MYC-Regulated Mevalonate Metabolism Maintains Brain Tumor–Initiating Cells

Glioblastoma is the most prevalent and lethal primary intrinsic brain tumor (1). Despite standard-of-care therapy consisting of maximal surgical resection followed by concurrent chemoradiation and adjuvant chemotherapy, glioblastomas remain universally fatal with median survival of less than 2 years, even in patients enrolled in clinical trials (2). Glioblastoma represents one of the most extensively characterized solid cancers, yet the molecular understanding of these cancers has not translated into clinically meaningful therapeutic benefit. Among molecularly targeted therapies, only bevacizumab (Avastin) has been approved by the FDA for glioblastoma therapy, but bevacizumab does not extend overall survival in an unselected patient population (3). Glioblastomas can be segregated based on mutational and transcriptional profiles, but patient outcome and response to therapy have been minimally informed based on these molecular features, beyond a favorable outcome for patients harboring mutations in IDH1 or IDH2, representing fundamentally distinct cancers. Treatment failure in glioblastoma is derived from myriad causes, but prominently stem-like tumor cells, brain tumor–initiating cells (BTIC), or glioma stem cells promote tumor growth through sustained proliferation, self-renewal, therapeutic resistance, angiogenesis, and invasion of normal brain (4, 5). Although BTICs remain controversial due to unresolved questions of cell-of-origin and identification methods, glioblastoma remains one of the most reliable solid cancers for the study of cancer stem cells (6–8). Cancer stem cells, like normal stem cells, are defined by their functional characteristics, including stem cell marker expression, self-renewal, and recapitulation of the tissue from which they are derived (in this case, a heterogeneous tumor; ref. 9). Due to their malignant behavior, therapies targeting cancer stem cells might have significant impact on patient outcome.

MYC is both a central regulator of stem cell biology and a contributor to the genesis of many human cancers, as one of the most common oncogenic events observed in cancer (10, 11). MYC mediates its effects in cancer through its roles as a transcription factor and chromatin regulator, with increasing evidence of a central role in metabolic reprogramming, the Warburg effect (10). Metabolic effects of MYC manifest in induction of glycolysis and glutaminolysis, as well as anabolic support of tumor growth (11). MYC promotes sustained metabolic changes with augmented mitochondrial and ribosome biogenesis (10). In glioma, MYC is over expressed in about 70% of tumors (12, 13). We previously showed that human BTICs preferentially express MYC and depend on its activity for self-renewal and tumor growth (14). Concurrently, the DePinho group showed that a genetically engineered mouse glioma model driven by mutation of Pten and p53 contained sphere-forming cells dependent on MYC (15). Thus, MYC appears to be an important target in BTIC biology. The role of MYC in glioblastoma biology is further supported by a role of TORC2 in controlling glycolytic metabolism through MYC upregulation (16). Despite its importance, or perhaps because of it, MYC has not proven amenable to effective therapeutic targeting, supporting the need for alternative targeting strategies.

The mevalonate metabolic pathway provides cells with bioactive molecules crucial for multiple cellular processes, including cell proliferation, differentiation, and survival (17). The mevalonate pathway produces cholesterol and coenzyme Q, as well molecules involved in signal transduction, including isoprenoid intermediates: geranylgeranyl pyrophosphate (GGPP), and farnesyl pyrophosphate (FPP; refs. 18–20). Hundreds of proteins, including ras family members, undergo geranylgeranylation or farnesylation, permitting proper subcellular localization required for function. The rate-limiting step in mevalonate biology is the enzyme, 3-hydroxyl-methylglutaryl coenzyme A reductase (HMGCR), which is inhibited by statins (18). As would be expected by its central role in metabolism, mevalonate has been implicated in numerous aspects of tumor development and progression (21). Some cancers have deficient feedback control of HMGCR or increased HMGCR expression and activity, leading to increased mevalonate metabolism (22). Epidemiology studies have suggested that patients prescribed statins for cholesterol control are at lower risk for cancer, and preclinical studies have shown that statins exhibit antitumor effects against various leukemia and solid tumor cells through HMGCR inhibition (19, 20). Thus, statins have been proposed as antineoplastic agents with acceptable toxicity based on millions of patients treated with these agents.

To date, metabolic crosstalk between the MYC and mevalonate pathways has been poorly understood. Based on the function of MYC in glioblastoma and the lack of effective therapies for this cancer, we interrogated the interplay between MYC and mevalonate in BTIC biology.

Simvastatin (S6196; Sigma-Aldrich), 6-Fluoromevalonate (F2929; Sigma-Aldrich), geranylgeranyl transferase I inhibitor (GGTI-298; Sigma-Aldrich), and U0126 (Cell Signaling Technology) were prepared in DMSO. A farnesyl transferase inhibitor (FTI-277; Sigma-Aldrich) was diluted in PBS. BMP4 (B2680; Sigma-Aldrich) was diluted in PBS. For in vivo studies, simvastatin (50 mg/kg) or vehicle control (DMSO) was administered intraperitoneally one time per day. Antibodies used for immunoblotting and chromatin immunoprecipitation (ChIP) included HMGCR (Abcam; ab174830), FDPS (Abcam; ab38854), MYC (N-262; Santa Cruz Biology #sc-764), total ERK (Cell Signaling Technology #9102), phospho-ERK (Cell Signaling Technology #9101), OLIG2 (Milipore; MABN50), Autophagy Antibody Sampler Kit (Cell Signaling Technology #4445), and α-tubulin (Sigma #T6074). ChIP was performed using the EZ-ChIP chromatin immunoprecipitation Kit (EMD Millipore #17-371). The apoptosis assay was performed using an Apoptosis/Necrosis Detection kit (blue, green, red; Abcam; ab176749). Senescence was analyzed by the Senescence β-Galactosidase Staining Kit (CST, 9860). In separate experiments to study the effects of miR-33b, BTICs were transduced with either MISSION microRNA Mimic hsa-miR-33b (Sigma-Aldrich; HMI0500) or, upon simvastatin treatment, MISSION synthetic human miRNA inhibitor has-miR-33b (Sigma-Aldrich; HSTUD0500), and the cells were collected after 48 hours for additional assays. The hsa-miR-33b primer set (Thermo Fisher Scientific) was used to detect miR-33b expression.

Glioblastoma tissues were obtained from excess surgical resection samples from patients at the Cleveland Clinic after neuropathology review with appropriate consent, in accordance with an Institutional Review Board–approved protocol (CCF 2259). The patient studies were conducted in accordance with the Declaration of Helsinki. The BTICs were derived from human specimens at Duke University around 2005 to 2006. The normal neural cell lines were derived from human specimens at Cleveland Clinic around 2015 to 2016. To prevent culture-induced drift, patient-derived xenografts were generated and maintained as a recurrent source of tumor cells for study. Immediately upon xenograft removal, a papain dissociation system (Worthington Biochemical) was used to dissociate tumors according to the manufacturer's instructions. Cells were then cultured in Neurobasal medium (Life Technologies) supplemented with B27, l-glutamine, sodium pyruvate, 1% penicillin/streptomycin, 10 ng/mL bFGF, and 10 ng/mL EGF (R&D Systems) for at least 6 hours to recover expression of surface antigens. As no marker is uniformly informative for BTICs, we employed a combination of functional criteria to validate BTICs. Both BTICs and differentiated glioma cells (DGC) were derived immediately after dissociation or after transient xenograft passage in immunocompromised mice using prospective sorting followed by assays to confirm stem cell marker expression, sphere formation, and secondary tumor initiation. For experiments using matched BTIC and DGC cultures, we segregated AC133 marker-positive and marker-negative populations using CD133/1 antibody-conjugated magnetic beads (Miltenyi Biotech), as previously described (5, 23). The BTIC phenotype was validated using stem cell marker expression (OLIG2 and SOX2), functional assays of self-renewal (serial neurosphere passage), and tumor propagation by in vivo limiting dilution. Cells were tested for Mycoplasma once every month. Cells were grown at 75% confluency for at least 48 hours. Media were then collected and PCR amplified using the following primers: MP1s- ACACCATGGGAGCTGGTAAT; MP1a- GTTCATCGACTTTCAGACCCAAGGCAT. Cells tested positive for mycoplasma contamination were either discarded or treated with Plasmocin (1:1,000; Invivogen, ant-mpp) for 2 weeks as described by the manufacturer, after which successful eradication of mycoplasma was again confirmed by PCR amplification. STR analyses were performed on each tumor model used in this article every 6 months for authentication. All cells were thawed within 1 month of these experimental procedures.

Lentiviral clones expressing two, nonoverlapping shRNAs directed against human c-Myc (TRCN0000039640, TRCN0000039642), human HMGCR (TRCN0000233114, TRCN0000233115), or a nontargeting control shRNA that has no targets in human genomes (shCONT: SHC002) were obtained from Sigma-Aldrich. shRNAs with nonoverlapping sequences that had the best relative knockdown efficiency were used for all experiments. A lentiviral overexpression plasmid for MYC was generated by cloning the ORF with the N-terminal Flag into the pCDH-MCS-T2A-Puro-MSCV vector (System Biosciences; ref. 24). Lentiviral particles were generated in 293FT cells in Neurobasal complete medium (Life Technologies) with cotransfection of the packaging vectors pCMV-dR8.2 dvpr and pCI-VSVG (Addgene) using a standard calcium phosphate transfection method.

Cell proliferation was measured using CellTiter-Glo (Promega). All data were normalized to day 0 and presented as mean ± SEM. Neurosphere formation was measured by in vitro limiting dilution, as previously described (23, 25). Briefly, decreasing numbers of cells per well (50, 20, 10, 5, and 1) were plated into 96-well plates. Seven days after plating, the presence and number of neurospheres in each well were recorded. Extreme limiting dilution analysis was performed using software available at http://bioinf.wehi.edu.au/software/elda, as previously described (23, 25).

Total cellular RNA was isolated using Trizol reagent (Sigma Aldrich), followed by reverse transcription into cDNA using the qScript cDNA Synthesis Kit (Quanta BioSciences). Real-time PCR was performed using Applied Biosystems 7900HT cycler using SYBR-Green PCR Master Mix (Thermo Fisher Scientific). Sequences for gene-specific primer sets were as follows: human HMGCR forward 5′-TGATTGACCTTTCCAGAGCAAG-3′ and reverse 5′-CTAAAATTGCCATTCCACGAGC-3′; human PMVK forward 5′- CCTTTCGGAAGGACATGATCC-3′ and reverse 5′-TCTCCGTGTGTCACTCACCA-3′; human MVK forward 5′-CATGGCAAGGTAGCACTGG-3′ and reverse 5′-GATACCAATGTTGGGTAAGCTGA-3′; human MVD forward 5′-CTCCCTGAGCGTCACTCTG-3′ and reverse 5′-GGTCCTCGGTGAAGTCCTTG-3′; human IDI1 forward 5′-AACACTAACCACCTCGACAAGC-3′ and reverse 5′-AGACACTAAAAGCTCGATGCAA-3′; human FDPS forward 5′-TGTGACCGGCAAAATTGGC-3′ and reverse 5′-GCCCGTTGCAGACACTGAA-3′; c-Myc forward 5′-GGCTCCTGGCAAAAGGTCA-3′ and reverse 5′-CTGCGTAGTTGTGCTGATGT3′; and 18S RNA forward 5′-AACCCGTTGAACCCCATT-3′ and reverse 5′-CCATCCAATCGGTAGTAGCG-3′. For miR-33b detection, total cellular RNA was isolated using Trizol reagent (Sigma Aldrich), followed by TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific, 4366596). Real-time PCR was performed using Applied Biosystems 7500HT cycler using TaqMan Universal PCR Master Mix, no AmpErase UNG (Thermo Fisher Scientific). Sequences for the miR-33b primer set are to be found with hsa-miR-33b (Thermo Fisher Scientific).

Cells were collected and lysed in RIPA buffer (50 mmol/L Tris-HCl, pH 7.5; 150 mmol/L NaCl; 0.5% NP-40; 50 mmol/L NaF with protease inhibitors) and incubated on ice for 30 minutes. Lysates were centrifuged at 14,000 rpm at 4°C for 10 minutes, and supernatant was collected. Protein concentration was determined using the Bradford assay (Bio-Rad Laboratories). Equal amounts of protein samples were mixed with SDS Laemmli loading buffer, boiled and electrophoresed using NuPAGE Bis-Tris Gels (Life Technologies), and then transferred onto PVDF membranes (Millipore). Blocking was performed for 45 minutes using TBST supplemented with 5% nonfat dry milk and blotting performed with primary antibodies at 4°C for 16 hours.

Cells (4 × 10⁶) per condition were collected, and 5 μg MYC antibody (N-262; Santa Cruz Biotechnology #sc-764) or rabbit IgG was used for the immunoprecipitation of the DNA–protein immunocomplexes. Crosslinking was reversed by heating for 6 hours at 65°C, followed by digestion with proteinase K. The purified DNA was subjected to quantitative PCR using the following primer sets: HMGCR forward 5′-AGATGGTGCGGTGCCTGTT-3′ and reverse 5′-AGGAGCCCTCACCTTACGC-3′; PMVK forward 5′-ACATGTGCCACAGCAAATC-3′ and reverse 5′-GCGGCTAGAAAAGCAATCA-3′; MVK forward 5′-CGTGTCACCTCTCTGCGTTC-3′ and reverse 5′-CCTGGGACAAACTGCGTAAC-3′; MVD forward 5′-ACCTGTCTGCTGGTGCTGT-3′ and reverse 5′-TCACACAGCCCTTCGCACAG-3′; IDI1 forward 5′- CCTTGAGTGTGTTGCTGTCA-3′ and reverse 5′-CGCAGGAGACTAAGAGTTGC-3′; FDPS forward 5′-CGCGACCTCTAGACTTCAGC-3′ and reverse 5′-AGTAGTTCCCGCTCTCCCCA-3′.

Right cerebral cortex injection at a depth of 3.5 mm of NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ, The Jackson Laboratory) mice was performed using BTICs derived from human glioblastoma specimens. Treatment of mice with simvastatin was started on the seventh day after intracranial implantation of BTICs. Mice were treated with simvastatin at the dose of 50 mg/kg for body weight daily using intraperitoneal injection throughout the period of tumor growth. The control group subjects were treated with the same volume of vehicle control (DMSO) per day. All murine experiments were performed under an animal protocol approved by the Cleveland Clinic Institutional Animal Care and Use Committee. Animals were maintained until neurological signs were apparent, at which point they were sacrificed. Brains were harvested and fixed in 4% formaldehyde, cryopreserved in 30% sucrose, and then cryosectioned. Sections were stained with hematoxylin and eosin for histologic analysis. In parallel survival experiments, animals were monitored until they developed neurological signs.

No statistical methods were used to predetermine sample sizes, but the choice of sample sizes was consistent with those reported in previous publications (23, 25). All grouped data were presented as mean ± SEM. Two-sided Student t test was used to assess differences between two groups. Kaplan–Meier survival curves were generated using Prism software, and log-rank test was performed to assess statistical significance between groups. Correlation between gene expression and patient survival was performed through analysis of The Cancer Genome Atlas (TCGA) and brain tumor datasets downloaded from the TCGA data portal or NCBI GEO database. Raw data from enhancer profiling of primary glioma tissues were deposited at GSE101148. ChIP-seq data from Suva and colleagues were accessed from NCBI GEO database at GSE54792 (26) and GSE17312 (27).

The central role of mevalonate in core cellular metabolism suggested that pathway function may be linked to cellular differentiation state. We, therefore, interrogated mRNA expression levels of several mevalonate pathway genes (HMGCR, PMVK, MVK, MVD, IDI1, and FDPS) in multiple BTIC models that we previously functionally validated as having a cellular hierarchy of self-renewal, stem cell marker expression, and tumor propagation (23, 25, 28–30). Each BTIC model expressed high baseline mRNA levels of all six mevalonate genes with substantial reduction of levels upon induction of differentiation (Fig. 1A–F). As serum could contain mevalonate that would possibly modulate mevalonate enzyme expression levels, we confirmed these data by inducing differentiation in the absence of serum. We treated T387, T3691, and T4121 BTICs with 100 ng/mL of BMP4 in the absence of EGF and FGF for 1 week, then examined six mevalonate genes' mRNA levels using real-time PCR. Mevalonate genes' mRNA levels were reduced upon BMP4-induced differentiation (Supplementary Fig. S1). To determine the potential upstream regulation of mevalonate pathway enzymes, we interrogated the comparative regulation of each gene promoter in normal brain, glioblastoma patient specimens, and three pairs of BTICs and matched DGCs derived from three independent glioblastoma patients (26). Consistent with a potential role for mevalonate metabolism in glioblastoma, the promoter of each key mevalonate gene showed increased levels of active promoters marked by histone 3 lysine 27 acetylation (H3K27ac) in tumors relative to normal brain (Fig. 1G–L). Further, BTICs showed increased activating promoter marks relative to DGCs (Fig. 1G–L). Coregulation of mevalonate enzymes was additionally supported by substantial coexpression in glioblastomas, as determined by pairwise correlations (Supplementary Fig. S2). Confirming these transcriptional differences translated into differences in protein levels, elevated protein levels of mevalonate enzymes in BTICs and reduction upon differentiation were analyzed by immunoblot (Fig. 1M–O). Collectively, these results support the preferential upregulation of the mevalonate pathway in BTICs.

HMGCR catalyzes the conversion of HMG-CoA to mevalonic acid. As BTICs express elevated HMGCR mRNA (Fig. 1A) and protein levels (Fig. 1M–O), we hypothesized that HMGCR promotes BTIC maintenance. We, therefore, interrogated HMGCR function in BTICs using two nonoverlapping shRNAs to knock down HMGCR, designated shHMGCR-1 and shHMGCR-2. The efficiency of HMGCR knockdown was confirmed at the mRNA level by real-time PCR (Fig. 2A). Targeting HMGCR decreased BTICs growth and cell viability (Fig. 2B and C). Sphere formation represents a surrogate of self-renewal, albeit imperfect. Using an in vitro limiting dilution assay with BTICs expressing shCONT, shHMGCR-1 or shHMGCR-2, knockdown of HMGCR impaired self-renewal capacity, as measured by neurosphere formation (Fig. 2D and E). Sphere size, considered a surrogate of proliferation, was also reduced in BTICs expressing shRNAs targeting HMGCR compared with control shRNA (Fig. 2F). As in vivo tumor growth represents the most important BTIC phenotype, BTICs expressing shHMGCR or nontargeting shCONT were implanted into the brains of immunocompromised mice to interrogate the effect of HMGCR in tumor initiation in vivo. Mice bearing BTICs expressing shHMGCR demonstrated increased survival relative to shCONT mice, establishing the role of HMGCR in maintaining tumorigenic capacity of BTICs in vivo (Fig. 2G and H). These results establish that the HMGCR is essential to maintain BTICs proliferation, self-renewal, and tumorigenicity.

To evaluate the clinical relevance of mevalonate pathway activation in BTICs, we interrogated the TCGA glioblastoma database for the rate-limiting mevalonate enzyme, HMGCR. Befitting a role in glioblastoma biology, HMGCR levels were associated with poor prognosis in glioblastoma patients in TCGA with a mesenchymal transcriptional profile, which has been associated with tumor progression after therapy (Fig. 3A; ref. 31). Further in silico analyses of the Ivy Glioblastoma Atlas Project (Ivy GAP) and TCGA revealed that the expression of mevalonate genes was enriched in areas of infiltrating tumor and in oligodendroglioma, respectively (Supplementary Fig. S3A and S3B). To leverage these results into clinically actionable findings, we treated BTICs with statins, which function predominantly through disruption of the mevalonate pathway, to examine pathway function in BTIC maintenance. Statin (simvastatin) treatment inhibited BTIC viability and cell proliferation (Fig. 3B and C). Cell-cycle analysis of statin-treated BTICs revealed arrest in G₁ and reduction of the S phase, as measured by flow cytometry (Supplementary Fig. S4A–S4C). Using an in vitro limiting dilution assay, we demonstrated that statin treatment reduced the frequency of sphere formation (Fig. 3D–F). In contrast, statin treatment did not significantly alter cellular proliferation in matched DGCs (Supplementary Fig. S5A–S5C), supporting a dispensable role of mevalonate pathway in differentiated tumor cells. To determine the cellular mechanism by which statin therapy reduced BTIC cell numbers, we ruled out effects on autophagy (Supplementary Fig. S6) and senescence (Supplementary Fig. S7A–S7C), but we detected induction of apoptosis in statin-treated BTICs (Supplementary Fig. S8A and S8B). As all drugs have off-target effects, we investigated the effects of another mevalonate pathway inhibitor, 6-fluoromevalonate, which has a mechanism of action distinct from statins. Similar inhibitory effects on BTIC proliferation were detected with 6-fluoromevalonate treatment, supporting BTIC dependence on mevalonate metabolism (Fig. 3G–I). Collectively, these results demonstrate that the mevalonate pathway is essential specifically for BTIC growth.

Mevalonate generates isoprenoid intermediates, including GGPP and FPP (21), which bind to Ras family members and other GTP-binding proteins to facilitate proper intracellular localization and function (32). We, therefore, investigated the roles of the prenylation transferases catalyzing geranylgeranylation and farnesylation in BTIC maintenance. Befitting the activation of key oncogenic pathways, including ras family members, by prenyl posttranslational modifications, treatment with either a FTI or geranylgeranyl transferase inhibitor (GGTI) impaired BTIC viability and sphere formation (Supplementary Fig. S9A–S9F).

Having elucidated mevalonate pathway function in BTICs in vitro, we examined mevalonate function in vivo. BTICs were implanted into the brains of immunocompromised mice, and tumor-bearing hosts were treated systemically with a mevalonate pathway inhibitor. In two BTIC models derived from different patients, statin treatment extended the survival of tumor-bearing mice relative to vehicle control (Fig. 4A and B). Reduced tumor size upon statin therapy was confirmed by histology (Fig. 4C and D). Although we expected inhibition of the mevalonate pathway to block prenylation and, thus, ras activation, tumors derived from statin-treated mice demonstrated a modest increase in activated ERK levels (Fig. 4E and F).

To more thoroughly elucidate the molecular mechanism involved in mevalonate metabolism, we examined the impact of targeting mevalonate pathway function on ras activation by treated BTICs with two inhibitors with distinct mechanisms, simvastatin and 6-fluoromevalonate, and found that ERK activity was increased upon treatment, confirming the in vivo findings (Fig. 5A and B). In contrast, simvastatin treatment of normal primary neural cells failed to activate ERK, supporting a BTIC-specific link between mevalonate and ERK (Fig. 5C). ERK activity progressively increased in simvastatin-treated BTICs over a 48-hour time course (Fig. 5D). To determine if activation of ERK was involved in the mechanism of statin treatment, MEK inhibitor (U0126) treatment rescued the decreased cell viability caused by simvastatin therapy (Fig. 5E–G). These results suggest that targeting the mevalonate pathway in BTICs attenuates tumor initiation through RAS-ERK pathway activation.

Based on our prior studies demonstrating specific expression of MYC in BTICs and the role of MYC in BTIC metabolism, we hypothesized that MYC may regulate the expression of the enzymes in the mevalonate pathway. Analyzing the promoters of the six major mevalonate enzymes revealed that each contained MYC-binding sites (data not shown). To determine the potential regulation of these genes by MYC, we targeted MYC expression in BTICs using lentiviral-transduced shRNA (Fig. 6A), revealing marked reduction in the expression levels of all six mevalonate pathway genes at the mRNA (Fig. 6B–G) and protein levels (Fig. 6H), compared with control, nontargeting shRNA. As further support to these loss-of-function studies, we performed gain-of-function studies in differentiated tumor cell progeny; MYC overexpression in DGCs (Fig. 6I) induced increased expression of all six mevalonate pathway genes at the mRNA (Fig. 6J–O) and protein levels (Fig. 6P). To confirm direct binding of MYC to the relevant gene promoters, we performed ChIP followed by PCR (ChIP-PCR) for all six genes, revealing an enrichment of binding compared with the IgG control (Fig. 6Q and R). MYC regulates the expression of numerous molecular targets, so we interrogated the contribution of mevalonate metabolism in the effects of targeting MYC expression. The addition of FPP, one of the isoprenoid intermediates generated in mevalonate metabolism, partially rescued the defect in cell viability, sphere number, and sphere size of BTICs induced by shMYC (Supplementary Fig. S10A–S10C). These results demonstrate that MYC directly regulates the transcription of mevalonate pathway genes through binding to the promoter regions of these genes to maintain BTIC growth.

Based on the regulation of oncogene activation by mevalonate, we investigated the coexpression of mevalonate genes and MYC mRNA levels in glioblastoma. In glioblastomas, MYC mRNA levels correlated with a mevalonate gene signature in the TCGA database (Fig. 7A). As mevalonate can activate oncogenes, we investigated the possibility of a feed-forward regulatory loop between mevalonate and MYC, as mevalonate may be both an upstream and downstream regulatory node of MYC. Indeed, targeting mevalonate pathway function through statin treatment of BTICs attenuated MYC expression at both the mRNA and protein levels (Fig. 7B and C). To validate these results as specific to mevalonate inhibition, treatment with 6-fluoromevalonate also reduced MYC mRNA and protein levels (Fig. 7D and E). To confirm these results in vivo, 3,691 BTICs were implanted into the immunocompromised mice, and tumor-bearing mice were treated with simvastatin for 2 weeks. Statin treatment decreased MYC mRNA and protein levels in vivo (Supplementary Fig. S11A and S11B). As mevalonate inhibition activates ERK to mediate its biological effects, we investigated the effects of the MEK inhibitor, U0126, on the statin effects on MYC expression in BTICs, rescuing MYC expression levels depleted by statin therapy (Fig. 7F and G). To further define the mechanistic link between mevalonate and MYC, we examined mevalonate regulation of an miR previously reported to regulate MYC levels, miR-33b (33, 34). After statin therapy, miR-33b expression levels increased in vitro (Fig. 7H) and in vivo (Supplementary Fig. S11C). Further, treatment with an MEK inhibitor (U0126) blocked statin-mediated increase in miR-33b expression, suggesting miR-33b regulation by MEK-ERK signaling (Fig. 7H). To determine the relative position of MYC and miR-33b downstream of mevalonate, we treated BTICs with a miR-33b mimic, finding decreased MYC mRNA and protein levels (Fig. 7I and J). Further supporting the linkage between miR-33b and MYC, transducing BTICs with an miR-33b inhibitor partially rescued simvastatin-mediated decrease in MYC (Supplementary Fig. S12A) and loss of proliferation in BTICs (Supplementary Fig. S12B). Collectively, these results demonstrate that the mevalonate pathway regulates MYC through a RAS–ERK–miR33b pathway, creating a feed-forward loop.

Metabolic dysregulation represents a hallmark of many advanced cancers. In gliomas, mutations in the metabolic enzymes IDH1 or IDH2 are tumor-initiating mutations that define distinct genetic tumor subgroups with favorable prognoses. Although IDH wild-type tumors may not share such a central link to metabolism in tumor initiation, many of the common genetic lesions (e.g., Pten deletion, PIK3CA mutations, EGFR mutation and amplification, etc.) have been directly linked to cellular metabolism. In the current studies, we focused our efforts on the mevalonate metabolic pathway, which creates essential metabolites including those involved in the posttranslational modification of signaling molecules, especially the RAS family members. MYC and RAS represent two of the most powerful canonical oncogenes. We and others previously reported that MYC is preferentially expressed in BTICs (14, 15). In our studies, we now connect MYC to RAS pathway regulation in BTICs through mevalonate metabolism.

RAS family members undergo prenylation to direct their localization to selected membrane compartments to permit participation in signaling. Although RAS mutations are rare in glioblastoma (35, 36), RAS activity is commonly increased in glioblastomas, potentially through alterations of growth factor signaling or NF1 mutations. Some pediatric brain tumors, notably pilocytic astrocytomas, harbor mutations or fusion in downstream effectors of RAS, specifically BRAF (37, 38). Mutant H-RAS has been a frequent tool in the generation of genetically engineered glioma models. Collectively, these results have stimulated the translation of RAS-targeting approaches into clinical trial for glioma patients, including FTIs and raf inhibitors. Unfortunately, the results of these trials, to date, have been disappointing, although trials with novel antagonists, including dabrafenib and selumetinib, are underway.

Our results suggest that mevalonate inhibition, which can be disrupted with statins or genetic targeting of HMGCR, could have high therapeutic indices that may be used as monotherapy or in combination with other antineoplastics. The therapeutic use of statins is an example of drug repositioning, which involves the use of agents developed for other indications for the treatment of cancer (39). Previous preclinical studies also indicate that statins could be used as anticancer agents in glioblastoma (40, 41). Statins induce apoptosis in C6 rat glioma cells when GGPP biosynthesis is inhibited and consequently decrease the level of phosphorylated ERK1/2 and AKT (42). In malignant glioma cells, statins induce apoptosis by the activation of JNK1/2 or by increasing the expression of BIM (41, 43). However, micromolar concentrations of statins can cause induction of apoptosis in normal cells, such as primary cultures of neuronal brain cortices (44, 45) Our studies are distinguished from the prior studies, which were focused on established cell lines, like U87, which poorly represent BTICs, so may underestimate the value of these agents. Cohort studies have suggested contradictory conclusions with regard to correlation between use of preoperative statin and decreased risk of glioma formation and mortality (46, 47). Due to the small number of patients, retrospective nature of these studies and more importantly, lack of sustained use of statins, further work is needed to understand the role of statins in glioblastoma therapy. The low toxicity of statins suggests that their addition to clinical paradigms may hold benefit with a large therapeutic index.

In our studies, the activation of ERK signaling with statin treatment appears counterintuitive initially, as ERK family members are downstream of the RAS family members, so might be expected to be inactivated with mevalonate pathway inhibition. However, ERK function in central nervous system development is complex and context-dependent with genetic models demonstrating key roles in fate determination between neuronal and astrocyte lineages (48). Further, statin-induced apoptosis may force surviving cells to activate ERK to promote survival. The interplay between RAS signaling and mevalonate also serves to feedforward onto MYC, with its myriad effects on cellular differentiation state and tumor growth through miR-33b, suggesting that miR-33b and ERK activation may represent early therapeutic response biomarkers to statin therapy.

Although statin treatment and other agents targeting mevalonate appear somewhat promising, our studies and the limited clinical trial data would suggest that clinical efficacy will require combination with other agents. As radiotherapy is the most effective nonsurgical therapy and we previously showed that BTICs are preferentially resistant to radiation (5), we expect that future studies will focus on the development of combinatorial efforts. In fact, statin use has been suggested in the context of other cancers in which cardiovascular effects of radiation may be partially ameliorated by statin use. As radionecrosis and radiation-induced stroke complicates brain tumor treatment, statins may represent agents that may have substantial efficacy against tumors, while reducing treatment toxicity. In conclusion, our results demonstrate that mevalonate metabolism may serve as an important therapeutic target against cancer stem cells with limited toxicity.

No potential conflicts of interest were disclosed.

Conception and design: X. Wang, S.C. Mack, W. Zhou, S. Bao, J.N. Rich

Development of methodology: X. Wang, Q. Wu, Z. Dong, J.N. Rich

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Wang, Q. Wu, K. Yang, Z. Zhu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Wang, B.C. Prager, K. Yang, L.J.Y. Kim, Y. Shi, T.E. Miller, C.G. Hubert, Z. Zhu

Writing, review, and/or revision of the manuscript: X. Wang, S.C. Mack, R.C. Gimple, A. Song, X. Fang, V. Mahadev, S. Bao, J.N. Rich

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Wang, Z. Huang, Q. Wu, S. Lai, Q. Xie, X. Fang, J.N. Rich

Study supervision: S. Bao, J.N. Rich

We appreciate flow cytometry assistance from C. Shemo and S. O'Bryant, and the glioblastoma tissue provided by M. McGraw and the Cleveland Clinic Tissue Procurement Service. We thank members of the Rich laboratory for critical reading of the article and helpful discussions.

This study was funded by the NIH grants CA197718, CA154130, CA169117, CA171652, NS087913, NS089272 (J.N. Rich), and CA184090, NS091080, and NS099175 (S. Bao).

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.