Control of endothelial quiescence by FOXO-regulated metabolites

Articles https://doi.org/10.1038/s41556-021-00637-6 1Angiogenesis and Metabolism Laboratory, Max Planck Institute for Heart and Lung Research, Bad Nauheim, Germany. 2Medical Research Council Cancer Unit, University of Cambridge, Cambridge, UK. 3Graduate School of Medical Science and Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea. 4Center for Vascular Research, Institute for Basic Science (IBS), Daejeon, Korea. 5Gene Editing Group, Institute of Biochemistry II, Goethe University, Frankfurt (Main), Germany. 6Department of Cardiac Development and Remodeling, Max Planck Institute for Heart and Lung Research, Bad Nauheim, Germany. 7Department of Anatomy, Keio University School of Medicine, Tokyo, Japan. 8UCIBIO–Unidade de Ciências Biomoleculares Aplicadas, Departamento Ciências da Vida, Faculdade de Ciências e Tecnologia–Universidade Nova de Lisboa Campus de Caparica, Caparica, Portugal. 9Instituto de Medicina Molecular, Faculdade de Medicina, Universidade de Lisboa, Lisboa, Portugal. 10Berlin Institute of Health (BIH) at Charité–Universitätsmedizin Berlin, Berlin, Germany. 11Max Delbrück Center for Molecular Medicine (MDC), Berlin, Germany. 12Present address: Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA. 13Present address: Department of Oncology and Ludwig Institute for Cancer Research, University of Lausanne and Centre Hospitalier Universitaire Vaudois, Epalinges, Switzerland. 14These authors contributed equally: Jorge Andrade, Chenyue Shi, Ana S. H. Costa. ✉e-mail: michael.pot[email protected] Homeostasis of the blood vasculature relies on a single layer of endothelial cells (ECs) forming the inner surface of all vessels. In most adult tissues, ECs reside in a non-cycling, quiescent state, which is critical for their function as a barrier and signalling interface1,2. This resting state is reversible, and ECs can (re-)activate, enter the cell cycle and expand to meet physiological or stress-induced demands. While the processes of endothelial activation and proliferation are being defined with increasing molecular resolution, the mechanisms leading to the acquisition of a quiescent phenotype remain poorly understood. Previous studies have demonstrated that the forkhead box O (FOXO) transcription factor FOXO1 is a critical driver of endothelial quiescence3,4. FOXO1 activity is controlled by the phosphatidylinositol-3-OH kinase–AKT pathway that inhibits FOXO1 transcriptional responses through AKT-mediated phosphorylation5,6. ECs are highly sensitive to changes in FOXO1 activity, as both EC-specific deletion and forced activation are early embryonic lethal in mice3,4,7,8. Loss of FOXO1 leads to unregulated endothelial proliferation and vascular overgrowth, whereas forced activation causes premature quiescence and vascular rarefaction. FOXO1 promotes quiescence, in part, by suppressing MYC signalling, which leads to a reduction in endothelial metabolic activity4. Yet, the FOXO1-regulated metabolic programmes in ECs are largely unknown. Given the importance of endothelial metabolic regulation for vascular growth and function9–19 and the sensitivity of ECs towards changes in FOXO1 activity, we sought to investigate the global metabolic profile of FOXO1-induced ECs. Results FOXO1 activation induces 2-hydroxyglutarate generation in ECs. We performed untargeted metabolomics of human umbilical vein endothelial cells (HUVECs) that were transduced with adenoviruses encoding a constitutively active FOXO1 mutant (AdFOXO1A3) or green fluorescent protein (GFP) as a control (AdCtrl). This mutant has the three AKT phosphorylation sites replaced by alanine (→FOXO1A3), which renders FOXO1 in the nucleus5,6. ECs expressing FOXO1A3 were in a reversible proliferation arrest and exhibited a molecular signature characteristic of cellular quiescence, including suppression of the proliferation markers MYC and proliferating cell nuclear antigen (PCNA), as well as induction of the cell cycle inhibitor p27 (CDKN1B) and the repressive histone mark histone H3 lysine 27 trimethylation (H3K27me3)20–23 (Extended Data Fig. 1a–c,f). Principal component analysis (PCA) of the intracellular metabolites collected 24 h post-transduction revealed a separate clustering of control and FOXO1A3-transduced ECs (Fig. 1a). ECs with nuclear Control of endothelial quiescence by FOXO-regulated metabolites Jorge Andrade 1,14, Chenyue Shi1,14, Ana S. H. Costa 2,12,14, Jeongwoon Choi3,4, Jaeryung Kim 4,13, Anuradha Doddaballapur 1, Toshiya Sugino 1, Yu Ting Ong 1, Marco Castro 1, Barbara Zimmermann1, Manuel Kaulich 5, Stefan Guenther6, Kerstin Wilhelm1, Yoshiaki Kubota 7, Thomas Braun 6, Gou Young Koh 3,4, Ana Rita Grosso 8,9, Christian Frezza 2 and Michael Potente 1,10,11 ✉ Endothelial cells (ECs) adapt their metabolism to enable the growth of new blood vessels, but little is known how ECs regulate metabolism to adopt a quiescent state. Here, we show that the metabolite S-2-hydroxyglutarate (S-2HG) plays a crucial role in the regulation of endothelial quiescence. We find that S-2HG is produced in ECs after activation of the transcription factor forkhead box O1 (FOXO1), where it limits cell cycle progression, metabolic activity and vascular expansion. FOXO1 stimulates S2HG production by inhibiting the mitochondrial enzyme 2-oxoglutarate dehydrogenase. This inhibition relies on branched-chain amino acid catabolites such as 3-methyl-2-oxovalerate, which increase in ECs with activated FOXO1. Treatment of ECs with 3-methyl-2-oxovalerate elicits S-2HG production and suppresses proliferation, causing vascular rarefaction in mice. Our findings identify a metabolic programme that promotes the acquisition of a quiescent endothelial state and highlight the role of metabolites as signalling molecules in the endothelium. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 413 Articles NaTurE CEll BIOlOGy FOXO1 exhibited increased levels of several branched-chain amino acid (BCAA) metabolites and of early tricarboxylic acid (TCA) cycle intermediates (Supplementary Table 1). Notably, the most increased metabolite (7.6-fold) was 2-hydroxyglutarate (2HG) (Fig. 1b and Extended Data Fig. 1d), which is a chiral molecule derived from 2-oxoglutarate (2OG) that functions as a competitive inhibitor of many 2OG-dependent dioxygenases. Among these are Jumonji C (JmjC) domain-containing lysine demethylases and hypoxia-inducible factor (HIF)-regulating prolyl hydroxylases (PHDs)24–29, which suggests that FOXO1 may exert profound effects on gene expression via regulation of this metabolite. We therefore confirmed the regulation of 2HG using a lentiviral system for doxycycline-inducible expression of FOXO1A3 (Fig. 1c and Extended Data Fig. 1e) and contact inhibition of endothelial proliferation—a stimulus for nuclear translocation and activation of endogenous FOXO1 (Fig. 1e–i and Extended Data Fig. 1g). Indeed, we observed an 8.4-fold increase in 2HG abundance in dense (contact-inhibited) culture conditions, whereby FOXO1 is an induced and predominantly nuclear protein and in which canonical FOXO1 target genes are highly transcribed (for example, MAX interactor 1 (MXI1), pyruvate dehydrogenase kinase 1 (PDK1), pyruvate dehydrogenase kinase 4 (PDK4) and CD36 molecule a PC2 (17.6%) 0 10 20 –10 –20 0 10 20–10 PC1 (50.5%) AdCtrl AdFOXO1A3 c db –log10(P value) 3 2HG 0 2 4 6 8 0 1 2–1–2 log2(fold-change) AdFOXO1A3 vs. AdCtrl e 50 Tubulin 75 Lamin A/C Mr (K) FOXO1 75 Cyto Cyto Nuc Nuc h fSparse Dense i j Nuc Nuc-cyto Cyto **** ECs per field (%) 0 Sparse Dense 25 50 75 100 125 NS g FOXO1 PECAM DAPI FOXO1FOXO1 50 µm Metabolite levels (AU) 0 5 10 15 20 iLentiCtrl iLentiFOXO1A3 iLentiCtrl iLentiFOXO1A3 2HG **** 25 0 5 10 15 Metabolite levels (AU) 20 R-2HG S-2HG **** **** 0 5 10 15 Metabolite levels (AU) 2HG 20 **** **** 0 2.5 5.0 7.5 Metabolite levels (AU) 10.0 R-2HG S-2HG 12.5 **** Mr (K) FOXO1 75 50 MYC 25 S6 25 p-S6 (Ser235/236) 25 PCNA Sparse Dense 50 + + – – – – + + Sparse Dense + + – – – – + + Tubulin 15 H3 15 15 H3K27me3 H3K9me3 100 HIF2α 25 P27 Sparse Dense Sparse Dense **** Fig. 1 | Activation of FOXO1 signalling induces the generation of S-2HG in eCs. a, PCA score plot showing distinct metabolic signatures in control (AdCtrl) and FOXO1A3-expressing (AdFOXO1A3) HUVECs (n = 4 independent samples). The percentage variance explained by each principal component (PC) is shown in parentheses. b, Volcano plot of metabolites showing 2HG as the most increased metabolite in FOXO1A3-expressing HUVECs. c, 2HG metabolite levels measured by LC–MS in HUVECs transduced with a doxycycline-inducible control-encoding (iLentiCtrl) or FOXO1A3-encoding (iLentiFOXO1A3) lentivirus (n = 8 independent samples). AU, arbitrary units. d, Chiral derivatization and enantioselective LC–MS measurement of Rand S-2HG levels in HUVECs transduced with iLentiCtrl or iLentiFOXO1A3 (n = 8 independent samples). e, Immunoblot analysis of HUVECs cultured in sparse and dense conditions. HUVECs in dense (contact-inhibited) conditions express markers linked to cellular quiescence. K, 1,000. f,g, Immunofluorescence analysis (f) and quantification (g) of the subcellular localization of FOXO1 (red) in HUVECs cultured in sparse and dense conditions. The isolated FOXO1 signal is shown on the right side of each image in grey (n = 14 (sparse condition) or 8 (dense condition) independent samples). DAPI (grey), nuclei; PECAM (cyan), intercellular endothelial junctions. Cyto, cytoplasmic; Nuc, nuclear; Nuc–cyto, nuclear–cytoplasmic. h, FOXO1 protein levels in cytoplasmic and nuclear fractions isolated from HUVECs cultured under sparse and dense conditions. Lamin A/C and tubulin were used as nuclear and cytoplasmic markers, respectively. i, 2HG metabolite levels in HUVECs cultured in sparse (FOXO1 inactive) and dense (FOXO1 active) culture conditions (n = 7 independent samples). j, Enantioselective LC–MS measurement of Rand S-2HG in sparse and dense HUVEC cultures (n = 7 independent samples). Western blot data in e and h are from the respective experiment, processed in parallel, and are representative of at least three independent experiments. For c, d, g, i and j, data represent the mean ± s.e.m.; two-tailed unpaired t-test, ****P < 0.0001, NS, not significant. The numerical data, unprocessed western blots and P values are provided as source data. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 414 Articles NaTurE CEll BIOlOGy (CD36)) (Fig. 1f–i and Extended Data Fig. 1g). Thus, ECs generate 2HG when FOXO1 signalling is activated. FOXO1-activated ECs produce S-2HG, which inhibits 2OG-dependent dioxygenases. Two conformations of 2HG exist, the R-2HG and S-2HG enantiomers, which have similar but not identical functions30–32. Chiral derivatization followed by liquid chromatography–mass spectrometry (LC–MS) revealed that FOXO1 induces the preferential generation of S-2HG, which is also the more abundant enantiomer in ECs (Fig. 1d,j). To explore whether 2HG regulates the activity of endothelial 2OG-dependent dioxygenases, we first analysed HIFs, which are marked by PHDs for proteasomal degradation under normoxic conditions. In line with an inhibitory effect of 2HG on PHDs30,31,33–35, incubation of HUVECs with cell-permeable forms of R-2HG or S-2HG increased HIF1α and HIF2α protein abundance and enhanced HIF target gene expression (Extended Data Fig. 2a–c). Interestingly, S-2HG had more profound effects on HIF responses than R-2HG (Extended Data Fig. 2a–c). Similar results were obtained for H3K27me3, which is targeted by the 2OG-dependent JmjC histone demethylases (Extended Data Fig. 2d), which suggests that S-2HG is the more potent inhibitor of 2OG-dependent dioxygenases in ECs. Therefore, we focused our further analysis on this enantiomer. S-2HG promotes a quiescent endothelial state. To characterize the role of S-2HG in ECs, we incubated HUVECs with different S-2HG concentrations and noted a timeand dose-dependent suppression of endothelial proliferation (Fig. 2a–c and Extended Data Fig. 3a). S-2HG-treated ECs were arrested in the G0/G1 phase of the cell cycle, and markers of cell growth and proliferation, including MYC, PCNA and phosphorylated ribosomal protein S6, were diminished (Fig. 2d,e and Extended Data Fig. 3b). S-2HG did not cause metabolic distress (depletion of a particular metabolite), senescence or apoptotic cell death, as reporters of these processes were unaltered or not detectable (Fig. 2f and Extended Data Fig. 3c–f). To further characterize this antiproliferative effect, we transduced HUVECs with a lentiviral vector encoding the fluorescent ubiquitination-based cell cycle indicator (FUCCI). This dual-colour reporter labels nuclei of cells in the G0/G1 phase in red and those in the S/G2/M phases in green (Extended Data Fig. 3g,h). S-2HG caused a progressive increase in ECs with red fluorescence, thereby confirming the G0/ G1 arrest (Fig. 2g and Supplementary Videos 1 and 2). However, S-2HG withdrawal led to the reappearance of ECs with green fluorescent nuclei (Fig. 2g and Supplementary Videos 1 and 2), which indicates that the S-2HG-induced proliferation arrest is reversible. RNA sequencing (RNA-seq) of HUVECs treated with vehicle or S-2HG for 24 h revealed that S-2HG regulates genes strongly linked to proliferation (Fig. 2h, Extended Data Fig. 4a–c and Supplementary Table 2). A total of 43 out of the 50 most downregulated genes are involved in cell cycle progression and cell division, including DNA topoisomerase II alpha (TOP2A), anillin (ANLN), marker of proliferation Ki-67 (MKI67) and cyclin-dependent kinase 1 (CDK1) (Extended Data Fig. 4b). Gene set enrichment analysis (GSEA) corroborated this notion (Fig. 2i and Extended Data Fig. 4c). Moreover, numerous quiescence-associated genes36, including those that are not directly linked to cell proliferation, were enriched in S-2HG-treated ECs (Fig. 2i). These data suggest that S-2HG regulates not only endothelial proliferation but also, more generally, promotes a quiescent endothelial state. Consistent with this idea, S-2HG-treated ECs showed a number of cellular phenotypes that are characteristic for quiescent cells23, including lower metabolic activity and reduced RNA and protein synthesis (Fig. 2j–m and Extended Data Fig. 4d). S-2HG limits the angiogenic behaviour of ECs. To seek further evidence of a quiescence-promoting function of S-2HG, we visualized endothelial angiogenic behaviour in an in vitro scratch assay and found that S-2HG-treated ECs were less motile (Fig. 3a,b and Supplementary Videos 3 and 4). Moreover, analysis of three-dimensional endothelial spheroid cultures revealed that S-2HG-treated spheroids formed fewer and shorter sprouts compared to controls (Fig. 3c and Extended Data Fig. 5a). We validated the relevance of these findings by studying blood vessel growth in the mouse retina, in which blood vessels develop postnatally in a highly stereotypical manner. Postnatal day 5 (P5) pups received a single intraocular injection of S-2HG in one eye and vehicle in the other (Extended Data Fig. 5b). Analysis of the retinal vasculature 2 days later (at P7) showed a poorly connected and hypocellular endothelial network in S-2HG-treated retinas (Fig. 3d,e,i and Extended Data Fig. 5c). This phenotype was particularly evident at the angiogenic front, where most of the endothelial growth and proliferation occurs, while central parts of the retinal vasculature that formed before the injection were less affected (Fig. 3d,e). The sparse network was not due to increased vessel pruning or apoptotic cell death, since analysis of endothelial-less basement membrane sleeves or cleaved caspase-3-positive ECs did not reveal significant changes (Fig. 3g–i). Instead, we found a marked suppression of EC proliferation, which was indicated by a reduction in 5-ethynyl-2′-deoxyuridine (EdU)-positive endothelial nuclei (Fig. 3f,i and Extended Data Fig. 5c). These data are consistent with an Fig. 2 | S-2HG supports a quiescent endothelial phenotype. a, Growth curves of HUVECs stimulated with DMSO (Ctrl) or cell-permeable S-2HG (n = 16 independent samples). b, EdU incorporation in control or S-2HG-treated HUVECs at 48 h. The percentage of EdU-positive ECs is shown (n = 14 independent samples). c, DNA synthesis is reduced in HUVECs treated with S-2HG for 48 h. Values represent the fold-change relative to DMSO-treated HUVECs (n = 8 independent samples). d, Cell-cycle analysis of control and S-2HG-stimulated HUVECs 48 h after treatment (n = 4 independent samples). e, Immunoblotting of proliferation and growth-associated proteins in HUVECs treated with S-2HG for 24 h. f, Immunoblot analysis of the apoptotic markers cleaved (cl.) caspase-3 (CASP3) and cleaved PARP showing that S-2HG does not cause cell death in HUVECs. TNFα (TNF) and cycloheximide (CHX) co-stimulation was used as a positive control. g, Time-course analysis of HUVECs transduced with the dual FUCCI reporter. ECs were treated with vehicle or S-2HG for 48 h, followed by withdrawal of the treatment and further analysis for 48 h (n = 3 independent samples). h, Volcano plot of differentially expressed genes in control or S-2HG-treated HUVECs at 24 h. Genes with a P value cut-off of ≤0.05 and a fold-change ≥ or ≤2 are shown (n = 3 independent samples). i, GSEA plots of quiescent versus dividing-down and quiescent versus dividing-up gene sets in the transcriptomes of control or S-2HG-treated HUVECs. ES, enrichment score; FDR, false discovery rate; NES, normalized enrichment score. j, OCRs in control or S-2HG-treated HUVECs (n = 9 independent samples). AA/R, antimycin A/rotenone; Oligo, oligomycin. k, ATP levels in HUVECs 48 h after treatment with vehicle or S-2HG (n = 4 independent samples). l, RNA synthesis is decreased, as measured by 14C-glucose incorporation into RNA, in HUVECs treated with S-2HG for 48 h. Values are represented as the fold-change relative to control (n = 8 independent samples). m, Protein synthesis is decreased, as measured by 3H-tyrosine incorporation into protein, in HUVECs treated with S-2HG for 48 h. Values are represented as the fold-change relative to control (n = 8 independent samples). Western blot data in e and f are from the respective experiment, processed in parallel, and are representative of at least three independent experiments. For a–d, g and j–m, the data represent the mean ± s.e.m.; two-tailed unpaired t-test, *P < 0.05, **P < 0.01, ****P < 0.0001. The numerical data, unprocessed western blots and P values are provided as source data. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 415 Articles NaTurE CEll BIOlOGy anti-angiogenic activity of S-2HG in ECs and suggest that S-2HG enforces the pro-quiescent function of FOXO1. FOXO1 induces S-2HG generation by inhibiting the OGDH enzyme. Previous studies have shown that S-2HG is generated in conditions of reduced mitochondrial function in which 2OG becomes promiscuously reduced to S-2HG33–35,37–41. To understand how ECs generate S-2HG in response to FOXO1 activation, we examined mitochondrial metabolism. This analysis revealed altered mitochondrial respiration in FOXO1A3-expressing ECs (Extended Data Fig. 6a), which suggests that there is diminished activity of the 2OG dehydrogenase (OGDH) enzyme—a multisubunit complex that catalyses the conversion of 2OG to succinyl-CoA (Extended Data Fig. 6b). Indeed, we observed that FOXO1A3-expressing ECs accumulated the OGDH substrate 2OG, while succinyl-carnitine— a surrogate metabolite of succinyl-CoA—was depleted (Fig. 4a and Supplementary Table 1). Of note, recent genetic studies identified OGDH inhibition as a mechanism for S-2HG generation33,42,43. We therefore used the clustered regularly interspaced short palindromic repeats (CRISPR)–Cas9 method to specifically inactivate OGDH in ECs (Fig. 4b). Transduction of HUVECs with lentiviral constructs encoding FLAG-tagged Cas9 (Cas9FLAG) and guide RNAs (gRNAs) targeting OGDH (gOGDH) led to the expected accumulation of 2OG and a decline in the mitochondrial oxygen consumption rate (OCR) (Fig. 4c and Extended Data Fig. 6c). Importantly, OGDH deficiency increased endothelial S-2HG levels and abrogated the ability of ECs to proliferate (Fig. 4d–g). The growth inhibitory effect was not caused by the increased 2OG levels, because treatment of HUVECs a 20 30 40 10 0 Cell number (×104) 20 40 60 80 100 Time (h) 0 Ctrl S-2HG ***** **** **** b S-2HG Ctrl Gated ECs (%) 0 25 50 75 100 G2/M S G0/G1 **** **** * d eMr (K) Mr (K) Ctrl S-2HG 50 + + – – – – + + + + – – – – + + – – – – – – + Tubulin 37 S6 37 p-S6 (Ser235/236) 37 PCNA 50 MYC 50 Tubulin 100 PARP 75 25 15 CASP3 cl. CASP3 Ctrl S-2HG TNF/CHX cl. PARP fg 0.4 0.6 0.8 0.2 0 Ratio of ECs in S/G2/M vs. G0/G1 20 40 60 80 100 Time (h) 0 Ctrl S-2HG 1.0 Withdrawal FUCCI i 0 –0.2 –0.4 –0.6 –0.8 Enrichment score S-2HG Ctrl ES: –0.88 NES: –3.12 FDR: 0.00 Quiescent vs. dividing - down 0 0.1 0.2 0.3 0.4 0.5 S-2HG Ctrl ES: 0.50 NES: 1.83 FDR: 0.02 Quiescent vs. dividing - up log2(fold-change) 0 4 8–4–8 3.0 1.0 2.0 3.5 2.5 1.5 0.5 0 –log10(P value, ×102) h Ctrl S-2HG Ctrl S-2HG 0 10 20 35 EdU+-ECs (%) **** 30 25 15 5 c 0 0.50 1.00 1.25 0.75 0.25 **** DNA synthesis j 100 150 200 50 0 OCR (pmol O2 min–1) 20 40 60 800 Time (min) Ctrl S-2HG FCCPOligo AA/R *** ** ** k Ctrl S-2HG Ctrl S-2HG Ctrl S-2HG 0 7.5 5.0 2.5 ATP (pmol per cell) * l **** 0 0.50 1.00 1.75 0.75 0.25 RNA synthesis 1.25 1.50 m 0 0.50 1.25 0.25 Protein synthesis 0.75 1.00 **** RNA-seq: S-2HG vs. Ctrl NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 416 Articles NaTurE CEll BIOlOGy with (cell permeable) 2OG did not influence their capacity to sprout or divide (Fig. 4h and Extended Data Fig. 6d,e). Moreover, it was not a generic response to TCA cycle perturbation, because inactivation of two other TCA cycle enzymes, succinate dehydrogenase (SDH) and fumarate hydratase (FH), neither increased S-2HG levels nor inhibited endothelial proliferation (Fig. 4i–k and Extended Data Fig. 6f–k). These data suggest a crucial role of OGDH in endothelial cell-cycle control, which relies on the signalling metabolite S-2HG. Endothelial OGDH is crucial for angiogenic growth. To determine the physiological relevance of these findings, we generated mice in which exon 3 and 4 of the Ogdh gene are flanked by loxP sites (Ogdhfl) (Extended Data Fig. 7a). Deleting these exons in ECs (and some haematopoietic cells) using a constitutive Tie2-cre transgene (OgdhEC-KO) caused severe defects in the yolk sac vasculature that led to developmental retardation and embryonic death (Fig. 5a,b and Extended Data Fig. 7b,c). To further characterize the functional consequences of Ogdh deletion, we analysed blood vessel growth in the postnatal retina. To this end, we bred Ogdhfl mice with mice expressing the tamoxifen-inducible recombinase creERT2 from the endothelial-restricted Cdh5 promoter (OgdhiECKO). 4-Hydroxytamoxifen (4-OHT)-induced depletion of OGDH caused a sparse vascular network that mimicked the defective vasculature in the constitutive Ogdh knockout embryos (Fig. 5c–e and Extended Phalloidin cS-2HGCtrl aCtrl 12 h S-2HG 12 h Ctrl S-2HG b 300 µm 200 µm200 µm 200 µm 200 µm 200 µm 100 µm ed S-2HGCtrl VA PECAM gf Ctrl S-2HG hi Ctrl S-2HG VAVA EdU ERG PECAM cl. CASP3 ICAM PECAM cl. CASP3 AF AF AF CP V A AF AF A F CP ICAM PECAM COL ERG PECAM 0 50 125 25 Scratch closure (%) 75 100 **** 0 25 100 EC area (%) 50 75 **** 0 20 50 EC proliferation (%) 30 40 10 0 5.0 10.0 Regression index 7.5 2.5 NS Proliferating ECs **** Fig. 3 | S-2HG restrains the angiogenic activity of eCs. a,b, S-2HG reduces the motile behaviour of cultured HUVECs in a scratch-wound assay. Quantification (a) and representative bright-field images (b) from control and S-2HG-treated HUVECs (n = 15 independent samples). c, Confocal images of phalloidin-labelled (grey) HUVEC spheroids showing reduced endothelial sprouting in S-2HG treated spheroids. Images were taken 24 h after treatment. d, PECAM1 (PECAM) immunofluorescence staining (grey) in P7 mouse retinas showing decreased vascular density after a single intravitreal injection of cell-permeable S-2HG at P5. Controls were obtained by injection of DMSO (vehicle, Ctrl) in the contralateral eye. A, artery; AF, angiogenic front; CP, capillary plexus; V, vein. e, Confocal images of P7 retinas from control and S-2HG-injected mice stained for ERG (cyan) and PECAM (red). f, Immunofluorescence staining for EdU (red), ERG (green) and PECAM (blue) in P7 mouse retinas of control and S-2HG mice. Proliferating (EdU and ERG double-positive) ECs are shown in yellow. g, Confocal images of retinas from control and S-2HG-injected mice stained for ICAM2 (ICAM, green), PECAM (blue) and collagen IV (COL, red). h, Immunofluorescence images of cleaved caspase-3 (yellow), ICAM (cyan) and PECAM (red) of P7 retinas from control and S-2HG-injected mice, excluding excessive apoptotic cell death in S-2HG-treated mice. Note that most of the cleaved caspase-3 signals come from nonvascular (ICAM/PECAM-negative) cells. i, Quantification of angiogenic parameters in retinas of control and S-2HG-injected P7 mice, as indicated (EC area: n = 21 samples each for control and S-2HG; EC proliferation: n = 24 each samples for control and S-2HG; regression index: n = 18 and 22 samples for control and S-2HG, respectively). For a and i, the data represent the mean ± s.e.m.; two-tailed unpaired t-test, ****P < 0.0001. The numerical data and P values are provided as source data. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 417 Articles NaTurE CEll BIOlOGy Data Fig. 7d–g). Ogdh mutant ECs had elevated S-2HG levels and exhibited diminished proliferation, but lacked signs of increased cell death (Fig. 5e–i and Extended Data Fig. 7f,g). This result underscores the importance of OGDH function for endothelial proliferation and vascular growth control. FOXO1 inhibits OGDH function via the generation of BCAA catabolites. We then investigated the mechanisms through which FOXO1 regulates OGDH function. First, we examined whether FOXO1 regulates the expression of components of the OGDH complex. For this purpose, we performed RNA-seq analysis in control and FOXO1A3-expressing HUVECs. While FOXO1 governed the expression of prototypical FOXO1 target genes in the expected manner, we failed to detect changes in components of the OGDH complex that would explain a reduction in enzymatic activity (Fig. 6a). Similar data were obtained at the protein level (Extended Data Fig. 8a). We therefore considered alternative mechanisms of regulation. Our initial metabolomics analysis showed that, besides 2HG, FOXO1 induces the accumulation of 3-methyl-2-oxovalerate (KMV), 3-methyl-2-oxobutyrate (KIV) and 4-methyl-2-oxopentanoate (KIC), and other intermediates of BCAA catabolism (Fig. 6b–d). Strikingly, these metabolites fd 20 30 50 10 0 Cell number (×104) 20 40 60 80 100 Time (h) 0 40 **** **** **** **** Metabolite levels (AU) 0 2.0 4.0 8.0 gCtrl gOGDH gCtrl + – – – – + – – – – + – – – – + gOGDH 2HG **** 6.0 e 0 2.5 Metabolite levels (AU) 7.5 R-2HG S-2HG 5.0 **** **** g EdU+-ECs (%) EdU+-ECs (%) EdU+-ECs (%) 0 5 10 20 gCtrl gOGDH 15 **** Ctrl 2OG 0 10 20 30 NS h i 50 FH OGDH 100 75 SDHA 50 Tubulin gCtrl gOGDH gSDHA gFH Metabolite levels (AU) 0 5.0 2.5 12.5 7.5 gCtrl gOGDH gSDHA gFH gCtrl gOGDH gSDHA gFH **** **** *** 10.0 S-2HG j 0 10 30 20 50 40 NS NS k a gCtrl gOGDH gCtrl gOGDH 50 Tubulin 150 100 FLAG (Cas9) OGDH Mr (K) Mr (K) cb 0 50 100 150 Basal FCCP OCR (pmol O2 min–1) **** **** gCtrl gOGDH Metabolite levels (AU) 0 + + – – – – + + 0.5 1.5 2.0 1.0 ** Succinylcarnitine Metabolite levels (AU) 0 0.5 1.0 2OG 1.5 *** AdCtrl AdFOXO1A3 AdCtrl AdFOXO1A3 Fig. 4 | OGDH inactivation induces S-2HG generation and limits endothelial proliferation. a, Metabolite levels of 2OG and succinyl-carnitine (a surrogate for succinyl-CoA) in HUVECs transduced with control (AdCtrl) or FOXO1A3-encoding (AdFOXO1A3) adenoviruses (n = 4 independent samples). b, Immunoblot analysis of OGDH protein abundance in control and OGDH-depleted HUVECs. Cells were generated by lentiviral transduction of FLAG-tagged Cas9 nuclease and control (gCtrl) or OGDH-targeting (gOGDH) gRNAs. c, OCRs in gCtrland gOGDH-transduced HUVECs under basal conditions and in response to FCCP (n = 8 independent samples). d, 2HG metabolite levels in gCtrl and gOGDH HUVECs measured by LC–MS (n = 7 (gCtrl) and 8 (gOGDH) independent samples). e, Enantioselective LC–MS measurement of Rand S-2HG levels in gCtrl and gOGDH HUVECs (n = 7 (gCtrl) and 8 (gOGDH) independent samples). f, Cell-proliferation curves comparing gCtrl and gOGDH HUVECs (n = 12 independent samples). g, EdU incorporation in control and OGDH-deficient HUVECs. The percentage of EdU-positive ECs is shown (n = 8 (gCtrl) and 12 (gOGDH) independent samples). h, EdU incorporation in control and cell-permeable 2OG-treated HUVECs. The percentage of EdU-positive ECs is shown (n = 5 independent samples). i, Immunoblot analysis of OGDH, SDHA and FH protein levels in control (gCtrl) and CRISPR–Cas9-engineered HUVECs (gOGDH, gSDHA or gFH). j, LC–MS measurement of 2HG enantiomers showing that OGDH-deficient HUVECs (gOGDH), but not SDHAor FH-deficient cells, have increased S-2HG levels (n = 8, 7, 8 and 8 independent samples for gCtrl, gOGDH, gSDHA and gFH, respectively). k, EdU incorporation in gCtrl, gOGDH, gSDHA and gFH HUVECs. The percentage of EdU-positive ECs is shown (n = 5, 4, 4 and 4 independent samples for gCtrl, gOGDH, gSDHA and gFH, respectively). Western blot data in b and i are from the respective experiment, processed in parallel, and are representative of at least three independent experiments. For a, c, d–h, j and k, the data represent the mean ± s.e.m.; two-tailed unpaired t-test, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, NS, not significant. The numerical data, unprocessed western blots and P values are provided as source data. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 418 Articles NaTurE CEll BIOlOGy are known inhibitors of OGDH34,38,44,45, which suggests that the FOXO1-induced increase in these branched chain α-keto acids (BCKAs) leads to inhibition of the OGDH complex. Consistent with this model, KMV reduced mitochondrial oxygen consumption and OGDH activity, and led to a 9.7-fold increase in endothelial S-2HG levels (Fig. 6e–h). Genes involved in BCAA metabolism are direct FOXO1 target genes. We probed further to elucidate how FOXO1 regulates BCKA generation. Interrogation of the transcriptional signature of FOXO1A3-transduced HUVECs demonstrated that FOXO1 induced several genes coding for enzymes of the BCAA catabolic pathway, including dihydrolipoamide branched chain transacylase E2 (DBT), branched chain keto acid dehydrogenase E1 subunit beta (BCKDHB), acyl-CoA dehydrogenase short/branched chain (ACADSB) and methylmalonyl-CoA mutase (MUT), among others (Fig. 7a,b and Extended Data Fig. 8b). Notably, DBT and BCKDHB are subunits of the branched-chain α-ketoacid dehydrogenase EdU ERG PECAM A V AV aOgdhEC-KO OgdhiEC-KO Ctrl 500 µm 200 µm 500 µm b eCtrl f g ERG PECAM A V cCtrl d VA A OgdhiEC-KO Ctrl OgdhiEC-KO A V VA A PECAM cl. CASP3 ICAM PECAM A VA V A V EC area (%) 0 20 60 40 **** **** Number of ECs 0 50 100 75 25 EC proliferation (%) 0 20 40 30 10 *** AdCtrl AdCre S-2HG Metabolite levels (AU) 0 2.0 3.0 4.0 ** 1.0 AdCtrl AdCre Mr (K) OGDH 100 37 Cre 100 + – + – – + – + PECAM h i Proliferating ECs 200 µm 200 µm 200 µm Fig. 5 | loss of endothelial OGDH restricts vascular growth. a,b, Control (Ogdhfl/fl, Ctrl) and endothelial-restricted Ogdh (Tie2-cre;Ogdhfl/fl, OGDHEC-KO) mutant yolk sacs (a) and embryos (b) at embryonic day 11. 5 (E11.5) showing reduced vascularization of the yolk sac and retarded development of the Ogdh knockout embryos. c, Confocal images of PECAM-stained (grey) P6 retinas isolated from control (Ogdhfl/fl) and inducible endothelial-specific Ogdh mutant mice (Cdh5-creERT2;Ogdhfl/fl; OgdhiEC-KO) following 4-OHT injection from P1 to P4. d, Immunofluorescence staining for ERG (cyan) and PECAM (red) in control and OgdhiEC-KO mice. e, Labelling of EdU (red), ERG (green) and PECAM (blue) of control and OgdhiEC-KO retinas at P6, revealing a decreased number of proliferating ECs (yellow) in OgdhiEC-KO mutants. f, Confocal images of cleaved caspase-3 (yellow), ICAM (cyan) and PECAM (red) labelling in retinas from control and OgdhiEC-KO mice. White arrowheads indicate apoptotic ECs. g, Quantification of retinal angiogenesis at P6 in control and OgdhiEC-KO mice (EC area: n = 10 (control) and 8 (OgdhiEC-KO) samples; number of ECs: n = 10 (control) and 9 (OgdhiEC-KO) samples; EC proliferation: n = 3 (control) and 5 (OgdhiEC-KO) samples). h, OGDH protein expression in ECs isolated from the lungs of Ogdhfl/fl mice followed by transduction with control (AdCtrl) or Cre-encoding (AdCre) adenoviruses. i, S-2HG metabolite levels in AdCtrl and AdCre-transduced lung ECs derived from Ogdhfl/fl mice (n = 3 (AdCtrl) and 5 (AdCre) independent samples). Western blot data in h are from the respective experiment, processed in parallel, and are representative of at least three independent experiments. For g and i, the data represent the mean ± s.e.m.; two-tailed unpaired t-test, **P < 0.01, ***P < 0.001, ****P < 0.0001. The numerical data, unprocessed western blots and P values are provided as source data. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 419 Articles NaTurE CEll BIOlOGy (BCKD) complex, which oxidizes KMV, KIC and KIV to branched-chain acyl-CoAs and constitutes the rate-limiting step in BCAA metabolism46. Moreover, expression of protein phosphatase, Mg2+/Mn2+-dependent 1K (PPM1K), a phosphatase that activates the BCKD complex46, was also increased by FOXO1, which suggests that FOXO1 coordinates the expression of genes involved in BCAA catabolism. Of note, FOXO1 regulates these genes independently of its repressive effects on MYC signalling4, since inactivation of MYC by CRISPR–Cas9 did not suppress the levels of these BCAA transcripts (Extended Data Fig. 8c–e). We therefore studied whether FOXO1 directly targets these genes, and conducted FOXO1 chromatin immunoprecipitation with sequencing (ChIP-seq) in FOXO1A3-expressing ECs. We found that FOXO1 peaks were enriched for the high-affinity FOXO1-binding motif and were preferentially located at gene promoters close to the transcriptional start site (Extended Data Fig. b 0 2 4 6 8 0123–1–2 –log10 (P value) log2 (fold-change) AdFOXO1A3 vs. AdCtrl BCAA metabolites KMV KIV KIC c Metabolite levels (AU) 0 iLentiCtrl iLentiFOXO1A3 KMV 1 2 3 4 5 ** *** *** ** *** ** **** 0 4 40 60 80 CD36 MXI1 PDK4 PDK1 DLST DLD OGDH MYC RNA-seq AdCtrl AdFOXO1A3 Fold-change 8 a h 0 1 Metabolite levels (AU) 3 R-2HG S-2HG 2 g 0 2.0 3.0 3.5 Metabolite levels (AU) 2.5 1.5 1.0 0.5 **** **** **** 2HG e Ctrl KMV Ctrl KMV Ctrl KMV Ctrl KMV 0 50 100 125 OGDH activity (%) 75 25 **** f 0 175 Basal FCCP ** ** 150 125 100 75 50 25 OCR (pmol O2 min–1) d 0 1 3 Fold-change BCAA metabolites 2 3-Methyl-2-oxovalerate 3-Methyl-2-oxobutyrate 4-Methyl-2-oxopentanoate Isovalerylcarnitine β-hydroxyisovalerate Isobutyrylcarnitine Ethylmalonate ****** **** * **** **** *** **** ***** ** α-hydroxycaproate 3-Hydroxyisobutyrate 2-Methylbutyrylglycine 3-Methylcrotonylglycine Methylsuccinate 2-Methylbutyrylcarnitine 3-Hydroxy-2-ethylpropionate N-acetylleucine α-hydroxyisovalerate Valine β-hydroxyisovaleroylcarnitine 4 AdCtrl AdFOXO1A3 ** Fig. 6 | FOXO1 induces S-2HG generation by regulating BCAA catabolism. a, Expression of canonical FOXO1 targets and OGDH complex subunits in HUVECs transduced with a control-encoding (AdCtrl) and FOXO1A3-encoding (AdFOXO1A3) adenovirus. Cells were collected 24 h after transduction and analysed by RNA-seq (n = 3 independent samples). b, Volcano plot showing increased levels of BCAA catabolites in FOXO1A3-expressing HUVECs (n = 4 independent experiments). c, KMV metabolite levels in HUVECs transduced with a doxycycline-inducible control-encoding (iLentiCtrl) or FOXO1A3-encoding lentivirus (iLentiFOXO1A3) (n = 6 (iLentiCtrl) and 10 (iLentiFOXO1A3) independent samples). d, Changes in BCAA metabolites in AdCtrl versus AdFOXO1A3 expressing HUVECs. Data represent the fold-change relative to control (n = 4 independent samples). e, OGDH activity assay in control (PBS, Ctrl) or KMV-treated HUVECs (n = 3 independent experiments). f, Decreased basal and maximal (FCCP) OCRs in HUVECs treated with KMV for 48 h compared to control (n = 5 (control) or 8 (KMV) independent samples). g, 2HG metabolite levels in control and KMV-treated HUVECs measured by LC–MS (n = 9 independent samples). h, 2HG chiral derivatization and enantioselective MS measurement of Rand S-2HG levels in control or KMV-treated HUVECs (n = 9 independent samples). For a–h, the data represent the mean ± s.e.m.; two-tailed unpaired t-test, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. The numerical data and P values are provided as source data. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 420 Articles NaTurE CEll BIOlOGy 9a,b). FOXO1 ChIP-seq identified bona fide FOXO1 targets, including several of the differentially regulated BCAA genes (DBT, DLD, PPM1K, MUT and OXCT1) (Fig. 7c,d, Extended Data Fig. 9c–f and Supplementary Table 3). These FOXO1-bound genes were also in an active state, as indicated by the acetylation of H3K27 and the trimethylation of H3K4 (Fig. 7c,d and Extended Data Fig. 9c–f). ab 0 5 10 45 Fold-change AdCtrl AdFOXO1A3 ** qPCR 15 40 35 ** ** ** ** ** * AOX1 OXCT1 ABAT ACADSB MUT DBT PPM1K BCKDHB 16 h 24 h 32 h 16 h 24 h 32 h AdCtrl AdFOXO1A3 Row min Row max AOX1 OXCT1 ACADSB ABAT MUT DBT ACADM BCKDHB PPM1K d c ChIP-seq signal 0 129.8 259.5 MUT 0 423.5 847.0 MUT 0 432.7 865.4 FOXO1 H3K27ac H3K4me3 ChIP-seq signal 0 123.7 247.4 DBT 0 492.2 984.5 DBT sfitom1OXOFsfitom1OXOF 0 611.8 1223.6 FOXO1 H3K27ac H3K4me3 e hi KMVCtrl g PECAM A V AVV EdU ERG PECAM (h) Cell number (×104) 20 80 100 60 0 20 40 60 80 100 0 Ctrl KMV 40 **** **** **** **** ** f Ctrl KMV Ctrl KMV 0 20 40 60 EC area (%) 0 10 20 35 EC branch points 30 25 15 5 0 20 40 50 EC proliferation (%) ** 30 10 0 20 30 35 EdU+-ECs (%) 25 10 *** 15 5 Proliferating ECs RNA-seq *** 200 µm 200 µm Fig. 7 | The FOXO1-regulated BCAA catabolite KMV limits endothelial proliferation. a, Heatmap of genes involved in BCAA metabolism that are induced in FOXO1A3-expressing (AdFOXO1A3) HUVECs compared to control (AdCtrl). Transcript levels were assessed at 16, 24 and 32 h post-transduction and analysed by RNA-seq. Genes with a fold-change ≥2 and a P value of <0.05 are shown (n = 3 independent samples). b, RT–qPCR analysis validating the increased expression of BCAA metabolism genes in AdFOXO1A3-transduced HUVECs. Values are normalized to β-actin levels and represented as fold-change relative to control (n = 3 independent samples). c,d, FOXO1, H3K27ac and H3K4me3 ChIP-seq signals at the genomic loci of the DBT (c) and MUT (d) gene. FOXO consensus motifs bound by FOXO1 are indicated in orange. Unbound FOXO motifs are shown in grey. ChIP-seq signals are represented as reads per kilobase million. e, Cell-proliferation curves of HUVECs treated with vehicle (PBS, Ctrl) or KMV for the indicated times (n = 12 independent samples). f, EdU incorporation is reduced in HUVECs treated with KMV for 48 h (n = 6 independent samples). g, Confocal images for PECAM (grey) labelling in P7 mouse retinas showing decreased vascular density after a single intraocular injection of KMV at P5. Mice injected with PBS were used as a control. h, Quantifications of vascular parameters in the retina of control and KMV-injected mice, as indicated (EC area: n = 20 (control) and 24 (KMV) samples; EC branch points: n = 21 (control) and 24 (KMV) samples; EC proliferation: n = 21 (control) and 23 (KMV) samples). i, EdU (red), ERG (green) and PECAM (blue) labelling of control and KMV-injected retinas at P7, revealing decreased endothelial proliferation following KMV treatment. For b, e, f and h, the data represent the mean ± s.e.m.; two-tailed unpaired t-test, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. The numerical data and P values are provided as source data. NATuRe Cell BiOlOGY | VOL 23 | APRIL 2021 | 413–423 | www.nature.com/naturecellbiology 421 Articles NaTurE CEll BIOlOGy Extended Data Fig. 1 | Foxo1 promotes a quiescent endothelial phenotype. a, Immunoblot analysis of quiescence-associated protein markers in HUVECs transduced with a FOXO1A3 (AdFOXO1A3) or control (AdCtrl) adenovirus. A FLAG antibody was used to validate the expression of the FLAG-tagged FOXO1A3 mutant. Tubulin served as loading control. b, Confocal images showing decreased EdU-incorporation in HUVECs transduced with AdFOXO1A3. The analysis was performed 24h after transduction. DAPI was used to identify endothelial nuclei. c, Quantification of EdU-incorporation in AdCtrl and FOXO1A3-expressing HUVECs. Values represent the percentage of EdU-labelled ECs, (n=6, 10 independent samples for AdCtrl and AdFOXO1A3). d, 2HG levels in AdCtrland AdFOXO1A3-transduced HUVECs measured by LC-MS, (n=4 independent samples). AU, arbitrary units. e, Immunoblot analysis of HUVECs transduced with a doxycycline-inducible control- (iLentiCtrl) or FOXO1A3-encoding lentivirus (iLentiFOXO1A3) showing expression of the FLAG-tagged FOXO1A3 mutant after doxycycline (Dox) treatment for 48h. The asterisk (*) denotes an unspecific protein detected by the FLAG antibody. f, Immunoblot analysis showing that the FOXO1-induced quiescence signature is reversible. HUVECs transduced with iLentiCtrl or iLentiFOXO1A3 were treated with Dox for 48h, after which Dox was removed from the culture media. HUVECs were then cultured for additional 48h. g, Quantitative RT-PCR (RT-qPCR) showing increased expression of canonical FOXO1 target genes in dense or sparse HUVEC cultures. Values are normalized to β-actin and represent fold-change regulation relative to control, (n=3 independent samples). Western blot data in a, e and f were from the respective experiment, processed in parallel, and are representative of at least three independent experiments. c, d and g, Data represent mean ± s.e.m.; a two-tailed unpaired t-test was used; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001. The numerical data, unprocessed western blots and P values are provided as source data. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 2 | Rand S-2HG reduce the activity of 2-oxoglutarate-dependent dioxygenases in eCs. a, Immunoblot analysis of HIF1α and HIF2α protein abundance in HUVECs stimulated with cell-permeable Ror S-2HG for 24h. HUVECs treated with vehicle (DMSO) were used as a control (Ctrl). The PHD inhibitor DMOG, which stabilizes HIF protein levels, was used as a positive control. b, Heatmap of hypoxia associated genes that are differentially regulated in control (Ctrl) versus R-2HG or S-2HG-treated HUVECs. DMSO was used as a vehicle control. Cells were stimulated for 24h before total mRNA was collected for RNA-seq analysis, (n=3 independent samples). c, Gene set enrichment analysis (GSEA) showing a HIF gene expression signature in HUVECs treated with Rand S-2HG for 24h when compared to Ctrl. ES, enrichment score; NES, normalized enrichment score; FDR, false discovery rate. d, Immunoblot analysis showing increased histone H3 lysine 27 tri-methylation (K27me3) levels in HUVECs treated with R-2HG or S-2HG when compared to Ctrl. Cells were analysed 24 or 48h after stimulation. Total levels of histone H3 are shown as protein loading control. Western blot data in a and d were from the respective experiment, processed in parallel, and are representative of at least three independent experiments. Unprocessed western blots are provided as source data. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 3 | See next page for caption. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 3 | S-2HG is a cell cycle-arresting metabolite. a, Proliferation curves comparing HUVECs treated with DMSO (Ctrl) or cell-permeable S-2HG, showing a dose-dependent reduction in S-2HG-stimulated HUVECs at the indicated concentrations and time points, (n=10 independent samples). b, Representative flow cytometry density plots showing the cell cycle phase distribution in HUVECs treated with vehicle (Ctrl) or S-2HG for 48 hours. c, Relative abundance of TCA cycle metabolites in Ctrl and S-2HG-treated HUVECs. LC-MS measurements were performed 48h after stimulation, (n=8 independent samples). d, Relative abundance of amino acids in Ctrl and S-2HG-treated HUVECs. LC-MS measurements were performed 48h after stimulation, (n=8 independent samples). e, Representative images and quantification of senescence-associated β-galactosidase (SA-β-Gal) staining showing that S-2HG does not induce cellular senescence. HUVECs were stimulated with DMSO (Ctrl) or S-2HG for 48h. Hydrogen peroxide (H2O2) stimulated cells were used as a positive control, (n=4 independent samples). f, LC3A/B immunoblot analysis showing that S-2HG treatment for 48h does not induce autophagy in ECs. Chloroquine (CQ) treated HUVECs were used as a positive control. g, Visualization of cell cycle progression using the dual fluorescence ubiquitination-based cell cycle indicator (FUCCI) reporter. Cells with red-labelled nuclei (expressing mCherry-hCdt1(30/120)) are in G0/G1 while cells with green labelled nuclei (expressing mVenus-hGeminin(1/110)) are in the S/G2/M cell cycle phase. Yellow nuclei indicate temporal co-expression of both reporters. h, Experimental timeline for the cell cycle analysis in HUVECs transduced with the FUCCI reporter. The ratio between green and red cells indicates the percentage of ECs in S/G2/M compared to G0/G1, respectively. Western blot data in f were from the respective experiment, processed in parallel, and are representative of at least three independent experiments. a and c-e, Data represent mean ± s.e.m.; a two-tailed unpaired t-test was used; *P<0.05; ** P<0.01; ***P<0.001; **** P<0.0001; NS, not significant. The numerical data, unprocessed western blots and P values are provided as source data. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 4 | See next page for caption. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 4 | S-2HG promotes a quiescent state in eCs. a, Gene ontology (GO) analysis showing top GO terms of genes that are downregulated in S-2HG-treated HUVECs as determined by RNA-seq analysis at 24h post treatment. DMSO-treated HUVECs were used as a control (Ctrl), (n=3 independent samples). b, Heatmap showing the top down-regulated genes in the transcriptome of HUVECs treated with S-2HG for 24h. Transcripts highlighted in red are cell-cycle and proliferation related genes. c, GSEA plots of cell division and cell cycle phase transition gene sets in the transcriptomes of HUVECs treated with S-2HG or solvent (Ctrl) for 24h. ES, enrichment score; NES, normalized enrichment score; FDR, false discovery rate. d, S-2HG decreases endothelial protein synthesis. Immunoblot analyses showing reduced incorporation of puromycin (PURO) into nascent polypeptide chains in whole-cell lysates of HUVECs treated with S-2HG or solvent (Ctrl) for 48h. Cycloheximide (CHX) stimulation was used as a positive control to block protein synthesis. Western blot data were from the respective experiment, processed in parallel, and are representative of at least three independent experiments. Unprocessed western blots are provided as source data. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 5 | S-2HG limits the angiogenic capacity of eCs. a, Quantifications of vascular parameters showing reduced endothelial sprouting capacity in S-2HG treated HUVEC spheroids, (n=5 independent samples). b, Experimental timeline for the retinal analysis after intraocular injection of a single dose of vehicle (DMSO, Ctrl) or S-2HG. The images of PECAM-labelled retinas on the right illustrate the extent of angiogenic growth between P5 and P7 in wild-type mice. c, Quantifications of vascular parameters in P7 Ctrl and S-2HG injected mouse retinas as indicated, (Number ECs: n=56, 51 samples for Ctrl and S-2HG; Branching frequency: n=24, 25 samples for Ctrl and S-2HG; EC proliferation for Artery, Capillary and Vein: n=10, 10 samples for Ctrl and S-2HG). a and c, Data represent mean ± s.e.m.; a two-tailed unpaired t-test was used; *P<0.05; ** P<0.01; **** P<0.0001; NS, not significant. The numerical data and P values are provided as source data. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 6 | See next page for caption. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 6 | Changes in endothelial mitochondrial metabolism upon FOXO1A3 expression or OGDH depletion. a, Oxygen consumption rate (OCR) in control (iLentiCtrl) and FOXO1A3-transduced (iLentiFOXO1A3) HUVECs 48h after doxycycline induction. Oligo, oligomycin; FCCP, fluoro-carbonyl cyanide phenylhydrazone; AA / R, Antimycin A / Rotenone, (n=4 independent samples). b, Schematic representation showing the metabolic substrates and products catalysed by the TCA cycle enzymes OGDH, SDH and FH. c, Metabolite levels of 2-oxoglutarate (2OG) and succinate in control (gCtrl) and OGDH-depleted (gOGDH) HUVECs, (2OG: n=8, 8 independent samples for gCtrl and gOGDH; Succinate: n=5, 5 for gCtrl and gOGDH). AU, arbitrary units. d, Confocal images of phalloidin- (grey) labelled HUVEC spheroids showing that 2OG does not affect endothelial sprouting. Images were taken 24h after treatment. Quantifications are shown on the right, (n=3 independent samples). e, Scratch-wound assay quantification of vehicle (Ctrl) and 2OG-treated HUVECs, (n=5 independent samples). f-i, TCA cycle metabolite levels in control (gCtrl), SDHA- (gSDHA), OGDH- (gOGDH) and FH-depleted (gFH) HUVECs, (n=8 independent samples). AU, arbitrary units. j, EdU-incorporation in gCtrl, gSDHA, gOGDH and gFH transduced HUVECs. DAPI was used to identify endothelial nuclei. k, Oxygen consumption rate (OCR) in gCtrl, gOGDH, gSDHA and gFH depleted HUVECs. Measurements were performed under basal conditions and in response to FCCP stimulation, (n=6 independent samples). a, c-i and k, Data represent mean ± s.e.m.; a two-tailed unpaired t-test was used; ** P<0.01; ***P<0.001; **** P<0.0001; NS, not significant. The numerical data and P values are provided as source data. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology Articles NaTurE CEll BIOlOGy Extended Data Fig. 7 | loss of endothelial Ogdh impairs vascular growth. a, Schematic illustration of the strategy to generate a conditional Ogdh knockout allele, in which exons 3 and 4 are flanked by loxP sites (red triangles). The Ogdh genomic locus, the targeting vector, the targeted allele and the SDAand cre-recombined loci are shown. SDA-Neo-SDA, neomycin resistance cassette flanked by SDA sites. b, Immunofluorescence staining for ERG (cyan) and PECAM1 (PECAM, red) of control (Ctrl, Ogdhfl/fl) and endothelial-restricted Ogdh knockout (OgdhEC-KO, Tie2-cre;Ogdhfl/fl) yolk sacs at E11.5, showing reduced vascular density after Ogdh loss. c, Quantifications of vascular parameters in E11.5 Ctrl and OgdhEC-KO yolk sacs, as indicated, (EC area: n=7, 4 samples for Ctrl and OgdhEC-KO; Number of ECs: n=7, 4 samples for Ctrl and OgdhEC-KO). d, PCR analysis of genomic DNA from control (Ogdh+/+, lane 2; Ogdhfl/+, lane 3; Ogdhfl/fl, lane5) and conditional Ogdh mutant mice (Cdh5-creERT2; Ogdhfl/+, lane4; Cdh5-creERT2;Ogdhfl/fl, lane 6). Lane 1, DNA marker (M). e, Immunoblot analysis of OGDH protein levels in ECs isolated from the liver of 4-OHT-injected Ctrl (Ogdhfl/fl) and OgdhiEC-KO (Cdh5-creERT2;Ogdhfl/fl) mouse mutants. GAPDH served as loading control. f, Quantifications of vascular parameters in P6 Ctrl and OgdhiEC-KO mouse mutants, as indicated (Filopodia per vessel segment: n=10, 8 samples for Ctrl and OgdhiEC-KO; Branching frequency: n=10, 8 samples for Ctrl and OgdhiEC-KO; Regression index: n=5, 3 samples for Ctrl and OgdhiEC-KO). g, Immunostaining showing ICAM2- (ICAM, green), PECAM- (blue), and collagen IV- (COL, red) labelling of P6 retinas from 4-OHT-injected Ctrl and OgdhiEC-KO mice. Western blot data in e were from the respective experiment, processed in parallel, and are representative of at least three independent experiments. c and f, Data represent mean ± s.e.m.; a two-tailed unpaired t-test was used; ** P< 0.01; ***P<0.001; ****P<0.0001; NS, not significant. The numerical data, unprocessed western blots and P values are provided as source data. NATuRe Cell BiOlOGY | www.nature.com/naturecellbiology