Itraconazole and arsenic trioxide inhibit hedgehog pathway activation and tumor growth associated with acquired resistance to smoothened antagonists

SUMMARY

Recognition of the multiple roles of Hedgehog signaling in cancer has prompted intensive efforts to develop targeted pathway inhibitors. Leading inhibitors in clinical development act by binding to a common site within Smoothened, a critical pathway component. Acquired Smoothened mutations, including SMO^D477G, confer resistance to these inhibitors. We report here that itraconazole and arsenic trioxide, two agents in clinical use that inhibit Hedgehog signaling by mechanisms distinct from that of current Smoothened antagonists, retain inhibitory activity in vitro in the context of all reported resistance-conferring Smoothened mutants and GLI2 overexpression. Itraconazole and arsenic trioxide, alone or in combination, inhibit the growth of medulloblastoma and basal cell carcinoma in vivo, and prolong survival of mice with intracranial drug-resistant SMO^D477G medulloblastoma.

INTRODUCTION

The Hedgehog (Hh) signaling pathway is critical for embryonic patterning (Varjosalo and Taipale, 2008) and functions postnatally in tissue homeostasis through its action on stem or progenitor cells (Beachy et al., 2004; Shin et al., 2011). Aberrant Hh signaling has been implicated in oncogenesis, maintenance of tumor progenitor cells, and tumor-stromal interaction in a diverse array of cancers including tumors that arise sporadically or from individuals with heritable pathway mutations (Teglund and Toftgard, 2010). This has led to large-scale efforts by multiple pharmaceutical companies to develop Hh pathway antagonists for therapeutic use.

Hh pathway activity is regulated by Patched (PTCH), a twelve pass transmembrane protein, that suppresses the activity of Smoothened (SMO), a seven pass transmembrane protein that traffics constitutively through the primary cilium (Kim et al., 2009; Ocbina and Anderson, 2008). Upon binding of a Hh protein ligand to PTCH this suppression is relieved, leading to SMO activation and accumulation in the primary cilium (Corbit et al., 2005; Kim et al., 2009; Rohatgi et al., 2007); the GLI2 transcription factor, which also traffics through the cilium, consequently is activated and translocates to the nucleus (Kim et al., 2009), where it activates transcription of Hh-dependent target genes including PTCH1 and GLI1 (Figure 1).

Figure 1. A simplified view of Hh signaling.

Binding of a Hh ligand to Patched (PTCH) de-represses SMO, causing activation and translocation to the nucleus of GLI2, which initiates the transcription of target genes including PTCH and GLI1. Cyclopamine, its derivative, IPI-926, and its mimics GDC-0449 and NVP-LDE225 all bind competitively to a site within the transmembrane portion of SMO; the SMO missense mutations indicated by yellow asterisks decrease binding and confer resistance to these drugs. Itraconazole inhibits SMO by a distinct mechanism and ATO inhibits the pathway at the level of GLI.

The first Hh-dependent tumors were described in patients with Gorlin Syndrome (Gorlin, 1987) (also called Basal Cell Nevus Syndrome), an autosomal dominant condition associated with germline loss of one copy of the PTCH1 gene (Gailani et al., 1992; Hahn et al., 1996; Johnson et al., 1996). The most common tumors arising in these patients are basal cell carcinoma (BCC) of the skin, medulloblastoma (MB), and more rarely rhabdomyosarcoma (Gorlin, 1987). Hh pathway upregulation has been found in essentially all cases of BCC (Epstein, 2008), including sporadic BCC, with ~90% containing PTCH1 mutations (Aszterbaum et al., 1998; Gailani et al., 1996) and ~10% containing activating mutations in SMO (Reifenberger et al., 1998; Xie et al., 1998). Hh-dependent MB (Goodrich et al., 1997; Mao et al., 2006) accounts for approximately one-third of all MB (Monje et al., 2011) and is associated with intermediate prognosis (Cho et al., 2011; Ellison et al., 2011; Northcott et al., 2011).

SMO, as a central regulator of the pathway and an accessible cell membrane component, has been the primary focus for development of small molecule Hh pathway inhibitors. Cyclopamine, the archetypical SMO antagonist, was first described as a steroidal alkaloid teratogen from the corn lily associated with cyclopic lambs (Keeler and Binns, 1966; Keeler and Binns, 1968) and later determined to be a SMO antagonist (Chen et al., 2002; Cooper et al., 1998; Taipale et al., 2000). Of the many SMO inhibitors in development, four have progressed into phase II trials, including vismodegib (GDC-0449; Genentech), NVP-LDE225 (Novartis), IPI-926 (Infinity), and XL-139 (BMS/Exelixis); IPI-926 is derived from cyclopamine, and all of these inhibitors compete with cyclopamine binding to SMO (Gendreau and Fargnoli, 2009; Pan et al., 2010; Robarge et al., 2009; Tremblay et al., 2009). GDC-0449 was recently approved for use as a first-line therapy in advanced unresectable basal cell carcinoma (Jefferson, 2012).

The limited mechanistic diversity represented by clinically developed Hh pathway inhibitors has become a focus of clinical concern due the emergence of resistant SMO mutants. The first case of SMO antagonist resistance was reported in a patient with metastatic MB initially highly responsive to GDC-0449 (Rudin et al., 2009). Gene sequencing of recurrent, drug resistant tumors from this patient identified a SMO missense mutation, D473H that decreased the binding affinity of GDC-0449 by 100-fold. A homologous mouse mutation, D477G, was found in resistant murine Ptch^+/−; p53^−/− MB generated in vivo by repetitive cycles of treatment with GDC-0449 (Yauch et al., 2009). Another GDC-0449 resistant mutant, E518K, subsequently was identified in human SMO (Dijkgraaf et al., 2011). Development of in vivo resistance to another SMO antagonist, NVP-LDE225, was demonstrated in murine MB with mutations occurring at residues L225, N223, S391, D388, and G457 (Buonamici et al., 2010). The latter reports also identified other putative mechanisms of resistance including Gli2 and Ccnd1 amplification, and activation of the PI3K-AKT-mTOR signaling pathway.

We have previously identified itraconazole, an FDA-approved triazole anti-fungal agent, and arsenic trioxide (ATO), FDA-approved for the treatment of acute promyelocytic leukemia (APL), as potent inhibitors of the Hh pathway (Kim et al., 2010a; Kim et al., 2010b). Itraconazole inhibits Hh pathway activation at the level of SMO at a site distinct from that of cyclopamine mimics currently in development and by a mechanism distinct from its anti-fungal target of lanosterol-14α demethylase (Kim et al., 2010b).

ATO directly binds to the zinc finger motif of the promyelogenous leukemia -retinoic acid receptor alpha fusion protein (PML-RARα), the causative factor of APL (de The et al., 1990; de The et al., 1991; Rowley et al., 1977), and promotes its degradation (Lallemand-Breitenbach et al., 2008; Zhang et al., 2010). Similarly, ATO inhibits Hh signaling by inhibiting GLI2 ciliary accumulation and promoting its degradation (Kim et al., 2010a). ATO also inhibited the growth of Ewing Sarcoma tumors overexpressing GLI1 due to direct transcriptional activation by the EWS-FLI1 fusion oncoprotein (Beauchamp et al., 2009; Beauchamp et al., 2011; Joo et al., 2009; Zwerner et al., 2008).

The occurrence of drug-resistant SMO mutations highlights the therapeutic need for agents capable of maintaining robust on-target clinical responses. Small molecule compounds that overcome resistance to murine SMO^D477G (Dijkgraaf et al., 2011; Tao et al., 2011) or human SMO^D473H (Dijkgraaf et al., 2011) have been reported recently, but with no clear time-line for clinical development. As FDA-approved drugs, itraconazole and ATO represent readily available agents with distinct modes of Hh pathway inhibitory activity. Therefore, we sought to test the efficacy of itraconazole and ATO, as single agents and in combination, to inhibit Hh pathway activity and growth of tumors with drug resistant SMO mutations.

RESULTS

Itraconazole inhibits GDC-0449 resistant pathway activity mediated by SMO^D477G

As itraconazole acts on SMO at a site distinct from that of cyclopamine and cyclopamine mimics (Kim et al., 2010b), we hypothesized that itraconazole may have activity against the GDC-0449 resistance mutant, SMO^D477G (Yauch et al., 2009). First we evaluated the ability of itraconazole to inhibit Hh pathway activity in the context of SMO^WT or SMO^D477G expression. Smo^WT or Smo^D477G were co-transfected with 8×-GLI binding site firefly luciferase and SV40-Renilla reporters (Taipale et al., 2000) into a Smo^−/− mouse embryonic fibroblast (MEF) cell line (Varjosalo et al., 2006) and treated with SHHN conditioned medium (CM) (Maity et al., 2005). As expected, GDC-0449 was active against SMO^WT, but unable to inhibit SMO^D477G (Figure 2A). In contrast, itraconazole potently inhibited both SMO^WT and SMO^D477G (Figure 2B) although the maximal inhibition of SMO^D477G activity up to a concentration of 5 μM itraconazole left a residual ~30% activity (Figure 2B). Itraconazole thus acts as a full antagonist of SMO^WT and as a partial antagonist of SMO^D477G. Similar to human SMO^D473H (Yauch et al., 2009), murine SMO^D477G was less susceptible to inhibition by KAAD-cyclopamine, as evidenced by the right-shift of the mutant dose-response curve (Figure S1A). However, unlike GDC-0449, KAAD-cyclopamine was able to fully inhibit SMO^D477G activity at higher doses.

Figure 2. Itraconazole inhibits the Hh pathway in the context of GDC-0449 resistant SMO^D477G.

(A & B) Relative Hh pathway activity as determined by expression of 8×-GLI-luciferase reporter in SHHN stimulated Smo^−/− MEFs expressing SMO^WT (blue) or SMO^D477G (red) treated with (A) GDC-0449 or (B) itraconazole. Data represent mean of triplicates ± SD. (C & D) Proliferation of Ptch^−/+; p53^−/− MB tumorspheres expressing endogenous SMO^WT (blue) or SMO^D477G (red) treated with increasing doses of (C) GDC-0449 and (D) itraconazole. Data represent mean of quadruplicates ± SEM. (E & F) Relative Gli1 mRNA transcription in (C) GDC-0449-treated and (D) itraconazole-treated MB tumorspheres expressing endogenous SMO^WT (blue) or SMO^D477G (red). Data represent mean of quadruplicates ± SEM. See also Figure S1.

To evaluate itraconazole activity against Hh-dependent tumors, we tested the ability of itraconazole to inhibit pathway activity in tumorsphere cultures established using MB cells with constitutive Hh pathway activity from Ptch^+/−; p53^−/− mice (Berman et al., 2002; Goodrich et al., 1997; Romer et al., 2004). When cultured in neural stem cell medium, these cells grow in a Hh pathway-dependent manner and form tumorspheres maximally enriched for the highly tumorigenic CD-15⁺ cell population (Ward et al., 2009) (Read et al., 2009; Shi et al., 2011). We utilized tumorsphere cultures derived from allografts of parental SMO^WT and derivative SG274 MB cells that express spontaneously acquired SMO^D477G (Dijkgraaf et al., 2011; Yauch et al., 2009) to examine the differential potency of itraconazole in an endogenous system of drug-resistance. The SMO^D477G-expressing tumorspheres were ~100-fold less sensitive than parental SMO^WT tumorspheres to anti-proliferative activity of GDC-0449 by MTS analysis (Figure 2C). In this system, itraconazole displayed equivalent anti-proliferative activity in the parental (IC₅₀ 55 nM; 50–60 nM CI) and SMO^D477G mutant MB cells (62 nM; 53–70 nM CI) (Figure 2D). Inhibition of Hh pathway activity, as monitored by Gli1 mRNA levels, correlated with the anti-proliferative potencies of GDC-0449 (Figure 2E) and itraconazole (Figure 2F). In contrast to the Smo^−/− MEF reporter-based signaling assay, itraconazole fully inhibited proliferation and Hh pathway activity in the SMO^WT and GDC-0449 resistant MB cultures. MB spheres also were more sensitive to itraconazole, with ~100 nM IC₅₀ for Gli1 mRNA transcription compared to ~600 nM IC₅₀ for the luciferase reporter assay in Smo^−/− MEFs. HhAntag (Gabay et al., 2003; Romer et al., 2004), a SMO antagonist structurally distinct from GDC-0449 and cyclopamine, has been reported to inhibit SMO^D473H at 1 μM (~25x IC₅₀ for SMO^WT) (Dijkgraaf et al., 2011). Pathway inhibition by HhAntag, as measured by Gli1 mRNA, showed a similar behavior to that of KAAD-cyclopamine (Figure S1A) in that its inhibition curve shifts to the right (Figure S1B). The decrease in potency of HhAntag and KAAD-cyclopamine in the context of SMO^D477G is consistent with previous reports for other inhibitors that compete for cyclopamine binding to SMO (Dijkgraaf et al., 2011; Lee et al., 2012).

Combination of itraconazole and cyclopamine can inhibit SMO^D477G

As we previously showed that itraconazole and KAAD-cyclopamine synergize to inhibit SMO^WT (Kim et al., 2010b) and that these drugs act at distinct sites (Kim et al., 2010b), we hypothesized that itraconazole might also act synergistically with cyclopamine mimics to inhibit the activity of SMO^D477G. We found, however, that addition of GDC-0449 did not enhance itraconazole inhibitory potency against SMO^D477G, suggesting that GDC-0449 does not contribute to pathway suppression in the combination (Figure 3A). Combination of itraconazole and KAAD-cyclopamine completely inhibited SMO^D477G activity at doses that are partially inhibitory for the individual agents (Figure 3B). The addition of itraconazole to KAAD-cyclopamine did not shift the IC₅₀ of KAAD-cyclopamine (Figure 3C, S2; Table S1) suggesting that the two drugs inhibit SMO^D477G additively.

Figure 3. Itraconazole combines with cyclopamine to inhibit GDC-0449-resistant SMO function.

(A) Relative 8×-GLI-luciferase expression in SHHN-stimulated Smo^−/− MEFs expressing SMO^WT or SMO^D477G, treated with GDC-0449, itraconazole, or both. (B) Effect of KAAD-cyclopamine 100 nM, itraconazole 2 μM, or both on SHHN-stimulated 8×-GLI-luciferase expression in Smo^−/− MEFs expressing SMO^D477G. (C) Dose-response curves of KAAD-cyclopamine for SHHN activated signaling in Smo^−/− MEFs expressing SMO^D477G, in the presence or absence of itraconazole. Data represent mean of triplicates ± SD. See also

Itraconazole and arsenic trioxide combine to inhibit Hh pathway activity and tumor growth

As ATO inhibits the Hh pathway at the level of GLI proteins (Kim et al., 2010a) and itraconazole acts on SMO, we hypothesized that the two drugs might synergize in pathway inhibition. We tested the drug combination against SMO^WT in transiently transfected NIH-3T3 cells using fixed doses of one drug while titrating the other. The combination of itraconazole and ATO improves Hh pathway inhibition (Figure 4A, 4B), but this effect appears additive rather than synergistic as there is no leftward shift of IC₅₀ for either drug (Figure 4A, 4B; Figure S3A, S3B; Table S2 and S3).

Figure 4. ATO and itraconazole combine to inhibit Hh pathway activation and tumor growth by SMO^WT.

(A & B) Effect of combination therapy on the IC₅₀ concentrations of itraconazole (A) or ATO (B) in NIH-3T3 cells expressing endogenous SMO^WT stimulated with SHHN in 8×-GLI-luciferase signaling assays. Data represent mean of triplicates ± SD. (C & D) Nude mice with SMO^WT hind-flank MB allografts were treated with vehicle control (40% cyclodextrin PO bid, N=7 tumors; black), GDC-0449 (100 mg/kg PO bid, N=7; red) ATO (7.5 mg/kg IP qd, N=7 tumors; green), itraconazole (75 mg/kg PO bid, N=7 tumors; blue), or both ATO and itraconazole (N=7 tumors; blue-green). Effect of GDC-0449, ATO, itraconazole, or the combination of ATO and itraconazole on (C) tumor growth and (D) Gli1 mRNA expression compared to vehicle control. Data represent group means ± SEM. *p < 0.01 vs. Vehicle; ** p < 0.01 vs. ATO alone; *** p < 0.01 vs. itraconazole alone. See also Figure S3, Table S2 and S3.

We next tested the itraconazole/ATO combination in tumor growth inhibition in vivo using a subcutaneous allograft model of Ptch^+/−; p53^−/− mouse MB. Nude mice with established tumors were treated with ATO 7.5 mg/kg intraperitoneal (IP) once daily (qd), or oral (PO) itraconazole 75 mg/kg twice daily (bid), or a combination of both drugs. Single-agent therapy with ATO or itraconazole significantly inhibited tumor growth (Figure 4C), and was associated with 78% (p < 0.005) and 96% (p < 0.001) growth inhibition, respectively, as compared to vehicle-treated tumors on day 9 of treatment. During the same vehicle-controlled time period, the combination of ATO and itraconazole not only inhibited tumor growth but also reduced tumor volumes by 72% from the start of treatment (p < 0.001 vs. vehicle; p < 0.001 vs. initial volumes). Combination treatment compared to ATO or itraconazole alone significantly improved anti-tumor efficacy, with 99% and 85% inhibition of tumor growth relative to ATO or itraconazole alone through the single agent controlled portions of study, days 13 and 17, respectively (Figure 4C; p < 0.001 for both). ATO and itraconazole were well tolerated, both as single-agents and in combination, as body weights were similarly maintained across cohorts (Figure S3C).

Inhibition of MB tumor growth was associated with parallel reductions in Hh pathway activity in treated tumors (Figure 4D). Compared to vehicle-treated tumors, single-agent ATO or itraconazole treatment resulted in 39% and 55% inhibition of tumor Gli1 mRNA expression, respectively, (p <0.01 for both), whereas combined ATO and itraconazole suppressed tumor Gli1 mRNA transcript levels by 77% (p <0.001 vs. vehicle and ATO alone; p <0.01 vs. itraconazole alone).

We also monitored the status of the PI3K pathway and GLI2 in the treated tumors (Figure S3D–F), as PI3K pathway activation and GLI2 amplification and overexpression have been reported as mechanisms of resistance to cyclopamine-competitive antagonists (Buonamici et al., 2010; Dijkgraaf et al., 2011). Tumors with Hh pathway inhibition (Figure 4D) displayed a low level of PI3K pathway activation at baseline, evidenced by phosphorylation of AKT, S6K, and 4E-BP1 (reviewed in (Cully et al., 2006)), that was largely unchanged in response to the single-agent or combination treatments (Figure S3D). Tumors that eventually progressed during drug therapy (Figure 4C) did not show increased levels of pAKT and pS6K (Figure S3E), nor did they display increased Gli2 mRNA (Figure S3F). These results suggest that the previously described mechanisms of resistance to GDC-0449 and LDE225are not the cause of eventual tumor growth with itraconazole and ATO treatment, which more likely is caused by incomplete, albeit substantial, pathway suppression.

Additionally, in a subcutaneous allograft model of BCC derived from Ptch^+/−; K14-Cre^ER2/+; p53^fl/fl mice (Tang et al., 2011), the combination of itraconazole and ATO significantly inhibited tumor growth compared to control (p < 0.001 at day 30 of treatment) (Figure 5). Mice treated with single agents displayed reductions in mean tumor growth that did not reach statistical significance compared to control.

Figure 5. Combination of ATO and itraconazole inhibits tumor growth of Hh-dependent basal cell carcinoma.

NOD/SCID mice with established K14-Cre^ER/+; Ptch^+/−; p53^fl/fl BCC allografts were treated with vehicle control (40% cyclodextrin po bid, N=4 tumors; black), ATO (7.5mg/kg IP qd, N=6 tumors; green), itraconazole (75 mg/kg po bid, N=10 tumors; blue), or both ATO and itraconazole (N=16 tumors; blue-green; p<0.01 for the combination compared to either itraconazole or ATO alone). Data represent group means ± SEM.

Combination of itraconazole and ATO can inhibit Hh-pathway activity mediated by GDC-0449 resistant SMO in vitro and in vivo

Having established efficacy of the itraconazole/ATO combination for antagonism of SMO^WT activity in vitro and in vivo, we tested the combination against the GDC-0449 resistant SMO mutant, SMO^D477G (Yauch et al., 2009). We previously showed that ATO can inhibit SMO^D477G activity (Kim et al., 2010a). In SHHN stimulated Smo^−/− MEFs transfected with Smo^D477G, addition of ATO to itraconazole inhibited Hh-pathway activation in a dose dependent fashion (Figure 6A) and led to improved inhibition by itraconazole at higher doses (Figure 6B), indicating a greater potency for the combination. Similar to SMO^WT (Figure 4A, 4B), ATO and itraconazole inhibited SMO^D477G activity additively (Figure 6C, S4A; Table S4), with no leftward IC₅₀ shift (Figure S4A; Table S4).

Figure 6. ATO and itraconazole combine to inhibit SMO^D477G activity and tumor growth, and improves survival in an orthotopic medulloblastoma model.

(A–C) Smo^−/− MEFs were transfected with either Smo^WT or Smo^D477G and stimulated with SHHN. Data represent mean of triplicates ± SD. (A) Relative Hh pathway activity as assessed by relative 8×-GLI-luciferase expression in the presence or absence of itraconazole, ATO, or the combination. Dashed line represents the level of pathway inhibition of SMO^WT by itraconazole 1.5 μM. (B) Effect of the addition of increasing doses of ATO to itraconazole in cells expressing SMO^D477G. (C) Effect of increasing doses of itraconazole on the IC₅₀ of ATO in cells expressing SMO^D477G. (D & E) Nude mice with established SMO^D477G MB allografts were treated with vehicle control (40% cyclodextrin PO bid, N=8 tumors; black), GDC-0449 (100 mg/kg PO bid, N=8; red); ATO (7.5mg/kg IP qd, N=8 tumors; green), itraconazole (75 mg/kg PO bid, N=8 tumors; blue), or both ATO and oral itraconazole (N=8 tumors; blue-green); (D) tumor growth over time and (E) relative Gli1 mRNA expression. Data represent group means ± SEM. *p < 0.01 vs. Vehicle; ** p < 0.01 vs. ATO alone; *** p < 0.01 vs. ITRA alone. (F) Kaplan-Meier survival analysis of an orthotopic model of SMO^D477G MB treated with vehicle control (N=9), GDC-0449 (N=9), itraconazole (N=9), ATO (N=9), or both ATO and itraconazole (N=9) using same doses as in (D). See also Figure S4 and Table S4.

We tested combination treatment in vivo in a GDC-0449 resistant Ptch^+/−; p53^−/− murine MB allograft model expressing endogenous SMO^D477G (Yauch et al., 2009). In contrast to the results observed in SMO^WT tumors (Figure 4C), twice-daily treatment with GDC-0449 did not significantly impact tumor growth or Hh pathway status compared to control (Figure 6D and 6E). Of note, a similar albeit less dramatic differential was associated with once-per-day treatment with HhAntag treatment in this model (Figure S4B–D). Unlike GDC-0449, ATO and itraconazole inhibited SMO^D477G MB growth with similar potency to that of SMO^WT tumors. In GDC-0449 resistant tumors, single-agent ATO and itraconazole resulted in 78% and 92% growth inhibition, respectively, versus control through the 9-day vehicle controlled period (p < 0.001 for both). Combination treatment over the same period completely inhibited tumor growth and caused 48% tumor regression from the start of study (p < 0.001 vs. vehicle; vs. initial volumes). Furthermore, the efficacy of the combination was superior to either agent alone, and resulted in 97% relative tumor growth inhibition over ATO alone (p < 0.001) and 85% improvement over itraconazole alone (p < 0.001) through the 12-day period controlled by the respective single-agent arms (Figure 6D).

Treated tumors revealed that single-agent itraconazole and ATO inhibited Hh pathway activity in GDC-0449 resistant allografts (Figure 6E) with similar magnitudes of 43% and 45% of control treated tumors, respectively, as measured by Gli1 mRNA transcript (p < 0.01 for both). Combination treatment resulted in further pathway suppression to 66% (p <0.01). Gli1 mRNA expression was not significantly inhibited (p >0.05) in response to GDC-0449.

The hind-flank allograft model of SMO^D477G MB serves as a valuable proxy for anti-tumor potential in the context of GDC-0449 resistance. However, the ectopic nature of this model fails to account for the challenges of treating an advanced intracranial malignancy. To further evaluate the potential clinical relevance of itraconazole and ATO in MB treatment, we assessed survival in an orthotopic model of SMO^D477G MB. Survival of mice bearing engrafted intracranial tumors did not improve with GDC-0449 as both treatment and control resulted in median survival times of 14 days (Figure 6F, red and black lines, respectively; p = 0.69). In contrast, survival of mice receiving single agent treatment with ATO or itraconazole (Figure 6F, green or blue lines, respectively) improved significantly over control, with median survival times of 18 (p = 0.0041) and 22 days (p = 0.0001). The combination of these drugs (Figure 6F blue-green line) further improved survival over either single-agent alone (p ≤ 0.0001) and doubled the survival time over GDC-0449 and vehicle arms, with a median survival of 29-days (p < 0.0001 compared to vehicle or GDC-0449 arms).

Combination of itraconazole and ATO remains active in all known drug-resistant SMO mutations and in the context of GLI2 overexpression

In addition to SMO^D473H (murine SMO^D447G), alanine scan mutagenesis identified potential alterations of SMO^E518 that confer resistance to GDC-0449 (Dijkgraaf et al., 2011). Five resistance mutants to another SMO antagonist, NVP-LDE225, have also been reported in Ptch^+/; p53^−/− and Ptch^+/−; Hic1^+/− MB models (Buonamici et al., 2010). We transiently transfected these drug resistant Smo mutants into Smo^−/− MEFs and induced pathway activation with SHHN CM. Cells were treated with GDC-0449 0.5 μM (~40× IC₅₀ (Robarge et al., 2009)), NVP-LDE225 0.5 μM (~80× IC₅₀ (Buonamici et al., 2010)), itraconazole 1.5 μM, ATO 2.5 μM, and a combination of itraconazole and ATO. We confirmed previously reported resistance of the various SMO mutants (Figure 7), including partial sensitivity of Smo^G457S to NVP-LDE225 (Figure 7C) consistent with an IC₅₀ of 400 nM (Buonamici et al., 2010). Smo^G457S was, however, resistant to 50 nM of NVP-LDE225 (~8× IC₅₀ of SMO^WT; Figure 7C).

Figure 7. Combination of ATO and itraconazole inhibits all other known drug-resistant SMO mutants and GLI2 overexpression.

SHHN stimulated Smo^−/− MEFs were transfected with (A) Smo^WT or mutant Smo constructs resistant to (B) GDC-0449 or (C–G) NVP-LDE225. The effects of itraconazole, ATO, and the combination of both agents on Hh pathway activity was assessed by relative 8×-GLI-luciferase activity. Labels of panels (B–G) indicate the mutant SMO that was expressed. (H) Effect of itraconazole, ATO, or the combination on relative 8×-GLI-luciferase activity in NIH-3T3 cells transfected with Gli2 +/− SHHN stimulation. Data represent mean of triplicates ± SD. See also Figure S5.

Itraconazole inhibited activity of all known SMO resistance mutants at similar levels to SMO^D477G (Figure 7) except for murine SMO^E522K, which corresponds to human SMO^E518. SMO^E522K was resistant to GDC-0449 and, unlike SMO^D477G, was almost completely inhibited by itraconazole (Figure 7B). Itraconazole 1.5 μM inhibited the NVP-LDE225 resistant mutants (Figure 7C–G) with a range of ~40% (Figure 7C) to ~60% (Figure 7D). The addition of ATO 2.5 μM to itraconazole was able to completely inhibit the activity of these mutants. As expected, higher doses of ATO completely inhibited all of the resistant mutants (Figure S5), as ATO acts downstream of SMO.

We previously showed that ATO inhibits the Hh pathway through inhibition of ciliary transport and degradation of the GLI proteins (Kim et al., 2010a). We tested itraconazole and ATO, alone and in combination, on NIH-3T3 cells that ectopically overexpressed GLI2 with and without the addition of SHHN CM (Figure 7H). GLI2 overexpression alone caused elevated pathway activity and the addition of SHHN CM further induced Hh pathway activation. As expected, itraconazole inhibited pathway activity induced by SHHN but not by GLI2 overexpression (Figure 7H), as itraconazole acts upstream of GLI2 (Kim et al., 2010b). ATO, however, was able to inhibit pathway activity induced by SHHN and GLI2. The combination of itraconazole and ATO inhibited pathway activity induced by both SHHN and GLI2 overexpression to a similar extent as was observed with ATO alone.

DISCUSSION

Itraconazole, cyclopamine-mimics and SMO

Prior to our studies of itraconazole, most or all SMO modulators, including GDC-0449 and NVP-LDE225, were found to compete with BODIPY-cyclopamine for binding to SMO, presumably by binding at closely overlapping sites. Recent studies (Dijkgraaf et al., 2011; Nachtergaele et al., 2012; Tao et al., 2011), however, reveal additional complexity. Tao et al. (2011) show that several recently identified Hh antagonists compete for binding with distinct subsets of BODIPY-cyclopamine competitive Smo modulators, suggesting that the BODIPY-cyclopamine binding pocket and nearby regions within SMO may accommodate two non-overlapping modulators simultaneously. Furthermore, oxysterol agonists that do not compete with BODIPY-cyclopamine (Dwyer et al. 2007) have been shown to bind SMO directly (Nachtergaele et al., 2012), presumably at a distinct site.

Itraconazole appears to act entirely outside the BODIPY-cyclopamine binding pocket, as it fails to compete with BODIPY-cyclopamine or the SMO agonist, SAG (Kim et al., 2010b). We cannot rule out the possibility that itraconazole acts on SMO indirectly via another molecule as we do not have a direct binding assay. Nonetheless, the ability of itraconazole to inhibit all of the SMO variants resistant to GDC-0449 and NVP-LDE225 is consistent with its action on SMO at a distinct site from that of cyclopamine and its mimics. Although compounds that overcome SMO^D477G activity and act within the BODIPY-cyclopamine pocket of SMO have been reported (Dijkgraaf et al., 2011; Tao et al., 2011), their inhibitory activity against other SMO mutations and their toxicity and pharmacological properties in patients remain unknown.

Interestingly, itraconazole only partially inhibits the activity of SMO^D477G (Figure 2B) and the other mutants except for SMO^E522K (Figure 7). Small molecule partial antagonists for the GPCR mGlu5 receptor have been described (Lamb et al., 2011; Rodriguez et al., 2010; Rodriguez et al., 2005) that act as non-competitive allosteric inhibitors of the orthosteric endogenous ligand, glutamate (Wood et al., 2011). By analogy, perhaps itraconazole acts as an allosteric inhibitor of a currently undefined endogenous SMO–interacting factor. The mutant SMO proteins may exist in conformations which allow itraconazole to exert its inhibitory function in an incomplete fashion.

Combination of itraconazole and ATO as therapy for Hh-dependent tumors

We have shown that itraconazole and ATO, alone and in combination, inhibit Hh pathway activity and growth of in vitro cultured cells and in vivo tumors bearing wild-type and drug-resistant SMO. The eventual, albeit significantly delayed, in vivo tumor growth from itraconazole and ATO treatment is not due to previously described resistance mechanisms of PI3K-mTOR activation or Gli2 amplification (Figures S3D–F). Induction of P-glycoprotein (PGP) and drug efflux, recently reported (Lee et al., 2012) as a resistance mechanism to IPI-926, is unlikely to be the case in our studies as itraconazole is a potent inhibitor of PGP (Wang et al., 2002). The tumor growth kinetics in our studies more likely reflect the incomplete albeit substantial degree of Hh pathway inhibition by itraconazole and ATO; this incomplete inhibition might be improved by adjusting the relative or overall dosages of itraconazole and ATO in the combination. Other mechanisms to account for the eventual tumor regrowth, however, cannot be fully excluded at this time.

Itraconazole also has been reported to inhibit mTOR activity (Xu et al., 2010) and ATO has been reported to upregulate PTEN (Redondo-Munoz et al., 2010) and degrade AKT (Mann et al., 2008). However, we did not observe significant changes in PI3K pathway activity in our short-term (Figure S3D) or long-term (Figure S3E) analyses. Several reasons may account for the apparent discrepancy. First, the studies cited for the activity of itraconazole and ATO against the PI3K pathway were all performed in cell culture systems with robust activation of the PI3K pathway and may not accurately reflect the in vivo environment examined here. Second, the marginal detection of PI3K pathway activation in our in vivo tumor models suggest that this pathway contributes little to tumor growth and further inhibition by pharmacologic means may not be possible.

The therapeutic regimen of itraconazole and ATO combination is active against all reported functional SMO resistance mutants and has several advantages over single-agent SMO inhibition. In addition to Hh pathway inhibition, itraconazole inhibits angiogenesis through VEGF and FGF blockade (Aftab et al., 2011; Chong et al., 2007; Nacev et al., 2011). Likewise, ATO induces apoptosis by caspase activation, BCL2 inhibition, and reactive oxygen species accumulation, NFκB inhibition (Emadi and Gore, 2010) and the promotion of GLI protein degradation (Kim et al., 2010a). The combination of itraconazole and ATO simultaneously targets two distinct and critical loci within the Hh pathway: SMO and GLI proteins. Such multi-focal targeting within the Hh pathway itself and other pathways necessary for tumor growth may prevent or delay the emergence of resistant tumor clones. Moreover, the additive inhibitory effect of the itraconazole/ATO combination allows for the use of lower doses of each drug to maintain or improve anti-tumor efficacy while decreasing the likelihood of adverse toxicities. Finally, our data suggest that ATO, alone or in combination with itraconazole, may also inhibit tumor growth associated with GLI2-mediated resistance (Figure 7H and Kim et al., 2010a).

With the recent FDA-approval of GDC-0449 for advanced BCC and with multiple other cyclopamine mimics in clinical trials, we anticipate that examples of clinical resistance and other drug-resistant SMO mutants will become more prevalent. This represents a substantial threat to patients and emphasizes the need to identify clinical candidates capable of maintaining on-target efficacy in the face of such mutations. Itraconazole and ATO are both FDA-approved drugs that are well tolerated, readily available and with well characterized pharmacokinetic and toxicity profiles. We have shown that the combination of itraconazole and ATO is effective against both SMO^WT, the recently reported SMO mutants resistant to cyclopamine-mimics and against GLI2 overexpression. These factors strongly support the clinical evaluation of itraconazole and ATO in the treatment of de novo Hh-dependent tumors or those with acquired resistance to cyclopamine-mimics.

EXPERIMENTAL PROCEDURES

Detailed procedures can be found in Supplemental Experimental Procedures

Reagents

Itraconazole (Sigma), vismodegib (GDC-0449; LC Laboratories) and NVP-LDE225 (LC Laboratories) were dissolved in DMSO for in vitro experiments. ATO powder (Sigma) for in vitro experiments was formulated as previously described (Zhu et al., 2002). Itraconazole oral solution (Sporanox, Ortho Biotech) and ATO (Trisenox; Cephalon Inc.) for in vivo experiments were obtained from the pharmacies of The Johns Hopkins and Stanford Cancer Centers. For in vivo studies, GDC-0449 100 mg/kg oral gavage (PO) twice-daily (bid), ATO 7.5mg/kg intraperitoneal (IP) once-daily (qd) and itraconazole 75 mg/kg PO bid were used both as single-agents and in combination.

Cell Based Signaling Assays

Plasmid constructs for Smo^WT or Smo^D477G, Gli2 (Kim et al., 2010a), GFP, GLI- dependent firefly luciferase reporter, and TK-Renilla luciferase reporter (Taipale et al., 2000) have been described previously. Constructs for other Smo resistance mutants were prepared using standard site mutagenesis techniques to pCEFL-Smo^WT and verified by DNA sequencing. Fugene 6 (Roche Applied Science) or TransIT (Mirus Bio) reagents were used for transient transfections.

NIH 3T3 cell line (ATCC) and Smo^−/− mouse embryonic fibroblast (MEFs) cell line (Varjosalo et al., 2006) were stimulated with SHHN conditioned medium (CM) (Maity et al., 2005) or control HEK 293 CM for Hh pathway signaling assays.

Cells were incubated for ~40 hrs with drug treatments. Relative luciferase (fold induction) units were obtained by normalization of luciferase signals from SHHN CM stimulated with control CM stimulated cells. Luminescence assays were performed on a Fluostar Optima (BMG Labtech) or Centro (Berthold Technologies)using Dual Luciferase Assay Reporter System (Promega).

In vivo Studies

All mouse studies were approved by and conformed to the policies and regulations of the respective Institutional Animal Care and Use Committees at the Johns Hopkins University and the Children’s Hospital Oakland Research Institute.

Basal Cell Carcinoma Allograft Model

Endogenous BCC tumors from Ptch1^+/−; K14-Cre^ER; p53^fl/fl mice (Tang et al., 2011) were subcutaneously established as flank xenografts in NOD/SCID mice (Jackson Laboratories). Oral itraconazole was given bid on weekdays and qd on weekends. IP ATO was given on a qd schedule.

Medulloblastoma Allograft Models

A parental murine MB model, expressing wild-type SMO, was derived from spontaneous, intracranial MB from Ptch^−/+; p53^−/− mice and maintained as subcutaneous hind-flank allografts in athymic nude mice. GDC-0449 resistant allografts expressing SMO^D477G (SG274) were derived from parental allografts as previously described (Yauch et al., 2009). Hind-flank or orthotopic xenograft models were generated from either parental or SG274 tissues.

Medulloblastoma Culture and Proliferation Assays

MB tumorsphere cultures were derived from either parental or SG274 hind-flank allografts using previously described methods(Shi et al., 2011). Tumorsphere proliferation was quantified by MTS assay following a 96-h treatment to the various drugs using exposure CellTiter 96 AQ_ueous One Solution Cell Proliferation Assay (Promega).

qPCR of Gli1 & Gli2 Transcripts in Medulloblastoma Systems

Established MB tumorspheres were treated for 24 h with the various drugs. Tissue from tumor bearing mice were treated for 3-days and harvested 4 h after the last-dose. Relative mRNA was quantified using Taqman gene probes for Actin, Gli1, and Gli2 as previously described (Shi et al., 2011) and reported as fold induction relative to control samples using the ΔΔCt (2^−ΔΔCt) method with actin as an internal control.

HIGHLIGHTS.

• Itraconazole inhibits the activity of GDC-0449-resistant SMO^D477G.

• Itraconazole & ATO combine to effectively block SMO^D477G activity in vitro.

• Itraconazole & ATO inhibit SMO^WT & SMO^D477G tumor growth and prolong survival.

• Itraconazole and ATO retain activity in all reported drug-resistant SMO mutants.

SIGNIFICANCE.

Small molecule Hedgehog pathway antagonists currently being evaluated in cancer clinical trials act by binding Smoothened in a manner competitive with cyclopamine. Resistance to these cyclopamine mimics caused by acquired mutations in Smoothened has been described in relevant animal models and in patients. Itraconazole and arsenic trioxide, two readily available FDA-approved drugs with well-defined pharmacokinetics and relatively mild toxicity profiles, have activity in the context of wild-type or resistant Smoothened mutants, in vitro and in vivo. Itraconazole and arsenic trioxide are active Hedgehog pathway inhibitors that could be introduced rapidly into clinical trials for the treatment of de novo Hedgehog-dependent tumors and those with acquired resistance to Smoothened inhibitors.