IDH1 Arg-132 mutant promotes tumor formation through down-regulating p53

Resistance to apoptosis and uncontrolled proliferation are two hallmarks of cancer cells. p53 is crucial for apoptosis triggered by a broad range of stresses and a well-known gatekeeper for neoplastic transformation. Here we show that oncogenic IDH1 R132H/R132Q mutants robustly inhibit p53 expression and such an effect is attributed to 2-HG production. Mechanistically, 2-hydroxyglutarate (2-HG) stabilizes hypoxia-inducible factor-2α, which in turn activates the expression of miR-380-5p, a characterized microRNA against p53 expression. Rescue expression of p53 can inhibit the proliferation rate and impair the resistance of apoptosis induced by doxorubicin in IDH1 R132Q mouse embryonic fibroblast cells. Furthermore, p53 protein levels correlates negatively with IDH1 R132H levels in human glioma samples. Our results thus shed a new light on how p53 is down-regulated by 2-HG and suggests that impairment of p53-mediated apoptosis contributes to the tumorigenesis driven by IDH1 mutants.

Resistance to apoptosis and uncontrolled proliferation are two hallmarks of cancer cells. p53 is crucial for apoptosis triggered by a broad range of stresses and a well-known gatekeeper for neoplastic transformation. Here we show that oncogenic IDH1 R132H/R132Q mutants robustly inhibit p53 expression and such an effect is attributed to 2-HG production. Mechanistically, 2-hydroxyglutarate (2-HG) stabilizes hypoxia-inducible factor-2␣, which in turn activates the expression of miR-380-5p, a characterized microRNA against p53 expression. Rescue expression of p53 can inhibit the proliferation rate and impair the resistance of apoptosis induced by doxorubicin in IDH1 R132Q mouse embryonic fibroblast cells. Furthermore, p53 protein levels correlates negatively with IDH1 R132H levels in human glioma samples. Our results thus shed a new light on how p53 is down-regulated by 2-HG and suggests that impairment of p53-mediated apoptosis contributes to the tumorigenesis driven by IDH1 mutants.
2-HG and ␣-KG are structurally similar except that the hydroxyl group in 2-HG is replaced by the C2 carbonyl group in ␣-KG (17,18). Accumulating lines of evidence ascribe the carcinogenicity of 2-HG to its competitive inhibition of dioxygenases with ␣-KG as a co-substrate due to their structural similarity. Elevated levels of 2-HG inhibits the methylcytosine dioxygenase TET2, leading to a hypermethylator phenotype in cells harboring various IDH1/2 mutations (16, 18, 20 -22). In addition, ␣-KG-dependent histone demethylases are also inhibited by 2-HG (18,23), which in turn results in hypermethylation of histone and the disruption of cell differentiation (23). Furthermore, several groups have reported that 2-HG could stabilize hypoxia-inducible factor-1␣ (HIF-1␣) by inhibiting HIF prolyl hydroxylase, which is responsible for HIF-1␣ hydroxylation, a process required for subsequent ubiquitination and degradation of HIF-1␣ via proteosome pathway (18,24).
Tumorigenesis is widely accepted as a multistep process resulting from abnormal activation of oncogenes and inactivation of tumor suppressor genes (25). p53 tumor suppressor is recognized as a gatekeeper for neoplastic transformation due to its critical role in triggering apoptotic cell death, cell cycle arrest, and senescence in response to diverse stressor including DNA damage, nutrient deprivation, and inappropriate mitogenic stimulation (26,27). The notion that p53 function has to be disrupted for tumor progression is supported by previous studies showing that restoring p53 function is sufficient to cause regression of several types of tumors in mice (28,29). The importance of p53 in preventing tumor initiation is also indicated by the presence of somatic mutations of p53 in ϳ50% of all human cancers (30). We questioned whether p53 inactivation is also involved in tumorigenesis caused by IDH1 mutations.
In this study, we report that IDH1 mutations robustly inhibit p53 expression in mouse embryonic fibroblasts (MEF) and other cell types. Such inhibition results from 2-HG-mediated inhibition of prolyl hydroxylase and subsequent stabilization of HIF-2␣. Increased HIF-2␣ transactivates the expression of miR-380-5p, which in turn down-regulates the p53 protein level. Consistently, p53 protein levels were decreased in human glioma samples with the IDH1 R132H mutation, implying that 2-HG-caused p53 deficiency may be a key component in tumorigenesis driven by IDH1 mutations.

Oncogenic IDH1 Arg-132 mutant robustly down-regulates p53
To find out whether the IDH1 mutation shows any inhibitory effect on p53, MEF cells with genotypes IDH1 WT/WT , IDH1 WT/LSL , and IDH1 LSL/LSL were isolated from the embryos of conditional IDH1 R132Q knock-in mice (22,31,32), followed by excision of lox-stop-lox (LSL) cassette with Cre recombinase to generate cell lines with five different genotypes, IDH1 WT/WT , IDH1 WT/LSL , IDH1 WT/Mut , IDH1 LSL/LSL , and IDH1 Mut/Mut (WT:WT; Mut:R132Q mutant). The genotypes and IDH1 protein levels of these cell lines were validated by polymerase chain reaction (PCR) and Western blotting (Fig. 1,  A and B). Next we determined p53 expression in these cell lines. As shown in Fig. 1B, p53 expression was dramatically suppressed in IDH1 WT/Mut and IDH1 Mut/Mut MEFs, but not altered in IDH1 WT/LSL and IDH1 LSL/LSL MEFs with reduced or without WT IDH1 expression indicating that mutant IDH1 rather than WT IDH1 was responsible for the down-regulation of p53 expression. Interestingly, the IDH1 R132Q mutant could also significantly suppress p53 accumulation induced by doxorubicin (DOX) (Fig. 1C). Consistently, the expression of p21, one of p53 target genes, was down-regulated in IDH1 mutant cells to the same extent as p53 (Fig. 1C). To test if the IDH1 Arg-132 mutant also suppresses p53 expression in human cancer cells, we expressed WT IDH1 or its R132H mutant in HCT116 cells (Fig. 1D) and U2OS cells (Fig. 1E) and observed the same results as in MEFs. Taken together, the oncogenic IDH1 Arg-132 mutant is capable of down-regulating p53 dramatically.

p53 down-regulation by IDH1 R132H/R132Q depends on 2-HG production
One of the most prominent feature of various IDH1 Arg-132 mutants is to produce extremely high levels of intracellular 2-HG, which disturbs a wide spectrum of biochemical reactions and thus leads to disorder of a broad range of cell biological functions (14,15,34). We spontaneously wanted to know whether IDH1 mutants induced down-regulation of p53 is the result of increased concentrations of 2-HG. First, we detected 2-HG by employing the LC-MS technique and found that dramatically high levels of 2-HG were produced in IDH1 WT/Mut MEFs relative to IDH1 WT/WT MEFs ( Fig. 2A). It is important to point out that we used heterozygous, but not homozygous

2-HG stimulates HIF-2␣ to suppress p53
IDH1 R132Q knock-in (KI) MEF cells in all of the following experiments, because of the observation that all somatic gainof-function mutations of IDH1 have been identified exclusively in one allele and heterozygous IDH1 mutation displayed the similar inhibitory effect on p53 as that of homozygous. Then we examined whether exogenous 2-HG could down-regulate p53. Treatment of U2OS cells with cell-permeable trifluoromethylbenzyl-(R)-2-HG (TFMB-2-HG) effectively increased the intracellular 2-HG level (Fig. S1A) and decreased the expression of p53 and p21 in a dose-dependent manner (Fig. 2B). Similar results were observed in the HCT116 cell line (Fig. 2C). To further bolster this conclusion, we treated HCT116 cells with another cell-permeable octyl-2-HG, whose permeability was confirmed by LC-MS (Fig. S1B). As expected, octyl-2-HG treatment markedly suppressed p53 expression in HCT116 cells even under exposure to 2.5 M DOX (Fig. 2D). In addition, AGI-5198, a powerful and selective inhibitor of IDH1 R132H for 2-HG production (35), could efficiently suppress 2-HG production (Fig. S1C) and restored p53 expression in HCT116 cells harboring the IDH1 R132H mutant (Fig. 2E). These data demonstrate that 2-HG inhibits p53 expression in a broad range of mammalian cell types.

2-HG down-regulates p53 at mRNA level
To dissect the mechanism underlying the p53 down-regulation mediated by the IDH1 R132 mutation, we examined whether this regulation occurs at mRNA or protein levels. We compared p53 mRNA levels in IDH1 WT/WT and IDH1 WT/Mut mice livers and observed that IDH1 R132Q KI mice livers had much less p53 mRNA than that in control mice livers (Fig. 3A). We also checked p53 mRNA levels in IDH1 WT/WT and IDH1 WT/Mut MEFs, and obtained exactly the same result (Fig.  3B). Furthermore, cell permeable octyl-2-HG treatment dramatically decreased the p53 mRNA levels in HCT116 cells (Fig.  3C). To our surprise, proteosome inhibitor MG132 and calpain inhibitor ALLN, but not lysosome inhibitors NH 4 Cl and chloroquine significantly blocked the inhibitory effect of 2-HG on p53 and restored its expression in HCT116 cells treated with octyl-2-HG (Fig. S2A) and MEF cells expressing the IDH1 R132Q mutant (Fig. S2B). One possibility is that 2-HG also down-regulated p53 by promoting its degradation in addition to the suppression of mRNA level. To validate this hypothesis, HCT116 cells were treated with cycloheximide (CHX), followed by detection of p53 protein levels and determination of p53 half-life. As shown in Fig. 3, D and E, both IDH1 R132H  ). B and C, TFMB-2-HG inhibits p53 expression in U2OS cells and HCT116 cells. U2OS (B) cells and HCT116 cells (C) were treated with the indicated amounts of TFMB-2-HG for 9 h, followed by Western blotting to detect p53 protein levels (left). Quantitation of signal intensities of Western blot bands of p53 (right) was performed by using ImageJ software. p53 levels were normalized to ␤-actin levels. Data are presented as the mean Ϯ S.D. of three independent experiments (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001, unpaired Student's t test). D, octyl-2-HG inhibit p53 accumulation induced by DOX in HCT116 cells. HCT116 cells were treated with DOX alone or in combination with octyl-2-HG (5 mM) for 8 h with octyl acetate as a control. Cell lysates were immunoblotted for p53 and p21. E, AGI-5198 restored p53 expression in HCT116 cells harboring IDH1 R132H mutant. HCT116 cells were infected with control lentivirus or lentivirus expressing FLAG-WT IDH1 or IDH1 R132H. At 48 h post-infection, cells were treated with or without 1.5 M AGI-5198 for 2 days, followed by treating with 2.5 M DOX (or not) for another 9 h in the presence of AGI-5198, as indicated.

2-HG stimulates HIF-2␣ to suppress p53
expression and 2-HG treatment failed to influence the half-life of p53. These results indicate that down-regulation of p53 expression by the IDH1 mutation appears to occur mainly at the mRNA level. A possible explanation to the observation that the proteosome inhibitor MG132 can partially antagonize 2-HG-induced p53 down-regulation is that 2-HG may also function to stimulate proteosome-mediated degradation of some positively transcriptional regulator of p53.

IDH1 Arg-132 mutations down-regulate p53 via promoting miR-380-5p expression
Next we asked whether 2-HG can diminish p53 transcription or p53 mRNA stability. To address this question, we constructed two chimeric luciferase reports containing p53-promoter and p53-3Ј UTR because the 3Ј UTR usually plays a very important role in the regulation of mRNA stability (36). 2-HG significantly suppressed the reporter activity of p53-3Ј UTRluc, but not p53-promoter-luc, implying that p53-3Ј UTR may be required for 2-HG down-regulation of p53 mRNA (Fig. 4, A  and B). miRNAs are 20 -25-nucleotide short RNA molecules that function to down-regulate gene expression by targeting almost the 3Ј UTR of mRNAs (37). To determine whether some miRNAs are involved in 2-HG caused p53 down-regulation, we utilized two independent short hairpin RNAs (shRNAs) to knockdown the expression of Dicer, one of the critical proteins in the maturation of miRNA (37). Loss of Dicer caused great up-regulation of p53 protein levels in HCT116 cells expressing mutated IDH1 (Fig. 4C), suggesting that p53 expression could be tightly regulated by miRNAs.
To find out which miRNA is responsible for the down-regulation of p53 by 2-HG, we performed a sequencing-based RNA profiling analysis using liver samples from IDH1 WT/WT and IDH1 WT/Mut mice. Several reports have revealed that a series of miRNAs, including miR-125b, miR-504, miR-25, miR-30d, miR-1285, and miR-380-5p directly target the 3Ј UTR of p53 mRNA to down-regulate p53 protein levels (37)(38)(39)(40)(41). We thus analyzed the expression of these miRNAs in our sequencing results and found that expression of miR-380-5p was significantly increased in IDH1 mutant mice liver (Fig. S3A). This observation was confirmed by qRT-PCR data (Fig. 4D). Consis- . C, octyl-2-HG down-regulates p53 mRNA level in HCT116 cells. HCT116 cells were treated with the indicated amounts of octyl-2-HG for 9 h, followed by qRT-PCR analysis of relative p53 mRNA levels. Data are presented as mean Ϯ S.D. of three independent experiments (**, p Ͻ 0.01, unpaired Student's t test). D, IDH1 Arg-132 mutant does not affect the turnover rate of p53 protein. HCT116 cells were transfected with the same amount of blank vector, FLAG-tagged IDH1, or IDH1 R132H. The levels of p53 at different time points after CHX treatment were determined by immunoblotting the total cell lysates (left panel) and quantified using ImageJ software (Bio-Rad) with ␤-actin as a loading control. Results plotted (right panel) are the amounts of p53 relative to that at time 0. Mean Ϯ S.D., n ϭ 3 independent experiments, are shown. E, 2-HG had no effect on the turnover rate of p53. HCT116 cells were treated with CHX alone or in combination with octyl-2-HG. Results were presented as in D.

p53 down-regulation is involved in IDH1 mutant-driven tumorigenesis
Because p53 plays a crucial role in triggering apoptotic cell death and cell cycle arrest in response to diverse stressors, and our data indicate that IDH1 R132H/R132Q mutations could dramatically suppress p53 expression, we examined whether such inhibition benefits cell proliferation and endow cells an ability to escape apoptosis induced by DNA damage agent. Rescue expression of p53 significantly attenuated the proliferation rate of IDH1 WT/Mut cells but not IDH1 WT/WT cells (Fig. 7A). Consistently, knockdown HIF-2␣ (Fig. S6A) or inhibition of

2-HG stimulates HIF-2␣ to suppress p53
endogenous miR-380-5p by anti-miR-380-5p oligonucleotides (Fig. S6B) in IDH1 mutant cells restored p53 expression and markedly suppressed the proliferation rate. On the contrary, 2-HG treatment (Fig. S6C) or knockdown of p53 (Fig. 7B) increased the proliferation rate of IDH1 WT/WT cells but not IDH1 WT/Mut cells. These findings suggest that inhibition of p53 by 2-HG confer IDH1-mutated cells a higher proliferating rate. Next we exposed IDH1 WT/WT and IDH1 WT/Mut MEFs to a lethal dose of DOX for 16 h and assessed their viability by Annexin V staining and flow cytometry. As shown in Fig. 7C, IDH1 WT/Mut MEFs displayed substantial resistance to DOX-induced apoptosis compared with IDH1 WT/WT cells. However, rescue expression of p53 rendered IDH1 WT/Mut MEFs sensitive to DOX again. Consistent with this result, knockdown of HIF-2␣ (Fig.  S6D) or anti-miR-380-5p oligonucleotides treatment (Fig. S6E) efficiently eliminated the resistance of IDH1 WT/Mut MEFs to DOX-induced apoptosis. In addition, 2-HG treatment (Fig.  S6F) or knockdown of p53 (Fig. 7D) desensitized IDH1 WT/WT cells but not IDH1 WT/Mut cells to DOX-induced apoptosis. Moreover, the p53 protein level was dramatically decreased in low grade glioma samples with the IDH1 R132H mutation compared with that expressing WT IDH1 (Fig. 7E), establishing a solid correlation between IDH1 R132H mutation and p53 down-regulation. We further determined the p53 genotype of human glioma samples by Sanger sequencing. As shown in Fig.  S6G, samples 1, 3, 7, 8, 9, 10, 11, and 12 harbor p53 mutations. Thus, IDH1 R132 mutations down-regulate p53 expression, despite the status of p53. It is important to point out that all presently available evidence indicates that IDH1 mutations are co-present with the p53 mutation and occur before the acquisition of p53 mutations (43,44). These evidences together with our data suggest that IDH1 R132 mutations induced downregulation of WT p53 may play an important role in the early stage of gliomagenesis. Collectively, IDH1 R132 mutations can remarkably stimulate cell proliferation and protect cells from DNA damage-induced apoptosis by suppressing p53 expression, thereby favoring tumor formation.
In conclusion, our data demonstrates that the high level of 2-HG produced by cancer-associated IDH mutations can stabilize HIF-2␣ and consequently stimulate miR-380-5p expres-  t test). B, knockdown of HIF-2␣ down-regulates miR-380-5p. HCT116 cells stably expressing WT IDH1 or its R132H mutant were infected with lentiviruses expressing two independent HIF-2␣ shRNAs or control shRNA, individually. 48 h after infection, the expression of miR-380-5p was analyzed and presented as in A (***, p Ͻ 0.001). C, knockdown of HIF-1␣ has no effect on miR-380-5p up-regulation caused by the IDH1 Arg-132 mutant. HCT116 cells stably expressing WT IDH1 or its R132H mutant were infected with lentiviruses expressing two independent HIF-1␣ shRNAs or control shRNA, individually. 48 h after infection, the expression of miR-380-5p was analyzed and presented as in A (N.S. ϭ not significant). D, HIF-2␣ binds to the promoter regions of miR-380-5p. HCT116 cells were transfected with HA-tagged HIF-1␣ or HIF-2␣. 36 h after transfection, ChIP assays were performed using IgG (control) or anti-HA antibodies. Proteins in precipitates and total cell lysates were determined by Western blot analysis (upper panel). Purified DNA was analyzed by standard PCR using primers targeting the indicated promoter regions (lower panel). Actin and p53 were used as negative controls.

2-HG stimulates HIF-2␣ to suppress p53
sion, which in turn down-regulates p53. Impaired p53 expression confer cells a higher proliferation rate and resistance to apoptosis, which contributes to the oncogenicity of the IDH1 mutations.

Discussion
It is well-accepted that aberrant genes' expression caused by epigenetic alterations and HIF-1␣ accumulation in IDH1-mutated cells contribute to tumor progression. 2-HG is believed to competitively inhibit ␣-KG-dependent dioxygenases, including prolyl hydroxylase 2, which disrupts HIF-1␣ hydroxylation and leads to aberrant accumulation of HIF-1␣ (18,24). Little is known about whether the HIF-2␣ protein level is also regulated by 2-HG and if so what role does HIF-2␣ play in tumorigenesis stimulated by 2-HG? We initially observed that the HIF-2␣ protein level was dramatically increased in cells harboring IDH1 mutations, similar to the case of HIF-1␣. Next we tried to find out whether HIF-2␣ plays any role different from that HIF-1␣ does in tumorigenesis.
The p53 tumor suppressor has long been recognized as the gatekeeper for tumor formation. The notion that p53 function has to be disrupted for the progression of some tumors is wellaccepted. Indeed, we observed that p53 protein levels were markedly decreased either in IDH1 WT/Mut MEFs or HCT116 cells expressing IDH1 R132H or treated with permeable 2-HG. Moreover, IDH1 WT/Mut MEFs took great advantage of suppression of p53 expression in proliferation and resistance to DOXinduced apoptosis. Importantly, the p53 protein level was abolished in low grade glioma samples with the IDH1 R132H mutation. Our data thus provided convincing evidence that p53 down-regulation is a key event in the IDH1 mutant that caused tumor formation. Our further observations indicated that down-regulation of p53 was at the mRNA level, and HIF-2␣, but not HIF-1␣, played a predominant a role in such regulation,

2-HG stimulates HIF-2␣ to suppress p53
raising an interesting question how was the p53 mRNA level reduced by HIF-2␣.
In ϳ50% of all human cancers, inactivation of p53 function is acquired by the presence of its somatic mutations (30). In addition, p53 is also inactivated through proteasome-mediated degradation caused by amplification of MDM2 or its homolog MDM4 in 10ϳ20% of total human cancer (45). In recent years, accumulating evidence has established that various miRNAs, including miR-125b, miR-504, miR-25, miR-30d, miR-1285, and miR-380-5p are involved in down-regulating p53 protein levels (37)(38)(39)(40)(41). These studies remind us that regulation of miRNAs against p53 by HIF-2␣ might be an intermediate event in 2-HG-induced down-regulation of p53. As expected, the expression of miR-380-5p was robustly activated by HIF-2␣ and required for down-regulation of p53 in IDH1 mutation cells. In summary, a high level of 2-HG produced by the IDH1 mutation stabilizes HIF-2␣, which further activates transcription of miR-380-5p. Elevated miR-380-5p levels effectively mediate degradation of the p53 mRNA and eventually benefits cell proliferation and tumor formation. In summary, we find a novel mechanism underlying the tumorigenesis driven by 2-HG-producing IDH1 mutations. Our observation that HIF-2␣ rather than HIF-1␣ is involved in the suppression of p53 will help us to further understand the delicate difference of HIF-2␣ and HIF-1␣ in promoting tumor formation.

Cell culture and transfections
HEK293T cells, HCT116 cells, U2OS cells, and MEF were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 4 mM L-glutamine at 37°C in a humidified incubator containing 5% CO 2 . HEK293T cells were transfected using polyethylenimine (catalog number 23966, Polysciences, Inc.) at a final concentration of 10 M. Total DNA for each plate was equalized using relevant empty vectors. Transfected cells were harvested at 24 h post-transfection. Lentivirus for infection of MEF and HCT116 cells was packaged in HEK293T cells using Turbofect (catalog number R0532, Thermo Scientific) transfection reagent. At 36 h posttransfection, virus-containing culture supernatant was collected by centrifugation (13,000 ϫ g, 5 min). After titration, an appropriate volume of the virus-containing supernatant was added to cells (1 ϫ 10 5 ) in the presence of Polybrene (Sigma) at a final concentration of 10 ng/l.

Constructs
WT IDH1 or its R132H mutant were cloned into BamHI and SmaI sites of the modified lentivirus vector FLAG-tagged pLV using the Exo III-assisted ligase-independent cloning method (33). For luciferase reporter assays, the human p53 promoter and p53 3Ј UTR were cloned into the pGL2-Basic vector and pmir GLO vector, respectively (Promega). The primer sequences used for amplifying the p53 promoter were: 5Ј-GCGTGCTAGCTCGAGTCGGCGAGAATCCTG-3Ј and 5Ј-CGGAATGCCAAGCTTTCTAGACTTTTGAGAAG-3Ј. The primer sequences used for amplifying the p53 3Ј UTR were: 5Ј-TGTTTAAACGAGCTCCATTCTCCACTTCTTGTTCC-3Ј and 5Ј-GACTCTAGACTCGAGGATGTTGACCCTTCC-AGCTGG-3Ј. p53-3Ј UTR-luc constructs with either one or both of two putative miR-380-5p-binding sites deleted were generated using a PCR-based, site-directed mutagenesis method employing Pfu polymerase. All plasmids were verified by DNA sequencing (sequences available upon request).

Immunoprecipitation and Western blotting
Cells were harvested in lysis buffer (20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerolphosphate, 1 mM sodium orthovanadate, 1 g/ml of leupeptin, 1 mM phenylmethylsulfonyl fluoride). For immunoprecipitations, lysates were incubated with antibodies indicated at 4°C for 3 h. Immunoprecipitates were washed three times in lysis buffer and boiled in SDS-PAGE loading buffer. Proteins in total cell lysates or immunoprecipitates were fractionated by SDS-PAGE and transferred to PVDF membranes. Blots were blocked in 5% nonfat milk or BSA and incubated with the appropriate antibodies.

Luciferase reporter assay
HCT116 cells transfected with chimeric luciferase reporter plasmids were washed with PBS and lysed by lysis buffer. The supernatants were collected by centrifugation (13,000 ϫ g, 5 min) and subjected to dual luciferase assays (Promega) by following the manufacturer's instructions. The transfection efficiency was normalized by co-expression of Renilla luciferase.

2-HG stimulates HIF-2␣ to suppress p53
Flow cytometric cell death assay Cells cultured in 6-well plates were left untreated or exposed to DOX for the times indicated in the figures. After treatment, suspended and trypsinized cells were centrifuged at 800 ϫ g for 5 min, washed once with PBS, and stained with FITC-conjugated Annexin V for 10 min at 37°C in the dark. Percentages of apoptotic cells were quantified by a fluorescence-activated cell using a flow cytometer.

Chromatin IP assay
The experiments were performed following the standard protocol of the SimpleChIP Enzymatic Chromation IP Kit (Cell Signaling, number 9002). Briefly, HCT116 cells transfected with HA-HIF-1␣ or HA-HIF-2␣ cultured in a 15-cm plate were cross-linked by 37% formaldehyde for 10 min at room temperature, and then glycine was added to stop the reaction. Cells were lysed, followed by nuclei preparation. 0.5 l of micrococcal nuclease were added per each sample then incubated for 20 min at 37°C with frequent mixing to digest DNA to ϳ150 -900 bp. After the nuclear membrane was broken by sonication, the same amount of chromatin was used for immunoprecipitation with anti-HA antibody or anti-IgG antibody (control). After the immunoprecipitation, chromatin was eluted from agarose beads with ChIP Elution Buffer and cross-links were reversed by incubating with NaCl and Proteinase K. DNA was purified using spin columns and quantification of DNA was determined by PCR or real-time quantitative PCR with primers targeting p53, actin, and the miR-380-5p TSS region. Primer sequences were: miR-380-5p forward: 5Ј-GTCAGTCATAGCACTAGT-TCC-3Ј, miR-380-5p reverse: 5Ј-CTGAGGCCTGATGTAGT-ATTG-3Ј; p53 forward: 5Ј-CCTGACTCTGCACCCTCCTC-3Ј, p53 reverse: 5Ј-CGAGGCTCCTGGCACAAAGC-3Ј; actin forward: 5Ј-GAGCACAGAGCCTCGCCTTT-3Ј and actin reverse: 5Ј-AGACAAAGACCCCGCCGGTT-3Ј.

2-HG measurement by LC-MS
LC-MS analysis of 2-HG levels was performed as described (14). Briefly, cells were cultured to ϳ80% confluence, washed in PBS, quenched in 1 ml of 80:20 methanol:water at Ϫ80°C, and detached from the culture dish using a cell scraper. Quenched cells were centrifuged at 12,000 ϫ g for 15 min at 4°C, and 0.8 ml of supernatant was dried under nitrogen gas and dissolved in 100 ml of aqueous LC buffer. This mixture was centrifuged at 12,000 ϫ g for 15 min and analyzed by LC-MS within 24 h. Total cell numbers and protein levels used for subsequent normalization of LC-MS signal intensities were determined from equivalently treated control plates. Sample separation and analysis were performed on a 50 ϫ 2 mm, 4-mm Synergi Hydro-RP 80 A column, using a gradient of buffer A (10 mM tributylamine, 15 mM acetic acid, 3% (v/v) methanol, in water) and buffer B (methanol) using multiple reaction monitoring transitions. 2-HG levels were quantified by comparing peak areas with pure metabolite standards at known concentration.

Animal experiments and patient samples
All animal experimental protocols were approved by the Institutional Animal Care and Use Committee at Xiamen University. Glioma samples were obtained with the approval of the research ethics boards of Xiamen University and Huanhu Hospital and in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all patients.

Statistics
Two-tailed Student's t test was used to compare differences between treated and control groups. Differences with p values Ͻ0.05 were considered statistically significant: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.