Arsenic antagonizes the Hedgehog pathway by preventing ciliary accumulation and reducing stability of the Gli2 transcriptional effector

Abstract

Aberrant Hedgehog (Hh) pathway activation has been implicated in cancers of diverse tissues and organs, and the tumor growth-inhibiting effects of pathway antagonists in animal models have stimulated efforts to develop pathway antagonists for human therapeutic purposes. These efforts have focused largely on cyclopamine derivatives or other compounds that mimic cyclopamine action in binding to and antagonizing Smoothened, a membrane transductory component. We report here that arsenicals, in contrast, antagonize the Hh pathway by targeting Gli transcriptional effectors; in the short term, arsenic blocks Hh-induced ciliary accumulation of Gli2, the primary activator of Hh-dependent transcription, and with prolonged incubation arsenic reduces steady-state levels of Gli2. Arsenicals active in Hh pathway antagonism include arsenic trioxide (ATO), a curative agent in clinical use for acute promyelocytic leukemia (APL); in our studies, ATO inhibited growth of Hh pathway-driven medulloblastoma allografts derived from Ptch^+/−p53^−/− mice within a range of serum levels comparable to those achieved in treatment of human APL. Arsenic thus could be tested rapidly as a therapeutic agent in malignant diseases associated with Hh pathway activation and could be particularly useful in such diseases that are inherently resistant or have acquired resistance to cyclopamine mimics.

Hedgehog (Hh) signaling specifies the normal growth and patterning of many embryonic structures (1–4). The Hh pathway also functions postembryonically in the homeostasis and repair of adult tissues and, when aberrantly activated, has been implicated in promoting the growth of cancers of diverse tissues and organs (5). Pathway activation is triggered upon binding of Hh ligand to its receptor Patched (Ptch), thus relieving Ptch inhibition of the seven-transmembrane protein Smoothened (Smo), which in turn permits Smo activation and, in mammals, accumulation of Smo within the primary cilium (6, 7). The primary activator of Hh-dependent transcription, Gli2, traffics continuously through the cilium (8) and thus is able to sense and transmit the activated state of Smo from the cilium to the nucleus by accumulating in the nucleus and activating target genes such as Ptch and Gli1. Gli1 contributes to full activation of Hh transcriptional targets, whereas negative feedback regulation by Ptch helps limit the spatial range and magnitude of pathway activity (4).

Roles for the Hh pathway in cancer growth have been attributed to activity within primary cancer cells or within supporting stromal cells (5). Pathway activation within cancer cells has been established for tumors of ectodermal or mesodermal origin, primarily cancers of skin, brain, and muscle. For example, Gorlin syndrome patients carry heterozygous mutations at the Ptch locus (9, 10), and pathway activation by loss of the remaining functional allele of Ptch can result in basal cell carcinoma and medulloblastoma. Activating mutations of Smo also have been reported in sporadic basal cell carcinoma (11, 12), and mutations in the intracellular regulator Suppressor of fused (Sufu) are associated with an increased prevalence of medulloblastoma and rhabdomyosarcoma (13–15). In addition, the Gli1 gene is amplified in many sarcomas (16, 17) and is expressed at high level in association with the EWS/FLI fusion oncogene that causes Ewing sarcoma (18, 19), thus causing activation of Hh pathway targets. Hh pathway activity also appears to be required for the maintenance of cancer stem cells in multiple myeloma (20), chronic myeloid leukemia (21, 22), and perhaps other hematologic malignancies (23).

Evidence for a contributory role of Hh pathway activation in stromal cells has been presented in cancers of endodermal organs such as pancreas and colon, which produce Hh ligands that act on surrounding stromal cells to elicit secretion of factors that in turn support the growth and progression of the tumor (24). Such activity within stromal cells is not necessarily exclusive of pathway activity within the tumor cells, however, as recent work in pancreatic cancer has implicated both Gli-mediated transcriptional activity within tumor cells (25) and stromal activation of the Hh pathway (26, 27) as contributory to growth.

From the standpoint of Hh pathway blockade for cancer therapy, it is the mechanism of Hh pathway activation that limits the types of pharmacological interventions that can be applied. Cyclopamine, the first known chemical antagonist of the Hh pathway, was identified as a plant-derived teratogen that causes cyclopia, a close phenocopy of the embryonic patterning defects noted in the Sonic hedgehog (Shh) mutant mouse (1, 28, 29). Cyclopamine has been found to antagonize Hh signaling by acting as an inverse agonist for Smo, that is, by binding and stabilizing Smo in its inactive conformation (30, 31). A number of additional Hh antagonists subsequently identified in vitro in signaling assays by screening of chemical libraries appear to mimic cyclopamine in binding to Smo (32, 33), and most compete with a fluorescent derivative of cyclopamine for binding to Smo. Smo thus appears to represent the major target in screens for Hh pathway antagonists, and cyclopamine mimics or cyclopamine derivatives have been the major focus of efforts to develop drugs that target the Hh signaling pathway (34–37); one of these candidates, GDC-0449, appears to be remarkably effective in recent clinical trials (38, 39).

One potential problem with cyclopamine mimics, however, is that mutations affecting Smo can produce resistance. Thus, mutations that constitutively activate Smo render it resistant to cyclopamine and many of its mimics (11, 30, 32). In addition, a metastatic medulloblastoma patient who initially responded well to GDC-0449 treatment subsequently succumbed to the disease that relapsed in association with a spontaneous mutation in Smo that did not affect its activity but impaired its binding to the drug (40). In addition, cancer-associated pathway activation in some cases is mediated by effects on components downstream of Smo, be it loss-of-function mutations in the intracellular regulator Sufu (13–15, 41) or non-Hh pathway-mediated increase of Gli1 expression (16–19, 25).

Cyclopamine mimics thus display potential limitations that might be circumvented by Hh antagonists that target downstream components of the pathway. Indeed, several groups recently have reported the identification of potential lead compounds in the development of such antagonists. GANT-58 and GANT-61 thus were identified as blocking the Hh pathway at the level of Gli transcription factors and were shown to inhibit xenograft tumor growth from the 22Rv1 prostrate cancer cell line (42). Zerumbone and Physalin F and B are natural products identified as inhibitors of Gli1-mediated transcription (43), and additional Hh antagonists that target Gli (HPI-1 for both Gli1 and Gli2 and HPI-2 for Gli2) were identified in a third screen (44).

In the context of Hh pathway antagonism for cancer therapy, we found it interesting that arsenic compounds induce a number of malformations in developing embryos, including defects of the axial skeleton and of the developing limbs (45–47), the patterning of which is dependent on Hh signaling. We therefore investigated the effects of arsenicals on Hh signal response.

Results

Arsenicals Are Specific Hh Antagonists.

We tested sodium arsenite, arsenic trioxide (ATO), and phenylarsine oxide (PAO) for their effects on Hh pathway activity using an NIH 3T3 cell-based Gli reporter assay. All these arsenicals potently inhibited response to the amino-terminal domain of Shh (ShhN) (Fig. 1A and Fig. S1). ATO, the therapeutic agent for acute promyelocytic leukemia (APL), inhibited ShhN-induced pathway activation in a dose-dependent manner with an IC₅₀ of about 0.7 μM (Fig. 1B). We excluded the possibility that apparent inhibition of Hh signaling in these assays was mediated by cytotoxicity, as arsenic concentrations below 10 μM did not decrease significantly the control Renilla luciferase activity driven by the constitutively active SV40 promoter in our NIH 3T3 cell-based Gli reporter assay (Fig. S1 B and D).

Fig. 1.

Arsenicals as potent and specific antagonists of the Hh pathway. (A) Sodium arsenite, ATO, and PAO inhibited Hh reporter activity. NIH 3T3 cells transfected with an Hh-responsive reporter (Gli-luciferase) were treated with the indicated concentrations of arsenicals in the presence of ShhN. Reporter activities in the presence of arsenicals were normalized to activity in the vehicle control. (B) ATO inhibited Gli-luciferase reporter activity in a dose-dependent manner with an IC₅₀ of 0.7 μM. (C and D) Arsenic inhibits the Hh pathway but not the Ras or Wnt pathways. (C) Ras pathway activity was assayed with 4× SRE-luciferase reporter in NIH 3T3 cells cotransfected with YFP or constitutively active K-rasG12V in the presence or absence of 10 μM of sodium arsenite. As controls, Gli-luciferase reporter activities in the presence or absence of sodium arsenite were included also. (D) For Wnt assay, NIH 3T3 cells were transfected with a Wnt reporter (7× TCF/LEF-luciferase) and treated with increasing concentrations of ATO in the presence or absence of Wnt3a. In A–D, specific reporter activities were normalized to activities of a cotransfected, constitutively active Renilla luciferase reporter. All experiments were repeated more than three times. Error bars indicate SD.

To examine further the specificity of arsenic action, we examined the effects of arsenic on the Ras and Wnt pathways in NIH 3T3 cells. We found that 10 μM sodium arsenite, a concentration that fully blocks ShhN response, did not inhibit activation of a serum response element (SRE) reporter induced by the expression of activated K-Ras G12V (Fig. 1C). Similarly, transcriptional response to stimulation by Wnt3A measured with a T-cell factor/lymphoid enhancer factor (TCF/LEF) reporter was not inhibited by ATO at concentrations that inhibit ShhN response (Fig. 1D). These results exclude a general suppressive effect on signal-dependent transcriptional events as the basis for arsenic inhibition of Hh pathway response.

In addition, because arsenic treatment has been shown to activate the JNK and p38 MAPK pathways (48, 49), we tested whether these pathways are required for arsenic inhibition of Hh response. We found that arsenic inhibition was sustained in the presence of the p38 MAPK inhibitor SB203580 or in the presence of the JNK inhibitor SP600125 (Fig. S2), thus ruling out these pathways as mediators of arsenic action on Hh signal response.

ATO Targets Gli Transcription Factors.

To determine the target of arsenic action within the Hh pathway, we tested whether arsenic inhibits ligand-independent pathway activity induced by manipulation of various downstream pathway components. We found, in contrast to cyclopamine, that arsenic suppresses pathway activity caused by treatment of cells with the Smo agonist, SAG1 (32, 33) (Fig. 2A) or by expression of the constitutively activated Smo variant, SmoA1 (corresponding to human SMO M2; refs. 11, 30) (Fig. 2B). These results suggest that arsenic acts within the Hh pathway at a point downstream of Smo, and we further examined this possibility by measuring pathway activity in Sufu^−/− mouse embryonic fibroblasts (MEFs) (50). Reporter activity in these cells is high in the absence of ShhN stimulation and is not suppressed by treatment with cyclopamine. We found, however, that treatment with arsenic reduced constitutive reporter activity to an extent similar to the reduction observed in ShhN-stimulated wild-type fibroblasts (Fig. 2C).

Fig. 2.

ATO targets Gli transcription factors. (A) ATO inhibited ShhN- or SAG-induced reporter activation, whereas cyclopamine blocked ShhN-induced reporter activity and was less effective in blocking SAG-induced reporter activity. (B) ATO but not cyclopamine inhibited constitutive reporter activity produced by transfection of mSmoA1 (corresponding to human SMOM2). (C) ATO but not cyclopamine inhibited high constitutive Gli-luciferase reporter activity in Sufu^−/− MEFs. Note that cyclopamine also blocked ShhN-inducible reporter activity in Sufu^−/− MEFs cotransfected with Sufu. (D) ATO but not cyclopamine inhibited reporter activity produced by Gli1 or Gli2 cotransfection. All experiments were repeated more than three times. Error bars indicate SD.

These results suggest possible arsenic action on the Gli transcriptional effectors, and we tested this possibility by examining reporter activity in cells transfected with low concentrations of Gli1 or Gli2 expression constructs. Such cells displayed a higher level of constitutive reporter activity as well as additional reporter activity that could be induced by treatment with ShhN. We found that cyclopamine blocked inducibility by ShhN and reduced this activity to a level similar to that produced constitutively by Gli2 or Gli1 expression; arsenic treatment, in contrast, reduced both constitutive and ShhN-induced activity to a considerably lower basal level (Fig. 2D). In sum, these results suggest that arsenic inhibits pathway activation by suppressing both the inducible and constitutive activity of Gli transcriptional effectors.

Arsenic Blocks ShhN-Induced Accumulation of Gli2 in the Primary Cilium.

Stimulation of cells by Shh causes accumulation of Smo within the primary cilium (6, 7), and this accumulation of Smo is followed by ciliary accumulation of Gli2, resulting ultimately in nuclear accumulation of Gli2 and activation of Hh target genes (8, 51). If arsenic acts at the level of Gli transcriptional effectors, we might expect that it would affect ciliary accumulation of Gli2 without affecting the ciliary accumulation of Smo. To examine this possibility, we made use of NIH 3T3/HA-Gli2 cells, which express very low levels of HA-tagged Gli2 that appears to behave like endogenous Gli2 (8). These cells are responsive to Shh stimulation and are sensitive to cyclopamine as well as to arsenic inhibition (Fig. S3A). We found by quantitative immunofluorescence microscopy that in ShhN-stimulated cells treated with ATO Smo accumulates normally in the primary cilium, whereas Gli2 fails to do so (Fig. 3 A and B). We further noted upon long-term incubation with ATO that the levels of HA-Gli2 in these cells, although still detectable, were dramatically reduced (Fig. 3C). The effect on ciliary trafficking occurs by 6 h (Fig. S3B), before the effect on overall Gli2 levels is apparent, indicating that the reduced ciliary accumulation of Gli2 in ShhN-stimulated cells treated with arsenic is not caused by an overall reduction in Gli2; both effects, however, could contribute to inhibition of Hh-induced target gene transcription. All these data are consistent with a primary effect of arsenic treatment on Gli transcriptional effectors.

Fig. 3.

ATO affects ciliary trafficking and stability of Gli2. (A and B) ATO blocks ShhN-induced ciliary accumulation of Gli2 but not Smo. NIH 3T3/HA-Gli2 cells were incubated in the presence or absence of ShhN and ATO for 20 h, as indicated. (A) Cells were stained to visualize Smo, Gli2, acetylated tubulin or detyrosinated tubulin, and DNA (DAPI). (B) The levels of ciliary Smo and Gli2 were quantified from immunofluorescence images. Error bars indicate SD. (C) Prolonged treatment with ATO destabilized Gli2 protein. NIH 3T3/HA-Gli2 cells were incubated in the presence or absence of ShhN and ATO for up to 30 h, as indicated. Gli2 repressor, present in the resting state, was dramatically reduced by 6 h of ShhN treatment, whereas the levels of full-length Gli2 remained stable. In the presence of ATO, Gli2 full-length protein levels began to decrease by 20 h. IP, immunoprecipitation; WB, Western blot.

Arsenic Inhibits Growth of Hh Activity-Induced Medulloblastoma Allografts.

Given the ability of arsenic to inhibit Hh pathway activity in vitro, we tested the in vivo applicability of arsenic as a therapeutic agent for a Hh pathway-dependent tumor in a flank allograft model. Primary medulloblastomas that arose spontaneously in Ptch^+/−p53^−/− mice were grafted into athymic nude mice, and tumor growth was measured with ATO or control treatments. Allografted tumors grew continuously in vehicle-treated animals, whereas ATO treatment inhibited tumor growth in a dose-dependent manner, blocking the growth almost completely at the highest dose used (Fig. 4 A and B). We also tested effects of ATO treatment on the growth of established tumors with treatment beginning when tumors reached sizes close to but smaller than 125 mm³. In comparison with control treatment, ATO significantly delayed the growth of tumors. Cyclopamine given i.p. daily at 25 mg/kg body weight also delayed tumor growth but was less efficient than ATO (Fig. S4). These results demonstrate that ATO inhibits growth of Hh pathway-dependent tumors in vivo.

Fig. 4.

Arsenic inhibits growth of Hh activity-induced medulloblastoma allografts in nude mice. (A and B) Primary medulloblastomas from Ptch^+/−p53^−/− mice were dissected and grafted into athymic nude mice. The mice were injected i.p. daily with PBS (control) or three different doses of ATO. Representative mice photographed at day 23 are shown in A. ATO suppressed tumor growth in a dose-dependent manner. The changes in average tumor volumes are shown as a function of time in B. (n = 4 per group; *P < 0.05; **P < 0.01). Error bars show SD.

To investigate further the feasibility of therapeutic inhibition of Hh pathway activity by arsenic treatment, we compared arsenic levels effective in the treatment of the mouse flank medulloblastoma model with levels achieved in human patients. Blood plasma levels of arsenic in human APL patients were reported to peak at 6.9 μM 4 h after i.v. ATO administration at a dosage of 0.15 mg/kg body weight infused over 2–3 h (52). In our experiments, arsenic levels were measured in mouse sera collected at various times after ATO administration by i.p. injection at 10 mg/kg body weight, the highest dosage used in treatment of mouse flank medulloblastoma allografts. We noted a sharp peak in concentration of about 18 μM 1 h after injection (Fig. S5), 2.6-fold higher than the peak plasma levels in human patients. The differences in kinetics of arsenic appearance probably are caused by i.p. injection in our mouse experiments compared with i.v. infusion over 2–3 h in patients. Area under the curve (AUC) calculations showed that the total exposure to arsenic in mice at the 10-mg/kg dosage (72.73 μmole·h/L) is 2-fold higher than that of the human APL patient (35.6 μmole·h/L) (52).

Arsenic Inhibits Pathway Activity Mediated by Smo Mutants Resistant to Cyclopamine and Cyclopamine Mimics.

As reported above (Fig. 2B), arsenic inhibits pathway activity produced by the oncogenic, constitutively active and cyclopamine-resistant SmoA1 (corresponding to human SMO M2). Other cyclopamine-resistant Smo mutations of interest have arisen in the context of treatment with the cyclopamine mimic, GDC-0449. Such a mutation (human SmoD473H) occurred in a metastatic medulloblastoma patient who initially responded well to GDC-0449 but subsequently succumbed when the disease relapsed in association with the Smo mutation, which does not affect Smo activity but impairs its binding to the drug (40). A resistance-conferring mutation also arose at the corresponding mouse residue (SmoD477G) in a mouse medulloblastoma treated with GDC-0449. We tested the response of SmoD477G to inhibition by GDC-0449, cyclopamine, and ATO using a Gli-luciferase reporter assay in the 4C20 Smo^−/− cell line established from MEFs (53, 54). We found that introduction of SmoD477G restored ShhN signal response to these cells (Fig. S6), but these cells displayed distinct sensitivities to Hh pathway antagonists. ShhN-stimulated 4C20 cells expressing SmoD477G thus displayed IC₅₀ values more than 40-fold greater for GDC-0449, 8.5-fold greater for cyclopamine, and 1.2-fold greater for ATO than 4C20 cells expressing wild-type Smo (Fig. 5). These results support the idea that Hh pathway antagonists acting at targets downstream of Smo have significant therapeutic potential against tumors with intrinsic or acquired resistance to cyclopamine mimics.

Fig. 5.

Arsenic suppresses ShhN-induced pathway activity mediated by the cyclopamine- and GDC-0449–resistant mutant SmoD477G. The effects of GDC-0449 (A), cyclopamine (Cyc) (B), and ATO (C) on ShhN-induced Gli-luciferase reporter activity were measured in Smo^−/− cells transfected with Smo or SmoD477G. Although both Smo and SmoD477G rescued ShhN-inducible reporter activity in Smo^−/− cells (Fig. S6), this activity was differentially sensitive to inhibition. GDC-0449 thus did not significantly suppress SmoD477G-mediated reporter activity at concentrations 40-fold above its IC₅₀ for Smo (A), whereas cyclopamine suppressed SmoD477G-mediated activity with an IC₅₀ 8.4-fold higher than that of Smo (B). ATO suppression of reporter activity mediated by Smo and SmoD477G was indistinguishable (C). Error bars indicate SD.

Combined Effect of Arsenic and Cyclopamine.

Given that arsenic acts on Gli transcriptional effectors, we considered the possibility that combined treatment with cyclopamine or a cyclopamine mimic might produce more potent pathway inhibition at lower drug doses. We found in cell-based signaling assays that the concentration of ATO required for 50% pathway inhibition is reduced from the IC₅₀ of ATO alone by 5-fold or 12-fold, respectively, in the presence of cyclopamine at 0.13 μM and 0.25 μM (Fig. S7A). Conversely, the presence of ATO at 0.5 μM or 1.0 μM reduced the concentration of cyclopamine required for 50% inhibition by 2-fold and 7-fold, respectively (Fig. S7B). Combinations of drugs that target distinct components within the Hh pathway thus appear to permit greater pathway inhibition at lower drug concentrations.

Discussion

Mechanism of Arsenic Action.

Treatment with various arsenicals inhibits Hh-dependent transcriptional activity at concentrations that do not affect transcription from a constitutively active promoter. Arsenicals also do not inhibit transcriptional activity triggered by activation of the Ras or Wnt signaling pathways and appear not to act on Hh response via previously established effects on JNK or p38 MAPK. Thus, although the effects of arsenic could be pleiotropic, its inhibition of Hh response is not caused by a general effect on cell health and is independent of several other signaling pathways examined.

Within the Hh pathway, we have shown that arsenic inhibits transcriptional activation of Hh targets whether induced by Hh ligand, by expression of a constitutively active Smo, by loss-of-function of Sufu, or by expression of Gli1 or Gli2. Given that arsenic treatment does not affect general transcription and either acts independently or fails to act on several other signaling pathways tested, the simplest interpretation of our data is that arsenic acts by targeting the Gli transcriptional effectors. Consistent with this possibility, we note that Hh pathway blockade by arsenic treatment is accompanied by a sharply reduced ciliary accumulation of Gli2. Gli2 normally undergoes continuous entry and exit from the cilium, and this trafficking appears to be required for its activation (8). The reduced ciliary accumulation upon Hh stimulation in the presence of arsenic suggests a reduced rate of ciliary trafficking and provides a mechanism for pathway inhibition. In addition, arsenic treatment in the long term causes a reduction of overall levels of Gli2.

The basis for arsenic action on the Gli transcriptional effectors could be a direct interaction with the Gli zinc fingers. Arsenic interacts readily with the sulfhydryl groups of diverse proteins and peptides (55), particularly di- and trithiols (56), and has been shown to interact with the sulfhydryls of peptides derived from the zinc fingers of estrogen receptor, PML, PML-RARα, and other zinc finger proteins (56, 57). Such an interaction with the zinc fingers of Gli proteins could affect their normal structure, thus disrupting their interaction with other proteins or DNA. Alternatively, arsenicals are known to interact with tubulin, and this interaction could antagonize Hh response (8, 58, 59) and affect the ability of Gli proteins to traffic to the cilium without perturbing Smo trafficking to the cilium (8). With longer-term treatment, arsenic causes a reduction in the levels of Gli2, and this effect also could be mediated by disruption of structure of Gli, leading to proteasomal degradation. Arsenic also causes degradation of a number of other proteins, including the viral transactivator Tax, the Akt kinase, PML-RARα oncoprotein, and other oncogenes such as AML1/MDS1/EVI1 (60–62). We note that the inhibition of Gli2 ciliary accumulation temporally precedes the overall reduction in Gli2 levels. This observation suggests that the primary Hh pathway-inhibitory mechanism of arsenic treatment may be a block of Gli2 ciliary trafficking, but that destabilization of Gli2 also may contribute to pathway inhibition in the long term.

Therapeutic Potential of Arsenic as a Hedgehog Antagonist in Cancer.

Cyclopamine mimics are currently in wide development in the pharmaceutical industry (36, 37, 63, 64) but are not expected to be effective in diseases in which pathway activation involves mutations in components downstream of Smo (13–15, 41) or in which pathway activity results from increased expression or activity of the Gli1 proteins (16–19, 25). In addition, mutations that cause constitutive Smo activity can render Smo resistant to the action of cyclopamine mimics (11, 30), or resistance can arise in the setting of treatment with a cyclopamine mimic, as has been documented for a human medulloblastoma patient (38, 40). In all these circumstances a Gli antagonist would have the virtue of inhibiting pathway activity.

Several other Hh pathway antagonists have been reported to act at the level of the Gli transcriptional effectors, including the GANT and HPI series of synthetic small molecules (42, 44) and several natural products (43). All these molecules show promise but will require significant additional development and preclinical testing before they can be used in patients. The arsenical ATO, which we have characterized here as a Gli antagonist, already has been used extensively in the setting of APL. Furthermore, the concentrations of ATO required for full inhibition of Hh pathway activity in an in vitro cell-based signaling assay (4–8 μM) are comparable to the peak levels measured in blood plasma (5.5–7.3 μM) of patients undergoing the standard regimen of ATO treatment for APL (52). Our measurement of serum arsenic levels in mice treated with ATO at 10 mg/kg showed peak levels 2.6-fold higher and AUC levels 2-fold higher than those achieved in patients undergoing ATO treatment for APL. In our tumor growth studies we noted significant inhibitory effects of ATO treatment not only at 10 mg/kg but also at 5 mg/kg and 2.5 mg/kg. It is also worth noting that we administered ATO in a single i.p. injection rather than by i.v. infusion over 2–3 h as done in human patients (52). The method of treatment we used in mice could be responsible for the very rapid accumulation observed in serum and perhaps also for more rapid clearance, thus necessitating higher dosages in mice than might be necessary with administration by i.v. infusion. It thus seems likely that blood arsenic levels comparable to levels achieved with the recommended therapeutic regimen in human patients may be adequate for Hh pathway inhibition, although it remains possible that a higher dosage may be necessary to achieve optimal Hh pathway suppression.

Exposure to arsenicals has been associated with increased cancer incidence (65), and a recent report has suggested that this increased incidence may result from arsenic-induced augmentation of Hh pathway activity (66). This latter report dealt with effects of arsenic concentrations similar to those used here but examined only effects on cells in the absence of Hh treatment. We were unable to reliably observe arsenic stimulation of reporter activity in the absence of Hh stimulation (e.g., Fig. S1), perhaps because of differences in our culture or assay conditions. It also is possible that arsenic treatment over the extended periods described in this work (66) could impact Hh pathway activity through mutagenesis and other indirect mechanisms. Of perhaps the greatest importance for the potential use of arsenic as a Gli antagonist in cancer therapy, all tested arsenicals always produced a dramatic reduction in pathway activity in cells with an activated Hh pathway, whatever the means of that activation. In addition, no increase in tumor incidence has been reported in APL patients undergoing ATO therapy (67).

As a final consideration for the potential use of ATO as a cancer therapeutic in Hh pathway-associated malignancies, we note that the combined effect of arsenic with cyclopamine in cultured cell-signaling assays permitted equivalent or more potent pathway inhibition at lower drug concentrations (Fig. S7). This effect may result from the action of these agents at distinct points within the pathway, i.e., Gli proteins in the case of ATO and Smo in the case of cyclopamine. In addition to the benefits of improved pathway inhibition, the lower drug levels that could be used in such combinations might be expected to reduce toxicities associated with effective use of a single drug. Such a combination strategy with ATO could be tested in patients once cyclopamine mimics emerge from the drug development process (64), or perhaps sooner using the FDA-approved drug itraconazole, which recently has been shown to antagonize pathway activity at the level of Smo (59).

Materials and Methods

Hh, Wnt, and Ras Reporter Assays.

For Hh signaling assay, NIH 3T3 cells transfected with Gli-luciferase reporter, control pRL-SV40 Renilla luciferase (Clontech), and other DNA constructs were incubated with ShhN-conditioned medium combined with additional treatments as indicated. For Wnt assay, NIH 3T3 cells transfected with a Wnt reporter (7× TCF/LEF-luciferase) and control Renilla luciferase were treated with Wnt3a-conditioned medium in the presence or absence of inhibitors. Ras activity was assayed in NIH 3T3 cells that were cotransfected with a control YFP DNA or constitutively active K-rasG12V with an SRE reporter (4× SRE-luciferase).

Growth Inhibition of Medulloblastoma Allograft.

Medulloblastomas from Ptch^+/−p53^−/− mice were grafted to athymic nude mice. PBS or indicated amounts of ATO or cyclopamine were injected i.p. daily. Tumor volumes were calculated by considering the average value of tumor width and length as a radius of sphere and using the volume of sphere formula V = 4/3πr³.

Further information is available in SI Materials and Methods.