REVIEWS

2-Oxoglutarate-dependent dioxygenases in cancer

Julie-Aurore Losman1,2, Peppi Koivunen3 and William G. Kaelin Jr. 1,4 ✉ Abstract | 2-Oxoglutarate-dependent dioxygenases (2OGDDs) are a superfamily of enzymes that play diverse roles in many biological processes, including regulation of hypoxia-inducible factor-mediated adaptation to hypoxia, extracellular matrix formation, epigenetic regulation of gene transcription and the reprogramming of cellular metabolism. 2OGDDs all require oxygen, reduced iron and 2-oxoglutarate (also known as α-ketoglutarate) to function, although their affinities for each of these co-substrates, and hence their sensitivity to depletion of specific co-substrates, varies widely. Numerous 2OGDDs are recurrently dysregulated in cancer. Moreover, cancer-specific metabolic changes, such as those that occur subsequent to mutations in the genes encoding succinate dehydrogenase, fumarate hydratase or isocitrate dehydrogenase, can dysregulate specific 2OGDDs. This latter observation suggests that the role of 2OGDDs in cancer extends beyond cancers that harbour mutations in the genes encoding members of the 2OGDD superfamily. Herein, we review the regulation of 2OGDDs in normal cells and how that regulation is corrupted in cancer.

Enantiomer It has been known for many decades that cancer cells specific metabolites, including 2OG and its structur- One of two molecules that have display characteristic alterations in metabolism and ally related metabolites. For example, EGLN prolyl the same atomic formula and epigenetics. Many cancers divert glucose carbons 4-hydroxylases (also known as PHDs) and the FIH1 the same sequence of atomic towards glycolysis (canonical anaerobic metabolism) asparaginyl hydroxylase are 2OGDDs that act as cellular bonds but that differ in their 3D orientations insofar as they are and away from oxidative phosphorylation (canonical oxygen sensors by regulating the hypoxia-inducible tran- mirror images of each other. aerobic metabolism) even when oxygen is available scription factors HIF1α and HIF2α (collectively referred (known as the ‘Warburg effect’)1, and cancer genomes to hereafter as ‘HIFα’)13. HIF transcriptionally regulates Hypoxia often display global DNA hypomethylation as well hundreds of genes, including genes that contribute to A deficiency in the amount of as focal locus-specific increases in DNA and histone the Warburg effect and genes linked to DNA and histone oxygen being supplied to body 2,3 tissues. methylation . The question of whether such changes methylation. Other 2OGDDs play direct roles in the con- in metabolism and epigenetics actually cause cancer trol of DNA (TET and ABH enzymes) and histone (KDM 1 Department of Medical was controversial until oncogenic driver mutations enzymes) methylation, as well as mRNA processing Oncology, Dana-Farber Cancer Institute and Brigham were identified in metabolic and epigenetic genes. It is (FTO) and protein translation (OGFOD1, MINA53 14 and Women’s Hospital, now clear that some cancers are caused by mutations and NO66) . Boston, MA, USA. in the genes encoding fumarate hydratase (FH)4, suc- Some 2OGDDs are directly dysregulated in cancer, 2Division of Hematology, cinate dehydrogenase (SDH)5 and isocitrate dehydro- by amplification, silencing, deletion or mutation of Department of Medicine, genase (IDH)6,7, which lead to the accumulation of the their encoding genes (Table 1). Other 2OGDDs appear Brigham and Women’s 2-oxoglutarate (2OG; also known as α-ketoglutarate) to be indirectly dysregulated in cancer, by hypoxia Hospital, Boston, MA, USA. analogues fumarate, succinate and the R enantiomer and/or by the action of aberrantly accumulated metab- 3Faculty of Biochemistry (Fig. 1a) and Molecular Medicine, of 2-hydroxyglutarate (R-2HG), respectively . olites that possess pro-oncogenic activities (so-called Biocenter Oulu, Oulu Center Similarly, some cancers are caused by mutations in ‘oncometabolites’). Several 2OGDDs promote or sup- for Cell-Matrix Research, genes encoding epigenetic regulators such as the press tumour growth in preclinical cancer models, fur- University of Oulu, Oulu, EZH2 H3K27 methyltransferase8, the KMT2A H3K4 ther implicating the dysregulation of 2OGDD activity in Finland. methyltransferase9, the TET2 DNA hydroxylase10,11 and oncogenesis. However, much remains unknown about 4 Howard Hughes Medical the KDM6A H3K27 lysine demethylase12. the roles of specific 2OGDDs in cancer. The study of Institute (HHMI), Chevy Chase, MD, USA. 2-Oxoglutarate-dependent dioxygenases (2OGDDs) these enzymes has been further motivated by the obser- ✉e-mail: William_Kaelin@ are a superfamily of enzymes that sit at the nexus of vation that their activities can be modulated by small dfci.harvard.edu cancer metabolism and cancer epigenetics (Box 1). molecules, suggesting that 2OGDDs could serve as ther- https://doi.org/10.1038/ These enzymes have the potential to sense oxygen, apeutic targets in cancer. Herein, we review our current s41568-020-00303-3 reactive oxygen species (ROS), iron availability and knowledge about the metabolic regulation of 2OGDD

Nature Reviews | Cancer Reviews

activity and discuss how dysregulation of these activi- The 2OGDD reaction ties contributes to tumorigenesis. We also discuss some 2OGDDs all share the same reaction mechanism but of the outstanding questions in the field that warrant act on different substrates, including proteins, DNA, further investigation. RNA, fatty acids and other small molecules15 (Box 1).

a Pyruvate O HO PDH PC Oxaloacetate O CS Citrate O OH HO OH HO OH O O O O HO O

MDH ACO

Malate Isocitrate OH O HO OH HO OH O O O OH HO O

FH IDH

Fumarate Succinate 2OG Glutamate

O O O NH2 HO HO HO OH HO OH OH OH OGDC GDH O SDH O O O O O SCS Mutant LDH IDH R-2HG S-2HG OH OH HO OH HO OH

O O O O b DHA Ascorbate

Fe2+ Fe3+

Substrate Substrate 2OGDD Substrate Substrate

OH

O O O HO Uncoupled 2OG decarboxylation HO OH + O OH + CO 2 2 2OG decarboxylation O O coupled with hydroxylation 2OG Succinate

Fig. 1 | 2OG and its analogues and a schematic of the 2OGDD reaction. a | The tricarboxylic acid (TCA) cycle intermediate 2-oxoglutarate (2OG) and its structurally similar analogues. b | The substrate of the 2-oxoglutarate-dependent dioxygenase (2OGDD) reaction becomes hydroxylated in a reaction utilizing three co-substrates: divalent iron (Fe2+), which is coordinated to the catalytic site by two conserved histidine residues and a positively-charged arginine or lysine residue; 2OG; and

molecular oxygen (O2), which provides the oxygen atom for the hydroxyl group. During catalysis, 2OG becomes decarboxylated

to succinate and CO2. The hydroxylated substrate can undergo further non-enzymatic modification, such as demethylation. Structural 2OG analogues, including fumarate, succinate, R-2-hydroxyglutarate (R-2HG) and S-2-hydroxyglutarate (S-2HG), have the potential to act as competitive inhibitors of 2OGDDs. Ascorbate is not a direct co-substrate of the 2OGDD reaction but supports the reaction by preventing inadvertent iron oxidation from occurring during an uncoupled reaction. ACO, aconitase; CS, citrate synthase; FH, fumarate hydratase; IDH, isocitrate dehydrogenase; LDH, lactate dehydrogenase; MDH, malate dehydrogenase; OGDC, oxoglutarate dehydrogenase complex; PC, pyruvate carboxylase; PDH, pyruvate dehydrogenase; SCS, succinyl-CoA synthase; SDH, succinate dehydrogenase.

www.nature.com/nrc Reviews

Box 1 | List of 2OGDDs and other proteins that utilize and regulate 2OG 2-Oxoglutarate-dependent dioxygenases • Protein hydroxylases: ASPH, EGLN1–3, FIH1, JMJD4–7, LEPRE1, LEPREL1, LEPREL2, MINA53, NO66, OGFOD1, P4HA1–3, P4HTM and PLOD1–3 • Histone demethylases: KDM2A, KDM2B, KDM3A, KDM3B, KDM4A, KDM4B, KDM4C, KDM4D, KDM5A, KDM5B, KDM5C, KDM5D, KDM6A, KDM6B, KDM6C, KDM7A, KDM7B and KDM9 • Nucleic acid oxygenases: ABH1, ABH2, ABH3, ABH5, ABH8, FTO, TET1–3 and TYW5 • Fatty acid and small-molecule oxygenases: BBOX1, PHYH and TMLHE • Unassigned catalytic function: ABH4, ABH6, ABH7, ASPHD1, ASPHD2, HSPBAP1, JARID2, JMJD8, KDM3C, OGFOD2, OGFOD3, PHF2 and PHYHD1 Other proteins involved in 2OG metabolism • Transaminases (see figure, part a): AADAT, ABAT, AGXT, AGXT2, BCAT1, BCAT2, CCBL1, CCBL2, GFPT1, GFPT2, GOT1, GOT2, GPT, GPT2, OAT, PSAT1 and TAT • Dehydrogenases (see figure, part b): AASS, ADHFE1, ALDH1B1, ALDH2, ALDH3A2, ALDH7A1, ALDH9A1, DHTKD1, DLD, GDH1, GDH2, IDH1, IDH2, IDH3A, IDH3B, IDH3G, OGDH, OGDHL, PHGDH, (d/R)-2HGDH and (l/S)-2HGDH • Transporters: SLC22A6, SLC22A7, SLC22A8, SLC22A11, SLC22A20, SLC22A25, SLC22A13 and SLC22A12 • Other: DLST and NIT2

a L-Leucine 2-Oxoglutarate 4-Methyl-2-oxopentanoate L-Glutamate

O O O NH2 + HO OH + HO OH OH OH

NH2 O O BCAT O O O

b L-Glutamate 2-Oxoglutarate NAD+ NADH + H+ NH2 O HO OH HO OH

O O + O O H2O NH4 GDH

2OG, 2-oxoglutarate; 2OGDD, 2-oxoglutarate-dependent dioxygenase.

2OGDDs all require the same co-substrates — dioxygen its enantiomer S-2HG, can act as competitive inhibitors 20–23 (O2), which provides the oxygen atom for hydroxylation, of 2OGDDs , modulating 2OGDD activity in both divalent iron (Fe2+) and 2OG — and yield a hydroxylated physiologic and pathophysiologic states (Figs 1a and 2; (Fig. 1b) Table 2 product, CO2 and succinate . Catalysis follows an ). Although other mechanisms of regulation ordered sequence. First, active site-bound Fe2+ coordi- of 2OGDDs have been reported, including alternative nates 2OG binding in a bidentate manner. Next, substrate splicing and translation initiation24–26, post-translational binding to the active site displaces an Fe2+-ligated water modification (for example, sumoylation27,28 and molecule, thereby allowing oxygen to bind16,17. Then, phosphorylation29–31) and differential expression of bind- 32,33 oxidative decarboxylation of 2OG forms succinate, CO2 ing partners , their contributions to the regulation of and a ferryl (Fe4+) intermediate that reacts with the sub- 2OGDD activity are less well understood. strate’s C–H bond, resulting in hydroxylation and reduc- tion of Fe4+ to Fe2+. Finally, the hydroxylated product is Oxygen released from the catalytic site, followed by the release of Pseudo-hypoxia as a driver of HIF-mediated oncogenesis. succinate. The hydroxylated product can then undergo The importance of dysregulated oxygen signalling in further non-enzymatic modifications such as demethyl- tumorigenesis is perhaps best exemplified by the role of ation. Although reducing agents (for example, ascorbate, HIF in tumours that harbour loss-of-function mutations glutathione and cysteine) are not direct cofactors of the in the gene encoding the von Hippel–Lindau tumour 2OGDD reaction, they support catalysis by preventing suppressor protein (pVHL), including clear cell renal cell inadvertent iron oxidation and, in the case of EGLN, by carcinoma, haemangioblastoma and paraganglioma34. In Oncometabolites Intermediates of metabolism preventing oxidation of intramolecular cysteine residues normal cells under normoxic conditions, the α-subunit 18,19 that abnormally accumulate in that are required for catalytic activity . of HIF (HIFα) is prolyl-hydroxylated by members of the cancer cells and that promote The affinity of specific 2OGDDs for oxygen, iron EGLN family of enzymes (EGLN1, EGLN2 or EGLN3). tumorigenesis. and 2OG varies, providing a mechanism of regulation This hydroxylation promotes HIFα binding to pVHL, of subsets of 2OGDDs at the level of co-substrate avail- which is the substrate recognition subunit of an E3 Bidentate (Fig. 2; Table 2) A bidentate ligand is a base ability . In addition, several endogenous ubiquitin ligase that ubiquitylates HIFα and targets it for that donates two pairs of 2OG analogues, including pyruvate, citrate, isocitrate, degradation. Oxygen is an indispensable co-substrate electrons to a metal atom. succinate, fumarate, malate, oxaloacetate, R-2HG and for EGLN enzymes, and their catalytic activity is highly

Nature Reviews | Cancer Reviews

Table 1 | 2OGDDs implicated in cancer 2OGDD Cancer types Genetic and/or transcriptional Functional evidence refs alteration in cancer Tumour-suppressive function EGLN1 PG Mutated and inactivated Loss induces PG-like lesions 44,45 JARID2a AML secondary to Lost by chromosomal deletion Loss promotes blast transformation of MPN to AML in PDX 206 MDS and MPN and GEMM models KDM2Ba Aggressive brain Silenced Negatively regulates ribosomal RNA synthesis to suppress 207 tumours cell proliferation KDM3B MDS and AML Lost by chromosomal deletion Overexpression represses AML colony formation 143,208 KDM5Aa AMKL, GBM Translocation partner with NUP98 Inhibits malignant phenotypes in glioma 68,69,209 in paediatric AMKL; silenced in GBM KDM5Ba AML –b Loss promotes malignant phenotypes 210 KDM5C Many cancer types Mutated and inactivated Maintains genomic stability; negative regulator 71,211–213 of enhancer activity KDM6Aa Many cancer types Mutated and inactivated Regulates super-enhancers; loss promotes malignant 12,63,70–72 phenotypes KDM6Ba PAAD Loss of heterozygosity Relieves differentiation arrest, promotes C/EBP-dependent 87,88 differentiation TET1 Many cancer types Mutated and inactivated, silenced Loss induces lymphoma, loss promoted malignant 89–91,214,215 phenotypes TET2 Many cancer types Mutated and inactivated Promotes malignant phenotypes 10,11,92,93,216 Oncogenic function ASPH HCC, PAAD, CHOL Overexpressed Promotes malignant phenotypes 217–219 FTO MLL-rearranged Overexpressed Promotes proliferation and migration, and inhibits 220,221 AML, advanced CESC apoptosis JARID2a HCC, BLCA Overexpressed Promotes EMT; promotes colony formation and 222,223 invasion JMJD6 BRCA, NBL Amplified and overexpressed Cooperates with MYC to enhance malignant phenotypes 224,225 KDM2A BRCA, LUAD Amplified and overexpressed Promotes malignant phenotypes and enhances 226,227 ERK signalling KDM2Ba PAAD Overexpressed Cooperates with KRAS; dependency in synovial sarcoma 228,229 KDM3C AML –b Dependency in AML1-ETO-positive and MLL-AF9-positive 230,231 AML KDM4A Many cancer types Amplified and overexpressed Promotes chromosomal copy gain and activation of 184,191,232,233 hormone receptors KDM4B OV, EGC Overexpressed Promotes malignant phenotypes; overexpression 234,235 promotes Jun function KDM4C BRCA, ESCA, Amplified (solid tumours); Promotes malignant phenotypes; dependency in AML 184–189 lymphoma, AML translocation partner with IGH locus (lymphoma) KDM5Aa Many cancer types Amplified and overexpressed Promotes malignant phenotypes and chemotherapy 169–174,176,202 resistance; dependency in Rb1-mutated and Men1-mutated tumours KDM5Ba Many cancer types Amplified and overexpressed Promotes malignant phenotypes and chemotherapy 83,236 resistance KDM6Aa Aggressive BRCA Overexpressed Promotes malignant phenotypes 182,237,238 KDM6Ba Haematologic Overexpressed Promotes haematopoietic stem cell self-renewal; 72,239,240 cancers dependency in T-ALL KDM7B LUAD, PRAD Overexpressed Promotes malignant phenotypes 241–243 MINA53 Many cancer types Overexpressed Promotes malignant phenotypes by activating cyclins and 244,245 cyclin-dependent kinases 2OGDD, 2-oxoglutarate-dependent dioxygenase; AMKL, acute megakaryoblastic leukaemia; AML, acute myeloid leukaemia; BLCA, bladder urothelial carcinoma; BRCA, breast cancer; C/EBP, CCAAT enhancer binding protein; CESC, cervical cancer; CHOL, cholangiocarcinoma; EGC, gastric cancer; EMT, epithelial-to-mesenchymal transition; ESCA, oesophageal cancer; GBM, glioblastoma multiforme; GEMM, genetically engineered mouse model; HCC, hepatocellular carcinoma; IGH, immunoglobulin heavy chain; LUAD, lung adenocarcinoma; MDS, myelodysplastic syndrome; MEN1, multiple endocrine neoplasia 1; MPN, myeloproliferative neoplasm; NBL, neuroblastoma; OV, ovarian cancer; PAAD, pancreatic cancer; PDX, patient-derived xenograft; PG, paraganglioma; PRAD, prostate cancer; Rb1, retinoblastoma 1; T-ALL, T cell acute lymphoblastic leukaemia. aEvidence for tumour-suppressive and oncogenic functions. bNone reported.

www.nature.com/nrc Reviews

2+ a O2 Fe 2OG BBOX1 BBOX1 ABH2 TET2 BBOX1 TET2 250 EGLN1 TET2 20 EGLN1 200 18 180 TET1 200 EGLN2 16 TET1 160 EGLN1 TET1 14 EGLN2 140 150 12 PLOD1 120 EGLN2 PLOD1 EGLN3 10 100 100 8 80 PLOD1 EGLN3 PHYH FIH1 6 PHYH 60 EGLN3 50 4 40 2 20 0 0 0 P4HA1 KDM4A PHYH FIH1 P4HA1 FIH1

KDM6B KDM4B KDM6B KDM2A P4HA1 KDM4A KDM6A KDM4C KDM6A KDM4A KDM5D KDM5A KDM6B KDM4B KDM5C KDM4B KDM5C KDM5B KDM6A KDM5B KDM5B KDM4C b Fumarate Succinate R-2HG S-2HG EGLN1 EGLN1 ABH2 ABH2 0 0 0 TET2 0 BBOX1 TET2 EGLN2 TET2 EGLN2 TET2 ABH3 200 2,000 2,000 2,000 4,000 TET1 400 EGLN1* 4,000 4,000 TET1 BBOX1 600 TET1 EGLN3 TET1 EGLN3 6,000 800 6,000 6,000 8,000 P4HA1 EGLN2* 10,000 1,000 8,000 P4HA1 FIH1 1,200 8,000 12,000 1,400 P4HA1* 10,000 FIH1 P4HA1* 10,000 FIH1 14,000 KDM6B 1,600 EGLN3* KDM6B KDM2A KDM6A FIH1 KDM6B KDM4A KDM6B KDM4A KDM6A KDM4A KDM5B KDM2A KDM6A KDM4B KDM6A KDM4B KDM5B KDM4B KDM4C KDM4A KDM5B KDM5B KDM4C KDM4B

Fig. 2 | Kinetic and inhibitory values of 2OGDD for co-substrates and 2OG analogues. a | Km (μM) values of oxygen, iron and 2-oxoglutarate (2OG). b | IC50 (μM) values of fumarate, succinate, R-2-hydroxyglutarate (R-2HG) and S-2-hydroxyglutarate

(S-2HG). *Ki value. High Km or low IC50 values, suggestive of potential regulatory roles for the corresponding co-substrates with respect to the corresponding 2-oxoglutarate-dependent dioxygenases (2OGDDs), are in the outer perimeter. In the

case of 2OGDDs for which divergent Km/IC50 values have been reported, the values presented are those determined by our laboratories or are an average of all reported values. For full details, see Table 2.

dependent on oxygen availability. Even a modest decline independent methods under conditions in which HIFα in cellular oxygen levels inhibits EGLN activity and is robustly hydroxylated did not detect hydroxylation of HIFα hydroxylation. As a result, HIFα accumulates, any non-HIFα proteins50. It should also be noted that dimerizes with its partner protein ARNT (also called roxadustat, the first-in-class clinical EGLN inhibitor, has HIF1β) and transcriptionally activates HIF-responsive been approved for the treatment of anaemia secondary genes. Loss-of-function VHL mutations likewise pre- to renal failure51,52, and several other EGLN inhibitors vent the degradation of hydroxylated HIFα, resulting in have advanced to the final phase of clinical develop- hypoxia-independent stabilization of HIFα and consti- ment with no major side effects. This suggests that tutive expression of HIF-target genes. pVHL likely has EGLN enzymes are not promiscuous hydroxylases with functions unrelated to HIF, and there is some evidence pleiotropic functions. that loss of HIF-independent pVHL function can con- The oncogenic consequences of constitutive HIF tribute to paraganglioma formation35,36. However, dys- activation clearly demonstrate that the transcriptional regulation of HIFα, and particularly HIF2α, appears response to hypoxia can promote tumorigenesis. to be a fundamental mechanism of pathogenesis in However, it is important to keep in mind that VHL VHL-mutant kidney cancers37–43. and EGLN1 mutations create a state of pseudo-hypoxia Interestingly, loss-of-function EGLN1 mutations and because, in VHL-mutant and EGLN1-mutant tumours, gain-of-function mutations in EPAS1 (the gene encoding HIF activation is independent of cellular oxygen levels. HIF2α), although rare, have been identified in human The questions remain whether hypoxia can directly tumour tissue samples of paragangliomas44–48 (Table 1). contribute to tumorigenesis and, if so, whether it does This observation and the finding that formation of so by dysregulating 2OGDDs. With respect to the paraganglioma-like lesions after inactivation of EglN1 in former question, numerous conditions are associated rodents is HIF2-dependent49 suggest that these tumours, with chronic hypoxaemia, including life at high altitude, like VHL-mutated paragangliomas, are driven by HIF. haemo­globinopathies and certain cardiopulmonary Whether loss-of-function EGLN1 mutations also pro- diseases, and these conditions are not associated with a mote tumorigenesis by HIF-independent mechanisms is conspicuously increased risk of cancer. not known. EGLN enzymes have been reported to have Hypoxaemia A state in which the level of many substrates other than HIF. However, a recent study Hypoxia as a driver of HIF-mediated oncogenesis. oxygen in the blood is lower assessing the hydroxylation of these putative non-HIFα Tissue oxygen concentrations are typically 1–10% than normal. substrates by recombinant EGLN enzymes using three whereas tumours can be profoundly hypoxic, with

Nature Reviews | Cancer Reviews

Table 2 | Km and IC50 values of 2OGDDs for co-substrates​ and 2-oxoglutarate analogues

Km (μM) IC50 (μM) 2+ 2OGDD Fe 2OG O2 Fumarate Succinate R-2HG S-2HG ABH2 – 4 (ref.149) – – – 420–500 150 (ref.149) (refs149,246) ABH3 – – – – – 500 (ref.246) – BBOX1 10 (ref.247) 100–300 (refs247,248) 55 (ref.247) – – 13,200 (ref.149) 140 (ref.149) EGLN1 0.05 (ref.94) 1–270 (refs23,249,250) 65–250 (refs17,24,250), 80 (ref.23) 510 (ref.23) 7 ,300 (ref.149), 420–1,150b >450 (ref.249) 300a (ref.139) (refs139,149) EGLN2 0.05 (ref.94) 2 (ref.23) 230 (ref.24) 120 (ref.23) 830 (ref.23) 210a (ref.139) 630b (ref.139) EGLN3 0.1 (ref.94) 10 (ref.23) 230 (ref.24) 60 (ref.23) 570 (ref.23) – 90b (ref.139) FIH1 0.5 (ref.251) 25–150 (refs250,251) 90–240 (refs250–252) >10,000 (ref.23) >10,000 (ref.23) 1,100–1,500 190–300 (refs139,149) (refs139,149) KDM2A – 6 (ref.149) – – – 110 (ref.149) 50 (ref.149) KDM4A <0.1 (ref.22) 6–25 (refs22,65,66,149) 55–170 (refs63,65,66) 1,500–2,300 800 (refs22,125) 2–160 25–290 (refs22,125) (refs22,149,253) (refs22,149) KDM4B <0.1 (ref.22) 6 (ref.22) 150 (ref.63) >5,000 (ref.22) 2,300 (ref.22) 150 (ref.22) 450 (ref.22) KDM4C – 4–10 (refs65,149) 160 (ref.65) – – 80 (ref.149) 95 (ref.149) KDM5A – – 90 (ref.63) – – – – KDM5B <0.1 (ref.22) 10 (ref.22) 40 (ref.63) >5,000 (ref.22) 1,400 (ref.22) 3,600–10,870 630–1,600 (refs22,138) (refs22,138) KDM5C – 5 (ref.254) 35 (ref.63) – – – – KDM5D – – 25 (ref.63) – – – – KDM6A <0.1 (ref.22) 8–10 (refs22,255) 200 (ref.63) 3,000 (ref.22) 270 (ref.22) 180 (ref.22) 180 (ref.22) KDM6B 6 (ref.22) 8–50 (refs22,255) 25 (ref.63) >5,000 (ref.22) 550 (ref.22) 350 (ref.22) 750 (ref.22) P4HA1 2 (ref.256) 20 (ref.256) 40 (ref.256) 190b (ref.23) 400b (ref.257) 1,800 (ref.139) 310 (ref.139) PHYH – 50–190 (refs250,258,259) 95 (ref.250) – – – – PLOD1 2 (ref.260) 100 (ref.260) 45 (ref.260) – – – – TET1 5 (ref.21) 55 (ref.21) 0.3–30 (refs21,67) 390 (ref.21) 540 (ref.21) 4,000 (ref.139) 1,000 (ref.139) TET2 4 (ref.21) 60 (ref.21) 0.5–30 (refs21,67) 400 (ref.21) 570 (ref.21) 5,000 (ref.139) 1,600 (ref.139) 2OG, 2-oxoglutarate;​ 2OGDD; 2-​oxoglutarate-dependent​ dioxygenase; R-2HG, R enantiomer of 2-hydroxyglutarate;​ S-2HG, S enantiomer of 2-hydroxyglutarate.​ a b Km values for R-2HG as a co-substrate.​ Ki value.

oxygen concentrations of less than 2% and, in some cases, substrates, and it is speculated that these alternative sub- less than 0.1%53. Under hypoxic conditions, specific strates function to sequester FIH1 from HIFα, thereby 2OGDDs with low affinity for oxygen (that is, with high promoting HIF activity57,58. It should be noted that

O2 Km values) are inhibited whereas 2OGDDs with although FIH1 can regulate the transcriptional activity

high affinity for oxygen (that is, with low O2 Km values) of HIF ex vivo, it is less clear whether FIH1 plays a phys- remain active (Fig. 2a; Table 2). EGLN enzymes are the iologic role in regulating the cellular response to hypoxia

major cellular oxygen sensors, having O2 Km values in vivo. Genetic deletion of Fih1 in mice does not affect above atmospheric oxygen concentrations13. On the canonical HIF-regulated phenotypes such as angiogen- opposite end of the spectrum are the collagen prolyl esis or erythropoiesis. Rather, mice lacking Fih1 display 4-hydroxylases (P4HA1–3), which have high affinities a generalized metabolic dysregulation that appears to be for oxygen and are therefore catalytically active even in neurogenically driven59. profound hypoxia. The asparaginyl hydroxylase FIH1 has intermediate affinity for oxygen54. It remains active Hypoxia and chromatin modifying enzymes. The under moderately hypoxic conditions that are sufficient extent to which other 2OGDDs can act as direct oxy- to affect EGLN activity but becomes inactive under gen sensors is an area of active investigation. Although Km (Michaelis constant). The more severe hypoxia. Hydroxylation of HIFα proteins hypoxia-associated histone hypermethylation is a substrate concentration by FIH1 blocks association of HIFα with the tran- well-described phenomenon60–62, until recently it was required for an enzymatic scriptional co-activators CREB-binding protein (CBP) not clear whether this hypermethylation is a direct reaction to achieve one-half of and histone acetyltransferase p300, thereby inhibiting consequence of inhibition of oxygen-sensitive JmjC its maximum velocity in vitro; HIF-mediated transcription55,56. FIH1 also hydroxylates domain-containing histone lysine demethylases a high Km value indicates low affinity and a low Km value asparagine, aspartate and histidine residues on a large (KDM), which are 2OGDDs, or an indirect effect of indicates high affinity. pool of ankyrin repeat domain-containing non-HIF hypoxia on chromatin structure. We recently showed

www.nature.com/nrc Reviews

that KDM6A, but not its paralogue KDM6B, is highly numerous technical factors. This is well-exemplified oxygen-sensitive (Fig. 2a; Table 2) and that KDM6A reg- by EGLN enzymes, for which longer, more ‘natural’, sub-

ulates cellular differentiation in an oxygen-dependent strates yield lower O2 Km values than do short synthetic 63 17 but HIF-independent manner . Mechanistically, we peptides . The differences in Km values for oxygen that identified structural differences in the JmjC catalytic have been observed with different length substrates are domains of KDM6A and KDM6B that likely explain likely because substrate binding precedes oxygen binding their differential oxygen-sensing capacities (Fig. 3). and affects the structure of the oxygen binding site16,17. It A companion study found that acute inactivation of is also possible that the oxygen-sensing capacities of spe- KDM5A rapidly induces H3K4 methylation that pheno­ cific 2OGDDs in vivo are affected by post-translational copies the effects of acute hypoxia64. This is consistent modifications, accessory proteins or metabolic factors

with the observation that KDM5A has a high O2 Km that are typically absent when performing in vitro assays value (Fig. 2a; Table 2). Specific KDM4 paralogues have with recombinant proteins. also been reported to have oxygen-sensing capabilities, Interestingly, numerous oxygen-sensing KDM

although there are conflicting data regarding their O2 Km enzymes are either genetically inactivated or overex- values, especially in the case of KDM4A63,65,66 (Table 2). pressed in cancer (Table 1). This raises the possibility In the case of the TET family of DNA hydroxylases, two that, in hypoxic tumours, metabolic dysregulation of

independent studies determined that the O2 Km values KDM enzymes impacts tumour behaviour. For example, of TET1 and TET2 are very low21,67 (Fig. 2a; Table 2), KDM5A has been shown to inhibit the migration and suggesting that these 2OGDDs do not act as oxygen invasion of glioma cells and is transcriptionally silenced sensors21,67. However, tumour hypoxia has been reported in aggressive gliomas68,69. Hypoxia-associated inhibi- to cause DNA hypermethylation via reduced TET tion of KDM5A catalytic activity could be an alternative activity67. The basis for these inconsistencies is not clear, mechanism by which KDM5A is downregulated as brain although it should be noted that the latter study was per- tumours outgrow their blood supply. Similarly, KDM6A formed at 0.5% oxygen and, as stated above, tumours is mutated and inactivated in numerous tumour types can be even more profoundly hypoxic. It is also known and has been shown to function as a tumour suppres- 12,63,70–72 (Table 1) that in vitro O2 Km measurements can be affected by sor in vitro and in vivo . It is possible that, in KDM6A wild-type tumours, the hypoxic tumour a microenvironment directly downregulates KDM6A KDM6A 1099 QLHELTKLPAFVRVVSAGNLLSHVGHTILGMNTVQLYMKVPGSRTPGHQENNNFCSVNIN activity and thereby promotes tumour progression. KDM6B 1343 QLQELLKLPAFMRVTSTGNMLSHVGHTILGMNTVQLYMKVPGSRTPGHQENNNFCSVNIN **:** *****:**.*:**:**************************************** Fine-tuning the hypoxic response in tumours. KDM6A IGPGDCEWFVVPEGYWGVLNDFCEKNNLNFLMGSWWPNLEDLYEANVPVYRFIQRPGDLV Profound intratumoural hypoxia should, theoreti- KDM6B IGPGDCEWFAVHEHYWETISAFCDRHGVDYLTGSWWPILDDLYASNIPVYRFVQRPGDLV *********.* * ** .:. **:::.:::* ***** *:*** :*:*****:******* cally, inhibit oxygen-sensing KDMs, including KDMs such as KDM5A, KDM6A and KDM4, that have KDM6A WINAGTVHWVQAIGWCNNIAWNV 1241 KDM6B WINAGTVHWVQATGWCNNIAWNV 1485 tumour-promoting functions in some tumour con- ************ ********** texts (Table 1). This raises the intriguing possibility that hypoxic tumours engage specific mechanisms to main- b KDM6A KDM6B tain the activity of tumour-promoting KDMs, such as through enhanced 2OG availability (discussed in fur- ther detail below) or through increased KDM expres- sion. Indeed, numerous 2OGDDs are transcriptional targets of HIF60. Given that transcriptional programmes induced by HIF vary across cell types73, it is possible that induction of specific 2OGDDs by HIF fine-tunes the 2OGDD response to hypoxia in a cell type-specific manner. Although it is certainly plausible that upreg- ulated expression of hypoxia-sensitive 2OGDDs repre- sents a mechanism by which cancer cells compensate for the loss of tumour-promoting 2OGDD activities under hypoxic conditions, it is important to bear in mind that aggressive tumours often outgrow their blood supply, become hypoxic and upregulate HIF. Caution must Fig. 3 | Structural basis for the differential oxygen-sensing capacities of KDM6a and therefore be exercised when ascribing cause-and-effect KDM6B JmjC domains. a | Alignment of KDM6A isoform 3 (NP_066963.2) and KDM6B relationships to correlations between HIF-responsive isoform 1 (NP_001073893.1) residues with ClustalW shows a high degree (79.2%) of 2OGDD overexpression and patient outcomes. sequence identity (NCBI protein database). b | The JmjC domain crystal structures for It should be noted that the contribution of oxygen KDM6A (PDB: 3avr)261 and KDM6B (PDB: 5oy3)262 were obtained from the RCSB protein databank, and composite images of the active sites, with added oxygen molecules (red) to 2OGDD regulation is not limited to those 2OGDDs to illustrate scale, were created using PyMOL (PyMOL Molecular Graphics System, that are direct oxygen sensors (Fig. 4). As discussed Version 2.0; Schrödinger, LLC). These structures show that the size of the oxygen-binding below, hypoxia can produce secondary effects on cellular cavity opening is smaller in KDM6A than KDM6B, offering a potential explanation for metabolism that result in the accumulation of metab- 74,75 76–79 the differential O2 sensing by the two closely-related enzymes. The bound substrate is olites, including succinate and S-2HG , that can indicated in yellow (KDM6A) or orange (KDM6B). inhibit 2OGDDs that are relatively oxygen-insensitive

Nature Reviews | Cancer Reviews

a Normoxia HIF regulation in normoxia Substrate 2OG O2 2OG HIFα High O -K O2 2 m OH 2OGDD EGLN Substrate OH Ubiquitylation and proteolytic degradation HIFα OH

2OG O2 Substrate HRE Low O2-Km 2OGDD OH Substrate

b Moderate hypoxia HIF regulation in moderate hypoxia

2OG 2OG HIFα High O2-Km Substrate 2OGDD EGLN HIF- responsive HIFα HIFβ 2OGDD Substrate HRE 2OG 2OG O2

Low O2-Km 2OGDD OH LDH Substrate S-2HG

c Severe hypoxia HIF regulation in severe hypoxia Inhibitory cofactor HIFα

EGLN High O2-Km Substrate 2OGDD HIF- Malate responsive HIFα HIFβ 2OGDD O TCA HRE 2OG 2 cycle O2 LDH pH Low O2-Km Substrate Succinate 2OGDD S-2HG

Active Decreased activity Strongly inhibited

Fig. 4 | Dysregulation of 2OGDD activity by hypoxia. The extent to which a given 2-oxoglutarate-dependent dioxygenase (2OGDD) is inhibited by hypoxia is a function of multiple factors, including their level of expression, their oxygen affinity and their sensitivity to inhibition by succinate and S-2-hydroxyglutarate (S-2HG). a | Under normoxic

conditions, hypoxia-sensitive (high O2-Km) and hypoxia-resistant (low O2-Km) 2OGDDs are active (indicated in dark blue). b | Moderate hypoxia results in decreased activity (indicated in medium blue) of only hypoxia-sensitive 2OGDDs, including EGLN enzymes. EGLN inhibition results in induction of HIF transcriptional activity. c | More profound hypoxia results in more profound inhibition (indicated in light blue) of hypoxia-sensitive 2OGDDs and increased expression of HIF target genes, including lactate dehydrogenase (LDH) and a subset of 2OGDD. Production of S-2HG by LDH is potentiated by hypoxia and by cellular acidosis, resulting in accumulation of high concentrations of S-2HG. In some cell types, severe hypoxia also dysregulates tricarboxylic acid (TCA) cycle function, which results in enhanced production of succinate. S-2HG and succinate that accumulate under hypoxia promote the inhibition of 2OGDDs that are both hypoxia-sensitive and hypoxia-resistant. 2OG, 2-oxoglutarate; HRE, HIF-responsive element.

(Fig. 2; Table 2). It should also be noted that 2OGDDs In normal cells, the complex interplay between oxy- are not the only oxygen sensors in cells. ADO, a gen availability and 2OGDD activity provides a criti- non-2OG-dependent cysteine oxidase, signals oxygen cal mechanism of cellular adaptation. We are just now availability by oxidizing amino-terminal cysteine res- starting to appreciate how these same mechanisms are idues in substrate proteins to cysteine sulfinic acid80. co-opted by cancer cells to promote their survival as they Cysteine sulfinic acid modifications are involved in outgrow their blood supply and become progressively regu­lating redox balance, circadian rhythms and protein more hypoxic. stability81. Situated as they are at the interface between oxygen Iron and diverse biological processes, 2OGDDs directly link Dietary intake is the body’s only source of iron, and changes in oxygen availability to changes in chromatin cellular iron availability is primarily regulated by structure, gene expression and other biological functions. recycling82. The hepcidin–ferroportin axis controls the

www.nature.com/nrc Reviews

in-flow of iron from the intestine, liver and iron-storing that it may be possible to augment the function of macrophages, and plasma transferrin distributes iron to specific 2OGDDs with specific reductants. tissues. Normal plasma iron concentrations are ~20 µM, Cancer cells have increased levels of ROS relative to whereas total iron binding capacity is ~threefold higher. normal cells, and it has been suggested that redox stress Conditions that dysregulate transferrin and hepcidin inhibits 2OGDDs that constrain tumour growth and expression, such as inflammation, hinder normal iron thereby promotes tumorigenesis97 (Table 1). Although recycling83. It is possible that functional iron deficiency the precise mechanisms by which cancer-associated inhibits 2OGDDs that require high iron concentrations redox stress leads to 2OGDD dysfunction are unclear, for catalysis, such as peroxisomal phytanoyl-CoA dioxy­ studies have suggested that oxidation of Fe2+ plays a genase (PHYH; also known as PHAX), γ-butyrobetaine role98. There is a growing body of evidence that subclin- dioxygenase (BBOX1; also known as BODG), KDM6B, ical ascorbate deficiency can promote tumorigenesis by TET1 and TET2 (ref.84) (Fig. 2; Table 2). Whether this inhibiting 2OGDD activity. Although ascorbate does not contributes to the well-established link between chronic appear to be required for the physiologic regulation of inflammation and cancer is not known85,86, although it EGLN enzymes in mice99, ascorbate supplementation is interesting to note that several highly iron-dependent downregulates HIF levels in tumour cells ex vivo100 and 2OGDDs, including KDM6B, TET1 and TET2, suppress suppresses tumour growth in vivo in a HIF-dependent tumour growth in specific cellular contexts87–93 (Table 1). manner101–103. In the case of TET enzymes, ascorbate Most of the body’s iron is utilized by erythrocytes deprivation in Gulo–/– mice impairs TET function and and, as an integral component of haemoglobin, iron expands the haematopoietic stem cell compartment plays a central role in oxygen delivery. Iron deficiency whereas ascorbate supplementation suppresses mutant can therefore indirectly lead to cellular hypoxia and, Tet2-mediated leukaemogenesis, presumably by aug- potentially, to inhibition of hypoxia-sensitive 2OGDDs. menting residual TET2 function and activating other Other divalent metals such as Zn2+, Cu2+, Mg2+, Mn2+, TET paralogues92,104. Ascorbate supplementation also Cd2+, Ni2+ and Co2+ can compete with iron and inhibit restores TET function and slows the growth of renal 2OGDDs in vitro and in cells94,95. Other metals, such as cell cancer cells in vitro and in vivo105,106. Interestingly, Cr6+, that do not compete with iron but do oxidize ascor- the antitumour effects of ascorbate supplementation bate (see below) can also inhibit 2OGDDs96. Whether have been observed even in mice that express wild-type these divalent metals play a role in vivo in regulating or GULO92, suggesting that ascorbate is limiting even in dysregulating 2OGDD activity is not known. tumour cells that are able to synthesize ascorbate. It is possible that the increased requirement for ascorbate Ascorbate under 21% oxygen ex vivo and in tumours in vivo is ROS are chemically reactive by-products of normal due to an increased rate of ascorbate oxidation in these cellular metabolism. The major sources of ROS in cells settings. are the mitochondrial electron transport chain and plasma membrane NADPH oxidases (NOX). ROS are 2-Oxoglutarate produced by the transfer of free electrons to molecular 2OG is a tricarboxylic acid (TCA) cycle intermediate oxygen. This results in the production of superoxide, that is produced by oxidative decarboxylation of iso­ which is converted to hydrogen peroxide by superoxide citrate by IDH enzymes (IDH1 in the cytoplasm, IDH2 dismutase (SOD). Hydrogen peroxide can react with and IDH3 in the mitochondria) and oxidative deamina- redox-sensitive cysteine residues on proteins or can tion of glutamate by glutamate dehydrogenase. The con- be metabolized to water by antioxidant proteins such centration of 2OG in cells is estimated to be in the high as glutathione peroxidases (GPX). Hydrogen peroxide micromolar to low millimolar range under physiologi-

can also react with reduced iron, resulting in oxidation cal conditions. Given that the measured 2OG Km values of iron molecules and production of highly reactive for 2OGDDs are in the low micromolar range (Fig. 2b; hydroxyl radicals that are damaging to cells. Table 2), it is generally assumed that 2OG is not limiting 2OGDDs require a reducing agent, typically ascor- for 2OGDD activity in cells. However, this assumption

bate, to prevent spurious iron oxidation and to prevent may be flawed. 2OG Km values are typically measured in oxidation of intramolecular cysteine residues (Fig. 1b). the absence of endogenous 2OG-competitive inhibitors

Humans depend on dietary ascorbate because they do (see below), and in vivo 2OG Km values may therefore be not express functional gulonolactone oxidase (GULO), quite different from those measured in vitro. Moreover, the enzyme that catalyses the terminal step in ascor- accurate measurement of metabolite concentrations in bate biosynthesis. The importance of ascorbate for living cells is very challenging. Cellular volumes fluctu- 2OGDD function is best exemplified by the disease ate dramatically and vary significantly from cell to cell, scurvy, wherein ascorbate deficiency impairs collagen even in clonal cell populations, and metabolites, includ- cross-linking due to decreased activity of the P4HA ing 2OG, turn over very rapidly and exist at different family of 2OG-dependent collagen prolyl hydroxylases. concentrations in different subcellular compartments, This leads to bleeding gums and poor wound healing. with a considerable amount of intracellular 2OG being Although other reducing agents, such as reduced glu- sequestered in mitochondria107. Finally, 2OG levels tathione and dithiothreitol (DTT), can substitute for may themselves be subject to regulation by, for exam- ascorbate in vitro, there appears to be significant varia- ple, regulation of intracellular transport of 2OG or the bility in the capacity of different reducing agents to sup- shuttling of 2OG between subcellular compartments. port the activities of specific 2OGDDs18. This suggests This all begs the question of whether 2OGDD activity

Nature Reviews | Cancer Reviews

is sensitive to changes in intracellular concentrations where it can inhibit the activity of succinate-sensitive of 2OG. We recently reported that acute inhibition of 2OGDDs (Fig. 2b; Table 2) and thereby amplify the EGLN activity, either by genetic deletion of Egln1 or by direct effects of hypoxia on 2OGDD activity21 (Fig. 4). It small-molecule EGLN inhibition, increases total levels has also been reported that maintenance of a high 2OG of intracellular 2OG in cell lines and in mouse liver and to succinate ratio stimulates TET and KDM activity in muscle108. These results suggest that, at least under mouse embryonic stem cells116–118. However, the spe- some conditions, 2OG levels in cells are limiting and cific impact of 2OG and succinate on embryonic stem are a function of 2OG utilization. Interestingly, another cell function is still somewhat unclear, with conflicting recent study found that physiologic changes in the levels reports suggesting that high 2OG levels promote either of intracellular 2OG can indeed modulate the activity of pluripotency116,117 or differentiation118. specific 2OGDDs109. Whether dysregulation of 2OG Loss-of-function mutations in FH and SDH have homeostasis directly contributes to tumorigenesis is not been identified in various cancers, including papillary known, although one recent study found that the p53 renal cell carcinoma and paraganglioma4,5. These muta- tumour suppressor promotes the activities of specific tions result in the accumulation of high levels of intracel- 2OGDD tumour suppressors, including TET enzymes, lular fumarate and succinate, respectively, which act as by upregulating 2OG levels110. oncometabolites at least in part by modulating the activ- ities of 2OGDDs. For example, fumarate and succinate Fumarate and succinate inhibit EGLN enzymes and induce aberrant HIF stabi- The ability of 2OG to act as a co-substrate to support lization and constitutive HIF activation in cell lines and 2OGDD reactions is not merely a function of 2OG in primary FH and SDH-mutant human tumours119–122. availability. It is also a function of the relative concen- As described above, kidney cancers and paragangliomas trations of 2OG and its endogenous analogues, includ- display constitutive activation of HIF secondary to loss ing the TCA cycle intermediates fumarate and succinate. of pVHL34, and gain-of-function mutations in EPAS1 Fumarate and succinate are structurally and chemi- have been identified in paragangliomas46–48. Based on cally similar to 2OG (Fig. 1a) and can therefore act as these observations, it would be reasonable to conclude competitive inhibitors of 2OG-dependent reactions111. that HIF similarly functions as an oncogenic driver Interestingly, similar to their oxygen sensitivities, the in FH and SDH-mutant tumours as in VHL-mutant sensitivities of 2OGDDs to fumarate and succinate vary tumours. However, a cautionary note is provided by dramatically (Fig. 2b; Table 2). studies of Fh-deficient mice, where the formation of The modulation of 2OGDD activity by fumarate pre-malignant renal cysts is unaffected by concurrent appears to play an important role in cellular physiol- loss of Epas1 and is worsened by concurrent loss of Hif1a ogy. Radiation-induced DNA damage and subsequent (ref.123). Cyst formation in Fh-deficient mice appears activation of DNA-PK has been shown to recruit cyto- to be driven by upregulation of antioxidant signalling plasmic FH to sites of double-strand DNA damage, as a consequence of the covalent attachment of fuma- where FH is phosphorylated and inhibited112,113. This rate to specific cysteine residues in KEAP1, thereby results in local accumulation of fumarate, which inhib- preventing KEAP1 from downregulating NRF2. It its KDM2B and promotes H3K36 methylation and the remains unknown whether FH and SDH-mutant human subsequent recruitment and repair of DNA breaks by tumours are dependent on HIF transcriptional activity. non-homologous end-joining machinery112. Although Besides EGLN enzymes, numerous other 2OGDDs these findings suggest that fumarate can suppress tumor- are inhibited by fumarate and succinate in vitro and in igenesis by enhancing DNA repair, another study found, FH and SDH-deficient cells124–126 (Fig. 2; Table 2), and sev- on the contrary, that fumarate and other 2OG-like onco- eral of these 2OGDDs, including TET1 and TET2, have metabolites inhibit DNA repair and enhance the sensi- tumour suppressor activity90–92 (Table 1). However, it tivity of tumour cells to inhibitors of poly(ADP-ribose) remains to be elucidated which, if any, of these 2OGDDs polymerase (PARP)114. This study found that oncometab- are functionally relevant tumour suppressors in FH and olite inhibition of KDM4B results in hypermethylation SDH-mutant tumours. of H3K9 loci at sites of DNA damage, which impairs the recruitment of factors that mediate homology-directed R-2-Hydroxyglutarate repair of DNA double-strand breaks. Fumarate has also In addition to fumarate and succinate, another struc- been implicated in promoting innate immune memory tural analogue of 2OG, R-2HG, is an oncometabolite. by inhibiting KDM5 enzymes and enhancing H3K4 R-2HG is a reduced form of 2OG and is produced at methylation115. very low levels in normal cells as a by-product of cel- Physiologic 2OGDD activity is also susceptible to lular metabolism127. IDH1 and IDH2 are homodimeric modulation by succinate. SDH, in addition to oxidiz- enzymes that catalyse the reversible oxidative decarboxy­ ing succinate to fumarate in the TCA cycle (Fig. 1a), also lation of isocitrate to 2OG (Fig. 1a). Cancer-associated plays a critical role in the respiratory electron transport mutations in IDH, at Arg-132 of IDH1 and Arg-140 or chain as Complex II (that is, succinate–ubiquinone Arg-172 of IDH2, alter the catalytic activity of IDH such reductase). Succinate levels are induced in hypoxic that the mutant enzymes catalyse an irreversible reac- cardiomyocytes by aspartate and glutamate anaple- tion in which 2OG is reduced but not carboxylated, rosis into the TCA cycle and by reversal of SDH and resulting in the production of R-2HG128. Complex II activity74,75. This accumulated succinate IDH1 and IDH2 mutations are present in a wide is transported out of the mitochondria to the cytosol, range of cancers, including gliomas, chondrosarcomas,

www.nature.com/nrc Reviews

cholangiocarcinomas and acute myeloid leukaemias (Table 2), and that R-2HG blunts hypoxia-induced sta- (AML)6,7. In AML, R-2HG is necessary and sufficient bilization of HIFα129,139. In keeping with these observa- to mediate the oncogenic effects of mutant IDH129–131, tions, HIF levels are low in IDH1-mutant gliomas157,158 and drugs that inhibit mutant IDH activity have been and in T cells exposed to R-2HG159. It should be noted approved for the treatment of IDH1 and IDH2-mutant that loss of Hif1a potentiates tumour growth in an IDH AML132,133. R-2HG is hypothesized to transform cells wild-type orthotopic brain tumour mouse model160, sug- by competitively inhibiting the binding of 2OG to gesting that HIF1α functions as a tumour suppressor in 2OGDDs that function as tumour suppressors. The brain tumours more generally, not just in the context of best-validated target of R-2HG in cancer is TET2, a DNA IDH-mutant glioma. hydroxylase that converts 5-methylcytosine (5mC) to How R-2HG activates EGLN mechanistically has not 5-hydroxymethylcytosine (5hmC)134,135. TET enzymes been definitively established. Although our biochemical can further oxidize 5hmC to generate 5-formylcytosine studies suggest that EGLN1 stimulates rapid oxidation of (5fC) and 5-carboxylcytosine (5caC)136,137. Several R-2HG to 2OG, which is then decarboxylated to succi- lines of evidence suggest that mutant IDH promotes nate by canonical EGLN1 activity139, others have reported leukaemogenesis at least in part by inhibiting TET2. that R-2HG can be converted to 2OG even in the absence The catalytic activity of TET2 is inhibited by R-2HG of EGLN1, by prolonged incubation of R-2HG with high in vitro138,139, and primary IDH-mutant AML cells dis- concentrations of iron and reducing agents161. Either play global loss of 5hmC (refs140,141). Furthermore, loss mechanism would explain the unexpected finding that of TET2, like accumulation of R-2HG, is sufficient to transformation by R-2HG is associated with enhanced promote cytokine-independence and block the differ- EGLN activity129,139. entiation of haematopoietic cells129,142. Finally, IDH1, IDH2 and TET2 mutations are common, but are largely S-2-Hydroxyglutarate mutually exclusive, in de novo AML11,142. Taken together, 2HG is a five-carbon dicarboxylic acid with a chiral these observations suggest that mutant IDH1, mutant centre at the second carbon atom, resulting in two pos- IDH2 and TET2 loss activate similar oncogenic path- sible enantiomers of 2HG, R-2HG (that is, d-2HG) and ways and that TET2 is a functionally important target S-2HG (that is, l-2HG) (Fig. 1a). Like R-2HG, S-2HG of R-2HG in IDH-mutant AML. Interestingly, R-2HG is is a by-product of a normal cellular metabolism. It is actually quite a poor TET inhibitor, with an IC50 value produced by malate dehydrogenase (MDH) upon con- of ~5 mM139,143 (Fig. 2b; Table 2). Nonetheless, the obser- version of oxaloacetate to malate in the TCA cycle162 vation that R-2HG levels in IDH-mutant tumours can be (Fig. 1a). S-2HG levels are normally very low in cells. as high as 10 mM144–146 and the wealth of genetic evidence However, under hypoxic and acidic conditions, pro- that IDH and TET2 mutations activate similar leukae- miscuous substrate utilization by lactate dehydrogenase mogenic pathways strongly implicate TET2 as a func- (LDH) can promote the reduction of 2OG to S-2HG76–78 tionally important R-2HG target in IDH-mutant AML. (Figs 1a and 4), resulting in the accumulation of low But what about other IDH-mutant tumour types, such millimolar concentrations of S-2HG77. as IDH-mutant chondrosarcoma, that do not show evi- Hypoxia-induced S-2HG appears to play an impor- dence of TET inhibition and 5hmC loss147, and tumours, tant role in antitumour immunity79. In response to T cell such as glioma, that do not harbour loss-of-function receptor triggering, CD8+ T cells switch to a glycolytic mutations in TET2 (ref.148)? It is possible that R-2HG metabolic programme and produce S-2HG. S-2HG promotes transformation by targeting 2OGDDs other inhibits EGLN enzymes and thereby induces HIF, which than TET enzymes in these IDH-mutant solid tumours. upregulates LDH expression, further amplifying S-2HG Numerous other 2OGDDs besides TET2 have been production. The accumulated S-2HG inhibits TET2 proposed as potential tumour suppressor targets of activity, which promotes CD8+ T lymphocyte effec- R-2HG in IDH-mutant tumours22,138,149–151, with atten- tor function. Although this process does occur under tion particularly focused on KDM enzymes. R-2HG normoxic conditions, S-2HG production is strongly can inhibit numerous KDM enzymes in vitro and potentiated by hypoxia (Fig. 4). Given that tumours in vivo150,152–156, and many KDM enzymes are actually and inflammatory tissues are generally hypoxic, and significantly more sensitive to inhibition by R-2HG given that TET enzymes do not appear to be direct than is TET2 (Fig. 2b; Table 2). Moreover, several KDM oxygen sensors (Fig. 2; Table 2), this hypoxia–S-2HG enzymes are recurrently mutated or otherwise down- feedback loop could be a mechanism by which CD8+ regulated in cancer (Table 1), suggesting that these T lympho­cytes mitigate the immunosuppressive effects enzymes can function as tumour suppressors and that of intratumoural hypoxia and enhance their antitumour their inhibition by R-2HG could contribute to mutant activity79,163,164. IDH-mediated transformation. However, little evi- There is evidence to suggest that S-2HG, like R-2HG, dence currently exists to directly functionally impli- can function as an oncometabolite. In renal cell car- cate specific KDM enzymes in mutant IDH-mediated cinoma, S-2HG levels are elevated and 5hmC levels IC50 transformation. are decreased due to downregulation of expression of (Half-maximal inhibitory EGLN enzymes were initially reported to be inhib- l-2HG dehydrogenase (L2HGDH)165, an enzyme that concentration). The ited by R-2HG, implying that HIF acts as an oncogenic metabolizes S-2HG to 2OG162. Ectopic expression of concentration at which an 149 inhibitor half-maximally driver in IDH-mutant tumours . However, we found L2HGDH results in decreased accumulation of S-2HG, inhibits a biochemical that, on the contrary, R-2HG can act as an alternative restoration of 5hmC and suppression of cell proliferation reaction in vitro. 2OG-like co-substrate to potentiate EGLN activity and colony formation.

Nature Reviews | Cancer Reviews

Other 2OG-dependent cellular processes inhibit the viability of breast cancer cells190, although it 2OGDDs are not the only enzymes that require 2OG. is not yet known whether these antitumour effects are Numerous other enzymes (Box 1), including amino on-target. KDM4 family members have also been linked acid transaminases and components of the 2OG dehy- to potentiation of oestrogen receptor and androgen drogenase complex, are involved in 2OG metabolism receptor signalling in breast cancer and prostate can- and can potentially be affected by 2OG availability and cer, respectively184,191, and KDM4 inhibition suppresses endogenous 2OG analogues. We recently reported that androgen receptor signalling and prostate cancer cell the 2OG-dependent transaminases BCAT1 and BCAT2 proliferation in preclinical models192,193. are potently inhibited by R-2HG166. BCAT enzymes cat- In addition to 2OGDD ‘onco-enzymes’, numerous alyse the reversible transamination of branched-chain other 2OGDDs are potential therapeutic targets in can- amino acids to produce branched-chain α-ketoacids cer by virtue of the fact that their activities oppose the in a reaction that interconverts 2OG and glutamate. functions of tumour suppressors that are lost in cancer. BCAT enzymes and glutaminase are the two principal For example, loss of the MEN1 tumour suppressor pro- sources of glutamate in glial cells, and inhibition of tein attenuates the function of the H3K4 histone lysine BCAT by R-2HG induces a dependence on glutaminase methyltransferase KMT2A (also known as MLL)170. in IDH-mutant glioma. Consistent with these findings, We found that murine tumours driven by Men1 loss a similar metabolic dependency can be induced by direct grew significantly slower when the H3K4 histone lysine inhibition of BCAT1 in IDH wild-type glioma cells167. It demethylase Kdm5a (also known as RBP2) was concur- is not known whether other 2OG-like oncometabolites rently inactivated, compared with those Men1-deficient affect non-2OGDD-mediated 2OG-dependent cellu- tumours in which KDM5A was still expressed. Given lar processes. However, given the critical importance that KMT2A and KDM5A have opposing functions, it of 2OG in cells, perturbations in 2OG metabolism are is possible that loss of KDM5A impairs Men1-deficient likely to have wide-ranging biological effects. tumour growth by restoring H3K4 methylation homeo- stasis. Along similar lines, it might be possible to target Target validation and therapeutic intervention the loss of tumour-suppressive 2OGDDs by inhibit- 2OGDDs can be inhibited with drug-like small mole- ing the enzymes that oppose them. For example, pre- cules, such as compounds that act as competitive inhibi- clinical and early clinical data support the use of DNA tors of 2OG. This has motivated efforts to therapeutically demethylating agents in TET2-mutant myelodysplastic target oncogenic 2OGDDs. For example, loss of the retino­ syndrome and AML194,195. blastoma tumour suppressor protein (pRB) dysregu- Past clinical trials have not found ascorbate sup- lates the H3K4 demethylase KDM5A, which blocks plementation to be efficacious for cancer prevention differentiation and cellular senescence168, and genetic or treatment196. However, there is renewed interest in ablation of Kdm5a significantly retards the develop- ascorbate as an anticancer agent due to recent advances ment of Rb1–/– murine tumours169,170. KDM5A and its in our understanding of the role of ascorbate in 2OGDD paralogue KDM5B have also been implicated in the function197. Numerous ascorbate clinical trials are cur- development of cancer drug resistance and cancer ‘stem- rently underway in patients with diverse cancers based ness’171–176 (Table 1). Studies suggest that KDM5 inhibitors on the rationale that ascorbate can reactivate 2OGDD will not only have antitumour activity but will also aug- tumour suppressors. However, these trials are using high ment the activity of other anticancer drugs. This has led doses of intravenous ascorbate that achieve plasma levels

to the development of numerous tool compounds that that are orders of magnitude above the ascorbate Km val- selectively inhibit KDM5 enzymes177–179. ues for 2OGDDs, which are in the range of 100–300 µM24. KDM6 H3K27 demethylases have also been pro- It should be noted that such high levels of ascorbate posed as therapeutic targets in cancer. A study found that can actually deplete reduced glutathione and paradoxi- KDM6B and KDM6A have opposing functions in T cell cally increase redox stress198. It is therefore possible that acute lymphoblastic leukaemia (T-ALL), with the for- enhanced redox stress, independent of 2OGDD reactiva- mer acting as an oncoprotein and the latter as a tumour tion, could contribute to any observed antitumour activ- suppressor72,180. A subsequent study argued that KDM6A ity of high-dose ascorbate, especially in tumours that are is an oncoprotein in T-ALL protein 1 (TAL1)-positive, susceptible to oncogene-induced redox stress199,200. On but not TAL1-negative, T-ALL181. These findings pro- the other hand, it is possible that high ascorbate levels vide a rationale for targeting KDM6A in TAL1-positive could result in cardiotoxicity similar to that observed in T-ALL and KDM6B in TAL1-negative T-ALL. There is clinical trials of high-dose vitamin E for the prevention also emerging interest in targeting KDM6A in breast of cardiovascular disease201. It will be important to define cancer182 and in histone 3.3 K27M-positive paediatric the patients most likely to benefit from ascorbate, such brainstem glioma183. as those with tumours that harbour loss-of-function Several members of the KDM4 family of H3K9 2OGDD mutations (for example, TET2), and to iden- demethylases are overexpressed and sometimes ampli- tify the optimal level of ascorbate repletion for those fied in various cancers, including breast cancer184–189 patients. Additionally, other approaches to augmenting (Table 1). Both gain-of-function and loss-of-function 2OGDD function, for example with allosteric activators, experiments suggest that KDM4C promotes mammary may provide a way to specifically enhance the activities transformation, which is associated with increased of 2OGDD tumour suppressors without incurring the expression of stem cell markers186. Tool compounds that risks of activating oncogenic 2OGDDs or the risks of inhibit KDM4 activity promote H3K9 methylation and indiscriminate dysregulation of redox homeostasis.

www.nature.com/nrc Reviews

Conclusions meaning that they can compensate for one another. The 2OGDDs are a superfamily of druggable enzymes most striking evidence for non-redundancy comes linked to various fundamental processes including the from examples where paralogues seemingly oppose one HIF-dependent transcriptional response to hypoxia, another in the same cellular context, such as KDM6A and extracellular matrix formation, DNA methylation, his- KDM6B in T-ALL and kidney cancer72,203. At the same tone methylation, RNA processing and protein transla- time, KDM paralogues can also have a certain degree of tion. 2OGDDs can potentially sense oxygen, redox stress, functional redundancy204,205. Understanding such redun- iron availability and 2OG. As such, they are poised to dancies is especially important because single guide transduce various signals into changes in gene expres- RNA-based and short hairpin RNA-based dependency sion and cellular behaviour. Oncometabolites transform studies, such as the Broad Institute DepMap, typically cells, at least in part, by modulating the activity of specific involve inactivation of single genes. In such settings, 2OGDDs. In addition, many 2OGDDs are recurrently paralogue compensation can obscure cancer depen­ amplified or mutated in cancer or are indirectly affected dencies. First-generation 2OGDD inhibitors typically by mutations that dysregulate the enzymes they oppose. inhibit multiple paralogues. Whether it will be desira- Our knowledge of the biochemical and biolog- ble or not to develop paralogue-specific inhibitors will ical functions of 2OGDDs and their roles in cancer is depend on understanding these factors. Hopefully, as we far from complete. For example, it is not clear how the gain a greater understanding of the role of 2OGDDs in same enzyme, such as KDM5A, can act as a tumour oncogenesis, we will advance the development of clinical suppressor in one context and an oncoprotein in 2OGDD modulators to treat patients with cancer. another68,69,169,170,174,176,202. It is also unclear to what extent Published online xx xx xxxx the functions of various KDM paralogues are redundant,

1. DeBerardinis, R. J. & Chandel, N. S. We need to talk that these reactions are strongly influenced by 29. Cheng, M. B. et al. Specific phosphorylation of histone about the Warburg effect. Nat. Metab. 2, 127–129 in vitro reaction conditions. demethylase KDM3A determines target gene (2020). 18. Flashman, E., Davies, S. L., Yeoh, K. K. & Schofield, C. J. expression in response to heat shock. PLoS Biol. 12, 2. Baylin, S. B. & Jones, P. A. Epigenetic determinants Investigating the dependence of the hypoxia-inducible e1002026 (2014). of cancer. Cold Spring Harb. Perspect. Biol. https:// factor hydroxylases (factor inhibiting HIF and prolyl 30. Jeong, J. J. et al. Cytokine-regulated phosphorylation doi.org/10.1101/cshperspect.a019505 (2016). hydroxylase domain 2) on ascorbate and other and activation of TET2 by JAK2 in hematopoiesis. 3. Feinberg, A. P., Koldobskiy, M. A. & Gondor, A. reducing agents. Biochem. J. 427, 135–142 (2010). Cancer Discov. 9, 778–795 (2019). Epigenetic modulators, modifiers and mediators in This study shows that hydroxylation of HIF proteins 31. Olsen, J. V. et al. Global, in vivo, and site-specific cancer aetiology and progression. Nat. Rev. Genet. by EGLN prolyl hydroxylases and FIH is stimulated by phosphorylation dynamics in signaling networks. 17, 284–299 (2016). reducing agents in an enzyme and substrate-specific Cell 127, 635–648 (2006). 4. Schmidt, C., Sciacovelli, M. & Frezza, C. Fumarate manner. 32. Foxler, D. E. et al. The LIMD1 protein bridges an hydratase in cancer: a multifaceted tumour 19. Briggs, K. J. et al. Paracrine induction of HIF by association between the prolyl hydroxylases and VHL suppressor. Semin. Cell Dev. Biol. 98, 15–25 (2020). glutamate in breast cancer: EglN1 senses cysteine. to repress HIF-1 activity. Nat. Cell Biol. 14, 201–208 5. Gill, A. J. Succinate dehydrogenase (SDH)-deficient Cell 166, 126–139 (2016). (2012). neoplasia. Histopathology 72, 106–116 (2018). 20. Tarhonskaya, H. et al. Studies on the interaction of the 33. Zhang, C. S. et al. RHOBTB3 promotes proteasomal 6. Dang, L., Yen, K. & Attar, E. C. IDH mutations histone demethylase KDM5B with tricarboxylic acid degradation of HIFα through facilitating hydroxylation in cancer and progress toward development of cycle intermediates. J. Mol. Biol. 429, 2895–2906 and suppresses the Warburg effect. Cell Res. 25, targeted therapeutics. Ann. Oncol. 27, 599–608 (2017). 1025–1042 (2015). (2016). 21. Laukka, T. et al. Fumarate and succinate regulate 34. Kaelin, W. G. Jr. The VHL tumor suppressor gene: 7. Losman, J. A. & Kaelin, W. G. Jr. What a difference a expression of hypoxia-inducible genes via TET insights into oxygen sensing and cancer. Trans. Am. hydroxyl makes: mutant IDH, (R)-2-hydroxyglutarate, enzymes. J. Biol. Chem. 291, 4256–4265 (2016). Clin. Climatol. Assoc. 128, 298–307 (2017). and cancer. Genes Dev. 27, 836–852 (2013). This study shows that fumarate and succinate 35. Lee, S. et al. Neuronal apoptosis linked to EglN3 8. Kim, K. H. & Roberts, C. W. Targeting EZH2 in cancer. competitively inhibit the TET family of 5-mc prolyl hydroxylase and familial pheochromocytoma Nat. Med. 22, 128–134 (2016). hydroxylases, providing a biochemical basis for genes: developmental culling and cancer. Cancer Cell 9. Rao, R. C. & Dou, Y. Hijacked in cancer: the KMT2 their function as oncometabolites. 8, 155–167 (2005). (MLL) family of methyltransferases. Nat. Rev. Cancer 22. Laukka, T., Myllykoski, M., Looper, R. E. & 36. Maxwell, P. H. A common pathway for genetic events 15, 334–346 (2015). Koivunen, P. Cancer-associated 2-oxoglutarate leading to pheochromocytoma. Cancer Cell 8, 91–93 10. Bowman, R. L. & Levine, R. L. TET2 in normal and analogues modify histone methylation by inhibiting (2005). malignant hematopoiesis. Cold Spring Harb. Perspect. histone lysine demethylases. J. Mol. Biol. 430, 37. Kondo, K., Kim, W. Y., Lechpammer, M. & Med. https://doi.org/10.1101/cshperspect.a026518 3081–3092 (2018). Kaelin, W. G. Jr. Inhibition of HIF2α is sufficient to (2017). This study shows that fumarate and succinate suppress pVHL-defective tumor growth. PLoS Biol. 1, 11. Gaidzik, V. I. et al. TET2 mutations in acute myeloid competitively inhibit the KDM family of histone E83 (2003). leukemia (AML): results from a comprehensive genetic lysine demethylases, providing a biochemical basis 38. Kondo, K., Klco, J., Nakamura, E., Lechpammer, M. and clinical analysis of the AML study group. J. Clin. for their function as oncometabolites. & Kaelin, W. G. Jr. Inhibition of HIF is necessary for Oncol. 30, 1350–1357 (2012). 23. Koivunen, P. et al. Inhibition of hypoxia-inducible tumor suppression by the von Hippel–Lindau protein. 12. Wang, L. & Shilatifard, A. UTX mutations in human factor (HIF) hydroxylases by citric acid cycle Cancer Cell 1, 237–246 (2002). cancer. Cancer Cell 35, 168–176 (2019). intermediates: possible links between cell metabolism 39. Maranchie, J. K. et al. The contribution of VHL 13. Kaelin, W. G. Jr. & Ratcliffe, P. J. Oxygen sensing by and stabilization of HIF. J. Biol. Chem. 282, substrate binding and HIF1α to the phenotype of VHL metazoans: the central role of the HIF hydroxylase 4524–4532 (2007). loss in renal cell carcinoma. Cancer Cell 1, 247–255 pathway. Mol. Cell 30, 393–402 (2008). 24. Hirsila, M., Koivunen, P., Gunzler, V., Kivirikko, K. I. (2002). 14. Herr, C. Q. & Hausinger, R. P. Amazing diversity in & Myllyharju, J. Characterization of the human prolyl 40. Raval, R. R. et al. Contrasting properties of biochemical roles of FeII/2-oxoglutarate oxygenases. 4-hydroxylases that modify the hypoxia-inducible hypoxia-inducible factor 1 (HIF-1) and HIF-2 in von Trends Biochem. Sci. 43, 517–532 (2018). factor. J. Biol. Chem. 278, 30772–30780 (2003). Hippel–Lindau-associated renal cell carcinoma. 15. Islam, M. S., Leissing, T. M., Chowdhury, R., 25. Cervera, A. M. et al. An alternatively spliced transcript Mol. Cell Biol. 25, 5675–5686 (2005). Hopkinson, R. J. & Schofield, C. J. 2-Oxoglutarate- of the PHD3 gene retains prolyl hydroxylase activity. 41. Shen, C. et al. Genetic and functional studies implicate dependent oxygenases. Annu. Rev. Biochem. 87, Cancer Lett. 233, 131–138 (2006). HIF1α as a 14q kidney cancer suppressor gene. 585–620 (2018). 26. Tian, Y. M., Mole, D. R., Ratcliffe, P. J. & Gleadle, J. M. Cancer Discov. 1, 222–235 (2011). 16. Schofield, C. J. & Ratcliffe, P. J. Oxygen sensing by HIF Characterization of different isoforms of the HIF prolyl 42. Zimmer, M., Doucette, D., Siddiqui, N. & Iliopoulos, O. hydroxylases. Nat. Rev. Mol. Cell Biol. 5, 343–354 hydroxylase PHD1 generated by alternative initiation. Inhibition of hypoxia-inducible factor is sufficient for (2004). Biochem. J. 397, 179–186 (2006). growth suppression of VHL–/– tumors. Mol. Cancer 17. Koivunen, P., Hirsila, M., Kivirikko, K. I. & Myllyharju, J. 27. Yang, W. S., Campbell, M. & Chang, P. C. SUMO Res. 2, 89–95 (2004). The length of peptide substrates has a marked effect modification of a heterochromatin histone demethylase 43. Gordan, J. D. et al. HIF-α effects on c-Myc distinguish on hydroxylation by the hypoxia-inducible factor prolyl JMJD2A enables viral gene transactivation and viral two subtypes of sporadic VHL-deficient clear 4-hydroxylases. J. Biol. Chem. 281, 28712–28720 replication. PLoS Pathog. 13, e1006216 (2017). cell renal carcinoma. Cancer Cell 14, 435–446 (2006). 28. Bueno, M. T. & Richard, S. SUMOylation negatively (2008). This study shows that peptide substrate length modulates target gene occupancy of the KDM5B, 44. Dahia, P. L. Pheochromocytomas and paragangliomas,

affects the O2 Km values of EGLN-mediated a histone lysine demethylase. Epigenetics 8, genetically diverse and minimalist, all at once! Cancer hydroxylation reactions, thereby demonstrating 1162–1175 (2013). Cell 31, 159–161 (2017).

Nature Reviews | Cancer Reviews

45. Ladroue, C. et al. PHD2 mutation and congenital 64. Batie, M. et al. Hypoxia induces rapid changes to 86. Greten, F. R. & Grivennikov, S. I. Inflammation and erythrocytosis with paraganglioma. N. Engl. J. Med. histone methylation and reprograms chromatin. cancer: triggers, mechanisms, and consequences. 359, 2685–2692 (2008). Science 363, 1222–1226 (2019). Immunity 51, 27–41 (2019). This study identifies a novel germline mutation in This study demonstrates that the H3K4 histone 87. Yamamoto, K. et al. Loss of histone demethylase EGLN1 that is associated with constitutive HIF lysine demethylase KDM5A is an oxygen sensor. KDM6B enhances aggressiveness of pancreatic cancer activation, pseudo-hypoxic erythrocytosis and 65. Cascella, B. & Mirica, L. M. Kinetic analysis of through downregulation of C/EBPα. Carcinogenesis paraganglioma. iron-dependent histone demethylases: α-ketoglutarate 35, 2404–2414 (2014). 46. Yang, C. et al. Novel HIF2A mutations disrupt oxygen substrate inhibition and potential relevance to the 88. Yu, S. H. et al. JMJD3 facilitates C/EBPβ-centered sensing, leading to polycythemia, paragangliomas, regulation of histone demethylation in cancer cells. transcriptional program to exert oncorepressor and somatostatinomas. Blood 121, 2563–2566 Biochemistry 51, 8699–8701 (2012). activity in AML. Nat. Commun. 9, 3369 (2018). (2013). 66. Hancock, R. L., Masson, N., Dunne, K., Flashman, E. 89. Cimmino, L. et al. TET1 is a tumor suppressor 47. Zhuang, Z. et al. Somatic HIF2A gain-of-function & Kawamura, A. The activity of JmjC histone lysine of hematopoietic malignancy. Nat. Immunol. 16, mutations in paraganglioma with polycythemia. demethylase KDM4A is highly sensitive to oxygen 653–662 (2015). N. Engl. J. Med. 367, 922–930 (2012). concentrations. ACS Chem. Biol. 12, 1011–1019 90. Neri, F. et al. TET1 is a tumour suppressor that 48. Zhuang, Z. et al. HIF2A gain-of-function mutations (2017). inhibits colon cancer growth by derepressing inhibitors detected in duodenal gangliocytic paraganglioma. 67. Thienpont, B. et al. Tumour hypoxia causes DNA of the WNT pathway. Oncogene 34, 4168–4176 Endocr. Relat. Cancer 23, L13–L16 (2016). hypermethylation by reducing TET activity. Nature (2015). 49. Fielding, J. W. et al. PHD2 inactivation in type I cells 537, 63–68 (2016). 91. Wu, B. K. & Brenner, C. Suppression of TET1- drives HIF-2α-dependent multilineage hyperplasia and 68. Dai, B. et al. Histone demethylase KDM5A inhibits dependent DNA demethylation is essential for the formation of paraganglioma-like carotid bodies. glioma cells migration and invasion by down KRAS-mediated transformation. Cell Rep. 9, J. Physiol. https://doi.org/10.1113/JP275996 regulating ZEB1. Biomed. Pharmacother. 99, 72–80 1827–1840 (2014). (2018). (2018). 92. Cimmino, L. et al. Restoration of TET2 function blocks 50. Cockman, M. E. et al. Lack of activity of recombinant 69. Kong, S. Y., Kim, W., Lee, H. R. & Kim, H. J. aberrant self-renewal and leukemia progression. HIF prolyl hydroxylases (PHDs) on reported non-HIF The histone demethylase KDM5A is required Cell 170, 1079–1095.e20 (2017). substrates. eLife https://doi.org/10.7554/eLife.46490 for the repression of astrocytogenesis and regulated This study demonstrates that stimulation of (2019). by the translational machinery in neural progenitor TET activity by reducing agents reverses the This study comprehensively tests the ability of cells. FASEB J. 32, 1108–1119 (2018). leukaemogenic effects of TET2 loss, thereby EGLN enzymes to hydroxylate a panel of putative 70. Andricovich, J. et al. Loss of KDM6A activates providing a rationale for the use of ascorbate non-HIF EGLN substrates using several super-enhancers to induce gender-specific in patients with TET2-mutant haematopoietic experimental approaches and finds that none of squamous-like pancreatic cancer and confers malignancies. the substrates is hydroxylated by EGLN enzymes sensitivity to BET inhibitors. Cancer Cell 33, 93. Moran-Crusio, K. et al. Tet2 loss leads to increased in vitro. Although these negative results could be 512–526.e8 (2018). hematopoietic stem cell self-renewal and myeloid due to technical factors, they challenge the idea 71. Dalgliesh, G. L. et al. Systematic sequencing of renal transformation. Cancer Cell 20, 11–24 (2011). that these non-HIF proteins are authentic EGLN carcinoma reveals inactivation of histone modifying 94. Hirsila, M. et al. Effect of desferrioxamine and metals substrates in vivo. genes. Nature 463, 360–363 (2010). on the hydroxylases in the oxygen sensing pathway. 51. Chen, N. et al. Roxadustat treatment for anemia in 72. Ntziachristos, P. et al. Contrasting roles of histone 3 FASEB J. 19, 1308–1310 (2005). patients undergoing long-term dialysis. N. Engl. J. lysine 27 demethylases in acute lymphoblastic 95. Li, Q., Ke, Q. & Costa, M. Alterations of histone Med. 381, 1011–1022 (2019). leukaemia. Nature 514, 513–517 (2014). modifications by cobalt compounds. Carcinogenesis 52. Chen, N. et al. Roxadustat for anemia in patients with This study demonstrates that the H3K27 histone 30, 1243–1251 (2009). kidney disease not receiving dialysis. N. Engl. J. Med. lysine demethylase paralogues KDM6A and 96. Kaczmarek, M. et al. The role of ascorbate in the 381, 1001–1010 (2019). KDM6B have opposing functions in T-ALL. modulation of HIF-1α protein and HIF-dependent VI II 53. Hammond, E. M. et al. The meaning, measurement 73. Ivan, M. & Kaelin, W. G. Jr. The EGLN–HIF O2-sensing transcription by chromium and nickel . Free. Radic. and modification of hypoxia in the laboratory and the system: multiple inputs and feedbacks. Mol. Cell 66, Biol. Med. 42, 1246–1257 (2007). clinic. Clin. Oncol. 26, 277–288 (2014). 772–779 (2017). 97. Sabharwal, S. S. & Schumacker, P. T. Mitochondrial 54. Dayan, F., Roux, D., Brahimi-Horn, M. C., Pouyssegur, J. 74. Chouchani, E. T. et al. Ischaemic accumulation of ROS in cancer: initiators, amplifiers or an Achilles’ & Mazure, N. M. The oxygen sensor factor-inhibiting succinate controls reperfusion injury through heel? Nat. Rev. Cancer 14, 709–721 (2014). hypoxia-inducible factor-1 controls expression of mitochondrial ROS. Nature 515, 431–435 (2014). 98. Kuiper, C. & Vissers, M. C. Ascorbate as a co-factor distinct genes through the bifunctional transcriptional 75. Zhang, J. et al. Accumulation of succinate in cardiac for Fe- and 2-oxoglutarate dependent dioxygenases: character of hypoxia-inducible factor-1α. Cancer Res. ischemia primarily occurs via canonical Krebs cycle physiological activity in tumor growth and 66, 3688–3698 (2006). activity. Cell Rep. 23, 2617–2628 (2018). progression. Front. Oncol. 4, 359 (2014). 55. Lando, D. et al. FIH-1 is an asparaginyl hydroxylase 76. Intlekofer, A. M. et al. Hypoxia induces production 99. Nytko, K. J. et al. Vitamin C is dispensable for oxygen enzyme that regulates the transcriptional activity of of l-2-hydroxyglutarate. Cell Metab. 22, 304–311 sensing in vivo. Blood 117, 5485–5493 (2011). hypoxia-inducible factor. Genes Dev. 16, 1466–1471 (2015). 100. Knowles, H. J., Raval, R. R., Harris, A. L. & Ratcliffe, P. J. (2002). This study demonstrates that hypoxia indirectly Effect of ascorbate on the activity of hypoxia-inducible 56. Mahon, P. C., Hirota, K. & Semenza, G. L. FIH-1: inhibits the activity of numerous histone lysine factor in cancer cells. Cancer Res. 63, 1764–1768 a novel protein that interacts with HIF-1α and VHL to demethylases by inducing the production of S-2HG. (2003). mediate repression of HIF-1 transcriptional activity. 77. Intlekofer, A. M. et al. l-2-Hydroxyglutarate production 101. Kuiper, C., Dachs, G. U., Currie, M. J. & Vissers, M. C. Genes Dev. 15, 2675–2686 (2001). arises from noncanonical enzyme function at acidic Intracellular ascorbate enhances hypoxia-inducible 57. Yang, M. et al. Factor-inhibiting hypoxia-inducible pH. Nat. Chem. Biol. 13, 494–500 (2017). factor (HIF)-hydroxylase activity and preferentially factor (FIH) catalyses the post-translational 78. Oldham, W. M., Clish, C. B., Yang, Y. & Loscalzo, J. suppresses the HIF-1 transcriptional response. hydroxylation of histidinyl residues within ankyrin Hypoxia-mediated increases in l-2-hydroxyglutarate Free. Radic. Biol. Med. 69, 308–317 (2014). repeat domains. FEBS J. 278, 1086–1097 (2011). coordinate the metabolic response to reductive stress. 102. Campbell, E. J. et al. Restoring physiological levels of 58. Cockman, M. E. et al. Posttranslational hydroxylation Cell Metab. 22, 291–303 (2015). ascorbate slows tumor growth and moderates HIF-1 of ankyrin repeats in IκB proteins by the This study demonstrates that cells mitigate the pathway activity in Gulo–/– mice. Cancer Med. 4, hypoxia-inducible factor (HIF) asparaginyl hydroxylase, redox stress induced by hypoxia by producing a 303–314 (2015). factor inhibiting HIF (FIH). Proc. Natl Acad. Sci. USA metabolite, S-2HG, that inhibits the electron 103. Gao, P. et al. HIF-dependent antitumorigenic effect 103, 14767–14772 (2006). transport chain and glycolysis. of antioxidants in vivo. Cancer Cell 12, 230–238 This study identifies numerous ankyrin repeat 79. Tyrakis, P. A. et al. S-2-Hydroxyglutarate regulates (2007). proteins as non-HIF substrates of FIH and CD8+ T-lymphocyte fate. Nature 540, 236–241 104. Agathocleous, M. et al. Ascorbate regulates suggested that they indirectly influence the (2016). haematopoietic stem cell function and amount of FIH available to hydroxylate HIF. 80. Masson, N. et al. Conserved N-terminal cysteine leukaemogenesis. Nature 549, 476–481 (2017). 59. Zhang, N. et al. The asparaginyl hydroxylase factor dioxygenases transduce responses to hypoxia This study demonstrates that ascorbate depletion inhibiting HIF-1α is an essential regulator of in animals and plants. Science 365, 65–69 phenocopies the leukaemogenic effects of TET2 metabolism. Cell Metab. 11, 364–378 (2010). (2019). loss, thereby reinforcing the in vivo role of reducing 60. Melvin, A. & Rocha, S. Chromatin as an oxygen sensor 81. Akter, S. et al. Chemical proteomics reveals new agents in maintaining 2OGDD activity. and active player in the hypoxia response. Cell Signal. targets of cysteine sulfinic acid reductase. Nat. Chem. 105. Shenoy, N. et al. Ascorbic acid-induced TET activation 24, 35–43 (2012). Biol. 14, 995–1004 (2018). mitigates adverse hydroxymethylcytosine loss in renal 61. Lu, Y., Chu, A., Turker, M. S. & Glazer, P. M. 82. Coffey, R. & Ganz, T. Iron homeostasis: an cell carcinoma. J. Clin. Invest. 129, 1612–1625 Hypoxia-induced epigenetic regulation and anthropocentric perspective. J. Biol. Chem. 292, (2019). silencing of the BRCA1 promoter. Mol. Cell Biol. 31, 12727–12734 (2017). 106. Ge, G. et al. Restoration of 5-hydroxymethylcytosine 3339–3350 (2011). 83. Roy, C. N. & Andrews, N. C. Anemia of inflammation: by ascorbate blocks kidney tumour growth. EMBO 62. Chervona, Y. & Costa, M. The control of histone the hepcidin link. Curr. Opin. Hematol. 12, 107–111 Rep. https://doi.org/10.15252/embr.201745401 methylation and gene expression by oxidative stress, (2005). (2018). hypoxia, and metals. Free. Radic. Biol. Med. 53, 84. Barman-Aksozen, J., Beguin, C., Dogar, A. M., 107. Pritchard, J. B. Intracellular α-ketoglutarate controls 1041–1047 (2012). Schneider-Yin, X. & Minder, E. I. Iron availability the efficacy of renal organic anion transport. 63. Chakraborty, A. A. et al. Histone demethylase KDM6A modulates aberrant splicing of ferrochelatase through J. Pharmacol. Exp. Ther. 274, 1278–1284 (1995). directly senses oxygen to control chromatin and cell the iron- and 2-oxoglutarate dependent dioxygenase 108. Olenchock, B. A. et al. EGLN1 inhibition and rerouting fate. Science 363, 1217–1222 (2019). Jmjd6 and U2AF65. Blood Cell Mol. Dis. 51, 151–161 of α-ketoglutarate suffice for remote ischemic This study demonstrates that the H3K27 histone (2013). protection. Cell 165, 497 (2016). lysine demethylase KDM6A is an oxygen sensor 85. Furman, D. et al. Chronic inflammation in the 109. Raffel, S. et al. BCAT1 restricts αKG levels in AML stem and links this to the well-established ability of etiology of disease across the life span. Nat. Med. 25, cells leading to IDHmut-like DNA hypermethylation. hypoxia to influence cellular differentiation. 1822–1832 (2019). Nature 551, 384–388 (2017).

www.nature.com/nrc Reviews

110. Morris, J. P. T. et al. α-Ketoglutarate links p53 to and that its effects can be reversed in preclinical 153. Sasaki, M. et al. d-2-Hydroxyglutarate produced by cell fate during tumour suppression. Nature 573, models. The latter finding has galvanized efforts to mutant IDH1 perturbs collagen maturation and 595–599 (2019). develop drugs that block 2-HG production by basement membrane function. Genes Dev. 26, This study demonstrates that accumulation of mutant IDH. 2038–2049 (2012). intracellular 2OG stimulates the differentiation 130. Wang, F. et al. Targeted inhibition of mutant IDH2 in 154. Sasaki, M. et al. IDH1(R132H) mutation increases of p53-deficient tumour cells. leukemia cells induces cellular differentiation. Science murine haematopoietic progenitors and alters 111. Yong, C., Stewart, G. D. & Frezza, C. Oncometabolites 340, 622–626 (2013). epigenetics. Nature 488, 656–659 (2012). in renal cancer. Nat. Rev. Nephrol. 16, 156–172 131. Yen, K. et al. AG-221, a first-in-class therapy targeting 155. Turcan, S. et al. IDH1 mutation is sufficient to (2020). acute myeloid leukemia harboring oncogenic IDH2 establish the glioma hypermethylator phenotype. 112. Jiang, Y. et al. Local generation of fumarate promotes mutations. Cancer Discov. 7, 478–493 (2017). Nature 483, 479–483 (2012). DNA repair through inhibition of histone H3 132. Stein, E. M. et al. Enasidenib in mutant IDH2 relapsed 156. Turcan, S. et al. Mutant-IDH1-dependent chromatin demethylation. Nat. Cell Biol. 17, 1158–1168 (2015). or refractory acute myeloid leukemia. Blood 130, state reprogramming, reversibility, and persistence. 113. Yogev, O. et al. Fumarase: a mitochondrial metabolic 722–731 (2017). Nat. Genet. 50, 62–72 (2018). enzyme and a cytosolic/nuclear component of the 133. DiNardo, C. D. et al. Durable remissions with 157. Williams, S. C. et al. R132H-mutation of isocitrate DNA damage response. PLoS Biol. 8, e1000328 ivosidenib in IDH1-mutated relapsed or refractory dehydrogenase-1 is not sufficient for HIF-1α (2010). AML. N. Engl. J. Med. 378, 2386–2398 (2018). upregulation in adult glioma. Acta Neuropathol. 121, 114. Sulkowski, P. L. et al. Oncometabolites suppress DNA 134. Tahiliani, M. et al. Conversion of 5-methylcytosine to 279–281 (2011). repair by disrupting local chromatin signalling. Nature 5-hydroxymethylcytosine in mammalian DNA by MLL 158. Polivka, J. Jr. et al. IDH1 mutation is associated with 582, 586–591 (2020). partner TET1. Science 324, 930–935 (2009). lower expression of VEGF but not microvessel 115. Arts, R. J. et al. Glutaminolysis and fumarate 135. Ito, S. et al. Role of Tet proteins in 5mC to 5hmC formation in glioblastoma multiforme. Oncotarget 9, accumulation integrate immunometabolic and conversion, ES-cell self-renewal and inner cell mass 16462–16476 (2018). epigenetic programs in trained immunity. Cell Metab. specification. Nature 466, 1129–1133 (2010). 159. Bottcher, M. et al. d-2-Hydroxyglutarate interferes 24, 807–819 (2016). 136. He, Y. F. et al. Tet-mediated formation of with HIF-1α stability skewing T-cell metabolism 116. Carey, B. W., Finley, L. W., Cross, J. R., Allis, C. D. 5-carboxylcytosine and its excision by TDG in towards oxidative phosphorylation and impairing

& Thompson, C. B. Intracellular α-ketoglutarate mammalian DNA. Science 333, 1303–1307 (2011). TH17 polarization. Oncoimmunology 7, e1445454 maintains the pluripotency of embryonic stem cells. 137. Ito, S. et al. Tet proteins can convert 5-methylcytosine (2018). Nature 518, 413–416 (2015). to 5-formylcytosine and 5-carboxylcytosine. Science 160. Blouw, B. et al. The hypoxic response of tumors is This study demonstrates that the intracellular 2OG 333, 1300–1303 (2011). dependent on their microenvironment. Cancer Cell 4, to succinate ratio contributes to the maintenance 138. Xu, W. et al. Oncometabolite 2-hydroxyglutarate is 133–146 (2003). of stem cell pluripotency. a competitive inhibitor of α-ketoglutarate-dependent 161. Tarhonskaya, H. et al. Non-enzymatic chemistry 117. TeSlaa, T. et al. α-Ketoglutarate accelerates the initial dioxygenases. Cancer Cell 19, 17–30 (2011). enables 2-hydroxyglutarate-mediated activation of differentiation of primed human pluripotent stem This study shows that R-2HG is able to inhibit 2-oxoglutarate oxygenases. Nat. Commun. 5, 3423 cells. Cell Metab. 24, 485–493 (2016). specific 2OGDDs. (2014). This study demonstrates that 2OG contributes 139. Koivunen, P. et al. Transformation by the (R)- 162. Rzem, R., Vincent, M. F., Van Schaftingen, E. & to the differentiation of pluripotent stem cells. enantiomer of 2-hydroxyglutarate linked to EGLN Veiga-da-Cunha, M. l-2-Hydroxyglutaric aciduria, a 118. Wu, H. et al. Dual functions of Tet1 in transcriptional activation. Nature 483, 484–488 (2012). defect of metabolite repair. J. Inherit. Metab. Dis. 30, regulation in mouse embryonic stem cells. Nature This study shows that R-2HG functions as a 681–689 (2007). 473, 389–393 (2011). co-substrate to activate EGLN enzymes. 163. Munn, L. L. & Jain, R. K. Vascular regulation of 119. Gimenez-Roqueplo, A. P. et al. The R22X mutation 140. Madzo, J., Vasanthakumar, A. & Godley, L. A. antitumor immunity. Science 365, 544–545 of the SDHD gene in hereditary paraganglioma Perturbations of 5-hydroxymethylcytosine patterning (2019). abolishes the enzymatic activity of Complex II in the in hematologic malignancies. Semin. Hematol. 50, 164. Damgaci, S. et al. Hypoxia and acidosis: immune mitochondrial respiratory chain and activates the 61–69 (2013). suppressors and therapeutic targets. Immunology hypoxia pathway. Am. J. Hum. Genet. 69, 1186–1197 141. Kroeze, L. I. et al. Characterization of acute myeloid 154, 354–362 (2018). (2001). leukemia based on levels of global hydroxymethylation. 165. Shim, E. H. et al. l-2-Hydroxyglutarate: an epigenetic 120. Isaacs, J. S. et al. HIF overexpression correlates with Blood 124, 1110–1118 (2014). modifier and putative oncometabolite in renal cancer. biallelic loss of fumarate hydratase in renal cancer: 142. Figueroa, M. E. et al. Leukemic IDH1 and IDH2 Cancer Discov. 4, 1290–1298 (2014). novel role of fumarate in regulation of HIF stability. mutations result in a hypermethylation phenotype, 166. McBrayer, S. K. et al. Transaminase inhibition by Cancer Cell 8, 143–153 (2005). disrupt TET2 function, and impair hematopoietic 2-hydroxyglutarate impairs glutamate biosynthesis 121. Pollard, P. J. et al. Accumulation of Krebs cycle differentiation. Cancer Cell 18, 553–567 (2010). and redox homeostasis in glioma. Cell 175, 101–116. intermediates and over-expression of HIF1alpha This study shows that IDH and TET2 mutations are e25 (2018). in tumours which result from germline FH and SDH mutually exclusive in AML and that R-2HG acts as 167. Tonjes, M. et al. BCAT1 promotes cell proliferation mutations. Hum. Mol. Genet. 14, 2231–2239 an oncometabolite at least in part by inhibiting through amino acid catabolism in gliomas carrying (2005). TET2. wild-type IDH1. Nat. Med. 19, 901–908 (2013). 122. Selak, M. A. et al. Succinate links TCA cycle 143. Xu, X. et al. KDM3B shows tumor-suppressive activity 168. Benevolenskaya, E. V., Murray, H. L., Branton, P., dysfunction to oncogenesis by inhibiting HIF-alpha and transcriptionally regulates HOXA1 through Young, R. A. & Kaelin, W. G. Jr. Binding of pRB to the prolyl hydroxylase. Cancer Cell 7, 77–85 (2005). retinoic acid response elements in acute myeloid PHD protein RBP2 promotes cellular differentiation. Together with Isaacs et al. (2005) and Pollard et al. leukemia. Leuk. Lymphoma 59, 204–213 (2018). Mol. Cell 18, 623–635 (2005). (2005), this study shows that fumarate and 144. Choi, C. et al. 2-Hydroxyglutarate detection by 169. McBrayer, S. K. et al. Autochthonous tumors driven by succinate can inhibit EGLN enzymes and stabilize magnetic resonance spectroscopy in IDH-mutated Rb1 loss have an ongoing requirement for the RBP2 HIF. patients with gliomas. Nat. Med. 18, 624–629 histone demethylase. Proc. Natl Acad. Sci. USA 115, 123. Adam, J. et al. Renal cyst formation in Fh1-deficient (2012). E3741–E3748 (2018). mice is independent of the Hif/Phd pathway: roles for 145. Longuespee, R. et al. Rapid detection of 170. Lin, W. et al. Loss of the retinoblastoma binding fumarate in KEAP1 succination and Nrf2 signaling. 2-hydroxyglutarate in frozen sections of IDH mutant protein 2 (RBP2) histone demethylase suppresses Cancer Cell 20, 524–537 (2011). tumors by MALDI-TOF mass spectrometry. Acta tumorigenesis in mice lacking Rb1 or Men1. Proc. This study shows that deletion of HIF1 actually Neuropathol. Commun. 6, 21 (2018). Natl Acad. Sci. USA 108, 13379–13386 (2011). worsens the pathology of Fh1-deficient kidneys, 146. Jezek, P. 2-Hydroxyglutarate in cancer cells. Antioxid. 171. Hinohara, K. et al. KDM5 histone demethylase underscoring the need for caution when inferring Redox Signal. https://doi.org/10.1089/ars.2019.7902 activity links cellular transcriptomic heterogeneity to causality based purely on correlations. (2020). therapeutic resistance. Cancer Cell 34, 939–953.e9 124. Letouze, E. et al. SDH mutations establish a 147. Cleven, A. H. G. et al. IDH1 or -2 mutations do (2018). hypermethylator phenotype in paraganglioma. not predict outcome and do not cause loss of 172. Hou, J. et al. Genomic amplification and a role in Cancer Cell 23, 739–752 (2013). 5-hydroxymethylcytosine or altered histone drug-resistance for the KDM5A histone demethylase 125. Xiao, M. et al. Inhibition of α-KG-dependent histone modifications in central chondrosarcomas. in breast cancer. Am. J. Transl Res. 4, 247–256 and DNA demethylases by fumarate and succinate Clin. Sarcoma Res. 7, 8 (2017). (2012). that are accumulated in mutations of FH and SDH 148. Kim, Y. H. et al. TET2 promoter methylation in 173. Sharma, S. V. et al. A chromatin-mediated reversible tumor suppressors. Genes Dev. 26, 1326–1338 low-grade diffuse gliomas lacking IDH1/2 mutations. drug-tolerant state in cancer cell subpopulations. (2012). J. Clin. Pathol. 64, 850–852 (2011). Cell 141, 69–80 (2010). 126. Aspuria, P. P. et al. Succinate dehydrogenase 149. Chowdhury, R. et al. The oncometabolite 174. Teng, Y. C. et al. Histone demethylase RBP2 promotes inhibition leads to epithelial–mesenchymal transition 2-hydroxyglutarate inhibits histone lysine lung tumorigenesis and cancer metastasis. Cancer and reprogrammed carbon metabolism. Cancer demethylases. EMBO Rep. 12, 463–469 (2011). Res. 73, 4711–4721 (2013). Metab. 2, 21 (2014). This study shows that R-2HG is able to inhibit 175. Liu, X. et al. KDM5B promotes drug resistance by 127. Struys, E. A. et al. Kinetic characterization of human specific 2OGDDs. regulating melanoma-propagating cell subpopulations. hydroxyacid-oxoacid transhydrogenase: relevance to 150. Lu, C. et al. IDH mutation impairs histone Mol. Cancer Ther. 18, 706–717 (2019). d-2-hydroxyglutaric and γ-hydroxybutyric acidurias. demethylation and results in a block to cell 176. Roesch, A. et al. A temporarily distinct subpopulation J. Inherit. Metab. Dis. 28, 921–930 (2005). differentiation. Nature 483, 474–478 (2012). of slow-cycling melanoma cells is required for 128. Dang, L. et al. Cancer-associated IDH1 mutations 151. Janke, R., Iavarone, A. T. & Rine, J. Oncometabolite continuous tumor growth. Cell 141, 583–594 (2010). produce 2-hydroxyglutarate. Nature 462, 739–744 d-2-hydroxyglutarate enhances gene silencing through 177. Miyake, Y. et al. Identification of novel lysine (2009). inhibition of specific H3K36 histone demethylases. demethylase 5-selective inhibitors by inhibitor-based 129. Losman, J. A. et al. (R)-2-Hydroxyglutarate is sufficient eLife https://doi.org/10.7554/eLife.22451 (2017). fragment merging strategy. Bioorg Med. Chem. 27, to promote leukemogenesis and its effects are 152. Kernytsky, A. et al. IDH2 mutation-induced histone 1119–1129 (2019). reversible. Science 339, 1621–1625 (2013). and DNA hypermethylation is progressively reversed 178. Horton, J. R. et al. Structural basis for KDM5A This study shows that R-2HG acts as an by small-molecule inhibition. Blood 125, 296–303 histone lysine demethylase inhibition by diverse oncometabolite at least in part by inhibiting TET2 (2015). compounds. Cell Chem. Biol. 23, 769–781 (2016).

Nature Reviews | Cancer Reviews

179. Johansson, C. et al. Structural analysis of human 203. Chakraborty, A. A. et al. HIF activation causes 228. Banito, A. et al. The SS18-SSX oncoprotein hijacks KDM5B guides histone demethylase inhibitor synthetic lethality between the VHL tumor suppressor KDM2B-PRC1.1 to drive synovial sarcoma. Cancer development. Nat. Chem. Biol. 12, 539–545 (2016). and the EZH1 histone methyltransferase. Sci. Transl Cell 33, 527–541.e8 (2018). 180. Van der Meulen, J. et al. The H3K27me3 demethylase Med. https://doi.org/10.1126/scitranslmed.aal5272 229. Tzatsos, A. et al. KDM2B promotes pancreatic UTX is a gender-specific tumor suppressor in T-cell (2017). cancer via Polycomb-dependent and -independent acute lymphoblastic leukemia. Blood 125, 13–21 204. Shpargel, K. B., Sengoku, T., Yokoyama, S. & transcriptional programs. J. Clin. Invest. 123, (2015). Magnuson, T. UTX and UTY demonstrate histone 727–739 (2013). 181. Benyoucef, A. et al. UTX inhibition as selective demethylase-independent function in mouse 230. Chen, M. et al. JMJD1C is required for the survival of epigenetic therapy against TAL1-driven T-cell acute embryonic development. PLoS Genet. 8, e1002964 acute myeloid leukemia by functioning as a coactivator lymphoblastic leukemia. Genes Dev. 30, 508–521 (2012). for key transcription factors. Genes Dev. 29, (2016). 205. Klose, R. J. et al. The retinoblastoma binding protein 2123–2139 (2015). 182. Kim, J. H. et al. UTX and MLL4 coordinately regulate RBP2 is an H3K4 demethylase. Cell 128, 889–900 231. Zhu, N. et al. MLL-AF9- and HOXA9-mediated acute transcriptional programs for cell proliferation and (2007). myeloid leukemia stem cell self-renewal requires invasiveness in breast cancer cells. Cancer Res. 74, 206. Celik, H. et al. JARID2 functions as a tumor JMJD1C. J. Clin. Invest. 126, 997–1011 (2016). 1705–1717 (2014). suppressor in myeloid neoplasms by repressing 232. Black, J. C. et al. KDM4A lysine demethylase induces 183. Hashizume, R. et al. Pharmacologic inhibition of self-renewal in hematopoietic progenitor cells. Cancer site-specific copy gain and rereplication of regions histone demethylation as a therapy for pediatric Cell 34, 741–756.e8 (2018). amplified in tumors. Cell 154, 541–555 (2013). brainstem glioma. Nat. Med. 20, 1394–1396 (2014). 207. Frescas, D., Guardavaccaro, D., Bassermann, F., 233. Guerra-Calderas, L., Gonzalez-Barrios, R., Herrera, L. A., 184. Berry, W. L., Shin, S., Lightfoot, S. A. & Janknecht, R. Koyama-Nasu, R. & Pagano, M. JHDM1B/FBXL10 Cantu de Leon, D. & Soto-Reyes, E. The role of the Oncogenic features of the JMJD2A histone is a nucleolar protein that represses transcription of histone demethylase KDM4A in cancer. Cancer Genet. demethylase in breast cancer. Int. J. Oncol. 41, ribosomal RNA genes. Nature 450, 309–313 (2007). 208, 215–224 (2015). 1701–1706 (2012). 208. Hu, Z. et al. A novel nuclear protein, 5qNCA 234. Wilson, C. et al. The histone demethylase KDM4B 185. Cheung, N. et al. Targeting aberrant epigenetic (LOC51780) is a candidate for the myeloid leukemia regulates peritoneal seeding of ovarian cancer. networks mediated by PRMT1 and KDM4C in acute tumor suppressor gene on chromosome 5 band q31. Oncogene 36, 2565–2576 (2017). myeloid leukemia. Cancer Cell 29, 32–48 (2016). Oncogene 20, 6946–6954 (2001). 235. Wu, M. C. et al. KDM4B is a coactivator of c-Jun and 186. Liu, G. et al. Genomic amplification and oncogenic 209. van Zutven, L. J. et al. Identification of NUP98 involved in gastric carcinogenesis. Cell Death Dis. 10, properties of the GASC1 histone demethylase gene in abnormalities in acute leukemia: JARID1A (12p13) as 68 (2019). breast cancer. Oncogene 28, 4491–4500 (2009). a new partner gene. Genes Chromosomes Cancer 45, 236. Yamamoto, S. et al. JARID1B is a luminal 187. Vinatzer, U. et al. Mucosa-associated lymphoid tissue 437–446 (2006). lineage-driving oncogene in breast cancer. Cancer Cell lymphoma: novel translocations including 210. Wong, S. H. et al. The H3K4-methyl epigenome 25, 762–777 (2014). rearrangements of ODZ2, JMJD2C, and CNN3. regulates leukemia stem cell oncogenic potential. 237. Paolicchi, E., Crea, F., Farrar, W. L., Green, J. E. & Clin. Cancer Res. 14, 6426–6431 (2008). Cancer Cell 28, 198–209 (2015). Danesi, R. Histone lysine demethylases in breast 188. Yang, Z. Q. et al. Identification of a novel gene, 211. Rondinelli, B. et al. Histone demethylase JARID1C cancer. Crit. Rev. Oncol. Hematol. 86, 97–103 (2013). GASC1, within an amplicon at 9p23–24 frequently inactivation triggers genomic instability in sporadic 238. Xie, G. et al. UTX promotes hormonally responsive detected in esophageal cancer cell lines. Cancer Res. renal cancer. J. Clin. Invest. 125, 4625–4637 (2015). breast carcinogenesis through feed-forward 60, 4735–4739 (2000). 212. Shen, H. et al. Suppression of enhancer overactivation transcription regulation with estrogen receptor. 189. Berry, W. L. & Janknecht, R. KDM4/JMJD2 histone by a RACK7–histone demethylase complex. Cell 165, Oncogene 36, 5497–5511 (2017). demethylases: epigenetic regulators in cancer cells. 331–342 (2016). 239. Mallaney, C. et al. Kdm6b regulates context-dependent Cancer Res. 73, 2936–2942 (2013). 213. Zhan, D. et al. Whole exome sequencing identifies hematopoietic stem cell self-renewal and 190. Ye, Q. et al. Genetic alterations of KDM4 subfamily novel mutations of epigenetic regulators in leukemogenesis. Leukemia 33, 2506–2521 (2019). and therapeutic effect of novel demethylase inhibitor chemorefractory pediatric acute myeloid leukemia. 240. Wei, Y. et al. Global H3K4me3 genome mapping in breast cancer. Am. J. Cancer Res. 5, 1519–1530 Leuk. Res. 65, 20–24 (2018). reveals alterations of innate immunity signaling and (2015). 214. De Keersmaecker, K. et al. Exome sequencing overexpression of JMJD3 in human myelodysplastic 191. Shin, S. & Janknecht, R. Activation of androgen identifies mutation in CNOT3 and ribosomal genes syndrome CD34+ cells. Leukemia 27, 2177–2186 receptor by histone demethylases JMJD2A and RPL5 and RPL10 in T-cell acute lymphoblastic (2013). JMJD2D. Biochem. Biophys. Res. Commun. 359, leukemia. Nat. Genet. 45, 186–190 (2013). 241. Bjorkman, M. et al. Systematic knockdown of 742–746 (2007). 215. Park, J. L. et al. Decrease of 5hmC in gastric cancers epigenetic enzymes identifies a novel histone 192. Duan, L. et al. KDM4/JMJD2 histone demethylase is associated with TET1 silencing due to with DNA demethylase PHF8 overexpressed in prostate cancer inhibitors block prostate tumor growth by suppressing methylation and bivalent histone marks at TET1 CpG with an impact on cell proliferation, migration and the expression of AR and BMYB-regulated genes. island 3′-shore. Oncotarget 6, 37647–37662 (2015). invasion. Oncogene 31, 3444–3456 (2012). Chem. Biol. 22, 1185–1196 (2015). 216. Nickerson, M. L. et al. TET2 binds the androgen 242. Shen, Y., Pan, X. & Zhao, H. The histone demethylase 193. Carter, D. M. et al. Identification of a novel receptor and loss is associated with prostate cancer. PHF8 is an oncogenic protein in human non-small cell benzimidazole pyrazolone scaffold that inhibits KDM4 Oncogene 36, 2172–2183 (2017). lung cancer. Biochem. Biophys. Res. Commun. 451, lysine demethylases and reduces proliferation of 217. Lavaissiere, L. et al. Overexpression of human 119–125 (2014). prostate cancer cells. SLAS Discov. 22, 801–812 aspartyl(asparaginyl)β-hydroxylase in hepatocellular 243. Wang, Q. et al. Stabilization of histone demethylase (2017). carcinoma and cholangiocarcinoma. J. Clin. Invest. 98, PHF8 by USP7 promotes breast carcinogenesis. 194. Shih, A. H. et al. Combination targeted therapy to 1313–1323 (1996). J. Clin. Invest. 126, 2205–2220 (2016). disrupt aberrant oncogenic signaling and reverse 218. Dong, X. et al. Aspartate β-hydroxylase expression 244. Huang, M. Y., Xuan, F., Liu, W. & Cui, H. J. epigenetic dysfunction in IDH2- and TET2-mutant promotes a malignant pancreatic cellular phenotype. MINA controls proliferation and tumorigenesis of acute myeloid leukemia. Cancer Discov. 7, 494–505 Oncotarget 6, 1231–1248 (2015). glioblastoma by epigenetically regulating cyclins and (2017). 219. Zou, Q. et al. Hydroxylase activity of ASPH promotes CDKs via H3K9me3 demethylation. Oncogene 36, 195. Bejar, R. et al. TET2 mutations predict response to hepatocellular carcinoma metastasis through 387–396 (2017). hypomethylating agents in myelodysplastic syndrome epithelial-to-mesenchymal transition pathway. 245. Lu, Y. et al. Lung cancer-associated JmjC domain patients. Blood 124, 2705–2712 (2014). EBioMedicine 31, 287–298 (2018). protein mdig suppresses formation of tri-methyl lysine 196. Jacobs, C., Hutton, B., Ng, T., Shorr, R. & Clemons, M. 220. Li, Z. et al. FTO plays an oncogenic role in acute 9 of histone H3. Cell Cycle 8, 2101–2109 (2009). Is there a role for oral or intravenous ascorbate myeloid leukemia as a N6-methyladenosine RNA 246. Wang, P. et al. Oncometabolite d-2-hydroxyglutarate (vitamin C) in treating patients with cancer? demethylase. Cancer Cell 31, 127–141 (2017). inhibits ALKBH DNA repair enzymes and sensitizes A systematic review. Oncologist 20, 210–223 (2015). 221. Zou, D. et al. The m6A eraser FTO facilitates IDH mutant cells to alkylating agents. Cell Rep. 13, 197. Ngo, B., Van Riper, J. M., Cantley, L. C. & Yun, J. proliferation and migration of human cervical cancer 2353–2361 (2015). Targeting cancer vulnerabilities with high-dose cells. Cancer Cell Int. 19, 321 (2019). 247. Lindstedt, G., Lindstedt, S. & Nordin, I. γ-Butyrobetaine vitamin C. Nat. Rev. Cancer 19, 271–282 (2019). 222. Lei, X. et al. JARID2 promotes invasion and hydroxylase in human kidney. Scand. J. Clin. Lab. Invest. 198. Chen, Q. et al. Pharmacologic doses of ascorbate act metastasis of hepatocellular carcinoma by facilitating 42, 477–485 (1982). as a prooxidant and decrease growth of aggressive epithelial–mesenchymal transition through PTEN/AKT 248. Lindstedt, S. & Nordin, I. Multiple forms of tumor xenografts in mice. Proc. Natl Acad. Sci. USA signaling. Oncotarget 7, 40266–40284 (2016). γ-butyrobetaine hydroxylase (EC 1.14.11.1). 105, 11105–11109 (2008). 223. Zhu, X. X. et al. Jarid2 is essential for the maintenance Biochem. J. 223, 119–127 (1984). 199. Qiu, B. & Simon, M. C. Oncogenes strike a balance of tumor initiating cells in bladder cancer. Oncotarget 249. Tarhonskaya, H. et al. Investigating the contribution between cellular growth and homeostasis. Semin. Cell 8, 24483–24490 (2017). of the active site environment to the slow reaction of Dev. Biol. 43, 3–10 (2015). 224. Aprelikova, O. et al. The epigenetic modifier JMJD6 is hypoxia-inducible factor prolyl hydroxylase domain 2 200. Nguyen, C. & Pandey, S. Exploiting mitochondrial amplified in mammary tumors and cooperates with with oxygen. Biochem. J. 463, 363–372 (2014). vulnerabilities to trigger apoptosis selectively in c-Myc to enhance cellular transformation, tumor 250. Ehrismann, D. et al. Studies on the activity of the cancer cells. Cancers https://doi.org/10.3390/ progression, and metastasis. Clin. Epigenetics 8, 38 hypoxia-inducible-factor hydroxylases using an oxygen cancers11070916 (2019). (2016). consumption assay. Biochem. J. 401, 227–234 201. Bjelakovic, G., Nikolova, D., Gluud, L. L., Simonetti, R. G. 225. Wong, M. et al. JMJD6 is a tumorigenic factor and (2007). & Gluud, C. Antioxidant supplements for prevention of therapeutic target in neuroblastoma. Nat. Commun. 251. Koivunen, P., Hirsila, M., Gunzler, V., Kivirikko, K. I. & mortality in healthy participants and patients with 10, 3319 (2019). Myllyharju, J. Catalytic properties of the asparaginyl various diseases. Cochrane Database Syst. Rev. https:// 226. Liu, H., Liu, L., Holowatyj, A., Jiang, Y. & Yang, Z. Q. hydroxylase (FIH) in the oxygen sensing pathway are doi.org/10.1002/14651858.CD007176.pub2 (2012). Integrated genomic and functional analyses of histone distinct from those of its prolyl 4-hydroxylases. J. Biol. 202. Oser, M. G. et al. The KDM5A/RBP2 histone demethylases identify oncogenic KDM2A isoform in Chem. 279, 9899–9904 (2004). demethylase represses NOTCH signaling to sustain breast cancer. Mol. Carcinog. 55, 977–990 (2016). 252. Tarhonskaya, H. et al. Kinetic investigations of the role neuroendocrine differentiation and promote small 227. Wagner, K. W. et al. KDM2A promotes lung of factor inhibiting hypoxia-inducible factor (FIH) as an cell lung cancer tumorigenesis. Genes Dev. 33, tumorigenesis by epigenetically enhancing ERK1/2 oxygen sensor. J. Biol. Chem. 290, 19726–19742 1718–1738 (2019). signaling. J. Clin. Invest. 123, 5231–5246 (2013). (2015).

www.nature.com/nrc Reviews

253. Carbonneau, M. et al. The oncometabolite 259. Mukherji, M. et al. Structure–function analysis of Competing interests 2-hydroxyglutarate activates the mTOR signalling phytanoyl-CoA 2-hydroxylase mutations causing W.G.K. Jr. declares the following competing interests: Lilly pathway. Nat. Commun. 7, 12700 (2016). Refsum’s disease. Hum. Mol. Genet. 10, 1971–1982 Pharmaceuticals (board of directors), Agios Pharmaceuticals 254. Upadhyay, A. K. et al. An analog of BIX-01294 (2001). (scientific advisory board), Cedilla Therapeutics (founder), selectively inhibits a family of histone H3 lysine 9 260. Passoja, K., Rautavuoma, K., Ala-Kokko, L., Kosonen, T. Fibrogen (scientific advisory board, royalties), Nextech Invest Jumonji demethylases. J. Mol. Biol. 416, 319–327 & Kivirikko, K. I. Cloning and characterization of a third (scientific advisory board, carried interest) and Tango (2012). human lysyl hydroxylase isoform. Proc. Natl Acad. Therapeutics (founder). J.-A.L. and P.K. declare no competing 255. Walport, L. J. et al. Human UTY(KDM6C) is a Sci. USA 95, 10482–10486 (1998). interests. male-specific N-methyl lysyl demethylase. J. Biol. 261. Sengoku, T. & Yokoyama, S. Structural basis for Chem. 289, 18302–18313 (2014). histone H3 Lys 27 demethylation by UTX/KDM6A. Peer review information 256. Myllyharju, J. & Kivirikko, K. I. Characterization Genes Dev. 25, 2266–2277 (2011). Nature Reviews Cancer thanks M. Ashcroft, C. Simon and of the iron- and 2-oxoglutarate-binding sites of human 262. Jones, S. E., Olsen, L. & Gajhede, M. Structural basis M. Vissers for their contribution to the peer review of this prolyl 4-hydroxylase. EMBO J. 16, 1173–1180 of histone demethylase KDM6B histone 3 lysine 27 work. (1997). specificity. Biochemistry 57, 585–592 (2018). 257. Majamaa, K., Hanauske-Abel, H. M., Gunzler, V. & Publisher’s note Kivirikko, K. I. The 2-oxoglutarate binding site of Acknowledgements Springer Nature remains neutral with regard to jurisdictional prolyl 4-hydroxylase. Identification of distinct subsites This work was supported by Academy of Finland grants claims in published maps and institutional affiliations. and evidence for 2-oxoglutarate decarboxylation in a 266719 and 308009 and grants from the S. Jusélius ligand reaction at the enzyme-bound ferrous ion. Foundation, the Finnish Cancer Organization and the Jane Eur. J. Biochem. 138, 239–245 (1984). and Aatos Erkko Foundation to P.K. J.-A.L. is supported by Related links 258. Jansen, G. A. et al. Characterization of the NIH. W.G.K. is a Howard Hughes Medical Investigator DepMap: https://depmap.org/portal/ phytanoyl-Coenzyme A hydroxylase in human and supported by the NIH. NCBi protein database: https://www.ncbi.nlm.nih.gov/protein liver and activity measurements in patients with RCsB protein databank: https://www.rcsb.org/ peroxisomal disorders. Clin. Chim. Acta 271, Author contributions 203–211 (1998). The authors contributed equally to all aspects of the article. © Springer Nature Limited 2020

Nature Reviews | Cancer 本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP 图书馆。 图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具