Myc-mediated transcriptional regulation of the mitochondrial chaperone TRAP1 controls primary and metastatic tumor growth

The role of mitochondria in cancer continues to be debated, and whether exploitation of mitochondrial functions is a general hallmark of malignancy or a tumor- or context-specific response is still unknown. Using a variety of cancer cell lines and several technical approaches, including siRNA-mediated gene silencing, ChIP assays, global metabolomics and focused metabolite analyses, bioenergetics, and cell viability assays, we show that two oncogenic Myc proteins, c-Myc and N-Myc, transcriptionally control the expression of the mitochondrial chaperone TNFR-associated protein-1 (TRAP1) in cancer. In turn, this Myc-mediated regulation preserved the folding and function of mitochondrial oxidative phosphorylation (OXPHOS) complex II and IV subunits, dampened reactive oxygen species production, and enabled oxidative bioenergetics in tumor cells. Of note, we found that genetic or pharmacological targeting of this pathway shuts off tumor cell motility and invasion, kills Myc-expressing cells in a TRAP1-dependent manner, and suppresses primary and metastatic tumor growth in vivo. We conclude that exploitation of mitochondrial functions is a general trait of tumorigenesis and that this reliance of cancer cells on mitochondrial OXPHOS pathways could offer an actionable therapeutic target in the clinic.

A hallmark of cancer is the reprogramming of cellular metabolism (1). Much work has been devoted to the preferential utilization of glycolysis by tumor cells, even when oxygen is present (2), and its impact on cancer progression (3). However, we now know that mitochondria remain functional in most malignancies and that oxidative bioenergetics fuel important tumor traits (4), including metastasis (5,6). Although both metabolic states are likely to coexist during tumor growth (7), a direct, mechanistic link between oncogenic signaling and mitochondrial functions has not been determined, and the question of how general is the exploitation of mitochondria in cancer (4) has remained unanswered.
In this context, Myc proteins comprise a family of ubiquitous, transforming oncogenes (8) that are amplified, deregulated, or translocated in most human cancers (9). Oncogenic Myc drives a plethora of transcriptional and nontranscriptional responses that promote tumor growth and proliferation (10), linked to worse disease outcome in the clinic (11). Although the Myc target gene(s) in tumor progression has not been completely elucidated, and controversy still exists about the requirements of Myc-directed gene expression (12), this pathway connects to multiple aspects of tumor metabolism (13), including mitochondrial functions (14,15).
In this study, we investigate a role of oncogenic Myc in exploiting mitochondria for cancer progression.

Control of mitochondrial protein folding by Myc
Next, we looked at the mechanism(s) of Myc regulation of oxidative phosphorylation. We found that Myc silencing in PC3 cells caused the accumulation of detergent-insoluble subunits, i.e. misfolded oxidative phosphorylation complex subunits, including succinate dehydrogenase B (SDHB, complex II) and cytochrome c oxidase II (CoxII, complex IV) (Fig. 1, D and E). Other oxidative phosphorylation subunits, ubiquinol-cytochrome c reductase core protein 2 (UQCRC2, complex III) or ATP synthase F1 subunit (ATP5A, complex V), were unaffected (Fig. S2A). Consistent with these results, Myc knockdown resulted in reduced mitochondrial complex II and complex IV activity, compared with control transfectants ( Fig. 1F;  Fig. S2B). Instead, complex I activity was unchanged in control or Myc-silenced cells (Fig. S2, C and D). In line with the defects of oxidative bioenergetics, Myc loss was accompanied by increased production of mitochondrial ROS (Fig. S2E), hyperoxidation of peroxiredoxin-3 (Prx3) (Fig. S2F), a marker of oxidative stress, and phosphorylation of the energy sensor, AMPK (Thr-172), indicative of nutrient deprivation (Fig. S2F).

Mitochondrial chaperone TRAP1 is a novel Myc target gene
Previous studies have shown that SDHB protein folding requires the activity of the mitochondrial Hsp90-like chaperone, TNFR-associated protein-1 (TRAP1) (17). Analysis of ChIP-Seq tracks demonstrated time-dependent accumulation of Myc at the TRAP1 promoter in Burkitt's lymphoma P493 cells as well as neuroblastoma BE2C, Kelly, and NGP cell lines ( Fig. 2A). Consistent with this, Myc bound to the promoter of TRAP1, as well as NPM1, a known Myc target gene, by ChIP in Bioinformatics analysis of four independent cancer cell line data bases demonstrated that high levels of Myc strongly correlated (p ϭ 2.6 ϫ 10 Ϫ20 -6.1 ϫ 10 Ϫ117 ) with TRAP1 expression (Fig. S3C), reinforcing the generality of this response across genetically disparate tumors.
We next asked whether the Myc homolog, N-Myc, also regulated TRAP1 in the model of neuroblastoma ( Fig. 2A). Conditional silencing of N-Myc in neuroblastoma SHEP21 cells reduced TRAP1 mRNA (Fig. S4A) and protein expression (Fig.  S4B). Conversely, 4OHT-regulated induction of N-Myc in SHEP or SKNAS cells increased TRAP1 mRNA (Fig. S4C) and protein levels (Fig. S4B). Sequence analysis of a human TRAP1 locus revealed the presence of putative c-Myc-and N-Mycbinding sites within ϳ1.4 kb of the first ATG (Fig. S4D). Transfection of this TRAP1 promoter region upstream of a luciferase reporter gene (TRAP1-Luc) produced luciferase activity in PC3 or P493 cells, in a response abolished by siRNA knockdown of Myc (Fig. S4E). In addition, mutagenesis of a putative Mycbinding site at position Ϫ60 abolished TRAP1-Luc luciferase activity in control and Myc-knockdown PC3 cells (Fig. S4F).
Next, we asked whether TRAP1 was required for mitochondrial bioenergetics in Myc-expressing tumors. We found that re-expression of non-siRNA-inhibitable TRAP1 in Myc knockdown cells rescued the defect of OCR (Fig. 2C) and corrected basal respiration in these settings (Fig. 2D). Consistent with improved bioenergetics, reconstitution with TRAP1 restored ATP production in Myc knockdown cells (Fig. 2E) and reversed the expression of markers of cellular starvation (AMPK phosphorylation) and oxidative stress (Prx3 hyperoxidation) (Fig. 2F). As a prerequisite of these responses, re-expression of TRAP1 in Myc knockdown PC3 cells corrected the folding of SDHB (complex II) and CoxII (complex IV) subunits ( Fig. 2G; Fig. S5A), restoring complex II (Fig. S5, B and C) and complex IV (Fig. S5, D and E) activity, quantitatively comparable with control transfectants.

Myc-TRAP1 signaling fuels tumor chemotaxis and invasion
TRAP1-directed bioenergetics has been implicated in tumor traits (18), including tumor cell motility (19), and this possibility was next investigated. We found that Myc silencing with two independent siRNA sequences (Fig. S6A) or pooled siRNA (Fig.  3A) potently inhibited tumor cell motility ( Fig. S6A; Fig. 3A). Quantitative analysis of time-lapse video microscopy demon- In reconstitution experiments, re-expression of TRAP1 in Myc-silenced cells was sufficient to correct the defect in tumor cell motility ( Fig. 3A; Fig.  S6A), restore the speed of cell movements and the total distance traveled by individual cells (Fig. 3B; Fig. S6B), and increase tumor cell invasion across Matrigel to levels of control transfectants (Fig. 3, C and D; Fig. S6, C and D). Biochemically, PC3 cells reconstituted with TRAP1 also exhibited phosphorylation of cell motility kinases, Src and focal adhesion kinase (FAK), which was suppressed by Myc knockdown (Fig. 3E).

Pharmacologic targeting of Myc-TRAP1 for cancer therapy
Analysis of neuroblastoma patient cohorts revealed that TRAP1 expression was strongly associated with N-Myc levels (p ϭ 1 ϫ 10 Ϫ63 -2 ϫ 10 Ϫ96 ) (Fig. S7A) and correlated with shortened overall survival (Fig. S7B), reinforcing the clinical relevance of this pathway. To test whether this could provide an actionable therapeutic target, we next treated P493 or neuroblastoma NLF or IMR cells with a mitochondria-targeted small molecule TRAP1 antagonist, Gamitrinib (18). In these experiments, Gamitrinib treatment caused concentration-dependent dissipation of mitochondrial inner membrane potential (Fig.  S7C) and complete loss of tumor cell viability (Fig. 4A). This response was specific because siRNA silencing of TRAP1 atten-uated Gamitrinib-mediated cell killing, whereas reconstitution with TRAP1 strongly increased cell death in these settings (Fig.  4B). Consistent with these data, Gamitrinib suppressed neuroblastoma NLF or IMR colony formation (Fig. 4, C and D) and inhibited P493 xenograft tumor growth in immunocompromised mice (Fig. 4, E and F). When analyzed in a model of disseminated disease, systemic administration of Gamitrinib suppressed liver metastases of PC3 cells (Fig. 4G), reducing both the surface area and number of metastatic foci (Fig. 4H), compared with vehicle-treated animals.
In sum, the identification of TRAP1 as a direct transcriptional target of a ubiquitous oncogene, i.e. Myc (11), demonstrates that mitochondrial reprogramming is a universal trait of cancer, required for primary and metastatic tumor growth. This is consistent with an expanding role of mitochondria as tumor drivers (4), and mechanisms of intra-organelle protein folding, oxidative bioenergetics, and ROS buffering (18) as key requirements of cancer progression. The regulation by Myc explains why TRAP1 is prominently overexpressed in genetically disparate malignancies, compared with normal tissues (20), and validates its role as a tumor driver. Data from genetically engineered mouse models support this conclusion, as transgenic expression of TRAP1 accelerated prostatic tumorigenesis (21), whereas homozygous deletion of TRAP1 delayed age-associated pathologies, including cancer (22). On the other hand, the Myc-TRAP1 axis of mitochondrial reprogramming is druggable, and selective disruption of intra-organelle protein

Cells and cell culture
Prostate adenocarcinoma PC3 or DU145 cells were obtained from the American Type Culture Collection and maintained in culture according to the supplier's specifications. A human Burkitt's lymphoma P493-6 cell line was described previously (23). Clones of P493 cells containing a doxycycline (Dox)-regulated Myc transgene induced after Dox removal (Dox-off system) were as described previously (24). Neuroblastoma SHEP21N, SHEP21-NMycER (25), and SKNAS-N-MycER (26) cells containing a conditionally-regulated N-Myc transgene were as described. In these cells, treatment with 50 ng/ml Dox for 48 h suppresses N-Myc expression, whereas addition of 4-hydroxytamoxifen (4OHT, 0.5 g/ml) results in strong N-Myc induction.

Chromatin immunoprecipitation (ChIP) assay
P493 cells treated with ␤-estradiol plus Dox for 48 h or PC3 cells transfected with control nontargeting siRNA or Myc-directed siRNA for 72 h were used for ChIP experiments as described previously (27)

TRAP1 promoter analysis
Genomic DNA sequences were downloaded from the University of California Santa Cruz Genome browser (http:// genome.ucsc.edu) 3 (37). Putative Myc-or N-Myc-binding sites were identified using Factorbook motif. A 1,397-bp region of the human TRAP1 gene (Ϫ1330 to ϩ66, ϩ1 corresponds to the transcription start site) containing two putative Myc-binding sites (5Ј-Ca/gCGTG-3Ј) was amplified and cloned in pGL4.15-promoterless vector upstream from a luciferase reporter gene. For analysis of promoter activity, PC3 or P493 cells transfected with control nontargeting siRNA or Myc-di- EDITORS' PICK: Myc regulation of mitochondrial protein folding rected siRNA were reconstituted with 2 g of TRAP1-luciferase reporter plasmid (TRAP1-Luc), and luciferase activity was quantified following the manufacturer's protocol (Promega).

Global metabolomics screening
PC3 cells (2 ϫ 10 6 ) were transfected with control siRNA or Myc-directed siRNA followed by incubation with 800 l of methanol for 30 s. The cell extract was further incubated with 550 l of Milli-Q water containing internal standards (H3304-1002, Human Metabolome Technologies, Inc., Tsuruoka, Japan) for 30 s and centrifuged at 2,300 ϫ g for 5 min at 4°C. Proteins were removed by centrifugation of an 800-l upper aqueous layer through a Millipore 5-kDa cutoff filter at 9,100 ϫ g for 120 min at 4°C, and the filtrate was centrifugally concentrated and suspended in 50 l of Milli-Q water for capillary electrophoresis-mass spectrometry analysis. Global metabolome measurements were carried out at Human Metabolome Technology Inc. (Tsuruoka, Japan) (28). Hierarchical cluster analysis and principal component analysis were performed using proprietary PeakStat and SampleStat software, respectively. Detected metabolites were plotted on metabolic pathway maps using VANTED (Visualization and Analysis of Networks containing Experimental Data) software (29).

Focused metabolite analysis
LC-MS analysis was performed on a Thermo Fisher Scientific Q Exactive HF-X mass spectrometer equipped with a HESI II probe and coupled to a Thermo Fisher Scientific Vanquish Horizon UHPLC system. Polar metabolites were extracted using 80% methanol and separated at 0.2 ml/min by HILIC chromatography at 45°C on a ZIC-pHILIC 2.1 inner diameter ϫ 150-mm column using 20 mM ammonium carbonate, 0.1% ammonium hydroxide, pH 9.2, and acetonitrile with a gradient of 0 min, 85% B; 2 min, 85% B; 17 min, 20% B; 17.1 min, 85% B; and 26 min, 85% B. Relevant MS parameters were as follows: sheath gas, 40; auxiliary gas, 10; sweep gas, 1; auxiliary gas heater temperature, 350°C; spray voltage, 3.5 kV for the positive mode and 3.2 kV for the negative mode; capillary temperature, 325°C; and funnel RF level at 40. A sample pool (quality control) was generated by combining an equal volume of each sample and analyzed using a full MS scan at the start, middle, and end of the run sequence. For full MS analyses, data were acquired with polarity switching at: scan range 65 to 975 m/z; 120,000 resolution; automated gain control (AGC) target of 1E6; and maximum injection time (IT) of 100 ms. Data-dependent MS/MS was performed without polarity switching; a full MS scan was acquired as described above, followed by MS/MS of the 10 most abundant ions at 15,000 resolution, AGC target of 5E4, maximum IT of 50 ms, isolation width of 1.0 m/z, and stepped collision energy of 20, 40, and 60. Metabolite identification and quantitation were performed using Compound Discoverer 3.0. Metabolites were identified from a mass list of 206 verified compounds (high confidence identifications) as well as by searching the MS/MS data against the mzCloud database and accepting tentative identifications with a minimum score of 50.

Mitochondrial protein folding
Protein lysates were prepared in RIPA buffer, and equal amounts of protein lysates were separated by SDS-gel electrophoresis as described previously (30). Changes in mitochondrial protein folding were assessed at increasing detergent (CHAPS) concentrations (0 -2.5%) as described previously (17).

Cell viability assay and colony formation
Changes in cell viability in control or Gamitrinib (0 -15 M)treated P493, NLF, or IMR cultures were quantified by a fluorogenic Alamar Blue dye, as described previously (32). NLF and IMR neuroblastoma cells were treated with vehicle or Gamitrinib (0 -10 M) and plated at 200 cells per condition, and macroscopically-visible colonies stained with 0.5% w/v crystal violet/methanol were manually counted after 14 days.

Tumor cell motility and invasion
PC3 cells transfected with control nontargeting siRNA or Myc-directed siRNA were analyzed for 2D motility in 4-well Phϩ chambers (Ibidi) by time-lapse video microscopy over 10 h, as described previously (30). Tracking data were exported into the Chemotaxis and Migration Tool version 2.0 (Ibidi) for graphing and calculation of mean and standard deviation of speed accumulated distance of movement. Tumor cell invasion was determined using growth factor reduced Matrigel-coated 8 m PET Transwell chambers (Corning), as described previously (30).

Bioinformatics analysis
Differential expressions for metabolomics data were tested using Student's t test, and p values were corrected for multiple testing using Benjamini-Hochberg method. Metabolites that passed a false discovery rate Ͻ5% cutoff were considered significant. Overlap between the two metabolomics experiments was performed using KEGG and HMDB identifications. All known interactions between overlapped metabolites were derived from the Ingenuity Knowledge Base. CHIP-seq data for c-Myc in P493 Burkitt's lymphoma cells at three independent time points after Dox addition were derived from NCBI GEO database set GSE36354 (24). CHIP-seq analysis of N-Myc dis-EDITORS' PICK: Myc regulation of mitochondrial protein folding tribution in neuroblastoma BE2C, Kelly, and NGP cells was from GEO database set GSE36354 (33). Data were aligned using bowtie (34) against hg19 genome version and analyzed using the HOMER (35) algorithm to generate bigwig files used for visualizing CHIP-seq tracks in the UCSC browser.

Animal studies
Studies involving mice were carried out in accordance with the Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee (IACUC) of The Wistar Institute (protocol nos. 112625 and 112610). P493 cells (2 ϫ 10 7 ) were injected s.c. in athymic nude mice. Once tumors reached 100 mm 3 in volume, animals were randomized into two groups to receive 20% Cremophor (vehicle) or Gamitrinib (10 mg/kg) daily i.p. for 10 days. For a liver metastasis model, PC3 cells (1 ϫ 10 6 ) were injected into the spleen of SCID/beige mice, and randomized animal groups were treated with vehicle or Gamitrinib for 11 days before histological quantification of liver metastases, as described previously (36).

Statistical analysis
Data are expressed as means Ϯ S.D. of multiple independent experiments or replicates of representative experiments out of a minimum of two or three independent determinations. Twotailed Student's t test was used for two-group comparative analyses. Statistical analyses were performed using the GraphPad software package (Prism 6.0). A p value of Ͻ0.05 was considered as statistically significant.