Targeting IDH1 as a Prosenescent Therapy in High-grade Serous Ovarian Cancer

This article is featured in Highlights of This Issue, p. 1595

Among all gynecologic cancers, epithelial ovarian cancer (EOC) is the most lethal due to dissemination into the peritoneal cavity and omental seeding in late-stage disease (1). High-grade serous carcinoma (HGSC) is the most common histosubtype of EOC and has a poor prognosis (2). The 5-year survival rate for patients with HGSC is approximately 47% due to limited screening options and late-stage diagnosis. This percentage decreases to 20%–35% in women diagnosed with stage III and IV HGSC. Because of their characteristic TP53 mutations, many patients with HGSC initially respond to standard-of-care platinum and taxol chemotherapies; however, a significant portion relapse with chemoresistant disease (3). PARP inhibitors were recently approved as a maintenance therapy for ovarian cancer (4). Although these inhibitors show promise, especially for homologous recombination–deficient ovarian cancer, some patients do not respond and resistance to these drugs has recently become evident. Therefore, there is an urgent need for novel therapeutic strategies for patients with HGSC.

Recent evidence suggests that therapy-induced senescence leads to a better 5-year survival rate for patients with HGSC (5). Cellular senescence is a state of stable cell-cycle arrest that is induced by multiple stimuli, including shortened telomeres, oncogene activation, DNA damage, and certain therapeutics (6). Senescent cells are characterized by many hallmarks such as increased β-galactosidase activity (termed senescence-associated β-galactosidase or SA-β-gal), increased repressive histone marks, such as repressive di- and tri-methylation marks of histone 3 lysine 9 (H3K9me2/3) at E2F target genes (termed senescence-associated heterochromatin foci, SAHF), decreased incorporation of the thymine analog bromodeoxyuridine (BrdU), and decreased proliferation (7–9). Because of the decrease in proliferation, therapy-induced senescence is considered a desirable therapeutic outcome (5, 10).

Tumor cells are characterized by metabolic reprogramming to maintain uncontrolled cell proliferation and growth (11). Tumors specifically reprogram metabolic pathways to allow for the increased need for biomass, such as nucleotides, lipids, and other macromolecules (12, 13). This metabolic reprogramming is increasingly thought to modulate epigenetic marks, including histone and DNA methylation and histone acetylation (14). Altered metabolism in cancer cells provides a unique opportunity to exploit these changes as a targeted therapy (11). Cancer cells undergo anaerobic glycolysis and generate lactate, even in the presence of oxygen, termed the Warburg effect (12). Although the Warburg effect is a common feature of many cancers, TCA cycle metabolism remains critical for both ATP production and macromolecule synthesis. Indeed, recent evidence suggests that multiple cancer types have increased TCA cycle metabolism (15), although less is known about TCA cycle activity in HGSC. Therefore, targeting the TCA cycle in cancer may serve as a novel therapeutic strategy.

Isocitrate dehydrogenase I (IDH1) catalyzes the oxidative decarboxylation of isocitrate to alpha-ketoglutarate (αKG) in a reversible reaction and produces NADPH. IDH1 is well-known for its mutation and the resulting production of the oncometabolite d-2-hydroxyglutarate [(d)-2HG, also known as (R)-2HG] in secondary glioblastoma, acute myeloid leukemia (AML), and other cancers (16, 17). However, recent work demonstrates that overexpression of wild-type IDH1 also promotes cancer progression in primary glioblastoma (18). Previous work identified a role for IDH1 in redox homeostasis and lipid metabolism (18, 19). In addition, IDH1 regulates histone marks through production of αKG, which is a cosubstrate for the Jumonji C (JmjC) histone demethylases (20, 21). Increased repressive histone methylation is a hallmark of senescence, and JmjC histone demethylases have been implicated in altering the epigenome of senescent cells (21). Therefore, depleting αKG pools through suppression of IDH1 may increase repressive histone methylation to induce senescence. While mutant IDH1 is well-characterized in several cancers (16, 17), the role of wild-type IDH1 in metabolism and epigenetics has never been investigated in the context of HGSC.

In this study, we determined that HGSC cells preferentially utilize glucose in the TCA cycle. While most TCA cycle enzymes were upregulated in adherent HGSC cells compared with the cell-of-origin fallopian tube (22), only wild-type IDH1 expression was both increased in HGSC cell lines under spheroid conditions and associated with decreased progression-free survival. Functionally, knockdown of IDH1 induced senescence by increasing repressive histone methylation at multiple E2F target gene loci. Together, these data indicate that targeting wild-type IDH1 in HGSC represents a novel strategy for this patient population by inducing senescence through epigenetic silencing.

The following human HGSC cells were used: Ovcar3 (ATCC, obtained in July 2017) and Ovcar10 (kind gift from Dr. Rugang Zhang, The Wistar Institute, Philadelphia, PA, obtained in November 2016). Both Ovcar3 and Ovcar10 cells are wild-type for IDH1 (Supplementary Fig. S1A). Ovcar3 and Ovcar10 cells lines were cultured in RPMI1640 medium (Gibco) supplemented with 10% FBS. HGSC cell lines were thawed, and experiments were performed within 30 passages. The FT282 fallopian tube cell line was a kind gift from Dr. Ronny Drapkin (University of Pennsylvania, Philadelphia, PA, obtained in July 2017). PFTE4, FT4-Tag, FT6-Tag, and FT33-Tag fallopian tube cell lines were a kind gift from Dr. Anna Loshkin (University of Pittsburgh, Pittsburgh, PA, obtained in November 2017). Fallopian tube cells were cultured in DMEM:F12 supplemented with 2% FBS and 1% penicillin/streptomycin under 2% oxygen conditions. Fallopian tubes cells were thawed, and experiments were performed within four passages. HEK-293FT cells were used for lentiviral packaging and were cultured in DMEM (Corning) supplemented with 10% FBS according to ATCC. For spheroid cultures, HGSC Ovcar3 and Ovcar10 cells lines were cultured in ultralow attachment (ULA) plates (Corning) for 4 days. Cells were treated with 15 μmol/L GSK864 (Sigma Aldrich) or 1 mmol/L αKG (Sigma Aldrich) where indicated. All cell lines were cultured in MycoZap (Lonza) and were routinely tested every 3 months for Mycoplasma as described previously (23). Cell lines were authenticated using STR profiling in November 2018 (Genetica).

Metabolites were quantified by liquid chromatography–high-resolution mass spectrometry after extraction of the cells by 80:20 methanol:water at −80°C, sonication, centrifugation of protein at 17,000 rcf for 10 minutes at 4°C, evaporation of the supernatant to dryness under N₂ gas, and resuspended in 50 μL of 5% 5-sulfosalicylic acid for analysis. For non-acyl-CoA polar metabolites ion pairing, reversed phase liquid chromatography-mass spectrometry was conducted by modification of a previously published method on a Ultimate 3000 Binary UHPLC coupled to a Q Exactive HF mass spectrometer (24). Data were processed in Xcalibur (Thermo Fisher Scientific). Peak areas were normalized to stable isotope-labeled internal standards spiked into each sample before extraction as follows; lactate, pyruvate to ¹³C₃-pyruvate, succinate, malate to ¹³C₄-succinate, (d/l)-2-hydroxyglutarate to ¹³C₅-hydroxyglutarate, αKG to ¹³C₅-αKG, fumarate to ¹³C₄-fumarate, citrate to ¹³C₆-citrate, acetyl-CoA, CoA, and succinyl-CoA to ¹³C₂-acetyl-CoA. For malonyl-CoA, reversed phase liquid chromatography-mass spectrometry was conducted by a previously published method on a ultimate 3000 binary UHPLC coupled to a Q exactive plus mass spectrometer (25). For isotope tracing experiments, analysis was similar except the exact mass corresponding to the detected ion of each ¹³C isotopologue was integrated in a 5-ppm window in Tracefinder v4.1 (Thermo Fisher Scientific), and then isotopologue enrichment was calculated by the open, web-based software FluxFix against unlabeled controls (26).

Total RNA was extracted from cells with TRizol (Life Technologies), DNase treated, and isolated using a RNA Clean and Concentrator Kit (Zymo) following the manufacturer's instructions. The cDNA libraries were prepared using the QuantSeq 3′mRNA-Seq Library Prep Kit FWD for Illumina (Lexogen) supplemented with UMI (unique molecular index) as per the manufacturer's instructions. Briefly, total RNA was reverse transcribed using oligo (dT) primers. The second cDNA strand was synthesized by random priming, in a manner that DNA polymerase is efficiently stopped when reaching the next hybridized random primer, so only the fragment closed to the 3′ end gets captured for later indexed adapter ligation and PCR amplification. UMIs were incorporated to the first six bases of each read, followed by four bases of spacer sequences. The processed libraries were assessed for its size distribution and concentration using Bioanalyzer High Sensitivity DNA Kit (Agilent Technologies). Pooled libraries were diluted to 2 nmol/L in EB buffer (Qiagen) and then denatured using the Illumina protocol. The libraries were pooled and diluted to 2 nmol/L using 10 mmol/L Tris-HCl, pH 8.5 and then denatured using the Illumina protocol. The denatured libraries were diluted to 10 pmol/L by prechilled hybridization buffer and loaded onto an Illumina MiSeq v3 flow cell for a 150 cycles using a single-read recipe according to the manufacturer's instructions. Demultiplexed sequencing reads were generated using Illumina BaseSpace.

UMI-specific workflows that were developed and distributed by Lexogen were used to extract reads that were free from PCR artifacts (i.e., deduplication). First, the umi2index tool was used to add the six nucleotide UMI sequence to the identifier of each read and trims the UMI from the start of each read. This generates a new FASTQ file, which is then processed through trimming and alignment. Second, after the quality and polyA trimming by BBDuk (https://jgi.doe.gov/data-and-tools/bbtools/) and alignment by HISAT2 (version 2.1.0; ref. 27), the mapped reads are collapsed according to the UMI sequence of each read. Reads are collapsed if they have the same mapping coordinates (CIGAR string) and identical UMI sequences. Collapsing reads in this manner removes PCR duplicates. Read counts were calculated using HTSeq (28) by supplementing Ensembl gene annotation (GRCh38.78). EdgeR (29) was used to fit the read counts to the negative binomial model along with generalized linear model and differentially expressed genes were determined by the likelihood ratio test method implemented in the edgeR package. Significance was defined to be those with q < 0.05 calculated by the Benjamini–Hochberg method to control the FDR. RNA-seq files are available in the Gene Expression Omnibus (GEO) database (GSE128700). The list of differentially expressed genes was analyzed with Ingenuity Pathway Analysis. Gene Cluster Text files (GTC) and Categorical Class files (CLS) were generated following the Gene Set Enrichment Analysis (GSEA) documentation indications (http://software.broadinstitute.org/gsea/index.jsp). GTC and CLS files were used to run GSEA hallmarks and reactome under default parameters (javaGSEA desktop application).

Protein lysates of patients' samples were provided by B.G. Bitler. The University of Colorado Gynecologic Tumor and Fluid Bank has an Institutional Review Board–approved protocol (COMIRB #07-935) in place to collect tissue from gynecologic patients with both malignant and benign disease processes. All participants are counseled regarding the potential uses of their tissue and sign a consent form approved by the Colorado Multiple Institutional Review Board, and studies were conducted in accordance with the U.S. Common Rule. The tissues were processed, aliquoted, and stored at −80 degrees. Benign fallopian tube tissue and primary HGSC tumors were homogenized with a Polytron Homogenizer (Brinkman Instruments) in RIPA buffer (150 mmol/L sodium chloride, Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 50 mmol/L Tris, pH 8.0) supplemented with complete EDTA-free protease inhibitor cocktail (Roche), sodium fluoride (10 mmol/L), and sodium orthovanadate (1 mmol/L).

pLKO.1-shIDH1 plasmids were obtained from Sigma-Aldrich. The TCRN are as follows: shIDH1 #1: TRCN0000027253; shIDH1 #2: TRCN0000027249 (Supplementary Fig. S1B). To overexpress IDH1, the IDH1 ORF was cloned into the pBABE-puro backbone. The following antibodies were obtained from the indicated suppliers: rabbit anti-IDH1 (Cell Signaling Technology), rabbit anti-Lamin B1 (Abcam), rabbit anti-Cyclin A (Abcam), mouse anti-PCNA (Cell Signaling Technology), mouse anti-MCM3 (Santa Cruz Biotechnology), mouse anti-Vinculin (Sigma-Aldrich), mouse anti-Beta Actin (Sigma-Aldrich), rat anti-BrdU (Abcam), mouse anti-PML (Santa Cruz Biotechnology), mouse anti-γH2AX (EMD Millipore), rabbit anti-53BP1 (Bethyl), Fluorescein donkey anti-rat IgG (Jackson ImmunoResearch), and Cy3 donkey anti-mouse (Jackson ImmunoResearch).

Lentivirus was packaged in 293FT cells using the ViraPower Kit from Invitrogen following the manufacturer's instructions. Cells infected with viruses encoding the puromycin-resistance gene were selected with 3 μg/mL puromycin for 7 days.

IDH1 cDNA was transfected into cells using jetOPTIMUS following the manufacturer's instructions (Polyplus Transfection).

For BrdU, cells on coverslips were incubated with 1 μmol/L BrdU for 30 minutes. Cells were fixed with 4% paraformaldehyde (PFA), permeabilized with 0.2% Triton X-100, and then postfixed with 1% PFA + 0.01% Tween-20. Coverslips were DNaseI treated for 10 minutes, and the DNaseI reaction was stopped using 20 mmol/L EDTA. Coverslips were then blocked with 3% BSA/PBS and incubated in anti-BrdU primary antibody (1:500) followed by incubation in FITC anti-Rat secondary antibody (1:1,000). Finally, coverslips were incubated with 0.15 μg/mL DAPI, mounted, and sealed.

For immunofluorescence, cells on coverslips were fixed, permeabilized, and blocked as above and then incubated with the corresponding primary antibodies followed by incubation in FITC anti-rabbit (1:2,000) or Cy3 anti-mouse (1:5,000) secondary. Finally, coverslips were counterstained with DAPI, mounted, and sealed. All images were acquired at room temperature using a Nikon Eclipse 90i microscope with a 20×/0.17 objective (Nikon DIC N2 Plan Apo) equipped with a CoolSNAP photometrics camera.

SA-β-Gal staining was performed as described previously (30). Cells were fixed with 2% formaldehyde/0.2% glutaraldehyde in PBS and incubated at 37°C in staining solution (40 mmol/L Na₂HPO₄, 150 mmol/L NaCl, 2 mmol/L MgCl₂, 5 mmol/L K₃Fe(CN)₆, 5 mmol/L K₄Fe(CN)₆, 1 mg/mL X-gal).

An equal number of cells was seeded in 6-well plates and cultured for 10–14 days. Cells were then fixed with 1% PFA/PBS, stained with 0.05% crystal violet, and destained with 10% acetic acid. Absorbance (590 nm) was measured using a spectrophotometer (Spectra Max 190).

Total RNA was prepared as described above. Relative expression of target genes listed in Supplementary Table S1 was analyzed using the QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific). Primers were designed using the Integrated DNA Technologies tool (https://www.idtdna.com/scitools/Applications/RealTimePCR/). Briefly, 5 ng of total RNA was used to One-Step qPCR (Quanta BioSciences) following the manufacturer's instructions in a final volume of 10 μL. Conditions for amplification were: 10 minutes at 48°C, 5 minutes at 95°C, 40 cycles of 10 seconds at 95°C and 7 seconds at the corresponding annealing temperature. Relative quantification was determined normalizing to multiple references genes (B2M, PUM1, PSMC4, and MRPL9) using the ΔΔC_t method. Design of IDH1 primers are shown in Supplementary Fig. S1B.

Chromatin immunoprecipitation (ChIP) was performed as described previously (31) using the ChIP-grade antibody mouse anti-H3K9me2 (Abcam). Cells were fixed in 1% PFA for 5 minutes at room temperature and then quenched with 1 mL of 2.5 mol/L glycine for 5 minutes at room temperature. After washing, cells were lysed in 1 mL ChIP lysis buffer (50 mmol/L HEPES-KOH, pH 7.5, 140 mmol/L NaCl, 1 mmol/L EDTA, pH 8.0, 1% Triton X-100, and 0.1% deoxycholate with 0.1 mmol/L PMSF and the EDTA-free protease inhibitor cocktail). Samples were incubated on ice for 10 minutes and then centrifuged at 3,000 rpm for 3 minutes at 4°C. The pellet was resuspended in 500 μL lysis buffer 2 (10 mmol/L Tris, pH 8.0, 200 mmol/L NaCl, 1 mmol/L EDTA, and 0.5 mmol/L EGTA with 0.1 mmol/L PMSF and the EDTA-free protease inhibitor cocktail) and incubated at room temperature for 10 minutes. Samples were centrifuged at 3,000 rpm for 5 minutes at 4°C. Next, the pellet was resuspended in 300 μL lysis buffer 3 (10 mmol/L Tris, pH 8.0, 100 mmol/L NaCl, 1 mmol/L EDTA, 0.5 mmol/L EGTA, 0.1% DOC, and 0.5% N-lauroylsarcosine with 0.1 mmol/L PMSF and the EDTA-free protease inhibitor cocktail). Cells were sonicated using a Branson Sonifier 250 for four cycles of 10 seconds on 50 seconds off. Next, 30 μL of 10% Triton X-100 was added to each tube, and then samples were centrifuged at maximum speed for 15 minutes at 4°C. Antibody–bead conjugate solution (50 μL) was added to the supernatant, and chromatin was immunoprecipitated overnight on a rotator at 4°C. The following washes were performed: ChIP lysis buffer, ChIP lysis buffer + 0.65 mol/L NaCl, wash buffer (10 mmol/L Tris-HCl, pH 8.0, 250 mmol/L LiCl, 0.5% NP-30, 0.5% deoxycholate, and 1 mmol/L EDTA, pH 8.0), and TE (10 mmol/L Tris-HCl, pH 8.0, and 1 mmol/L EDTA, pH 8.0). DNA was eluted with TES (50 mmol/L Tris-HCl, pH 8.0, 10 mmol/L EDTA, pH 8.0, and 1% SDS) for 30 minutes at 65°C. Reversal of cross-linking was performed by incubating samples overnight at 65°C. Proteins were digested using 1 mg/mL proteinase K and incubating at 37°C for 5 hours. Finally, the DNA was purified using the Wizard SV Gel and PCR Clean Up Kit (Promega). Immunoprecipitated DNA was analyzed by qPCR using iTaq Universal SYBR Green Supermix (Bio-Rad). Conditions for amplification were: 5 minutes at 95°C, 40 cycles of 95°C for 10 seconds and 30 seconds with 62°C annealing temperature. Enrichment of H3K9me2 was determined by normalizing to a gene desert region (31). Primer sets used for ChIP-qPCR are detailed in Supplementary Table S1.

Cells lysates were collected in 1× sample buffer (2% SDS, 10% glycerol, 0.01% bromophenol blue, 62.5 mmol/L Tris, pH 6.8, 0.1 mol/L DTT), boiled, and sonicated. Protein concentration was determined using the Bradford assay. An equal amount of total protein was resolved using SDS-PAGE gels and transferred to nitrocellulose membranes (Thermo Fisher Scientific). Membranes were blocked with 5% nonfat milk followed by overnight incubation in primary antibodies. Membranes were washed, incubated with horseradish peroxide–conjugated secondary antibodies (Cell Signaling Technology), and washed again. Proteins were visualized on film after incubation with chemiluminescent substrate (Thermo Fisher Scientific).

GraphPad Prism version 8.0 was used to perform statistical analysis. t Test and one-way ANOVA followed by post hoc Tukey HSD tests were applied as appropriate. When indicated, P values were adjusted according to Benjamini–Hochberg FDR. The significance level was established at P < 0.05. Heatmaps were generated using GraphPad Prism. Kaplan–Meier curves were generated using publicly available ovarian cancer mRNA gene chip data (32). Patients with ovarian cancer were filtered by a serous histosubtype and TP53 mutation to signify HGSC.

To determine whether changes in metabolism correlate with ovarian cancer disease progression, we performed metabolomics on primary fallopian tube cells and HGSC cell lines. We observed a significant upregulation in the levels of all TCA cycle metabolites in HGSC cells (Fig. 1A). Metabolites in other pathways such as glycolysis or the pentose phosphate pathway (PPP) were not universally upregulated in HGSC cells versus fallopian tube cells (Supplementary Fig. S2A and S2B). In addition, metabolite profiling of cell media showed decreased lactate production in HGSC cells compared with fallopian tube cells (Supplementary Fig. S2C), suggesting that glucose is not being consumed by aerobic glycolysis. To determine whether upregulation of enzyme expression is responsible for the observed increase in TCA cycle metabolism, we performed qRT-PCR analysis of all 27 TCA cycle enzymes in five normal fallopian tube cell lines, including two primary and three immortalized lines, and two HGSC cell lines (Fig. 1B). Both HGSC cell lines tested exhibited a significant increase in expression of all TCA cycle enzymes, except IDH2, when compared with fallopian tube cells. Together, these data suggest that TCA cycle metabolism is upregulated in HGSC.

HGSC disseminates to the peritoneal cavity in the form of spheroids during late-stage disease (1). To mimic these conditions in vitro, ULA plates were used to induce the formation of spheroids. We compared TCA cycle enzyme expression of Ovcar3 and Ovcar10 cells from adherent and spheroid conditions (Fig. 1C). Under these conditions, only three TCA cycle enzymes showed increased expression: PDK1 (pyruvate dehydrogenase kinase 1), SUCLG2 (succinate-CoA ligase GDP-forming beta subunit), and IDH1 (P < 0.001). PDK1 is an inhibitor of the TCA cycle (33); however, we found that HGSC cell lines produced a decreased amount of lactate (Supplementary Fig. S2C) suggesting that PDK1 upregulation may not correlate with enzymatic activity. Further analysis of patients with HGSC using mRNA gene chip data (32) demonstrated that only high IDH1 expression was associated with a significant decrease in progression-free survival (Fig. 1D; Supplementary Fig. S2D and S2E). Consistently, wild-type IDH1 protein is increased in patients with primary HGSC tumors when compared with normal fallopian tube tissues (Fig. 1E and F). While TCGA datasets lack normal fallopian tube samples, we found a significant increase in IDH1 expression using a publicly available dataset of 12 normal fallopian tube and 13 HGSC patient samples (GSE10971; Fig. 1G; ref. 34). Data from TCGA demonstrate that no patients with HGSC harbor an IDH1 mutation, and we confirmed that Ovcar3 and Ovcar10 HGSC cells express wild-type IDH1 (Supplementary Fig. S1A). Together, these data demonstrate that wild-type IDH1 is significantly upregulated in both cell lines and patient samples of HGSC compared with fallopian tube.

IDH1 catalyzes a reversible reaction to convert isocitrate to αKG (oxidative decarboxylation from glucose) or αKG to isocitrate (reductive carboxylation from glutamine), which is NADP⁺/NADPH dependent (Fig. 2A). To determine whether HGSC cells utilize IDH1 for oxidative decarboxylation or reductive carboxylation, we knocked down IDH1 (Supplementary Fig. S3A) and determined αKG and citrate abundance. Knockdown of IDH1 decreased αKG while citrate was increased (Fig. 2B and C). This suggests that IDH1 mainly functions in the oxidative decarboxylation of isocitrate to αKG in HGSC. To further confirm these results, we performed stable isotope labeling using ¹³C₆ glucose or ¹³C₅ glutamine. We observed a decrease in αKG produced through oxidative decarboxylation (M+2) upon IDH1 knockdown (Fig. 2D). Citrate produced through reductive carboxylation (M+5) was undetectable for both control and IDH1-knockdown cells in multiple experiments, suggesting that the reductive carboxylation pathway is not highly active in these cells. Interestingly, knockdown of IDH1 did not alter the NADPH/NADP⁺ ratio, suggesting that other NADPH pathways are contributing to the total NADPH pool (Supplementary Fig. S3B). This is consistent with a recent study showing that NADPH is mainly produced by the PPP (35). Lipid metabolism may be influenced by IDH1 expression (18). Production of the lipid precursors acetyl-CoA and malonyl-CoA from glutamine was not changed in either Ovcar3 or Ovcar10 cells (Supplementary Fig. S3C), suggesting that IDH1 may not play a major role in lipid biosynthesis in HGSC. Together, these data suggest that HGSC cells preferentially utilize glucose for the oxidative decarboxylation reaction of IDH1 to produce αKG.

Next, we sought to determine whether upregulation of IDH1 is critical for HGSC proliferation. Toward this goal, we performed RNA-sequencing analysis of Ovcar3 cells upon IDH1 knockdown (GEO accession no. GSE128700). Indeed, knockdown of IDH1 was associated with a significant decrease in pathways related to the cell cycle and proliferation (Supplementary Fig. S4A). Importantly, other isocitrate dehydrogenases (IDH2 and IDH3α/β/γ) were not increased upon IDH1 knockdown (Supplementary Fig. S4B), suggesting there is no compensation from these enzymes. This is also consistent with our metabolomics data (Fig. 2B–D), indicating that IDH1 may be the main isocitrate dehydrogenase enzyme in HGSC cells.

To further confirm that knockdown of IDH1 reduces HGSC cell proliferation, we examined additional proliferation markers upon knockdown of IDH1 in both Ovcar3 and Ovcar10 HGSC cell lines using two independent short hairpin RNAs (shRNA; Fig. 3A; Supplementary Fig. S4C). Knockdown of IDH1 decreased BrdU incorporation, colony formation, and cyclin A expression in both Ovcar3 and Ovcar10 in adherent conditions (Fig. 3B–F; Supplementary Fig. S4D–S4G). To limit the potential off-target effects of shRNA knockdown, we overexpressed IDH1 cDNA (Supplementary Fig. S4H and S4I). Indeed, overexpression of IDH1 rescued the inhibition of proliferation induced by shIDH1 (Fig. 3G and H). We also assessed proliferation after treatment with an IDH1 inhibitor (IDH1i). There are currently no commercially available inhibitors of wild-type IDH1; therefore, we used a mutant IDH1i that targets the wild-type enzyme at higher concentrations (18, 36). Under adherent conditions, inhibition of IDH1 in both Ovcar3 and Ovcar10 cells decreased BrdU incorporation and colony formation (Fig. 3I–L; Supplementary Fig. S4J–S4M). Overexpression of IDH1 in IDH1i-treated cells (Supplementary Fig. S4N and S4O) rescued this phenotype, indicating the observed phenotype is specifically due to inhibition of IDH1 and not off-target effects (Fig. 3M and N). Similar to adherent conditions, proliferation was also decreased by IDH1 suppression in spheroid conditions as indicated by decreased cyclin A, PCNA, and MCM3 (Fig. 3O and P; Supplementary Fig. S4P). Taken together, we conclude that knockdown or inhibition of wild-type IDH1 suppresses proliferation of HGSC cells.

Next, we sought to determine the mechanism by which inhibition or knockdown of IDH1 suppresses HGSC cell proliferation. In adherent conditions, knockdown of IDH1 did not induce cell death (Supplementary Fig. S5A). Interestingly, analysis of our RNA-Seq revealed that the top hit in the GSEA “Hallmark Gene Sets” was E2F target genes (Supplementary Fig. S5B), which is a characteristic of senescence (37). Indeed, upon suppression of IDH1, cells exhibited a large and flat morphology, which are also hallmarks of senescence (Supplementary Fig. S5C); therefore, we investigated whether suppression of proliferation was due to senescence induction. Toward this goal, we examined the expression of SA-β-Gal activity. Knockdown or inhibition of IDH1 increased SA-β-Gal activity in both Ovcar3 and Ovcar10 cells (Fig. 4A–E; Supplementary Fig. S5D–S5H). Other senescent markers such as decreased lamin B1 and increased PML bodies were also observed when IDH1 was knocked down or inhibited (Fig. 4A, F–J; Supplementary Fig. S5D, S5I–S5M). Interestingly, the senescence-associated secretory phenotype (SASP; ref. 6) was not robustly increased in IDH1-knockdown cells when compared with etoposide-treated cells (Supplementary Fig. S5N). Overexpression of IDH1 in IDH1-knockdown or IDH1i-treated (Supplementary Fig. S4H, S4I, S4N, and S4O) cells rescued the senescent phenotype (Fig. 4K–P). In addition, knockdown or inhibition of IDH1 induced senescence in spheroid conditions as indicated by decreased LMNB1 (Fig. 4Q). Our metabolomics data indicate that knockdown of IDH1 inhibits the oxidative decarboxylation of isocitrate, thereby decreasing αKG levels (Fig. 2B–D). Supplementation of IDH1-knockdown cells with exogenous αKG, but not citrate, partially rescued the senescent phenotype (Fig. 4R–U; Supplementary Fig. S5O and S5P), again indicating the important role of IDH1 in oxidative decarboxylation in HGSC cells. These data suggest that inhibition or knockdown of IDH1 induces senescence in both adherent and spheroid conditions and may not lead to the detrimental side effects of senescence that are associated with the SASP (6).

Next, we aimed to determine the mechanism by which targeting IDH1 induces senescence. Inhibition of IDH1 has previously been shown to increase reactive oxygen species (ROS) through decreased NADPH production (38), and ROS are a known inducer of senescence (6). However, we did not observe an increase in ROS upon IDH1 knockdown (Supplementary Fig. S6A), and treatment with the antioxidant N-acetyl cysteine did not rescue senescence due to IDH1 knockdown (Supplementary Fig. S6B). This is consistent with our results showing no change in the NADPH/NADP⁺ ratio upon IDH1 knockdown (Supplementary Fig. S3B). Similarly, we did not observe an increase in expression of other NADPH-producing enzymes such as ME3 and G6PD in IDH1-knockdown cells (Supplementary Fig. S6C). Moreover, we did not observe an increase in DNA damage, another well-known mechanism of therapy-induced senescence (ref. 6; Supplementary Fig. S6D). These data suggest another mechanism is inducing senescence of HGSC cells upon IDH1 knockdown.

We observed a significant decrease in E2F target genes (Supplementary Fig. S5B). E2F target genes are downregulated by increased repressive H3K9me2 in senescent cells (7, 8, 39). Increased histone methylation may be due to decreased histone demethylase activity. The JmjC family of histone demethylases requires αKG as a cosubstrate (20). As we observed a decrease in αKG levels upon IDH1 knockdown (Fig. 2B), we hypothesized that this would increase repressive histone methylation of E2F target genes to induce senescence. ChIP assays using an antibody specific for H3K9me2 demonstrated increased H3K9me2 occupancy at well-established E2F target genes PCNA and MCM3 in both Ovcar3 and Ovcar10 cells with IDH1 knockdown compared with controls (Fig. 5A). Consistently, PCNA and MCM3 mRNA and protein expression was decreased in these cells (Fig. 5B and C). Addition of exogenous αKG rescued H3K9me2 occupancy at these loci (Supplementary Fig. S6E), indicating that decreased αKG upon IDH1 depletion was responsible for histone methylation changes at PCNA and MCM3. To expand upon these findings, we used our RNA-Seq analysis to find additional E2F targets known to play a role in proliferation and senescence that may be regulated by H3K9me2 upon IDH1 knockdown. From that analysis, we identified six additional gene loci that show increased H3K9me2 occupancy, which correlated with decreased expression in our RNA-Seq analysis (CCNA2, MKI67, MCM2, CDC45, STMN1, and CDC25A; Supplementary Fig. S6F and S6G). Together, these data suggest that knockdown of IDH1 increases histone methylation at E2F target gene loci to decrease their expression and induce senescence.

In this study, we found that the TCA cycle enzyme, IDH1, is significantly increased in HGSC cells compared with fallopian tube cells. Patient HGSC samples also indicated a significant increase in IDH1 when compared with normal fallopian tube, which correlated with worse progression-free survival. Inhibition or knockdown of wild-type IDH1 suppressed HGSC cell proliferation and induced senescence, a stable cell-cycle arrest. Mechanistically, targeting IDH1 induced senescence by increasing repressive H3K9me2 at the loci of E2F gene targets. Taken together, these data indicate that high expression of IDH1 is a poor prognostic factor for HGSC and inhibiting its activity represents a novel strategy for HGSC progression and dissemination.

IDH1 is well-known for its mutation in secondary glioblastoma and AML, which produces (R)-2HG, an oncometabolite (16, 17). Indeed, we observed an increase in 2HG in HGSC (Fig. 1A); however, this methodology cannot distinguish between (R) and (S)-2HG. While both Ovcar3 and Ovcar10 cells have wild-type IDH1 (Supplementary Fig. S1A), we cannot definitively rule out the possibility of (S)-2HG function in these cells. It is known that IDH1 functions in a reversible reaction between isocitrate and αKG (40). Although other studies have shown IDH1 undergoes the reductive carboxylation reaction (38), we found that IDH1 in HGSC cells is preferentially utilized in the forward oxidative decarboxylation reaction under adherent conditions (Fig. 2). The oxidative reaction produces NADPH, and previous studies have shown that IDH1 upregulation protects cells from oxidative stress (38, 41). Interestingly, we did not observe a change in NADPH/NADP⁺ ratio, which is consistent with a recent study demonstrating that PPP activity is necessary and sufficient for NADPH pools and not IDH1 (35). In addition, previous studies have also identified a role for IDH1 in lipid metabolism (18, 19). Our results indicate that precursors to lipid metabolism are unaltered by IDH1 knockdown, suggesting that this is a context- and cell-type–dependent effect. Because we observed an increase in acetyl-CoA in HGSC cells compared with fallopian tube cells, future studies will need to be performed using lipidomics approaches to determine the exact pathways leading to lipid synthesis in HGSC cells.

Senescence is characterized by marked chromatin remodeling and increased repressive histone methylation, collectively termed the SAHF (7, 8). Specifically, di- and tri- methylation of histone H3 lysine 9 at E2F target gene loci is implicated in senescence (7). We found that knockdown of IDH1 induced H3K9 methylation at multiple E2F target gene loci (Fig. 5A; Supplementary Fig. S6F), leading to decreased expression of those genes (Fig. 5B; Supplementary Fig. S6G). Previous publications found that mutant IDH1, through the oncometabolite (R)-2HG, alters histone methylation at DNA damage response and differentiation gene loci through competitive inhibition of histone demethylases that require αKG for their activity (42, 43). Indeed, supplementation of shIDH1 cells with αKG rescued the repressive histone phenotype (Supplementary Fig. S6E), suggesting that αKG loss is affecting histone demethylase activity. Multiple histone demethylases have been connected to senescence induction, although very little is known about the role of αKG in senescence (21). Our data suggest that IDH1 and αKG levels maintain H3K9 in a demethylated state that is critical for HGSC cells to proliferate, and inhibition of αKG production increases repressive histone marks to induce senescence. Future work is needed to determine the specific histone H3K9 demethylase responsible for the senescent phenotype upon IDH1 knockdown.

Approximately 70% of HGSC receiving platinum-based standard-of-care therapy will relapse with chemoresistant disease (3). Currently, there are few therapies available for these patients after relapse, ultimately leading to patient mortality. Aberrant metabolism is a global characteristic of cancer cells (11), and altered metabolism of cancer cells affects the response to many chemotherapies (44). Therefore, targeting metabolism may serve as a novel therapeutic in many cancer types. Previous studies have implicated altered metabolic pathways in ovarian cancer, and multiple studies have targeted glycolytic enzymes for ovarian cancer treatment (45–47). However, our study demonstrates that fallopian tube cells also undergo aerobic glycolysis (Supplementary Fig. S2A and S2C), suggesting that targeting this pathway may result in toxicity to normal tissue. We found that TCA cycle metabolites and enzymes are universally and significantly upregulated in HGSC cells compared with fallopian tube, which suggests that targeting this pathway may lead to less toxicity. Interestingly, PDK1 was also upregulated in HGSC cells, especially under spheroid conditions (Fig. 1B and C). PDK1 inhibits the TCA cycle (33); however, we did not observe an increase in lactate production in HGSC when compared with fallopian tube, suggesting no increase in PDK1 activity (Supplementary Fig. S2C). We also observed changes in other pathway metabolites, such as glycolysis and the PPP (Supplementary Fig. S2A and S2B). Future work will determine whether these pathways are critical for HGSC proliferation and survival. Notably, we identified wild-type IDH1 as a potential target for HGSC. Recently, the IDH1i tibsovo (ivosidenib, AG-120) was approved for relapsed AML with mutant IDH1. This inhibitor is also effective against wild-type IDH1, but not wild-type IDH2 (48). Our data and others suggest that targeting wild-type IDH1 is a rational therapy for multiple cancers (18).

Late-stage HGSC primarily disseminates to the peritoneal cavity in the form of spheroids (1). These cells then seed on organs within the peritoneal cavity, which in part leads to the morbidity and mortality of late-stage ovarian cancer (1). We found that inhibition and shRNA-mediated knockdown of IDH1 induced senescence in both adherent and spheroid conditions (Fig. 4; Supplementary Fig. S5). This observation suggests that targeting IDH1 may be a therapeutic strategy for HGSCs at various stages including early- and late-stage diagnoses. Moreover, HGSCs are universally characterized by p53 mutations and approximately 20% have decreased RB1 (encoding for retinoblastoma protein, RB; ref. 49). While these two pathways are implicated in the senescence-associated cell-cycle arrest (6), our data suggest that targeting IDH1 downregulates proliferation promoting genes through epigenetic modifications independent of p53 and RB status.

Senescence is considered a tumor suppressive mechanism, and senescence induction is considered a beneficial therapeutic outcome (5, 10). However, senescence may also promote cancer and chemoresistance through the SASP, which increases inflammation and alters the surrounding microenvironment milieu (6). We found that SASP gene expression was not increased upon IDH1 knockdown (Supplementary Fig. S5N), suggesting that targeting IDH1 may lead to a sustained cell-cycle arrest without the harmful side effects of the SASP. Our results, in addition to others, suggest that senescence induction in the cell-type–specific context of ovarian cancer may overall be tumor suppressing (5, 50). Altogether, we propose that targeting IDH1 may act as a novel prosenescent therapy for patients with HGSC.

No potential conflicts of interest were disclosed.

Conception and design: E.S. Dahl, K.M. Aird

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.S. Dahl, R. Buj, K.E. Leon, J.M. Newell, B.G. Bitler, N.W. Snyder, K.M. Aird

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.S. Dahl, R. Buj, Y. Imamura, N.W. Snyder, K.M. Aird

Writing, review, and/or revision of the manuscript: E.S. Dahl, J.M. Newell, B.G. Bitler, N.W. Snyder, K.M. Aird

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Imamura

Study supervision: K.M. Aird

This work was partially supported by grants from the NIH (F31CA236372 to E.S. Dahl, R00CA194309 to K.M. Aird, R00CA194318 to B.G. Bitler, R03CA211820 to N.W. Snyder, and the use of core facilities under P30ES013508), the W. W. Charitable Trust (to K.M. Aird), and Penn State Cancer Institute Postdoctoral Fellowship (to R. Buj). We would like to thank the members of the Aird Lab for their thoughtful contributions. We would like to thank Drs. Ronny Drapkin (University of Pennsylvania) and Anna Loshkin (University of Pittsburgh) for the fallopian tube cells. We would like to thank Dr. George-Lucian Moldovan's laboratory (Penn State College of Medicine) for the PCNA and MCM3 antibodies.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.