The metabolomic landscape plays a critical role in glioma oncogenesis

Abstract Cancer cells depend on metabolic reprogramming for survival, undergoing profound shifts in nutrient sensing, nutrient uptake and flux through anabolic pathways, in order to drive nucleotide, lipid, and protein synthesis and provide key intermediates needed for those pathways. Although metabolic enzymes themselves can be mutated, including to generate oncometabolites, this is a relatively rare event in cancer. Usually, gene amplification, overexpression, and/or downstream signal transduction upregulate rate‐limiting metabolic enzymes and limit feedback loops, to drive persistent tumor growth. Recent molecular‐genetic advances have revealed discrete links between oncogenotypes and the resultant metabolic phenotypes. However, more comprehensive approaches are needed to unravel the dynamic spatio‐temporal regulatory map of enzymes and metabolites that enable cancer cells to adapt to their microenvironment to maximize tumor growth. Proteomic and metabolomic analyses are powerful tools for analyzing a repertoire of metabolic enzymes as well as intermediary metabolites, and in conjunction with other omics approaches could provide critical information in this regard. Here, we provide an overview of cancer metabolism, especially from an omics perspective and with a particular focus on the genomically well characterized malignant brain tumor, glioblastoma. We further discuss how metabolomics could be leveraged to improve the management of patients, by linking cancer cell genotype, epigenotype, and phenotype through metabolic reprogramming.

Cancers of the brain have been particularly illuminating in providing insight into altered tumor metabolism. The brain is one of the most metabolically active organs in the body, using glucose, lactate, ketone bodies, fatty acids, and amino acids as fuel sources. 3 The reciprocal interaction among brain constituents including neurons and glial cells (such as astrocytes and oligodendrocyte precursor cells), heavily influences brain metabolic homeostasis. 4 Malignant brain tumors including glioblastoma (GBM) usurp the repertoire of metabolic networks in the brain for supporting their aggressive tumor growth.
For example, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) mediate synaptic contacts on glioma cells and neurons for their proliferation and invasion, 5 and perisynaptically located N-methyld-aspartate receptors (NMDARs), another type of ionotropic glutamate receptor, facilitate the growth of brain macro-metastases of breast cancer. 6 Furthermore, a small number of treatment-resistant glioma stem cells depend on distinct metabolic paths to form a niche within the intricate metabolic network in the brain. 7 Moreover, the unique dependencies in lipid metabolism formed by oncogene amplification in GBM may generate actionable metabolic vulnerabilities. [8][9][10][11] Recent metabolomic approaches involving the systematic measurement of metabolic enzymes and metabolites, have proven to be powerful tools to identify cancer biomarkers as well as drivers of tumorigenesis. 12 Furthermore, advanced technologies for in vivo metabolic analysis have been developed, including isotopelabeled metabolite tracing and noninvasive metabolic imaging, and these have permitted in vivo measurement of metabolic fluxes and abundances in tumor cells. 13 Interestingly, studies of the metabolomic landscape of cancer have unraveled the reciprocal interaction among each metabolic path, driven by cancer-specific alterations of genetics and epigenetics.
In this review, we primarily consider the metabolic landscape of cancer that has been derived from large-scale proteomic and metabolomic analyses, to provide a better understanding of the biology of cancer as well as to improve the diagnosis, monitoring, and treatment of cancer. An oncogenic phenotype could be formed by cancer metabolic reprogramming, the importance of which has been clarified by "multi-omics" research approaches including genetics, epigenetics, transcriptomics, proteomics and metabolomics ( Figure 1).

| Aberrant oncogenic signaling in cancer metabolic reprogramming
The complexities of neoplastic disease may be understood through the fundamental principles of the hallmarks of cancer. 14 Metabolic reprogramming is one such emerging core hallmark of cancer, 15,16 and comprehensive genomic studies are clarifying the regulators of cancer metabolism. 1 Constitutively activating mutations of phosphoinositide 3-kinase (PI3K)-Akt-mechanistic target of rapamycin (mTOR) signaling components are particularly prominent, and occur through several mechanisms including receptor tyrosine kinase (RTK) amplification and mutations, phosphatidylinositol 4,5-bisphosphate 3-kinase catalytic subunit alpha isoform (PIK3CA) mutations, and phosphatase and tensin homolog deleted from chromosome 10 (PTEN) loss. 17,18 mTOR is a serine/threonine kinase that merges growth factor receptor signaling into cell growth, proliferation and survival through two distinct multiprotein complexes: mTOR complex 1 (mTORC1), a well established protein translation and metabolism regulator, 19 and mTOR complex 2 (mTORC2), which was recently demonstrated to promote tumor growth and chemotherapy resistance in cancer cells independent from canonical Akt signaling. 20 F I G U R E 1 Multi-omics approaches to study the metabolic landscape of cancer cells. Genotype of the cancer cells is translated into a metabolic phenotype to facilitate cancer cell survival. This circuit can be studied at multiple omics levels including genomics, epigenomics, transcriptomics, proteomics, and metabolomics. Ac, acetyl-group; K, lysine residues; Me, methyl-group; Pol II, RNA polymerase II; P-TEFb, positive transcription elongation factor b; TF, transcription factor One of the master regulators of cancer metabolism is the oncogenic transcription factor, c-Myc. 21  converge the neoplastic phenotype of GBM cells into more aggressive behavior via reprogramming of cellular metabolism. 27,28 In addition to reprogramming metabolic circuits, the ability to sense nutrients in the microenvironment is necessary for cancer cells to exploit energy from metabolism. Of note, mTOR complexes play an important role in sensing these nutrients. 29 mTORC1 responds to a range of amino acids and relevant metabolites, including leucine and arginine. 19 Furthermore, we unraveled the novel role of mTORC2 in responding to glucose and acetate in the microenvironment through acetyl-coenzyme A (acetyl-CoA)-mediated acetylation of Rictor, the main component of mTORC2. 30 Using an additional, unbiased proteomics approach, we also showed that mTORC2 could suppress the activity of the cystine-glutamate antiporter, system X c transporter-related protein (xCT), indicating a new role for mTORC2 as a potential regulator of ROS metabolism. 31 This suggests that mTORC2 senses the availability of amino acids including glutamate and cystine, enabling tumor cells to buffer oxidative stress through glutathione, as necessary. These data lead to the proposal that glucose and amino acid metabolism interact in mTOR-activated cancer cells as dictated by the availability of nutrients. Apart from mTOR-dependent nutrient sensing, the adenosine monophosphate (AMP)-activated protein kinase (AMPK) pathway and hexosamine biosynthetic pathway (HBP)-hypoxia-inducible factor (HIF) axis are critical sensors of energy and nutrient status in cancer stem cells, 32 and nutrient sensing could therefore be the essential function to maximize the survival of cancer cells in various metabolic niches.

| Comprehensive view of the landscape of metabolic enzymes in cancer
A comprehensive approach for the evaluation of cellular metabolism is now based upon mass spectrometry (MS)-based proteomics, effectively linking cellular genotype and phenotype. Recent advancement in instrumentation as well as in bioinformatics has made it possible to quantify a repertoire of proteins simultaneously. 33 Following several studies on the large-scale generation of synthetic peptides, 34 and a more comprehensive project called ProteomeTools, 35 Matsumoto et al. have established an absolute quantitative approach to assess the metabolic landscape of cells. 36 This approach to define human proteomes relies on the generation of more than 18,000 recombinant proteins from human cDNA libraries to obtain proteotypic peptides for most human proteins, and the analytic platform is called iMPAQT (in vitro proteomeassisted multiple reaction monitoring (MRM) for protein absolute quantification). 36 Despite its mTRAQ approach with limited availability, the platform enables absolute quantification of the human proteome with internal peptide standards at known concentrations.
An absolute quantification approach with targeted proteomics including iMPAQT is a powerful tool to reveal the pathogenesis of various human cancer from a metabolic standpoint. In addition to its applicability to unravel a novel metabolic network of glutamine fate in malignant progression of cancer, 37 (Figure 3), and we detected the convergence of the path to the production of acetyl-CoA, which is the substrate closely associated with GBM biology through acetylation of nucleosomal histone tails, as well as nonhistone proteins through F I G U R E 2 Analysis on mTOR-dependent cancer metabolism by molecular-genetic versus proteomics approaches. mTORC1 promotes the glycolytic metabolism by activating hnRNPA1-dependent alternative splicing of a Myc-binding partner Delta Max, whereas mTORC2 signaling controls c-Myc transcription, translation, and protein level through the regulation of FoxO and microRNA. Quantitative proteome (iMPAQT) reveals that mTORC2 governs the Warburg effect in a comprehensive manner, including glycolysis, TCA cycle, and oxidative phosphorylation. FoxO, forkhead box O; hnRNPA1, heterogeneous nuclear ribonucleoprotein A1; iMPAQT, in vitro proteomeassisted MRM for protein absolute quantification; Max, myc-associated factor X; OXPHOS, oxidative phosphorylation; TCA, tricarboxylic acid a variety of metabolic pathways. 58 Furthermore, combined with iM-PAQT proteome data on the expression of metabolic enzymes that showed a significant upregulation of phosphoglycerate dehydrogenase (PHGDH) by mTORC2 (Figure 3), which coordinates serine synthesis and one-carbon unit fate, 59 our metabolome analysis also identified a shift of metabolites from one-carbon metabolism, the methyl-donor S-adenosylmethionine (SAM), which profoundly affects epigenetic changes including DNA and histone methylation for the survival of cancer cells (Figure 3). 60,61 Therefore, by combining proteomic and metabolomic data, we were able to obtain a more accurate spatio-temporal map of cancer metabolic activity.

F I G U R E 3
Metabolism-dependent epigenetic shifts in GBM analyzed by molecular-genetic versus metabolomic/proteomic approaches. Comprehensive metabolome approaches are useful for the identification of a specific hub of metabolites such as acetyl-CoA. Furthermore, combined with our iMPAQT proteome data on PHGDH upregulation which coordinates serine synthesis and one-carbon unit fate, the methyl-donor SAM was identified as mTORC2 targets, which can profoundly affect DNA and histone methylation status. Ac, acetyl-group; C-SCOPE, metabolome analysis by HMT Inc.; EGFRvIII, constitutively active form of EGFR mutant; iMPAQT, in vitro proteome-assisted MRM for protein absolute quantification; K, lysine residues; PDH, pyruvate dehydrogenase; PHGDH, phosphoglycerate dehydrogenase; PKM2, pyruvate kinase; M2; SAM, S-adenosylmethionine An important question is how metabolism and epigenetics are reprogrammed in IDH-wildtype tumors, especially the most malignant GBM. Genetic mutations of the histone H3 gene itself (e.g., H3K27M, H3G34R/V/D) has been reported to globally shift the epigenetic status of histone protein (e.g., H3K27me3, H3K36me3) in certain types of malignant brain tumors. 72 H3K27me3 is particularly important in the biology and diagnostics of brain neoplasms as its methylation status shifts in various types of tumors. 73 Our studies demonstrated that metabolic reprogramming, potentially thorough the aforementioned EGFR-mTOR axis, could significantly affect the metabolism-dependent epigenome in cancer cell by regulating key metabolic enzymes as well as multiple intermediary metabolites. 62 Essential histone modifications are represented by acetylation on the in the maintenance of GBM stem cells. 75 Alteration in the expression of epigenetic-modifying genes themselves including lysine/arginine methyltransferases, as well as acetyltransferases/deacetylases, contributes to GBM pathogenesis. 76 Future studies are needed to unravel the mechanisms by which cancer cells survive in various niches through EGFR/mTOR-and other metabolic regulator-dependent dynamic shifts in their epigenetic landscapes.

| ALL ROADS LE AD TO ME TABOLIS M: ME TABOLIS M A S AN ON COG ENI C PHENOT YPE
Cancer mutations reprogram intracellular metabolism through facilitated expression of metabolic enzymes as well as intermediary metabolites, which subsequently shifts the epigenome of cancer cells.
Intriguingly, the epigenetic reprogramming via metabolic change exerts its oncogenic effect by modulating cancer-specific metabolism, meaning that cancer metabolism could be reprogrammed, both genetically and epigenetically. We recently unraveled cancer-specific metabolism that is epigenetically driven by global shifts in the histone landscape through metabolic reprogramming. 77,78 One of the major marks in actively transcribed promoters is acetylation at the ninth lysine residue of histone H3 (H3K9ac). We recently demonstrated that GBM cells with activated EGFR-mTORC2 signaling increased H3K9ac through metabolic reprogramming in cooperation with histone-modifying enzymes including pyruvate dehydrogenase (PDH) and class IIa histone deacetylases (HDACs) (Figure 4). 77 Comprehensive studies with RNA-seq and ChIP-seq analyses revealed that the mTORC2-dependent increase in H3K9ac was uniquely induced at the promoter of iron metabolism genes including ferritin, transferrin receptor, divalent metal transporter 1, and hepcidin ( Figure 4). 77 The mechanisms by which intracellular iron accumulation leads to cell survival await investigation, 79 but our data and others suggest that epigenetic regulation of iron metabolism is essential for the induction of stemness in cancer cells (Figure 4). 80 Of F I G U R E 4 Epigenetic regulations of cancer-prone metabolism as a central oncogenic phenotype. Oncogene signaling reprograms central carbon metabolism at the outset. The effect of carbon metabolism reprogramming is far reaching, and globally shifts the epigenome of cancer cells including histone modifications and DNA methylation. Genome-wide epigenetic changes eventually govern each specific, effector metabolism for tumor cell survival through the induction of cancer cell stemness. Ac, acetyl-group; ac, acetylation; K, lysine residues; Me, methyl-group; met, methylation note, comprehensive metabolomic analyses revealed that, in addition to histone acetylation, DNA methylation and histone methylation (H3K27me3) could contribute to tumor aggressiveness by rewiring intracellular metabolic pathways such as those for glutamine, methionine, and ROS metabolism, leading to the maintenance of cancer stem cells. 81 Furthermore, other epigenetic modulators of metabolism include a member of the sirtuin families, which was found mutated in different human cancer, suggesting its tumor suppressive function. 82,83 Therefore, epigenetic regulation of metabolism could be a prevalent phenomenon in cancer, supported by EGFR/mTORsignaling as well as other epigenetic regulators. These findings lead to the fascinating proposal that oncogene signaling first reprograms far reaching phenomena such as central carbon metabolism, which eventually governs each specific, effector metabolism to adapt to a variety of environments through a genome-wide epigenetic shift ( Figure 4).

| CON CLUS I ON AND FUTURE PER S PEC TIVE: IN S I G HTS FROM MULTI -OMI C S ANALYS E S IN C AN CER
Cancer development, progression, and therapy response are profoundly influenced by intracellular metabolism and the exogenous microenvironment. This variably shifts the epigenetic landscape, including DNA methylation and histone modifications.
Interestingly, multi-omics analyses of cancer cells revealed that all pathways including genomics, epigenomics, transcriptomics, and proteomics converge on cancer metabolism, which is potentially the most important executioner of the oncogenic phenotype.
Importantly, cancer research is moving from a genotype-based static picture to a dynamic view in which genotype and tissue context interact to define the metabolic repertoire of tumor cells. 84 A complete "metabolic catalog" of human cancer should also be established through the development of the Cancer Cell Line Encyclopedia (CCLE), integrating quantitative analyses of 225 metabolites in 928 cell lines from more than 20 cancer types by liquid chromatography-mass spectrometry (LC-MS). 85 This effort, associated with unbiased and quantitative approaches of proteome and metabolome, enables association analyses linking the cancer metabolome to genetic alterations, epigenetic features, and gene dependencies. Cooperative, multidisciplinary, and translational approaches will be needed to translate metabolic insights into better treatments for cancer patients.

ACK N OWLED G M ENTS
We thank the Department of Neurosurgery, Tokyo Women's Medical University for biospecimen and biorepository support. This work is supported by Japan Society for the Promotion of Science KAKENHI Grant JP19K07649 (KM). PSM is supported in part by grants U24CA264379 and RO1 CA238249 from the National Institutes of Health (NIH) and a grant from the National Brain Tumor Society.

CO N FLI C T S O F I NTE R E S T
PSM is a co-founder of Boundless Bio, Inc. He has equity in the company and chairs the scientific advisory board, for which he is compensated. PSM is also a consultant for Sage Therapeutics, Asteroid Therapeutics, and Autobahn Therapeutics, and scientific co-founder and consultant for Pretzel Therapeutics, Inc. WKC is co-founder of Interleukin Combinatorial Therapeutic, Inc., InVaMet, Inc., and io0, LLC. KM and NS do not have any conflict of interest.