Statins Synergize with Hedgehog Pathway Inhibitors for Treatment of Medulloblastoma

Cholesterol is a key component of cellular membranes and a precursor for steroid hormones and bile acids. The brain is the most cholesterol-rich organ, containing approximately 20% of total body cholesterol. In the brain, cholesterol is primarily generated by de novo synthesis, as the blood–brain barrier in vertebrates prevents its uptake from the bloodstream (1). Almost all cholesterol in the brain is present in an unesterified, free form. Cholesterol is synthesized through a series of approximately 30 enzymatic reactions, starting with the substrate acetyl CoA (Fig. 1A). The enzyme 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCR) is rate limiting for cholesterol biosynthesis. HMGCR converts HMG-CoA to mevalonate and is the target of statins (2). NAD(P) steroid dehydrogenase-like (NSDHL), a 3β-hydroxysterol dehydrogenase, is involved in the removal of two C-4 methyl groups in one of the later steps of cholesterol biosynthesis. Deficiency in NSDHL dramatically impairs cholesterol biosynthesis in both humans and mice (3). In particular, conditional deletion of Nsdhl gene in cerebellar granule neuronal precursors inhibits cholesterol synthesis and dramatically represses proliferation (4). The final step in cholesterol synthesis is catalyzed by the enzyme 24-dehydrocholesterol reductase (DHCR24), which converts desmosterol to cholesterol by saturating the C-24,25 double-bond in the side chain. Triparanol, the first synthetic cholesterol-lowering drug, works by antagonizing DHCR24 (5).

In addition to its function as a component of cell membranes, cholesterol regulates many signaling pathways, including the hedgehog (Hh) pathway. The Hh pathway plays a critical role in brain development and function. In the absence of Hh ligand, the antagonizing receptor Patched1 (Ptch1) suppresses the seven-transmembrane protein, Smoothened (Smo), thus precluding downstream signaling. The interaction between Hh and Ptch1 relieves inhibition of Smo, which then triggers a cascade of events culminating in activation of the glioma-associated oncogenes 1 and 2 (Gli1 and Gli2). Gli proteins translocate into cell nuclei and promote transcription of Hh pathway target genes, including Ptch1 and Gli1 (6, 7). Extensive evidence suggests that some exogenous oxysterols, such as 20 (S)-hydroxycholesterol and 25-hydroxycholesterol, generated by cholesterol oxidation, can activate Smo by binding to a site located in the extracellular cysteine-rich domain (CRD). This indicates that cholesterol derivatives are capable of triggering Hh pathway activation in recipient cells (8–11). Two recent studies revealed that cholesterol itself synergizes with Hh ligand to activate Smo through binding to the CRD (12, 13). The physical interaction between Smo and cholesterol was further confirmed in studies of the crystal structure of Smo (14). These findings suggest that cholesterol functions as a native ligand for Smo activation and that Hh signaling may be influenced by the local availability of cholesterol.

Insufficient Hh signaling can lead to birth defects, whereas excessive activity of the Hh pathway is associated with the formation of human malignancies, including medulloblastoma, the most common malignant brain tumor in children. The Hh pathway is activated in approximately 30% of human medulloblastoma. Although the cellular and molecular basis for medulloblastoma tumorigenesis has been widely studied, the role of cholesterol in medulloblastoma tumorigenesis has yet to be elucidated.

Previously, several studies claimed that inhibition of cholesterol synthesis promoted apoptosis in medulloblastoma cells, and this effect was greater during cotreatment with the naturally occurring Hh pathway inhibitor cyclopamine (15–18). However, these effects required the use of statins at levels 1,000-fold greater than the physiologic concentrations required to block cholesterol biosynthesis (15–18). In addition, these studies relied on the use of established medulloblastoma cell lines in which the Hh pathway is now known to be downregulated (19). An additional confounding problem with these studies is that they utilized cyclopamine to inhibit the Hh pathway in cultured tumor cells, and it has now been demonstrated that, at the concentrations used, cyclopamine causes apoptosis by promoting ceramide production independently of any effects on the Hh pathway (20). Thus, the question remains open as to whether cholesterol pathway inhibitors, alone or in combination with Hh pathway inhibitors, have potential as therapeutic agents in the treatment of Hh-medulloblastoma.

Here, we demonstrate that there is enhanced cholesterol biosynthesis in Hh pathway–associated medulloblastoma (Hh-MB) and that cholesterol is required for proliferation of tumor cells in vitro as well as tumor growth in vivo. Furthermore, we show that cholesterol is necessary for maintaining Smo activity in the absence of Ptch1. Statins, antagonists of cholesterol biosynthesis, exhibited promising therapeutic effects in mouse medulloblastoma models. Moreover, statins synergized with the clinically approved Smo antagonist, vismodegib, to prevent medulloblastoma progression in vivo. These findings elucidate an important role for endogenous cholesterol in medulloblastoma tumorigenesis and demonstrate that cholesterol biosynthesis represents a novel therapeutic target for the treatment of medulloblastoma and, potentially, other Hh pathway–associated malignancies.

All mice were maintained under barrier conditions in the Laboratory Animal Facility at Fox Chase Cancer Center (Philadelphia, PA), and all experiments were performed in accordance with procedures approved by the Fox Chase Cancer Center Animal Care and Use Committee. Nsdhl^fl/fl mice were kindly provided by Dr. Gail Herman (The Ohio State University, Columbus, OH). Math1-Cre mice, Math1-CreER^T2 mice, GFAP-CreER^T2 mice, Ptch1^fl/fl mice, Ptch1^+/− mice, R26R-eYFP mice, and SmoM2 (R26SmoM2) mice were purchased from The Jackson Laboratory. CB17/SCID mice were generated by the animal facility laboratory at Fox Chase Cancer Center.

Tamoxifen (Sigma-Aldrich) was dissolved in corn oil and administrated by oral gavage at a dosage of 200 mg/kg daily during 3 subsequent days. Animals were sacrificed for further analyses after 7 days from initial tamoxifen treatment.

The following compounds were used for our in vitro and in vivo studies: simvastatin (Cayman Chemicals), atorvastatin (Cayman Chemicals), triparanol (Sigma-Aldrich), vismodegib (Chemleader Biochemical), and water-soluble cholesterol (cholesterol-methyl-beta-cyclodextrin, Sigma-Aldrich).

Hh conditioned media (Hh-CM) was prepared by transfection of pcDNA3.1 Shh-N (Addgene) plasmid into 293FT cells. Supernatant collected at 48 hours after transfection was used at 20% for treating NIH 3T3 cells.

CB17/SCID male mice of age 6 to 8 weeks were used for subcutaneous allograft establishment. Cells isolated from tumor-bearing mice in a single-cell suspension in PBS at concentration 2 × 10⁶ cells were mixed with reduced growth factor Matrigel (Corning) in a 20:80 ratio and injected subcutaneously on the flank of CB17/SCID mice. Four weeks following injection, tumor-bearing mice that satisfied initial tumor volume requirement (200–400 mm³) were randomized into experimental groups. Simvastatin DMSO stock was dissolved in a mixture of ethanol, 0.1 mol/L NaOH, and PBS and administrated in designated dose by intraperitoneal injection. Atorvastatin and vismodegib DMSO stock was dissolved in 0.5% methylcellulose 0.2% Tween-80 and administrated per os. Mixture of DMSO, ethanol, 0.1 mol/L NaOH, and PBS or DMSO and 0.5% methylcellulose 0.2% Tween-80 was used as a vehicle substance. Compounds were administrated in half of the indicated dose twice a day and tumor size measurements were taken every 2 days and tumor volume calculated as described previously (21).

Medulloblastoma cells were suspended in NB-B27 culture medium (Neurobasal, 1 mmol/L sodium pyruvate, 2 mmol/L l-glutamine, B27 supplement, and 1% penicillin/streptomycin, all from Invitrogen) and plated on poly-d-lysine–coated coverslips (Millipore). For bromodeoxyuridine (BrdUrd) incorporation assay, cells were pulsed with 10 μmol/L BrdUrd (Millipore) for 2 hours, fixed in 4% PFA, and incubated in 4.2 mmol/L MgCl₂/0.15 mol/L NaCl pH 5.0 buffer in the presence of 50 U/mL of DNase I (Sigma-Aldrich) at 37°C for 30 minutes.

Primary mouse embryonic fibroblasts (MEF) were isolated from E13.5 of Ptch1^fl/fl and SmoM2 mouse embryos. Dissociated MEFs were cultured in DMEM supplemented with 15% FBS (SH30071.03, HyClone, General Electric), 1 mmol/L sodium pyruvate, 2 mmol/L l-glutamine, 1 g/L glucose, penicillin/streptomycin 100 U/mL, and 1% nonessential amino acids (all from Life Technologies, unless noted otherwise).

NIH 3T3 and HEK293T cell lines were acquired from ATCC and cultured in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin, and 2 mmol/L l-glutamine (all from Life Technologies). All of the above cell lines were authenticated and tested for mycoplasma by the Cell Culture Facility at Fox Chase Cancer Center. NIH 3T3 cells and MEFs were starved in serum-free media for 24 hours before virus infection or Hh treatment. All in vitro studies were performed in serum-free environment.

For synergistic effects experiments, NIH 3T3 cells were starved for 24 hours before treatment with 20% Hh-CM. Hh-treated NIH 3T3 cells and freshly isolated Ptch1^+/− medulloblastoma cells were then incubated with 0.001 to 1 μmol/L water-soluble cholesterol (WSC) for 18 to 20 hours to saturate Smo, and 1 nmol/L vismodegib for additional 48 hours.

Lentiviral production was performed by utilizing Nsdhl shRNAs and scrambled shRNA (Genocopiea, MSH029806), pLUT Gli1-HA, pMD2.G, and psPAX2 (Addgene) plasmids in accordance with the standard protocol.

For histologic analyses, cerebella were removed from mice and fixed in 10% neutral-buffered formalin overnight. Samples were embedded in paraffin; 5 μm sagittal sections were prepared and stained for ABCA1 (Abcam, ab18180, 1:500), NSDHL (Proteintech, 15111-1-AP, 1:100), and SREBP2 (Abcam, ab30682, 1:100).

Frozen sections were prepared according to standard protocol (22). Briefly, tissue was fixed in 4% PFA overnight, cryoprotected in 30% sucrose overnight, and frozen in O.C.T. (Thermo Fisher Scientific). Sagittal sections (12 μm) were prepared with a cryostat (Leica). Cultured medulloblastoma cells were fixed in 4% PFA for 15 minutes.

Tissue sections or cultured cells were permeabilized, blocked with 10% normal goat serum in PBS, 0.1% Triton X-100, and incubated overnight in primary antibody mix against Ki67 (Abcam, ab15580, 1:1,000), NeuN (Abcam, ab104224, 1:200), cleaved caspase-3 (Cell Signaling Technology, CST9661, 1:500), BrdUrd (Sigma-Aldrich, B8434, 1:500), mCherry (Abcam, ab167453, 1:500), Zic1 [generous gift from Dr. Rosalind Segal at Dana Farber Research Institute (Boston, MA), 1:200], S100β (Sigma-Aldrich, S2532, 1:500), GFP (Life Technologies, A11122, 1:500), GFAP (BD, 556330, 1:500). After 2 hours of incubation in secondary antibodies, samples were counterstained with DAPI and mounted with Fluoromount-G. For fillipin staining experiments, medulloblastoma sections were incubated with 2 mg/mL filipin (Sigma-Aldrich) for 2 hours, counterstained with DRAQ5, and mounted with Fluoromount-G.

Cells and tissues were lysed in RIPA buffer (Thermo Fisher Scientific) supplemented with protease (Roche) and phosphatase (Thermo) inhibitors cocktail. Total protein lysates were separated in 8% SDS-PAGE and transferred on PVDF membrane. Membranes were probed with antibodies against Gli1 (Cell Signaling Technology, CST2534, 1:1,000), NSDHL (Proteintech, 15111-1-AP, 1:1,000), Ptch1 (Novus Biologicals, MAB41051, 1:1,000), GAPDH (Sigma-Aldrich, G8795, 1:1,000). Western blot signals were detected with SuperSignal West Pico Substrate and exposed on films.

RNA was isolated using TRI reagent (Sigma-Aldrich) in RNase-free conditions. cDNA was synthesized using oligo(dT) and Superscript II reverse transcriptase (Invitrogen). Quantitative PCR reactions were performed in triplicates using iQ SYBR Green Supermix (Bio-Rad) and the Bio-Rad iQ5 Multicolor Real-Time PCR Detection System. Primer sequences are available upon request.

Cells or tissue samples were washed with ice-cold DPBS, and lipids were extracted using a Bligh/Dyer procedure, in which 6 mL of 1:2 (v/v) chloroform:methanol was added to the sample. ²H₆-26,26,26,27,27,27-cholesterol (Sigma-Aldrich) was used as the internal standard spiked in individual samples. The cholesterol-containing chloroform layer was collected and evaporated under a gentle, dried nitrogen stream at 35°C. The dried samples were redissolved in 200 μL methanol and analyzed by ultraperformance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS). The peaks of samples were processed and analyzed using Xcalibur, version 2.1.

All studies were performed at least in three independent replicates with outcomes reflected on photographs, graphs, or film scans. Material collection time points were designed in accordance to previously published literature. During data analysis, no exclusion criteria were applied. Two-sided Student t tests were performed to determine the statistical significance of the difference in means between samples in the experiments reported. P < 0.05 was considered statistically significant. Error bars represent the SEM. Data handling and statistical processing was performed using Microsoft Excel.

The raw gene expression data used for analyzing subtype-specific pathway alterations in medulloblastoma were obtained from GEO (GSE21140) and normalized using RMA. Gene set variation analysis (GSVA; ref. 23) on the RMA normalized expression data was applied to identify pathways that are enriched in a single sample (method arguments: function=‘gsva’; mx.diff=‘FALSE’; verbose=FALSE). Over 4,900 gene sets compiled from all gene sets available in MsigDB, including KEGG and Biocarta pathways, were used for enrichment analysis. To identify pathways that differ significantly among subtypes, Kruskal–Wallis test was used to analyze the enrichment scores resulting from GSVA.

To compare the progression rates (Figs. 5A and 6A and C), we used generalized linear regression models assuming gamma family and log links. We used the following growth curve model to test for synergy:

where Y = tumor volume, D = day, V = vismodegib, S = statin, and E[Y|V,S,D] = conditional expectation (i.e., average). β₅ and β₆ test for additive drug growth effects, and the interaction term, β₇, tests for synergistic/antagonistic drug growth effects. Synergism was concluded if the direction of β₇ enhanced drug effects and P < 0.05 (24).

Human medulloblastoma comprises at least four subgroups, including Hh, Wnt, group 3, and group 4, based on distinct gene expression profiles (25, 26). Hh-MB arises from aberrant activation of the Hh pathway in the developing cerebellum. To examine pathway alterations among medulloblastoma subgroups, we analyzed transcriptomes of 103 primary human medulloblastoma by GSVA (23). This analysis uncovered 363 gene sets and pathways (GSP) that exhibit selective enrichment across the four subtypes. The significant GSPs ranking high in the list included the Hh and Netrin pathways known to play roles in Hh-MB tumorigenesis (Supplementary Table S1; Fig. 1B–D; ref. 27), In addition, we observed that cholesterol and sterol biosynthesis GSPs were associated with Hh-MB but not the other 3 medulloblastoma subgroups. The enrichment scores of both steroid and cholesterol biosynthesis pathways were higher in Hh-MB than the other 3 subtypes (Fig. 1C). Most of the genes in these two pathways are identical, and their expression levels are remarkably elevated in Hh-MB (Fig. 1D). These data suggest an unexpected association between cholesterol biosynthesis and Hh-MB.

Mice heterozygous for Ptch1 (hereafter referred as Ptch1^+/− mice) are widely used in preclinical studies for Hh-MB as 15% to 20% of mice spontaneously develop medulloblastoma in cerebella after 3 to 4 months of age (28), and this is accelerated to 100% in the absence of p53 (29). Constitutive activation of the Hh pathway as a consequence of loss of Ptch1 heterozygosity is critical for medulloblastoma formation in Ptch1^+/− mice. To investigate the possible association between cholesterol biosynthesis and tumorigenesis in Hh-MB, we examined the expression of proteins involved in cholesterol homeostasis in medulloblastoma tissue from Ptch1^+/− mice by IHC. Cholesterol homeostasis is tightly regulated by a complex protein network responsible for its import, synthesis, and export (30). As shown in Fig. 1E, expression of NSDHL was significantly upregulated in medulloblastoma tissue compared with adjacent normal cerebellar tissue, suggesting there is increased cholesterol synthesis in tumor tissue. Transcription of ATP-binding cassette transporter 1 (ABCA1), an established regulator of cellular cholesterol homeostasis, is normally driven by increased accumulation of intracellular cholesterol (31) and sterol regulator element-binding protein 2 (SREBP2) is well known to upregulate genes involved in cholesterol biosynthesis and uptake (32). The levels of both ABCA1 and SREBP2 were substantially elevated in tumors, compared with adjacent normal cerebellum (Fig. 1F–G). These data imply that there is enhanced synthesis of cholesterol in Hh-MB.

Having observed upregulation of cholesterol biosynthesis genes in Hh-MB, we investigated the possible function of cholesterol in medulloblastoma tumorigenesis. Medulloblastoma cells isolated from Ptch1^+/− mice were infected with a lentivirus carrying mCherry-tagged Nsdhl shRNAs or scrambled shRNA. At 48 hours following viral infection, NSDHL protein expression was effectively repressed by Nsdhl shRNAs (Fig. 2A). Expression of Gli1 was significantly reduced in NSDHL-deficient medulloblastoma compared with scrambled shRNA-infected medulloblastoma cells (Fig. 2A), indicating that Nsdhl knockdown impaired Hh pathway activity. We next examined proliferation of medulloblastoma cells at 48 hours following viral infection. As shown in Fig. 2B–D, approximately 15% of scrambled shRNA-infected medulloblastoma cells were positive for BrdUrd, compared with only 4% to 6% BrdUrd-positive cells among NSDHL-deficient medulloblastoma cells. These data suggest that cholesterol synthesis is critical for proliferation and Hh pathway activity in medulloblastoma cells.

We next examined proliferation of medulloblastoma cells in vivo after genetic ablation of cholesterol biosynthesis in tumor cells. Hh-MB cells specifically express the helix-loop-helix transcription factor Math1 (22). To establish an appropriate assay system, Math1-CreER^T2 mice, capable of expressing a tamoxifen-inducible Cre recombinase only in Math1-expressing cells, were crossed with Ptch1^+/− mice and R26R-eYFP mice carrying a loxP-flanked STOP cassette upstream of the eYFP gene (33). After tumors were established, Math1-CreER^T2/R26R-eYFP/Ptch1^+/− mice were treated with tamoxifen by oral gavage. As shown in Fig. 2E, most of the tumor cells positive for Zic1 (34) also expressed YFP, but no astrocytes (S100β positive) were positive for YFP (Fig. 2F). These data indicate that Math1-CreER^T2 mice can be utilized to target medulloblastoma cells in vivo.

To disrupt cholesterol biosynthesis in medulloblastoma cells, we utilized conditional Nsdhl^fl/fl knockout mice, in which a portion of the Nsdhl gene is flanked by loxP sites (4), crossed with Math1-CreER^T2 mice and Ptch1^+/− mice. Math1-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice and Nsdhl^fl/fl/Ptch1^+/− mice were orally treated with tamoxifen after tumors were fully established. At 7 days after tamoxifen treatment, medulloblastoma tissues from Math1-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice and Nsdhl^fl/fl/Ptch1^+/− littermates were examined for expression levels of NSDHL and Gli1 as well as the amount of intracellular cholesterol present in the tumor. As shown in Fig. 2G, after tamoxifen treatment, NSDHL was depleted in medulloblastoma tissues obtained from Math1-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice. Moreover, Gli1 expression was decreased in medulloblastoma tissue from Math1-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice compared with tumor tissue from Nsdhl^fl/fl/Ptch1^+/− mice, suggesting that Hh signaling was compromised in Nsdhl-deficient medulloblastoma. In addition, the amount of cholesterol in tumor tissue was reduced by 50% ± 10% after tamoxifen treatment, as revealed by mass spectrometry analysis (Fig. 2H), indicating that Nsdhl deletion effectively suppressed cholesterol biosynthesis in tumor cells. We next examined the proliferation and survival of cholesterol-deficient medulloblastoma cells in vivo. As shown in Fig. 2I–I″, cell proliferation was reduced by 50% ± 5% in Nsdhl-deficient medulloblastoma tissue compared control medulloblastoma tissue from Nsdhl^fl/fl/Ptch1^+/− mice, as shown by immunostaining for Ki67. Medulloblastoma cells that remained proliferating were predominately located in regions of tumor tissue in which cholesterol was not completely depleted based on filipin staining (Supplementary Fig. S1). It appeared that most of the Nsdhl-deficient tumor cells differentiated based on the expression of neuronal differentiation marker NeuN (Fig. 2J–J″). Increased apoptosis was also detected by cleaved caspase-3 (CC3) immunostaining of tumor tissue after Nsdhl deletion compared with that in control tumor tissue (Fig. 2K–K″). Consistent with our findings in NSDHL-deficient medulloblastoma tissues, expression of Hh pathway target genes, including Gli1, Myc-N, and Ptch1, was significantly repressed after Nsdhl deletion (Fig. 2L). The above data demonstrate that de novo biosynthesis of cholesterol in medulloblastoma cells is required for proliferation and tumor growth.

Astrocytes expressing glial fibrillary acidic protein (GFAP) were abundantly present and intermingled with tumor cells in medulloblastoma tissue (Zic1 positive, Supplementary Fig. S2A). To test whether cholesterol from astrocytes is required for medulloblastoma progression, we examined medulloblastoma cell proliferation after disrupting cholesterol synthesis in astrocytes by using GFAP-CreER^T2 mice (35). For this purpose, we generated GFAP-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice, and Nsdhl^fl/fl/Ptch1^+/− mice as a control. These mice were treated with tamoxifen by oral gavage after tumors were established at approximately 30 weeks of age. As shown in Supplementary Fig. S2B–S2D″, no obvious alterations in the proliferation, differentiation, or apoptosis were found in medulloblastoma tissue after Nsdhl deletion in astrocytes in GFAP-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice, compared with that in Nsdhl^fl/fl/Ptch1^+/− mice. These data suggest that astrocyte-derived cholesterol is dispensable for medulloblastoma growth.

To investigate the role of cholesterol in Hh pathway activation, we treated medulloblastoma cells from Ptch1^+/− mice with simvastatin in short-term cultures. Under these conditions, the Hh pathway remains active, whereas it is repressed in long-term cultures (19). At 48 hours after treatment, cells were harvested to determine the level of Gli1 expression by Western blot analysis. As shown in Fig. 3A, the expression levels of Gli1 declined in medulloblastoma cells as the concentration of simvastatin was increased. To confirm that compromised Hh pathway activity in medulloblastoma cells after treatment with simvastatin is a consequence of reduced cholesterol levels, we added WSC to the tumor cell cultures. As shown in Fig. 3A, the addition of exogenous cholesterol rescued Gli1 expression in simvastatin-treated medulloblastoma cells. These data suggest that simvastatin inhibits Hh pathway activity in medulloblastoma cells by disrupting cholesterol biosynthesis.

To identify which component of the Hh pathway is regulated by cholesterol, we investigated the effect of cholesterol on Hh ligand signaling in NIH 3T3 cells. As expected, Hh treatment significantly upregulated Gli1 expression in NIH 3T3 cells. The induction of Gli1 by Hh ligand was blocked by simvastatin or triparanol, and this was overcome by the addition of exogenous WSC to the culture medium (Fig. 3B). These data confirm that cholesterol is necessary for transduction of Hh pathway activity from ligand to target gene expression.

We next generated primary MEFs from Ptch1^fl/fl mice carrying exons 1 and 2 of the Ptch1 gene flanked by loxP sites, and from SmoM2 mice carrying loxP-flanked STOP cassette placed upstream of constitutively active Smo (36). MEFs were then infected with a lentivirus carrying Cre recombinase to delete Ptch1 or activate Smo, or with a control lentivirus carrying an empty vector. At 48 hours after infection with a Cre-bearing virus, Hh pathway activity was elevated in both models, as determined by increased levels of Gli1 expression (Fig. 3C and D). Importantly, treatment with simvastatin or triparanol dramatically inhibited Gli1 expression in Ptch1-deficient MEFs, and this inhibition was overcome by the addition of exogenous cholesterol, indicating that cholesterol is required for Hh pathway activation resulting from Ptch1 deletion. In contrast, treatment of cells harboring an activated Smo gene with simvastatin or triparanol treatment failed to reduce Gli1 expression, indicating that cholesterol is dispensable for Hh pathway signaling resulting from constitutively active Smo. Consistent with the results described above, we found that increased levels of Ptch1 caused by overexpression of Gli1 were not altered by treatment with simvastatin or triparanol (Fig. 3E). Taken together, our data demonstrate that cholesterol is required for Smo function in the Hh signaling pathway (Fig. 3F).

Because cholesterol is necessary for Hh signaling, we hypothesized that inhibition of cholesterol synthesis might suppress medulloblastoma cell proliferation. To test this idea, we examined medulloblastoma cell proliferation in the presence or absence of simvastatin or triparanol. Medulloblastoma cells isolated from Ptch1^+/− mice were briefly cultured in the presence of cholesterol inhibitors or vehicle for 48 hours, and examined for proliferation by immunocytochemistry. As shown in Fig. 4A, about 60% of medulloblastoma cells were proliferating (Ki67 positive) in control cultures, but the percentage of proliferating cells declined significantly in a dose-dependent manner as the concentration of simvastatin was increased (Fig. 4B–D). Reduced proliferation of medulloblastoma cells was also observed after treatment with triparanol (Fig. 4E). As shown in Fig. 4G–J, addition of exogenous cholesterol rescued the proliferation repressed by simvastatin or triparanol treatment (Fig. 4K), suggesting that reduced proliferation of medulloblastoma cells treated with simvastatin or triparanol was indeed due to cholesterol deficiency. We also noted that proliferation of vehicle-treated cells increased upon exposure to exogenous cholesterol, indicating that the endogenous levels are rate limiting under the experimental conditions used (Fig. 4F). These data indicate that antagonists of cholesterol biosynthesis significantly inhibit medulloblastoma cell proliferation. We next tested whether pharmaceutical inhibition of cholesterol synthesis could be utilized to treat medulloblastoma. For this purpose, we generated a medulloblastoma model by subcutaneous injection of CB17/SCID mice with medulloblastoma cells isolated from Ptch1^+/− mice, as described previously (19). This model was validated during the development of vismodegib and shown to accurately predict clinical responses in humans (37, 38). In contrast, human tumor cell lines fail to sustain Hh pathway activity; therefore, they cannot be used to test the efficacy of inhibitors of Hh pathway activity (19). After the allograft tumor volume reached 200–400 mm³, mice were treated by intraperitoneal injection of 20 mg/kg/day simvastatin, 40 mg/kg/day simvastatin, or vehicle. No significant difference in tumor growth was observed between the control and 20 mg/kg/day simvastatin treatment (P = 0.32; Fig. 5A). The volume of tumors from the group of mice treated with 20 mg/kg/day simvastatin was comparable with that from the control group after 21 days of treatment (Fig. 5B). However, tumor volume was dramatically reduced after treatment with 40 mg/kg/day simvastatin, compared with either vehicle (P < 0.001) or with 20 mg/kg/day (P < 0.001), suggesting that simvastatin at a dose of 40 mg/kg/day effectively suppressed medulloblastoma growth in vivo. To determine the mechanism responsible for suppressed tumor growth by simvastatin, we examined the proliferation and survival of tumor cells following simvastatin treatment by IHC. Proliferation of medulloblastoma cells (Ki67 positive) was markedly reduced after simvastatin treatment (Fig. 5C–C″) compared with vehicle. Moreover, simvastatin treatment promoted significant differentiation, evidenced by upregulation of NeuN in medulloblastoma tissue (Fig. 5D–D″). In addition, an increased number of apoptotic cells (CC3 positive) was observed in tumor tissues after simvastatin treatment, indicating increased apoptosis (Fig. 5E–E″). On the basis of qPCR and Western blot analyses, Gli1 mRNA and protein expression were significantly downregulated in tumors treated with 40 mg/kg/day simvastatin compared with the vehicle control (Fig. 5F–H), suggesting simvastatin compromised Hh signal transduction in medulloblastoma cells. In addition, Ptch1^+/− mice with established medulloblastoma were treated with 40 mg/kg/day simvastatin for 2 weeks. As shown in Supplementary Fig. S3A–S3D, simvastatin markedly inhibited tumor cell proliferation and promoted tumor cell differentiation in medulloblastoma tissue. Moreover, activation of Hh pathway in medulloblastoma tissue was repressed following simvastatin treatment. Together, these data demonstrate that simvastatin treatment can reduce medulloblastoma growth in vivo.

The SmoM2 mutation results in an amino acid substitution (W535L) in the seventh transmembrane domain, creating a constitutively active form of Smo by disrupting G-protein coupling (39). The SmoM2 mutation has been observed in human medulloblastoma and in basal cell carcinoma (39, 40). Hh signaling in MEF cells carrying a SmoM2 mutation was not affected by simvastatin treatment (Fig. 3D), suggesting that activation of Hh pathway by SmoM2 is not dependent on cholesterol. To test the effect of simvastatin on the growth of medulloblastoma resulting from Smo activation, we generated medulloblastoma using Math1-Cre/SmoM2 mice as described previously (41). As shown in Supplementary Fig. S4A–S4C, proliferation of medulloblastoma cells in vitro was not altered after simvastatin treatment. In the medulloblastoma allograft mouse model generated using medulloblastoma cells from Math1-Cre/SmoM2 mice, no obvious inhibition of medulloblastoma growth was observed after simvastatin treatment (Supplementary Fig. S4D–S4E). These data indicate that medulloblastoma cells carrying the SmoM2 mutation are resistant to simvastatin treatment and that the SmoM2 mutation uncouples Smo from cholesterol dependency.

Previous studies reported that oxysterols and cholesterol bind to Smo at CRD-binding site, which is distinct from the binding pocket of the conventional Smo antagonist, vismodegib (42). This led us to investigate the possible synergistic effects of statins and vismodegib in the treatment of medulloblastoma. For this purpose, mice bearing subcutaneous allograft tumors from Ptch1^+/− mice were treated with vismodegib, simvastatin, or the two in combination. Consistent with previous reports (43), treatment with vismodegib at 5 mg/kg/day reduced Hh-MB growth (Fig. 6A and B). However, combined treatment with vismodegib and simvastatin resulted in a complete inhibition of Hh-MB progression. Although 20 mg/kg/day simvastatin exhibited no inhibitory effect on medulloblastoma growth (P = 0.32), 5 mg/kg/day vismodegib in combination with 20 mg/kg/day simvastatin resulted in an additional reduction in tumor growth compared with vismodegib alone (P < 0.001). These data suggest that simvastatin and vismodegib act synergistically to prohibit tumor growth. The synergistic effect of simvastatin and vismodegib in medulloblastoma treatment was confirmed statistically as explained in the Materials and Methods section.

CB17/SCID mice carrying allografts of Ptch1^+/− medulloblastoma were also treated with another HMG-CoA inhibitor, atorvastatin, alone or combined with vismodegib. Atorvastatin treatment at 10 mg/kg/day (P < 0.001) or 40 mg/kg/day (P < 0.001) significantly repressed tumor growth compared with vehicle treatment, confirming inhibitory effects of statins on medulloblastoma growth (Fig. 6C and D). Moreover, 10 mg/kg/day atorvastatin dramatically enhanced the inhibitory effects of vismodegib on medulloblastoma growth, when compared with vismodegib treatment alone (P < 0.001), further confirming the synergistic effects of statins and vismodegib in inhibiting medulloblastoma growth.

Having observed the synergistic effects of statins and vismodegib in medulloblastoma treatment, we next investigated whether cholesterol could affect the interaction between Smo and vismodegib. For this purpose, we pretreated NIH 3T3 cells and Ptch1^+/− medulloblastoma cells with WSC and then examined the inhibitory effects of vismodegib on Hh signaling based on Gli1 protein expression. As expected, vismodegib significantly inhibited Gli1 expression in Hh-treated 3T3 cells (Fig. 6E) and Ptch1^+/− medulloblastoma cells (Fig. 6F). However, pretreatment with cholesterol abrogated vismodegib-induced inhibition of Hh signaling in a dose-dependent manner. These data suggest that at high concentrations, binding of cholesterol to Smo may prevent its interaction with vismodegib, perhaps through an allosteric effect, providing a potential basis for the synergistic effect of statins and vismodegib in medulloblastoma treatment.

Cholesterol is an essential component of plasma membranes, and it plays an important role in cell membrane organization, dynamics, and function. During normal development, cholesterol homeostasis is tightly regulated by a complex protein network involving importation, biosynthesis, exportation, metabolism, and esterification. Altered expression levels, and mutations of genes involved in the cholesterol homeostasis pathways, have been identified in many human malignancies, including glioblastoma, pancreatic cancer, and melanoma (44). Impaired cholesterol homeostasis in tumor cells influences tumor progression and dictates the sensitivity of tumor cells to chemotherapy (45), although it is still debated whether cellular cholesterol levels influence tumor initiation. Our studies reveal enhanced expression of genes associated with cholesterol biosynthesis in Hh-MB, suggesting that the intracellular cholesterol levels in tumor cells are elevated. Inactivating mutations of Ptch1, the most common mutation in Hh-MB, may also contribute to increased cholesterol levels in tumor cells, in that Ptch1 is involved in cholesterol efflux from cells (46).

Extensive evidence indicates that some exogenous oxysterols, such as 20 (S)-hydroxycholesterol and 25-hydroxycholesterol generated by cholesterol oxidation, activate Smo by binding to a site located in the extracellular CRD, demonstrating that cholesterol derivatives are capable of triggering Hh pathway activity (8–11). However, it is unclear whether these oxysterols are physiologic activators of Smo, given that the endogenous levels are significantly lower than the EC₅₀ for Hh pathway activation. In this study, we found that, upon inhibition of cholesterol biosynthesis, using proximal and distal inhibitors (simvastatin and triparanol, respectively), Hh pathway activity in medulloblastoma cells and fibroblasts was significantly repressed. Proliferation of medulloblastoma cells and tumor growth in vivo was repressed after pharmacologic or genetic inhibition of cholesterol biosynthesis. These data suggest that endogenous cholesterol biosynthesis is important for Hh pathway activity. Compromised Hh pathway activity was observed in cholesterol-deficient fibroblasts after loss of Ptch1 but not following constitutive activation of Smo or after overexpression of Gli1. These data indicate that cholesterol biosynthesis is required for Smo to function in Hh signaling. Recent findings that cholesterol synergizes with Hh to activate Smo through binding to the extracellular CRD (12, 13) are consistent with the conclusion that cholesterol, but not oxysterols, functions as the endogenous ligand responsible for Smo activation during Hh pathway signaling, and that Ptch1 inhibits Smo by counteracting its activation by cholesterol. Thus, cholesterol plays a unique role as a second messenger in vertebrate Hh signaling, by mediating the functional interaction between Ptch1 and Smo. These findings resolve a longlasting question of how Ptch1 suppresses Smo activity, and they identify cholesterol as a critical component for Smo activation by Ptch1-mediated Hh signaling. Simvastatin also inhibits proliferation of group 3 medulloblastoma, which are associated with activating mutations in c-Myc and dysfunctional p53 (47). These findings imply that cholesterol may regulate many other signaling pathways in medulloblastoma in addition to Hh signaling.

The blood–brain barrier efficiently blocks transport of cholesterol from the bloodstream into the brain. Therefore, nearly all brain cholesterol is produced via de novo synthesis (1). Recently, it has been demonstrated that tumor cells in glioblastoma multiforme, the most malignant brain tumor in adults, exclusively acquire cholesterol from astrocytes (45). In our studies, conditional inhibition of cholesterol biosynthesis in tumor cells dramatically inhibited medulloblastoma cell proliferation and promoted the differentiation of tumor cells, whereas disruption of cholesterol synthesis in astrocytes had no effects in medulloblastoma cell proliferation and survival. These observations suggest that medulloblastoma cells predominately rely on de novo synthesis of cholesterol. In our studies, cholesterol was not completely depleted after deletion of the Nsdhl gene in medulloblastoma cells in Math1-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice following tamoxifen treatment. On the basis of filipin staining, some regions of the tumor tissue were enriched with cholesterol. Such incomplete depletion of cholesterol in tumor tissue could be due to insufficient Cre recombination and/or inadequate tamoxifen penetration. However, we cannot completely exclude the possibility that tumor cells at least partially acquire cholesterol from their microenvironment. Nevertheless, proliferating medulloblastoma cells predominantly accumulated in cholesterol-enriched tumor regions of Math1-CreER^T2/Nsdhl^fl/fl/Ptch1^+/− mice, further confirming that cholesterol is critical for medulloblastoma growth.

The observation of Hh pathway dysregulation in tumors has caused it to be an attractive pharmacologic target for cancer treatment for many years. All the available Hh pathway inhibitors being studied in clinical trials target Smo. Our studies demonstrate that cholesterol biosynthesis is essential for medulloblastoma progression. Statins effectively inhibited medulloblastoma cell proliferation and suppressed tumor growth. The excellent safety profile, together with low cost and ready availability of statins, makes cholesterol biosynthesis a promising target for medulloblastoma treatment and perhaps other malignancies involving Hh signaling.

Smo inhibitors cause developmental bone abnormalities in young mice (43). Therefore, they are not recommended for use in young children. Potentially, the combined use of vismodegib and statins could permit the use of lower doses of Smo inhibitors, potentially reducing the extent of bone toxicity in the pediatric Hh-MB patient population. Currently, statins are being used to manage dyslipidemia in some children and are among the drugs approved for use in children (48, 49). So far, there have been no reports of bone toxicities or other phenotypes that could be ascribed to Hh pathway inhibition. Nevertheless, further studies are warranted to test the potential for combined toxicities of Smo inhibitors and statins in developing animal models prior to their use in children.

The potential efficacy of statins in Hh-MB therapy depends on which underlying mutation the tumor harbors in the Hh pathway. Among Hh-MB patients, approximately 80% of adult together with 50% of infant and childhood patients harbor mutations in Hh, Ptch1, or Smo, who are most likely to respond to Smo antagonists including statins (40). However, medulloblastoma carrying the SmoM2 mutation are predicted to be resistant to Smo inhibitors as well as cholesterol inhibitors. Similarly, around 45% of infant and childhood Hh-MB patients frequently harbor mutations downstream of Smo, such as Sufu or Gli2 (40), rendering these tumors intrinsically resistant to any drugs targeting Smo.

Our studies reveal the synergistic effects of statin and vismodegib in treating medulloblastoma. It has been confirmed that cholesterol activates Smo through interaction with the CRD region, which is distinct from the binding site of vismodegib. Recent structural studies found that binding with vismodegib precludes the interaction of Smo with cholesterol (14). In our studies, pretreatment with cholesterol overcame the capacity of vismodegib to inhibit Hh signaling in NIH 3T3 cells and medulloblastoma cells, implying that cholesterol compromises the interaction between vismodegib and Smo. These data suggest an indirect competition between cholesterol and vismodegib, possibly through allosteric effects, in binding to Smo.

Treatment with Smo antagonists results in dramatic regression of tumor burden and reduction of symptoms, but there is a rapid emergence of drug resistance (50). The resistance often arises as a consequence of mutations in the transmembrane domain of Smo that prevent binding of antagonists. Therefore, cholesterol inhibitors could not only synergize with vismodegib in treating medulloblastoma, but also potentially be effective in treating some drug-resistant Hh-MB.

No potential conflicts of interest were disclosed by the authors.

Conception and design: R.E. Gordon, J.M.Y. Ng, A.J. Andrews, I. Astsaturov, T. Curran, Z.-J. Yang

Development of methodology: R.E. Gordon, F. Du, J.M.Y. Ng, Z.-J. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.E. Gordon, L. Zhang, Y.-M. Kuo, J.M.Y. Ng, Z.-J. Yang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.E. Gordon, S. Peri, Y.-M. Kuo, B.L. Egleston, J.M.Y. Ng, T. Curran, Z.-J. Yang

Writing, review, and/or revision of the manuscript: R.E. Gordon, L. Zhang, Y.-M. Kuo, B.L. Egleston, J.M.Y. Ng, T. Curran, Z.-J. Yang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.E. Gordon, F. Du, J.M.Y. Ng, T. Curran, Z.-J. Yang

Study supervision: R.E. Gordon, J.M.Y. Ng, T. Curran, Z.-J. Yang

The authors would like to thank Drs. G.E. Herman and D. Cunningham (The Ohio State University, Columbus, OH) for providing Nsdhl^fl/fl mice, Drs. T. Yen and E. Golemis for thoughtful comments, and A. Efimov, E. Nicolas, F. Jin, and J. Oesterling for technical assistance. This work was also supported by NCI (CA178380, CA185504), ACS RSG (RSG1605301NEC), PA CURE Health Research Fund, Bucks County Chapter, and National Natural Science Foundation of China (81572724, 81573449). This research was supported by the subsidy of the Russian Government to support the Program of competitive growth of Kazan Federal University to Renata E. Gordon.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.