The cell cycle regulator 14-3-3σ opposes and reverses cancer metabolic reprogramming

Summary Extensive reprogramming of cellular energy metabolism is a hallmark of cancer. Despite its importance, the molecular mechanism controlling this tumour metabolic shift remains not fully understood. Here we show that 14-3-3σ regulates cancer metabolic reprogramming and protects cells from tumourigenic transformation. 14-3-3σ opposes tumour-promoting metabolic programs by enhancing c-Myc poly-ubiquitination and subsequent degradation. 14-3-3σ demonstrates the suppressive impact on cancer glycolysis, glutaminolysis, mitochondrial biogenesis and other major metabolic processes of tumours. Importantly, 14-3-3σ expression levels predict overall and recurrence-free survival rates, tumour glucose uptake and metabolic gene expression in breast cancer patients. Thus, these results highlight that 14-3-3σ is an important regulator of tumour metabolism, and loss of 14-3-3σ expression is critical for cancer metabolic reprogramming. We anticipate that pharmacologically elevating the function of 14-3-3σ in tumours could be a promising direction for targeted anti-cancer metabolism therapy development in future.


Introduction
Tumourigenesis is characterized by 10 hallmarks described by Hanahan and Weinberg in their seminal paper 1 . Among these, deregulation of cellular energetics (also called metabolic reprogramming) involves tumour cells "rewiring" their metabolic pathways to support rapid proliferation, continuous growth, metastasis, survival, and resistance to therapies [1][2][3][4] . Increases in glycolysis, glutaminolysis, lipid metabolism, mitochondrial biogenesis, and energy production are among the most prominent metabolic alterations in cancer [1][2][3][5][6][7][8][9] . In fact, these processes provide tumours with not only energy but also essential precursors to support their biosynthesis and proliferation 2,5,10 . Cancer metabolism is regulated mainly by c-Myc (Myc), HIF1α, and p53 2,3,5,9,11,12 . The interplay between these master regulators determines the status of tumour metabolism and has a decisive impact on tumourigenesis 2 . However, the regulation of cancer bioenergetics is not fully understood, suggesting that more regulators remain to be identified 3,13 . c-Myc (Myc) is frequently overexpressed in many human cancers. Myc is a major oncogenic transcription factor that can induce tumorigenesis by promoting cell proliferation, causing genome instability, and blocking cell differentiation 14 . More importantly, Myc can also upregulate glycolytic genes, thereby promoting glucose consumption and glycolysis 2,15 . The upregulation of these glycolytic enzymes is due to Myc's binding to its target genes for transcriptional activation 2,12,16,17 . In addition to glycolysis, Myc is the primary inducer of glutaminolysis in cells 11,17,18 . The metabolic switch to aerobic glycolysis and glutaminolysis is crucial to support Myc-mediated proliferation, growth, survival and metastasis of tumor cells. Despite the significant roles of Myc in many signaling pathways and cellular processes, the mechanism behind Myc regulation is not fully understood.
In this study, we show that the frequent loss of 14-3-3σ in cancer leads to the metabolic reprogramming phenotype that aids cancer growth and correlates with poor cancer survival. We demonstrate that 14-3-3σ mitigates tumour-promoting metabolic programs by promoting c-Myc poly-ubiquitination and subsequent degradation, thereby reversing Myc-mediated cancer glycolysis, glutaminolysis and mitochondrial biogenesis in cancer. Our study discovers 14-3-3σ as an important regulator of cancer cellular energetics and holds the potential to unlock a door to new cancer treatment therapies.

Loss of 14-3-3σ in cancer results in metabolic reprogramming
Immunohistochemical analysis of breast tumour tissue microarrays and retrospective comparison with patient clinical data revealed that a low level of 14-3-3σ protein expression in breast tumours was significantly associated with poor survival (Fig. 1a, Supplementary Table 1). Bioinformatics analysis revealed marked increases in major cancer hallmarks and corresponding biological processes when 14-3-3σ was down-regulated ( Fig. 1b,  Supplementary Fig. 1-2). 14-3-3σ down-regulation in breast cancer was correlated with deregulation of cancer hallmarks, especially, deregulated cellular energetics, genomic instability, sustained proliferative signaling, resistance to cell death as illustrated by a Circos map (Fig. 1b). Deregulation of cellular energetics (large size of symbol) and genome instability were among the cancer hallmarks most enhanced upon 14-3-3σ silencing. These observations suggest that 14-3-3σ could have broad control of tumourigenesis. Notably, elevations in genes associated with deregulation of cellular energetics (i.e., glucose uptake, glycolysis, glutaminolysis, energy production, and mitochondrial biogenesis) were among the most significant changes in tumour gene expression profiles associated with 14-3-3σ down-regulation ( Fig. 1b-c, Supplementary Fig. 3). Accordingly, breast cancer patients whose tumours expressed low levels of 14-3-3σ had higher uptake of [ 18 F]fluorodeoxyglucose ( 18 FDG), indicating that loss of 14-3-3σ increases tumour glucose import (Fig. 1d, Supplementary Table 2). Likewise, 14-3-3σ silencing was correlated with elevated expression of glycolytic and glutaminolytic genes in breast tumours ( Fig. 1e-f). These data suggest that the regulatory role of 14-3-3σ in tumourigenesis may include inhibiting metabolic reprogramming and other important cancer hallmarks.
We also showed that deletion and knockdown of 14-3-3σ increased extracellular acidification rate (ECAR), an indicator of glycolysis, in colorectal and breast cancer cells (Fig. 2b, Supplementary Fig 5a). In contrast, induced expression of 14-3-3σ reduced ECAR (Fig. 2c, Supplementary Fig. 5b). 14-3-3σ also suppressed mitochondrial respiration as indicated by oxygen consumption rate (OCR; Fig. 2d-e, Supplementary Fig. 6). In the OCR assay, oligomycin, which prevents ATP synthesis, is added to inhibit oxygen consumption. But oxygen consumption can be stimulated again when oxidative phosphorylation is uncoupled by addition of FCCP (p-trifluoromethoxy carbonyl cyanide phenylhydrazone), which destroys the mitochondrial membrane potential and causes a short-circuit in the respiratory chain. Following this, oxygen consumption can be completely stopped when rotenone (inhibitor of mitochondrial complex I) and antimycin A (inhibitor of mitochondrial complex III) are added to block the electron transport chain.

14-3-3σ is also regulating metabolism under hypoxia
Interestingly, our data showed that 14-3-3σ also decreased extracellular acidification rates of HCT116 and MDA-MB-231 cells in hypoxia ( Fig. 8a-b). Loss and knockdown of 14-3-3σ promoted glycolytic gene expression of HCT116 cells cultured in a hypoxic condition ( Fig.  8c-d). These observations further highlight the potential of 14-3-3σ in regulating cancer metabolism. We additionally observed that HCT116 14-3-3σ −/− proliferated much slower when glucose was replaced by galactose in culture medium ( Supplementary Fig. 14). Loss of 14-3-3σ also rendered HCT116 cells addicted to glutaminolysis, a commonly seen phenomenon in cancer cells that have high levels of Myc expression ( Supplementary Fig.  15). Furthermore, we noticed that induction of Flag-14-3-3σ expression significantly induced cell death of HCT116 14-3-3σ colon cancer cells, MDA-MB-231 metastatic breast cancer cells and H1299 lung cancer cells ( Supplementary Fig. 16), which indicates the potential of 14-3-3σ in anti-cancer therapies.

14-3-3σ mitigates tumour metabolic reprogramming in vivo
To validate the function of 14-3-3σ in vivo, we established an orthotopic xenograft mouse model using MDA-MB-231 cells carrying an inducible tet-On 14-3-3σ expression system ( Supplementary Fig. 17). The induction of 14-3-3σ reduces cell proliferation and cell viability ( Supplementary Fig. 17). MDA-MB-231 is a highly metastatic breast cancer cell line that loses 14-3-3σ expression because of hypermethylation of the CpG islands of the SFN promoter area ( Supplementary Fig. 18). The experimental mouse group received doxycycline for 2 weeks to induce 14-3-3σ expression, while the control group received no doxycycline. We then used 18 FDG microPET scans to assess the impact of 14-3-3σ on tumour glucose uptake in vivo. Our data show that restoring 14-3-3σ expression caused a 50% reduction in glucose uptake by the xenograft tumours (Fig. 9a). This observation is interesting because glucose is a major source of energy and nutrients for cancer cells 1-3 .
In summary, we discovered that 14-3-3σ opposed and reversed cancer metabolic reprogramming by promoting Myc degradation (Fig. 9i, Supplementary Fig. 22). We also showed that loss of 14-3-3σ was associated with the upregulation of many major cancer hallmarks. Our results additionally elucidated the crosstalk among central cellular processes such as cell cycle, energy metabolism, p53 signaling and Myc network in cancer. Our findings also pointed out that elevating 14-3-3σ's function in tumours is a potential direction for targeted anti-cancer metabolism therapies.

Discussion
Metabolic reprogramming has emerged as a new hallmark of cancers. Many oncogenes or tumor suppressor genes are deregulated, which leads to abnormal cellular bioenergetics that give cancer cells a growth advantage either by providing fuel for cancer cell growth or creating microenvironments that facilitate cancer invasion. Our clinical data, based on PET SUVs, indicates that loss of 14-3-3σ leads to increased glucose uptake in tumors. This finding suggests a functional role for 14-3-3σ in controlling cancer glycolysis. In searching for a potential mechanism behind 14-3-3σ-mediated suppression of the glycolytic pathway, we found that Myc to be the main culprit. Our data shows that 14-3-3σ is able to downregulate Myc stability by enhancing Myc polyubiquitination. The 14-3-3σ-facilitated destabilization of Myc could regulate cancer energy metabolism through the following mechanisms: (1) by blocking Myc-mediated expression of glycolytic genes; (2) by suppressing Myc-mediated upregulation of genes involved in glutaminolysis and mitochondrial biogenesis. Thus, our results reveal a new regulatory pathway in cancer cell bioenergetics. Additionally, 14-3-3σ-Myc axis regulation also provides a new example of the important balance between oncogenes and tumor suppressors in cellular metabolism regulation.
Our data indicates that the loss of 14-3-3σ leads to elevated lactate production, and that reexpression of 14-3-3σ reduces lactate generation. This elevation of lactate production is caused by upregulation of Lactate Dehydrogenase A when 14-3-3σ expression is lost. The increase in lactate production due to 14-3-3σ loss enhances the acidification of extracellular environment as shown by our ECAR measurements. Lactate is important for cancer cells invasion and metastasis. It was documented that lactate can significantly enhance cancer cell motility and facilitate the breaking down of extracellular matrix 30 . It has also been shown that lactate stimulates the migration of epithelial cancer cells and can promote cancer metastasis to distant organs. For example, lactate administration fuels lung metastasis in a breast cancer mouse model 31 . Therefore loss of 14-3-3σ may give cancer cells an advantage in terms of cell migration and metastasis as indicated in our cancer hallmarks characterization. In addition, the fact that 14-3-3σ null cells have a high concentration of ATP means that cancer cells have an edge when it comes to cell growth and drug resistance. This is because ATP is both a major DNA precursor often used for repairing genetic damage and a key energy source for multidrug resistance efflux pumps that discard toxic chemotherapy agents. It has also been reported that glycolytic enzymes play important roles in cancer cell survival. For example, Hexokinase 2 blocks cytochrome c release from mitochondria 45,46 , thereby protecting cancer cells from apoptosis. Additionally, Glucose-6phosphate isomerase has been shown to stimulate cell motility, an important step in the metastatic process 4748 . Therefore, the fact that loss of 14-3-3σ expression elevates the levels of these glycolytic enzymes could have a tremendous impact not only on tumor bioenergetics but also other hallmarks of cancers.
Myc can drive oxygen consumption, mitochondrial biogenesis, and glycolysis. This indicates that Myc stimulates both aerobic glycolysis and mitochondrial oxidative phosphorylation. This phenomenon allows cells with high levels of Myc to adjust to microenvironments with various oxygen levels (both hyper-and hypoxic), freeing them from dependence on oxygen from blood vessels. We have found that loss of 14-3-3σ leads to increased oxygen consumption. More importantly, re-expressing 14-3-3σ can suppress oxygen consumption and mitochondrial biogenesis; this is consistent with the 14-3-3σ-Myc axis regulation. 14-3-3σ expression markedly reduced mitochondrial mass, TFAM level and the expression of mitochondrial genes. This again is the result of 14-3-3σ-Myc axis regulation. This observation is important because mitochondria are not only cellular power stations but also factories that produce many important precursors for amino acid (oxaloacetate, α-ketoglutarate), lipid (citrate), heme group (succinyl CoA), and nucleotide (oxaloacetate transaminated to aspartate) biosynthesis. Therefore, a strong decline in mitochondrial mass could have a deleterious impact on cancer cell growth. Indeed, our Nuclear Magnetic Resonance analysis demonstrated that 14-3-3σ expression correlates with a reduction of many key amino acids (e.g., methionine, glutamate, glutamine, and arginine), phospholipid metabolites and precursors for nucleotide synthesis. It is still unclear whether the reduction in amino acid and phospholipid levels is a direct effect of 14-3-3σ, a consequence of 14-3-3σ-mediated suppression of glycolysis, glutaminolysis and mitochondrial mass, or both. These observations, however, imply that 14-3-3σ has a multifunctional role in cancer cell metabolic control.
Furthermore, glutaminolysis is both a major source of metabolites for large-scale biosynthesis and a vital energy source for cancer cells. Myc promotes glutaminolysis by activating the expression of glutamine transporters (ASCT2, SN2) and Glutaminase 1 (GLS1). Glutamine is transported by glutamine transporters and is converted to αketoglutarate (α-KG) as a result of an anaplerotic reaction catalyzed by glutamatedehydrogenase to replenish biosynthetic intermediates of the TCA cycle. It is important to point out that 14-3-3σ re-expression in 14-3-3σ null cells leads to the reduction of glutamate concentration levels, indicating that 14-3-3σ may also be involved in glutaminolysis. Indeed, our biochemical data shows that re-expression of 14-3-3σ in 14-3-3σ null cells can suppress the Myc-activated genes involved in glutaminolysis. Thus, our data indicates that loss of 14-3-3σ leads to Myc activation and widespread derailing of cancer bioenergetics.
Importantly, all the roles of 14-3-3σ in limiting glucose uptake, pyruvate-lactate flux, mitochondrial biogenesis and glutaminolysis can be recapitulated in vivo. We used both microPET and Magnetic Resonance Spectroscopy Imaging (MRSI) to investigate the role of 14-3-3σ in glycolysis in xenograft tumor models. For the MRSI, we used a hyperpolarized 13 C pyruvate tracer to accurately measure the real time metabolic flux of pyruvate-to-lactate reaction in xenograft tumor models. This advanced tumor imaging method is sensitive, safe, non-invasive, non-radioactive, and able to provide deep insight into glycolytic regulation during tumorigenesis 37,4950,51 . These two independent techniques confirmed the regulatory role of 14-3-3σ in cancer energy metabolism in vivo. In conclusion, our study contributes to further elucidate the complicated regulation of cancer bioenergetics.
However, further study is needed to fully elucidate the function of 14-3-3σ. Our mass spectrometry analysis of 14-3-3σ-interacting partners revealed that 14-3-3σ could interact with hundreds of targets. The influences of those interactions on metabolic reprogramming including in the condition of hypoxia are still largely uncharacterized. This could possibly explain the various observations about 14-3-3σ's impact on tumorigenesis. We believe that more studies are needed to fully understand the complex and multi-layered role of 14-3-3σ in normal and cancer cells. In conclusion, our study contributes to further elucidate the complicated control of cancer bioenergetics by identifying 14-3-3σ-Myc as a new regulatory axis of cancer metabolism. Furthermore, these findings suggest that 14-3-3σ could be used as a target in the design of cancer metabolism-targeted therapies 3 .

Clinical relevance of 14-3-3σ in breast cancer
Breast cancer patients' tumour and normal breast tissues were stained with anti-14-3-3σ antibodies (1:200 dilution, Fitzgerald Industries International). Immunohistochemical staining results of 14-3-3σ were quantified by a Dako ChromaVision ACIS III system and analysed by the Department of Pathology at MD Anderson Cancer Center (MDACC). Immunohistochemical staining results of 14-3-3σ in normal breast tissue were used as a reference to classify breast cancer patients into low and normal-high 14-3-3σ groups. There were 60 patients in the low 14-3-3σ group and 25 patients in the normal-high 14-3-3σ group. Overall and relapse-free survival curves were built based on corresponding cancer patients' clinical records at MD Anderson Cancer Center.

Immunohistochemical staining
Breast cancer tissue microarrays were provided by the Department of Pathology at MD Anderson Cancer Center. Xenograft breast tumour slides were constructed by the Histology Core Facility at MD Anderson. Immunohistochemical, hematoxylin and eosin, and cleaved caspase-3 staining was done by the Tissue Core Facility at MD Anderson. Other immunohistochemical staining was performed according to protocols provided by the Department of Pathology 29,52,53 . Tissue microarrays and tumour slides were stained with anti-14-3-3σ (1:200 dilution) and anti-c-Myc (1:50 dilution). Immunohistochemical staining results were analysed by pathologists of the Department of Pathology and quantified by using a Dako ChromaVision System ACIS III in the Department of Molecular and Cellular Oncology. Immunohistochemical staining score (arbitrary unit, a.u.) is the product of immunhistochemical staining intensity multiplied by percentage of brown area. The IHC expression score is defined by this formula: IHC expression score = Intensity of IHC staining (within tumor area) x Area of IHC staining (within tumor area). The intensity and area of IHC staining were calculated by Dako ACIS III Imaging Station slide scanner. The whole tumor tissue microarrays were scanned by this slide scanner. Tumour locations were determined based on pathological observations. The intensity and area of IHC staining within tumour areas were calculated using Dako ACIS III Imaging Station software.

Bioinformatics analysis
Breast cancer patient data sets GSE20194, GSE5847, GSE2109, GSE11121, among others were downloaded from the Oncomine, The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus databases. These cohorts were analysed by Gene Set Enrichment Analysis (Broad Institute), Nexus Expression 3 (BioDiscovery), Gene Spring 12 (Agilent Technologies), and Ingenuity Pathway Analysis (Ingenuity Systems). A Circos map of the data (Circos Software, Circos.ca) analysis 54 revealed that downregulation of 14-3-3 was associated with an increase in cancer hallmarks and their related biological processes. The GSE20194 cohort, which includes 255 patients with stage I-III breast cancer who had not received any treatment at the time of sample collection, was chosen for tumour analysis. Breast tumour biopsies from this cohort were collected by fine-needle aspiration, with minimal contamination by normal tissues. 14-3-3σ expression in the tumours was quantified by microarray analysis. 14-3-3σ expression of normal breast tissues was used as a reference to select the tumours that were in the highest (high) or lowest (low) 14-3-3σ quartile. The cohort included 64 patients with high 14-3-3σ and 63 with low 14-3-3σ.
Metabolomics analysis by nuclear magnetic resonance HCT116 14-3-3σ +/+ , HCT116 14-3-3σ −/− , HCT116 14-3-3σ −/− infected with Ad-βgal, and HCT116 14-3-3σ −/− infected with Ad-14-3-3σ were grown under culture conditions 26,55 for 48 hours. A sample of cells (20×10 6 ) from each cell line was collected and snap-frozen with liquid nitrogen. Total metabolites were then extracted from cell pellets using freeze/thaw extraction and filtration protocols provided by Chenomx, Inc. Metabolite extracts were snapfrozen with liquid nitrogen and shipped to Chenomx on dry ice for NMR analysis. Data were analyzed by Chenomx NMR software and are displayed as heat maps generated by the Cluster and Java Treeview programs. Cell pellet collection procedure: For each plate, 20 million cells were rapidly collected by trypsinization and centrifuged at 280g for 5 min at 4°C. Each cell pellet was washed with 40 mL of ice-cold phosphate-buffered saline (PBS), followed by centrifugation at 280g for 5 min at 4°C. PBS was removed by vacuum. Each cell pellet was resuspended in 1ml of ice-cold PBS and transferred to a 1.5ml conical tube. Centrifugation was done at 1700 g (maximal speed without causing damage to cellular structures) for 5 min at 4°C.PBS was removed by vacuum using fine tips. All cell pellets were then snap-frozen using liquid nitrogen and stored at −80°C until metabolite extraction steps.
Cell lysis procedure using freeze/thaw method: Cell pellets were removed from −80°C freezer and thawed on ice. Thawed cell pellets were resuspended in 500 μl of 500μM phosphate buffer (pH 7.0). Cell samples were then snap-frozen in liquid nitrogen and then slowly thawed on ice. Cell samples were sonicated for 30 seconds on ice. Cell samples were snap-frozen in liquid nitrogen and then slowly thawed on ice. Cell samples were centrifuged at 13,793g for 10 minutes at 4°C. 10 μl of cell lysates were used for protein concentration measurement (for normalization purpose). Metabolite extraction procedure: Cell samples were filtered through a prewashed Nanosep 3K Omega microcentrifuge filter tube (Pall Corporation, Catalog # OD003C33) by centrifuging for 15-30 min at 13,793g at 4°C for 30 minutes or until the maximum amount of sample was accumulated. Filtered samples were transferred to new vials and labeled and snap frozen by liquid nitrogen. Samples were shipped to analysis site on dry ice. NMR data were analyzed using Chenomx NMR Suite (www.chenomx.com). Heat map was built using Cluster and Treeview software.

Establishment of retroviral tet-On 14-3-3σ systems
Flag-14-3-3σ cDNA was amplified by PCR from a pCMV5-Flag-14-3-3σ plasmid in Dr. Lee's laboratory and cloned into a pRetro-CMV/TO retroviral plasmid (kindly provided by Dr. L.M. Staudt, National Cancer Institute). After sequence verification, the pRetro/CMV-TO-Flag-14-3-3σ plasmid was co-transfected with VSVg and Gag/Pol plasmids into 293T cells to produce 14-3-3σ tet-On retroviruses. Tet repressor retroviruses were produced from the 293T cells after co-transfection of p-BMN-TetR-BSR, VSVg, and Gag/Pol plasmids (also provided by Dr. Staudt). Cells of interest were first infected with TetR-BSR retroviruses and then selected with appropriate culture medium containing 20 μg/ml blasticidin for 1 week. Selected cells were infected with retroviruses carrying Flag-14-3-3σ and subjected to a second selection with 5 μg/ml puromycin for 1 week. The culture media for cells with the retroviral tet-On system are DMEM High glucose (Hyclone Laboratories, Utah, USA) (Catalog #SH30243.01) and I-1640 (Hyclone Laboratories, Utah, USA) (Catalog # SH30027.01) supplemented with 2 mM L-glutamine, and 1% antibioticantimycotic solution. The Fetal Bovine Serum (FBS) for growing cells carrying tet-On system is tetracycline-free (tetracycline tested) FBS from Atlanta Biologicals (Georgia, USA) (Catalog # S10350). All culture media for maintaining cancer cells infected with tet-On Flag-14-3-3σ retroviruses contain 10 μg/ml Blasticidine (Invivogen, California, USA) (Catalog # ant-bl-1) to ensure the continuous expression of the fusion protein Tet repressor -Blasticidine resistant protein, which is important for tight control of Flag-14-3-3σ expression, and 2 μg/ml Puromycin (Invivogen, California, USA) (Catalog # ant-pr-1) to keep Flag-14-3-3σ gene presence. For in vivo studies, MDA-MB-231 tet-On 14-3-3σ breast cancer cells were injected into the mammary fat pad of female nude mice. Tumor growth was monitored for 4 weeks. Tumor-bearing mice were divided into 2 groups of 5-10 mice, an experimental and a control group. To induce 14-3-3σ expression in vitro, doxycycline (a stable derivative of tetracycline) 56 was added to the culture medium at a final concentration of 5 ng/ml for 48, 72 or 96 hours. The culture medium with doxycycline was replaced daily to induce Flag-14-3-3σ expression. To induce 14-3-3σ expression in xenografts in vivo, mice in the experimental group were given doxycycline in their drinking water (final concentration 200 μg/ml) for 2 weeks. Because of the sensitivity to light and short half-life of doxycycline, the drinking water with doxycycline was contained in opaque bottles and replaced with a freshly prepared solution every 48 hours. The control group received plain drinking water without doxycycline.

Lentiviral shRNA knockdown of 14-3-3σ and c-Myc
14-3-3σ and c-Myc shRNA lentiviral plasmids were purchased from Sigma-Aldrich. 14-3-3σ and c-Myc shRNA lentiviruses were then produced from 293T cells following a protocol similar to that used for the tet-On Flag-14-3-3σ system. Infection efficiency was enhanced by adding 8 μg/ml Polybrene and subjecting the cells to centrifugation at 2000 rotations per minute for 60 minutes at 32°C. Knockdown efficiency was verified by Western blot.

Extracellular acidification and oxygen consumption rate
Cancer cell lines containing the retroviral 14-3-3σ tet-On system were maintained in the growth medium already described. The other cancer cell lines were maintained in a different growth medium: high-glucose DMEM containing 10% normal FBS, 2 mM L-glutamine, and 1% antibiotic-antimycotic solution. Oligomycin, an inhibitor of ATP synthase, was prepared from 1000X stock at a concentration of 10 mM in dimethyl sulfoxide (DMSO). FCCP, an ionophore and strong mitochondrial depolarizer, was prepared from 1000X stock at a concentration of 5 mM in DMSO. Rotenone, a potent inhibitor of mitochondrial complex I, and antimycin A, a strong suppressor of mitochondrial complex III, were solubilized from 1000X stock solutions at concentrations of 10 mM in DMSO. To measure ECAR and OCR, 3×10 4 cells from each cell line were seeded into each well of a XF24 microplate 16 hours before the experiment. Right before the assays, these cells were changed from a culture medium to an assay medium (low-buffered DMEM containing 25 mM D-glucose, 1 mM sodium pyruvate, and 1 mM L-glutamine) and incubated for 1 hour at 37°C. OCR and ECAR were measured by a Seahorse Extracellular Flux Analyser (Seahorse Biosciences). After baseline measurements, 75 μl of the above-mentioned inhibitors prepared in assay medium were sequentially injected into each well to reach the final working 1× concentrations. After 5 minutes of mixing to equally expose cancer cells to the chemical inhibitors, OCR and ECAR were measured again. Data were analysed by Seahorse XF software. OCR is reported in pmole/minute, ECAR in mpH/minute. OCR and ECAR measurements were normalized to either final cell number or protein concentration.

ATP production
ATP concentration was measured by the ATP Bioluminescence Kit CLS II (Roche) following the manufacturer's protocol. The principle of ATP Bioluminescence Kit relies on the following luciferase-catalyzed reaction: ATP + D-luciferin + O 2 → oxyluciferin + PPi + AMP + CO2 + light. The bioluminescent signal from this reaction is proportional to the ATP concentration and measured by a Clarity Luminescence Microplate Reader (BioTek). A standard curve with known ATP concentrations was used to determine the actual ATP levels in samples.

Glutaminolytsis assay
Glutaminolytic activity in cells was measured by using a glutaminolysis assay kit (Sigma Aldrich, Catalog# AA0100-1KT) that measures ammonia production as a surrogate for glutaminolysis. In this assay, L-glutamate dehydrogenase specifically catalyzes the reaction to convert ammonia and α-ketoglutaric acid to form L-glutamate. This biochemical reaction rapidly reduces nicotinamide adenine dinucleotide phosphate (NADPH), and also oxidizes nicotinamide adenine dinucleotide phosphate (NADP + ). The oxidation of NADPH, which is correlated and associated with the ammonia concentrations, leads to a reduction in absorbance at 340 nm. This decrease in absorbance at 340 nm is detected and quantified by spectrophotometers.

Mitochondrial mass measurement assays
Mitochondrial mass was quantified by staining with MitoTracker Green FM (Molecular Probes, Invitrogen) and analysis by a FACS Canto flow cytometer (BD Biosciences). MitoTracker Green FM is a green fluorescent dye associated with functional mitochondria. Flow cytometry data was analyzed using a FlowJo X software (FlowJo).

Real-time polymerase chain reaction
For quantitative real-time PCR, total RNA was collected by using Trizol Reagent (Invitrogen), and cDNA was synthesized by using an iScript cDNA Synthesis Kit (BioRad). Real-time PCR was performed using iQ-SYBR Green Supermix (BioRad), StepOnePlus Real-time PCR platform (Applied Biosystems, Inc.), and iCycler CFX96 Real-time PCR Detection System (BioRad). Actin and 18S rRNA were used as internal references for normalization. The nucleotide sequences of the primers used for real-time PCR were listed in Supplementary Table 4.

Dual luciferase reporter assay
c-Myc transactivational activity was measured by using a Dual-Luciferase Reporter Assay System (Promega) with a luciferase reporter plasmid containing a TERT (telomerase reverse transcriptase) promoter. TERT is a c-Myc target gene and its promoter contains an E-box c-Myc binding site. microPET Scan 7-week-old female nude mice bearing MDA-MB-231 tet-On 14-3-3σ breast cancer xenografts were imaged with a MicroPET Rodent-R4 scanner 30 minutes after injection with 400 μCi 18 FDG according to an imaging protocol used by the Small Animal Imaging Facility of MD Anderson. Tumours were also visualized with a Burker 7.0 Tesla magnetic resonance imager according to the facility's standard imaging procedures. The details of microPET Scan were described as below: MicroPET R4 system was used for these mice imaging studies. The PET images were acquired with energy windows set at 350-650 keV (for better scatter rejection). Coincidence timing window was set at 6 ns. Animals were in sequentially scanned. Each animal was injected around 0.4 μCi F-18 FDG radiotracers, with around 2.7 μCi/cm 3 concentration. It was followed by a 30-minute uptake time while animal was put in cage with little movement to reduce the muscle update.Each image was acquired 30 minutes per bed position. Image data were analyzed with ASIPro VM ™ mciroPET imaging analysis software. The image-quantitation calibration was measured with a uniform phantom with known activity concentration. A representative chart of imaging workflow is included (Supplementary Fig. 23). Three different ratios were calculated based on the maximal, minimal and mean values of image intensities measured within the selected region-of-interest (ROI). In this imaging experiment, when we compared the control mice's (without Flag-14-3-3σ induction) images to those of the treated mice (with Flag-14-3-3σ induction), there was a consistent 2:1 reduction of lesion/background image intensity ratios. The mean-calculated lesion (tumor) / background (normal tissue) ratios, which are more representative with less bias of single pixel value domination, were used to build our graphs.

C hyperpolarization
A mixture containing 20 μL (26 mg) of [ 13 C]pyruvic acid (Cambridge Isotopes) with 15 mM of trityl radical OX063 (GE Healthcare) was combined with 0.6 μL 50 mM Prohance (Bracco Diagnostic, Inc.) and polarized using a Hypersense DNP system (Oxford Instruments). In brief, the solution was frozen to approximately 1.4°K and irradiated at 94.154 GHz for 45-60 minutes in a 3.35T magnetic field. When the solid-state polarization reached a plateau, the solution was dissolved in 4 mL of a heated (180°C) solution containing 80 mM NaOH and 50 mM NaCl. The final solution containing 80 mM hyperpolarized pyruvate at a nominal temperature of 37°C and pH of 7.6 was flushed into a receptacle from which 200 μL was drawn for injection. The injection was administered to tumour-bearing and control mice over 10 seconds beginning within 12 seconds of sample dissolution.

Magnetic resonance spectroscopy
Animals were anesthetized using 0.5%-5% isoflurane in oxygen and then placed on a positioning sled with built-in channels for distribution of anesthesia gases and circulation of warm water to maintain animal temperature. The chemical fate of the hyperpolarized [ 13 C]pyruvate was monitored using a 7T Biospec USR7030 system (Bruker Biospin MRI, Inc.) with a dual-tuned 1 H-13 C volume coil with 72-mm inner diameter and a custom-built 13 C surface coil (12 mm inner diameter) placed over the tumour. Animal position was confirmed by using a 3-plane gradient-echo proton MRI sequence (TE = 3.6 ms, TR = 100 ms, 30° excitation angle), and the location of the tumour was confirmed in T2-weighted axial and coronal spin-echo images (TE = 50 ms, TR = 2500 ms, 1-mm slice thickness, 4 cm × 3 cm field-of-view encoded over a 256 × 192 matrix in 1-mm sections). Field homogeneity was maximized over the tumour volume using a point-resolved spectroscopy sequence. For 13 C measurements, a signal was generated by the volume coil and detected using the surface coil. Dynamic spectra were acquired by using a slice-selective single-pulse sequence (SW = 5000 Hz over 2048 points, 10° excitation, 8-mm slice thickness) that was repeated every 2 seconds for 3 minutes beginning just prior to injection of 13 C via tail-vein catheter. The spectra were phase-adjusted, and the area under the spectral peaks corresponding to the C1 of pyruvate and lactate were integrated to generate curves describing the arrival of hyperpolarized pyruvate and its conversion into hyperpolarized lactate. These curves were fit to kinetic models that account for chemical conversion. The model we used is a modification from the one described by Harris et al. 30 . Our model has two physical compartments to separate the effects of tracer circulation, extravasation, and activity. All modeling was performed by using Matlab (The Mathworks).

Statistical analyses
All bar graphs and line graphs represent means ± 95% confidence intervals or standard deviations unless otherwise stated. The error bars of ECAR and OCR line graphs from Seahorse Extracellular Flux Analyser are standard deviations. Student t-test was used to compare differences between two groups. One-way analysis of variance (ANOVA) was employed for comparison if there were more than two groups, unless otherwise stated. A p value less than 0.05 was considered to represent a statistically significant difference. Pearson correlation was used to identify positive and negative correlations between 14-3-3σ expression and expression of other genes. The log-rank (Mantel-Cox) test was used to compare survival times of breast cancer patients. Data were analysed by Prism (GraphPad software) and Sigma Graph (Systat Software) unless otherwise noted. All experiments were independently repeated at least 4 times.

Compliance information
All experiments involving cancer patient samples and information were done under clinical protocols approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center. All experiments and procedures involving animals were approved by the Institutional Animal Care and Use Committee of MD Anderson, which is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.  (a) High 14-3-3σ expression was associated with prolonged breast cancer patients' survival. 14-3-3σ protein expression levels in breast cancer patients at MD Anderson Cancer Center were evaluated by immunohistochemistry staining and matched with clinical data. 14-3-3σ protein expression in normal breast tissues was used as a reference to stratify patients. (b) Down-regulation of 14-3-3σ is associated with upregulation of major cancer hallmarks especially deregulation of cellular energetics (i.e., metabolic reprogramming). Transcriptomic profiles of breast cancer patients (cohort GSE20194, n=255, Gene Expression Omnibus database) were analysed by Nexus Expression 3.0 software (BioDiscovery). The gene expression profiles of the lowest 14-3-3σ quartile were compared to those of the highest 14-3-3σ quartile and matched with corresponding biological processes and cancer hallmarks. A Circos map (circos.ca) was used to display the biological processes and cancer hallmarks that were upregulated upon 14-3-3σ down-regulation. The size of cancer hallmarks' symbols indicated the magnitude of their upregulation upon 14-3-3σ silencing. The bar graph on the upper right corner shows a significant increase in cancer energy metabolism when 14-3-3σ expression is down-regulated. Enrichment scores were calculated based on transcriptomic analysis using Nexus Expression 3.0 program (BioDiscovery). Details about transcriptomic changes associated with 14-3-3σ downregulation are described in Supplementary Figure X. (c) Low 14-3-3σ expression is associated with increased oncogene expression and reduced tumour suppressors' levels. This Venn diagram highlights genes and biological processes with significantly changed expression when 14-3-3σ is down-regulated in breast cancer patients' tumours (cohort GSE20194, n= 255, Gene Expression Omnibus database). (d) High 14-3-3σ expression is accompanied by a decrease in 18 FDG uptake by human breast tumours. PET scan data and gene expression profiles of patients' breast tumours (collected at MDACC) are combined in this analysis to study the impact of 14-3-3σ on tumour glucose uptake in patients. Average±95% confidence interval (CI).
(e-f) Gene Set Enrichment Analysis (GSEA, Broad Institute) of breast cancer patients showed upregulation of metabolic genes in breast tumours that expressed a low level of 14-3-3σ compared to breast tumours with high 14-3-3σ expression.
were analyzed using a BD Biosciences FACS Canto flow cytometer. FlowJo X software was used to build Mitochondrial Tracker Green FM histograms. Non-induced MDA-MB-231 TetR 14-3-3σ cells were used as a control.