L-2hydroxyglutaric acid rewires amino acid metabolism in colorectal cancer via the mTOR-ATF4 axis

Oncometabolites, such as D/L-2-hydroxyglutarate (2HG), have directly been implicated in carcinogenesis; however, the underlying molecular mechanisms remain poorly understood. Here, we showed that the levels of the L-enantiomer of 2HG (L2HG) were specifically increased in colorectal cancer (CRC) tissues and cell lines compared with the D-enantiomer of 2HG (D2HG). In addition, L2HG increased the expression of ATF4 and its target genes by activating the mTOR pathway, which subsequently provided amino acids and improved the survival of CRC cells under serum deprivation. Downregulating the expression of L-2-hydroxyglutarate dehydrogenase (L2HGDH) and oxoglutarate dehydrogenase (OGDH) increased L2HG levels in CRC, thereby activating mTOR-ATF4 signaling. Furthermore, L2HGDH overexpression reduced L2HG-mediated mTOR-ATF4 signaling under hypoxia, whereas L2HGDH knockdown promoted tumor growth and amino acid metabolism in vivo. Together, these results indicate that L2HG ameliorates nutritional stress by activating the mTOR-ATF4 axis and thus could be a potential therapeutic target for CRC.


INTRODUCTION
Cancer cells exhibit distinct metabolic rewiring compared to normal cells as they have adapted to a harsh microenvironment characterized by hypoxia and a minimal nutrient supply. D/L-2hydroxyglutaric acid (2HG) is a characteristic metabolite of cancer cells that has been shown to directly contribute toward the malignant progression of cancer, and has thus attracted considerable attention as an oncometabolite.
In contrast, it has been reported that the levels of the L-enantiomer of 2HG (L2HG) are increased by changes in the tumor microenvironment, including hypoxia and a low pH [13][14][15][16].
The mechanisms underlying this increase in L2HG levels include an increase in the expression or activity of lactate dehydrogenase A (LDHA) and malate dehydrogenase (MDH), and a decrease in the expression or activity of L2HG dehydrogenase (L2HGDH) [13][14][15][16]. Notably, L2HG is important for adaptation to hypoxic stress and may be involved in redox homeostasis and the activity of the hypoxia-inducible factor-1 pathway [17]. Although L2HG and D2HG have been predicted to have many similar functions, L2HG has higher lysine demethylase (KDM) and ten-eleven translocation (TET) inhibitory activities than D2HG [5] and the underlying molecular mechanisms are not yet fully understood.
Previously, we performed a comprehensive metabolic analysis and identified specific metabolic changes in human colorectal cancer (CRC), including increased levels of S-adenosylmethionine, a donor molecule for DNA and histone methylation, decreased glucose levels [18], and significantly increased 2HG levels in CRC tissues without IDH mutation. However, the roles of 2HG and the mechanisms underlying its accumulation in CRC remain unclear. In this study, we identified a specific increase in L2HG levels in CRC and investigated its biological significance and the underlying molecular mechanisms.

L2HG levels are increased in colorectal tumors
To determine the relative contribution of the two 2HG enantiomers in CRC, we quantified their expression using liquid chromatography-time of flight mass spectrometry (LC-TOFMS). We found that the levels of both 2HG isomers were upregulated in CRC tumor tissues compared to non-neoplastic tissues (Fig. 1A, B). In particular, L2HG levels were very low in non-neoplastic tissues but were markedly increased in CRC tumors (Fig. 1C). L2HG levels were consistently higher than those of D2HG in the eight CRC cell lines tested (Fig. 1D). Furthermore, 13 C-flux analysis using 13 C 1 -glutamine ( 13 C 1 -Gln) revealed that 13 C-labeled L2HG was produced at a much higher rate than 13 C-labeled D2HG in most CRC cell lines (Fig. 1E).
Previous studies have demonstrated that 2HG contributes toward cell survival under serum-free stress conditions [19,20]. Therefore, we examined the effects of the two 2HG enantiomers on survival of CaR1 and HT29 cells. When cells were treated with membrane-permeable octyl esters of L2HG (OC-L2HG) under serum-free stress, cell survival was significantly higher than when treated with membrane-permeable octyl esters of D2HG (OC-D2HG; Fig. 1F; Supplementary Fig. S1A, B). In addition, L2HG suppressed the expression of markers of apoptosis (CL-PARP and CL-caspase 3) and autophagy (LC3B-II; Fig. 1G; Supplementary Fig.  S1C). Together, these findings suggest that L2HG production is upregulated in CRC tumors and contributes toward cell survival.

L2HG induces the expression of ATF4 and its target genes in CRC cells
To elucidate the molecular mechanisms through which L2HG contributes toward cell survival, we performed transcriptome analysis in L2HG-treated CRC cells and using data previously obtained from CRC tissues [18]. A total of 26 and 21 genes were up-and downregulated, respectively ( Fig. 2A). Interaction analysis for transcription factors revealed that the upregulated genes were enriched for activating transcription factor 4 (ATF4) target genes (e.g., ADM2, DDIT4, ASNS, and SLC7A5; Fig. 2B). Furthermore, L2HG increased the expression of ATF4 target genes in HT29 and HCT116 cells in a dose-and time-dependent manner ( Fig. 2C; Supplementary Fig. S2A-C), as well as ATF4 mRNA expression. The levels of ATF4 target genes were also upregulated by L2HG in CaR1 cells ( Supplementary Fig. S2D).
Next, we focused on the mechanism of ATF4 activation by L2HG in CRC cells ( Fig. 2D-H). L2HG increased ATF4 expression in both the whole and nuclear fractions of CRC cells (Fig. 2D, E) and enhanced the activity of the ATF4-responsive reporter (Fig. 2F). Conversely, siRNA-mediated ATF4 knockdown suppressed the induction of ATF4 target genes by L2HG in CRC cells ( Fig. 2G; Supplementary Fig. S2E, F) and abrogated the effect of L2HG on cell survival under serum-free stress (Fig. 2H). Thus, L2HG appears to contribute toward the survival of CRC cells by promoting the expression of ATF4 and its target genes.
When we further examined the ability of 2HG enantiomers to induce ATF4 expression, we found that L2HG induced the expression of ATF4 and its target genes more strongly than D2HG after ATF4 nuclear translocation and ATF4-dependent transcription ( Fig. 2I-K; Supplementary Fig. S2G).
2HG is known to be a competitive inhibitor of αKG-dependent dioxygenases (αKGDs) [17]. We examined whether exogenous αKG could override the effect of L2HG and prevent activation of mTOR/ ATF4. When cells were treated with membrane-permeable octyl esters of αKG (OC-αKG), the inductions of S6K phosphorylation and ATF4 expression by L2HG were downregulated (Supplementary Fig. S3B). L2HG-induced ATF4 target genes expression decreased as well ( Supplementary Fig. 3C). Together, these results suggest that L2HG increases the expression and transcriptional activity of ATF4 via mTOR signaling in dependence of the inhibitory activity of αKGDs.
L2HG mediates ATF4 induction to regulate amino acid metabolism in CRC cells ATF4 is a critical transcription factor for amino acid metabolism that induces the expression of various amino acid transporters and enzymes (e.g., SLC7A5, SLC7A11, ANSN, and PHGDH) and contributes toward metabolic properties, including redox balance, autophagy, energy production, and nucleotide synthesis. To determine whether L2HG affects amino acid metabolism via ATF4 activation, we performed metabolome analysis in CRC cells treated with OC-L2HG (Supplementary Table 1). L2HG significantly altered the levels of 23 metabolites, including many amino acids (Fig. 4A, B). As expected, L2HG increased the levels of Tyr, Leu, Ile, Val, Phe, and Cys, which are amino acids imported through SLC7A5 or SLC7A11 [24], increased the levels of Ser and Gly, which are synthesized by PHGDH and PSAT1 [25], and elevated the ratio of Asn to Asp, which is controlled by ASNS ( Fig. 4B; Supplementary Fig. S4A, B). We also examined the effect of the two 2HG enantiomers on amino acid metabolism and found that L2HG had a larger impact than D2HG, consistent with the observed ATF4 induction (Fig. 4B). In contrast, ATF4 knockdown decreased the levels of amino acids upregulated by L2HG (Fig. 4C). These results suggest that L2HG rewires amino acid metabolism by inducing ATF4.

L2HGDH knockdown increases tumor growth and alters amino acid metabolism in vivo
To investigate the function of L2HG in vivo, we evaluated the effect of L2HGDH knockdown on tumor growth using a mouse xenograft model. Briefly, we subcutaneously injected L2HGDHknockdown CRC cell lines transfected with L2HGDH shRNA#1 (shL2HGDH#1) or #2 (shL2HGDH#2; Supplementary Fig. S6) into nude mice. Sixteen days after injection, L2HG levels were significantly higher in shL2HGDH#2 tumors than in control tumors (shLacZ; Fig. 7A). Notably, the knockdown efficiency of shL2HGDH#2 was higher than that of shL2HGDH#1 in tumors (Fig. 7C), consistent with the increased levels of L2HG (Fig. 7A). shL2HGDH#2 cells significantly increased tumor growth (Fig. 7B) as well as the expression of ATF4 target genes, including amino acid metabolism genes (Fig. 7C), and altered amino acid metabolism compared to the control (shLacZ; Fig. 7D). The levels of Lys and Met were significantly increased in the L2HGDH knockdown (shL2HGDH#2) tumors. Furthermore, we investigated the relationship between 2HG (sum of D2HG and L2HG) and amino acid metabolism in human CRC tumors and adjacent non-neoplastic tissues using metabolomic data presented in one of our previous studies [18] and found that levels of Phe, Gly, Tyr, Val, Ile, Ser, and Leu, regulated by ATF4 target genes such as PHGDH, SLC7A5, and SLC7A11, were significantly correlated with 2HG levels (Fig. 8). In summary, these results suggest that L2HG regulates amino acid metabolism by activating ATF4 to promote tumor growth in vivo.

DISCUSSION
The oncometabolites D2HG and L2HG have directly been implicated in carcinogenesis; however, the underlying molecular mechanisms remain poorly understood. Here, we demonstrated that L2HG levels were significantly increased in human CRC tissues and subsequently upregulated amino acid levels by activating the mTOR-ATF4 pathway, thereby improving cell survival under serum deprivation.
In addition, we found that both L2HGDH and OGDH were downregulated in human CRC tissues, leading to increased L2HG levels and mTOR-ATF4 signaling activation. Hypoxia-induced L2HG production also contributed toward amino acid metabolism via the mTOR-ATF4 axis and L2HGDH knockdown in a mouse xenograft model increased L2HG levels and promoted CRC tumor growth. Regarding L2HG concentrations in vitro and in vivo, L2HG concentrations in tissues and cultured cells cannot be simply compared because tissues include not only cells but also extracellular matrix, and normalization methods differ. L2HG concentrations averaged 54 nmol/g tissue in human colon cancer tissue (Fig. 1A) and 217 nmol/g tissue in L2HGDH knockdown tumors (Fig. 7A). Assuming a specific gravity of 1 for the tumors, these values were 60 μM and 200 μM, respectively, suggesting that the concentrations of L2HG in vitro (250 or 500 μM) were 1.2 to 8.3 times higher than those in vivo. On the other hand, the intracellular D/L-2HG concentration in HT29 cells treated with 500 μM L2HG for 24 h was approximately 6 fmol/cell (Supplementary Table 1), which was higher than the concentration induced by hypoxia (approximately 1 fmol/cell). The treatment concentration of L2HG may not reflect the biological L2HG production level and further investigation is needed. However, when endogenous L2HG production was increased by L2HGDH knockdown or hypoxia (Figs. 5 and 6), ATF4 target gene and amino acids levels increased, suggesting that the increased L2HG levels that can occur in living cells have the same effect of activating ATF4 as L2HG treatment.
We found that downregulation of L2HGDH and OGDH contributes to L2HG production and activates mTOR/ATF4 signaling in CRC cells. However, since OGDH knockdown inhibits the metabolic flux of the tricarboxylic acid cycle, the observed effect should be interpreted with caution. We thus used L2HGDH single knockdown cells for the in-vivo experiments (Fig. 7). Whether the activation of mTOR/ATF4 by OGDH knockdown fully depends on L2HG production requires further investigation.
It has previously been reported that D2HG activates the mTOR signaling pathway [30], consistent with our results. However, we found that L2HG increased the levels of P-S6K (mTORC1), P-4EBP (mTORC1), and P-AKT S473 (mTORC2) much more potently than D2HG in CRC cells under serum deprivation. D/L-2HG has also been shown to inhibit ATP synthase and mTOR signaling, and extend the life span of C. elegans [10]. Under culture conditions with abundant serum, the basal activity of mTOR was high, and further L2HG-induced activation of the mTOR signaling was not clearly observed in CRC cells. The effect of L2HG on mTOR signaling might therefore depend on nutrient conditions. In addition, the effect of L2HG might vary among different cell types, species, or genetic backgrounds.
D2HG has been reported to activate mTOR signaling by inhibiting KDM4, an αKG-dependent enzyme of the Jumonji family of lysine demethylases, which is involved in the protein stability of DEPTOR, a negative regulator of mTORC1/2 [30]. However, we found that DEPTOR expression was not altered by L2HG in CRC cells ( Supplementary Fig. S3A), suggesting that other as a scaling ratio relative to unstimulated cells. D Immunofluorescence staining and quantification of ATF4 in HT29 cells treated with 500 μM OC-L2HG. Data represent the mean ± SD of three independent experiments. Scale bar, 200 μm. E ATF4 protein expression in the whole and nuclear fractions of HT29 and HCT116 cells treated with 500 μM OC-L2HG. F Transcriptional activity of ATF4 in HT29 and HCT116 cells treated with 500 μM OC-L2HG (see Methods). Data represent the mean ± SD of three independent experiments. G Effect of ATF4 knockdown on the mRNA expression of ATF4 target genes in HT29 cells treated with 500 μM OC-L2HG. mRNA levels in each sample were converted to a log2 fold-change relative to the control. Red and blue indicate higher and lower levels, respectively, compared to the control (white). H Effect of ATF4 knockdown on the viability of CaR1 cells treated with L2HG (250 μM) under serum-free conditions for 6 d. Viable cells were counted using a trypan blue exclusion assay. Data represent the mean ± SD of three independent experiments. I, J Expression levels of nuclear ATF4 in HT29 cells treated with 500 μM OC-L2HG or 500 μM OC-D2HG for 24 h. Nuclear ATF4 expression was measured using western blotting (I) and immunofluorescence staining (J). Data represent the mean ± SD of three independent experiments. K Transcriptional activity of ATF4 in HT29 and HCT116 cells treated with OC-L2HG or OC-D2HG. Data represent the mean ± SD of three independent experiments.
pathways might be involved. Notably, L2HG increased the amino acid pool (Supplementary Fig. S4), including several amino acids (Gln, Val, Leu, Met, and Arg) that are known to activate mTOR signaling [31,32], suggesting the existence of a positive feedback mechanism. Further studies are needed to fully determine the relationship between mTOR signaling and L2HG in CRC cells.
ATF4, a member of the CREB/ATF family of bZIP transcription factors, can regulate both pro-survival and pro-apoptotic signaling pathways [33,34] and plays a critical role in various biological processes, including the redox response, amino acid metabolism, autophagy, senescence, and apoptosis. The functions of ATF4 depend on the cell type, nutrient conditions, stressor properties, and the duration of stress. Although previous studies have shown that ATF4 is a crucial mediator of the integrative stress response (ISR) that is controlled by the phosphorylation of the translation factor eIF2α [33,34], emerging evidence has suggested that mTORC1 signaling also activates ATF4 by regulating its mRNA translation and stability [21][22][23]. Importantly, the mTORC1-ATF4 pathway also increases the expression of specific ATF4 target genes, including those involved in amino acid uptake, synthesis, and tRNA charging, compared to the wide range of genes expressed during ISR-induced ATF4 activation [35]. In this study, we found that the L2HG-induced activation of mTOR-ATF4 signaling contributed toward altered amino acid metabolism and survival under serum deprivation in CRC. Moreover, L2HG suppressed autophagy in CRC cells cultured under serum-free conditions for three days ( Fig. 1G; Supplementary Fig. S1C). Since mTORC1 activation and amino acid supply abrogate autophagy [36,37], L2HG could be an essential regulator of autophagy. In addition, a recent study showed that mitochondrial stress can cause neuronal dysfunction through ATF4dependent increases in L2HG levels in Drosophila brains [38]. Together with our findings, these studies suggest the existence of bidirectional feedback between L2HG and ATF4, which might have biological significance in CRC and the nervous system.
In RCC, decreased L2HGDH levels have been reported to lead to L2HG accumulation, contributing toward cancer malignancy [39][40][41]. In addition, the loss of heterozygosity of L2HGDH, located on chromosome 14q, was shown to be correlated with downregulated L2HGDH expression [39,40]. High L2HG levels have also been observed in metastatic RCC tumors, conferring a poor prognosis. Here, we found that ectopic L2HGDH expression reduced intracellular L2HG levels and inhibited CRC tumor growth in vivo. Recently, an L2HG-sensing FRET sensor was developed to evaluate the biological functions of L2HGDH and hypoxia-induced L2HG accumulation [42]. Although the functions of L2HG have been reported in RCC, its roles in other types of cancer remain largely unknown. Our study revealed the pathological significance of L2HG in CRC and proposed that L2HG regulates amino acid metabolism through the mTOR-ATF4 axis.
Furthermore, we found that both L2HG and D2HG were upregulated in human CRC tumor tissues (Fig. 1A), consistent with previous reports that urinary D2HG levels correlate positively with the number of subsequent polyps and dysplasia severity in a mouse model of colitis-associated CRC [43] and that D2HG contributes toward epithelial-mesenchymal transition and tumor metastasis in CRC [44]. However, L2HG levels increased more in tumors than D2HG and L2HG was a more potent mediator of mTOR-ATF4 signaling activation. Thus, L2HG and D2HG may act synergistically as oncometabolites to mediate the malignant phenotype of CRC and could serve as targets for CRC therapy. The CRC cell lines, HT29, HCT116, and SW620, were obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA). LOVO, CaR1, and RCM1 cells were obtained from the Japanese Collection of Research Bioresources (JCRB; Osaka, Japan). The RKO cell line was a gift from Dr. M. Tsujii at Osaka University (Osaka, Japan). All cells were grown in RPMI 1640 medium (Wako Pure Chemical Industries, Ltd., Osaka, Japan) containing 10% (v/v) fetal bovine serum (FBS; Sigma-Aldrich), and antibiotics (100 U/mL penicillin, 100 mg/mL streptomycin, and 0.25 mg/mL amphotericin B; Nakarai Tesque, Kyoto, Japan) at 37°C in a humidified atmosphere with 5% CO 2 .

Animal care
All animal experiments were carried out according to protocols approved by the Animal Care and Use Committee of Aichi Cancer Center Research Institute (Nagoya, Japan). BALB/c nu/nu mice were purchased from CLEA Japan (Tokyo, Japan) and acclimated for at least one week before any experimental procedures. Mice were housed in a specific-pathogen-free facility at room temperature with standard day-night cycles and were provided with commercial laboratory chow (CLEA Rodent Diet CE-2; CLEA Japan, Tokyo, Japan) and autoclaved tap water ad libitum.

Tumor growth assay
Six-week-old female BALB/c nu/nu mice were subcutaneously injected with 1 × 10 6 HT29 cells expressing shRNA against L2HGDH or LacZ under anesthesia (medetomidine hydrochloride (0.75 mg/kg), midazolam (4 mg/kg), butorphanol tartrate (5 mg/kg). The implanted tumors were measured using a caliper on the indicated days and their volumes were calculated using the following formula: V = (L × W 2 ) / 2, where V is the volume (mm 3 ), L is the largest tumor diameter (mm), and W is the smallest tumor diameter (mm).

Human colorectal tumor samples for D-and L-2HG measurements
As described previously [18], we conducted all experiments according to a study protocol approved by the Institutional Ethics Committee of Kagawa University (Heisei 24-040) after obtaining informed consent from all subjects. Tumor and surrounding non-neoplastic tissues were surgically obtained from 47 colorectal cancer patients as described previously [18] (Supplementary Table 2).

Analysis of D-and L-2HG
LC-TOFMS was performed using an Agilent 1200 series HPLC system (Agilent Technologies, Palo Alto, CA, USA) and an Agilent 6220 TOFMS. D-and L-2HG were separated using an Astec CHIROBIOTIC R column (2.1 mm i.d. × 250 mm, 5 µm; Supelco, Bellefonte, PA, USA) that was maintained at 20°C. Isocratic elution was performed using 75% ethanol/ methanol (3:1, v/v) and 25% water containing 0.1% trimethylamine acetate (pH 4.5) as the mobile phase at a flow rate of 0.1 mL/min. The sample injection volume was 1 µL. TOFMS was conducted in negative ion mode with the capillary voltage set at 3.5 kV. The flow rate of heated dry nitrogen gas (heater temperature, 300°C) was maintained at 12 L/min and the nebulizer gas was set at 20 psi. The fragmentor, skimmer, and Oct RFV voltages were set at 100 V, 50 V, and 200 V, respectively. Each acquired spectrum was automatically recalibrated using an Agilent G1969-85001 API-TOF reference mass solution kit (m/z 112.9856 and m/z 1033.9881). Quantification was performed by comparing the D-and L-2HG peak areas to the calibration curve generated using internal standardization techniques with 2-morpholinoethanesulfonate.

CE-TOFMS metabolite quantification
Intracellular metabolites were detected and quantified using CE-TOFMS (Agilent Technologies) as described previously [45][46][47]. Metabolite Fig. 3 L2HG increases ATF4 expression via mTOR signaling in CRC cells. A Representative western blot analyses of signaling components in the mTOR pathway. HT29 cells were treated with 500 μM OC-L2HG for 24 h under serum-free conditions (see Methods). B Effect of rapamycin (mTORC1 inhibitor) and Torin1 (mTORC1 and mTORC2 inhibitor) on the phosphorylation of S6K, AKT, and 4EBP1 in HT29 cells. Cells were treated with 500 μM OC-L2HG, 100 nM rapamycin, and/or 250 nM Torin1 for 24 h under serum-free conditions. C Differences between 2HG enantiomers in the phosphorylation of S6K, AKT, and 4EBP1 in HT29 cells. D-F Effect of rapamycin and Torin1 on the mRNA expression of ATF4 target genes (D), the expression of nuclear ATF4 (E), and the transcriptional activity of ATF4 (F). qPCR data are the average of three technical replicates and are representative of at least two independent experiments with similar results. G Effect of rapamycin and Torin1 on the viability of CaR1 cells treated with L2HG (250 μM) under serum-free conditions for 6 d. Viable cells were counted using a trypan blue exclusion assay. Data represent the mean ± SD of three independent experiments. identities were assigned by matching their m/z values and migration times to standard compounds.

Microarray analysis
Microarray analysis was conducted as described previously [48]. Total RNA was isolated from HT29 and HCT116 cells treated with or without 500 μM OC-L2HG for 24 h using the RNeasy Mini Kit (Qiagen, Venlo, Netherlands). RNA quality was assessed using an Agilent 2100 Bioanalyzer (Agilent Technologies). cRNA amplified from 100 ng total RNA was labeled using a Low Input Quick Amp Labeling Kit, One-Color (Agilent Technologies), hybridized to a SurePrint G3 Human GE 8x60K v2 microarray (Agilent Technologies) and then scanned using an Agilent scanner according to the manufacturer's instructions. Relative  hybridization intensity and background hybridization values were calculated using Agilent Feature Extraction Software (Agilent Technologies). Microarray data were analyzed using GeneSpring software (Agilent Technologies).

Hypoxic stress
To expose cells to hypoxic stress, they were cultured for 24

Real-time PCR analysis
Real-time PCR analysis was conducted as described previously [50]. Briefly, RNA was isolated from cells using TRIzol reagent (Thermo Fisher Scientific) according to the manufacturer's protocol and 2 μg of RNA was used for reverse transcription with a First Strand cDNA Synthesis kit (ReverTra Aceα; Toyobo Co., Osaka, Japan). Quantitative real-time PCR was carried out on a StepOne Plus Real-Time PCR system (Applied Biosystems, Foster City, CA, USA) using SYBR premix Ex Taq (Takara, Shiga, Japan) according to the manufacturer's instructions. The ΔΔCq method was used to quantify gene expression, using RPL27 expression as an internal reference [51]. All experiments were performed in triplicate. The primers used for real-time PCR are described in Supplementary Table 3. siRNA transfection siRNA duplexes for ATF4, L2HGDH, OGDH, and a negative control were purchased from Sigma-Aldrich. Cells were seeded in 6-well plates overnight and then transfected with 100 pmol of siRNA oligomer mixed with Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific) in serumreduced Opti-MEM (Thermo Fisher Scientific) according to the manufacturer's instructions. After 4-5 h, complete culture medium was added to each well and cells were incubated at 37°C in a CO 2 incubator for another 24 h.

Cell viability assay
Cell viability was measured using the 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide (MTT, Sigma-Aldrich) assay or trypan blue exclusion assay. MTT assays were performed as described previously [52]. Briefly, cells (5 × 10 3 cells/well) were seeded in each well of a 96-well plate and incubated for 24 h. Cells were treated with OC-D2HG or OC-L2HG under serum-free conditions for 2, 4, 6, and 8 d. Next, 50 μL MTT (2 mg/mL in PBS) was added to each well and plates were incubated for a further 2 h. After aspiration of the culture medium, the resulting formazan crystals were dissolved in 100 μL dimethyl sulfoxide (DMSO) and plates were placed on a plate shaker for 1 min before being read immediately at 570 nm using a TECAN micro-plate reader with Magellan software (Tecan Group Ltd.). For the trypan blue exclusion assay, 2 × 10 5 cells/well were seeded in 12-well cell culture plates and incubated at 37°C. After treatment with reagents under serum-free conditions, cells were disaggregated in 500 μL medium and 10 μL of the suspension was mixed with 10 μL trypan blue (Thermo Fisher Scientific). Viable cells were counted using a Countess Automated Cell Counter (Thermo Fisher Scientific).

Western blotting
Immunoblot analysis was performed as described previously [48]
For DOX-inducible expression vectors, L2HGDH cDNA with a BamHI restriction site at the 5′ end and an EcoRI restriction site at the 3′ end was synthesized by Eurofins Japan (Tokyo, Japan) and subcloned into the pRetroX-TetOne vector (TaKaRa) using the BamHI and EcoRI restriction sites. Retroviral vectors were generated and used according to the manufacturer's instructions.

Analysis of correlations between amino acids and 2HG levels in human CRC tumors and adjacent non-neoplastic tissues
We reanalyzed the levels of amino acids and 2HG in paired tumors and adjacent non-neoplastic tissues from 275 CRC patients using Fig. 6 Ectopic L2HGDH expression suppresses L2HG levels and induces ATF4 under hypoxic conditions. A Expression of doxycycline (DOX)-inducible V5-tagged L2HGDH in HT29 cells. Cells were treated with 1 μg/mL DOX and hypoxia (1% O 2 ) for 24 h. B, C Effect of L2HGDH overexpression on the levels of L2HG (B) and nuclear ATF4 (C) in HT29 cells under hypoxic conditions. Data represent the mean ± SD of three independent experiments. D Effect of L2HGDH overexpression on the mRNA expression of ATF4 and ATF4 target genes in HT29 cells under hypoxic conditions. Data are the average of three technical replicates and are representative of at least two independent experiments with similar results. E Effect of Torin1 (250 nM) on the hypoxic stress-induced expression of nuclear ATF4 in HT29 cells. F Effect of L2HGDH overexpression on amino acid levels in HT29 cells under hypoxic conditions. Data represent the mean ± SD of four independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001.
our previously reported metabolomic data [18]. The correlations between amino acids and 2HG levels were analyzed using Spearman's correlation.
(GraphPad Software, Inc., La Jolla, CA, USA). For in vitro experiments, data from two or more groups were analyzed using Student's t-tests and oneway analysis of variance (ANOVA), respectively. For in vivo experiments, data from two or more groups were analyzed using Mann-Whitney U and Kruskal-Wallis tests, respectively. The correlations between amino acids and 2HG levels in Fig. 8 were analyzed using Spearman's correlation. Data are represented as mean ± SEM or ±SD; P values < 0.05 were considered statistically significant.

Fig. 8
Amino acids and 2HG levels in human CRC tumors. Correlation between amino acids (Phe, Gly, Tyr, Val, Ile, Ser, and Leu) and 2HG levels in paired tumors and adjacent non-neoplastic tissues from 275 CRC patients. The metabolites levels were measured using capillary electrophoresis time-of-flight mass spectrometry.