring Hh pathway activity in suFU−/− cells. Wnt activity was measured following co-transfection of a Top-flash and renilla luciferase reporter. In both Hh and Wnt activity assays, renilla luciferase reporter activity, or mass of protein, was used to normalize expression values. Luciferase signal was read by TopCount NX Microplate Scintillation and Luminescence Counter(Perkin Elmer). Quantitative PCR probes for Ptch1, Gli1, and β-actin were purchased from Applied Biosystems. Reactions and measurements were performed using on an Applied Biosystems 7900HT at Harvard FAS Center of System Biology. β-actin was used to normalize Ptch1 and Gli1 values.

Cos7 cells were transfected with a plasmid that co-expresses Smo and a nuclear localized tagRFPT marker (pCIT-Smo). The empty parental construct (pCIT) and a construct that co-express SmoM2 were used as controls to assess specificity and background signal. Three days after transfection, cells were incubated with 5nM Bodipy-cyclopamine, with or without additional compounds, for 1 hour at 37°C. Cells were then fixed and stained with Hoechst. Images were collected with the Opera High Content Screen System. Fluorescence values were assessed in transfected cells (red nuclei) with a program developed by the authors using Acapella 2.0 software. All of images were scanned with identical microscopic setting and analyzed with the same input parameters.

CGNP primary cells were isolated from P7 Ptch1+/− mice as previously reported(Chan, et al., 2009). Cells were seeded in poly-D-lysine coated imaging plates (Greiner Bio-one), treatments were applied 2 hours thereafter and last for 36 hours. Cells then were fixed with 4% paraformaldehyde (Electron Microscopy Sciences) , and stained with anti-pH3 antibody(Upstate; 1:100) followed by a secondary antibody(Invitrogen) and Hoechst(Invitrogen). Images were collected and cell proliferation quantified with a program developed by the authors utilizing Acapella 2.0 software. All of the images in each experiment were collected with identical microscopic settings and analyzed with identical input parameters.