IDH1 Inhibition Potentiates Chemotherapy Efficacy in Pancreatic Cancer

Recent investigations by our lab and others identified wild-type isocitrate dehydrogenase 1 (wtIDH1) as a key therapeutic target against cancer, including pancreatic ductal adenocarcinoma (PDAC; refs. 1–9). The protein is a cytosolic enzyme that is similar in function to the mitochondrial enzymes IDH2 and 3A. It catalyzes the interconversion of isocitrate and α-ketoglutarate (αKG) using NADP(H) as a cofactor (1–3). Under nutrient limitation, commonly present in the PDAC microenvironment, oxidative decarboxylation of isocitrate by IDH1 is favored (over reductive carboxylation), to produce NADPH and αKG. These products directly support antioxidant defense (providing reducing power) and mitochondrial function [through anaplerosis into the tricarboxylic acid (TCA) cycle], respectively (6, 8, 10). Our recent studies in PDAC revealed that pharmacologic inhibition of wtIDH1 can shrink tumors in some cases and improve survival in multiple murine models (8). Moreover, we showed that all off-the-shelf inhibitors of the neomorphic and oncogenic variant of IDH1, actually become potent inhibitors of the wild-type isoenzyme under tumor-associated conditions. Specifically, we observed that low Mg²⁺ found in tumors allows stronger target affinity and inhibition because the cation is a competitor of the inhibitor at the wtIDH1 allosteric site. Additionally, low glucose in the PDAC microenvironment augments dependency of PDAC on wtIDH1 for survival (8). Together, these observations point to a potential therapeutic window to treat wtIDH1 PDAC using allosteric IDH1 inhibitors.

Ivosidenib (AG-120) is approved by the United States Food and Drug Administration for the treatment of mutant IDH1 cholangiocarcinoma and acute myelogenous leukemia (AML; refs. 7, 11–14). It performed particularly well as a single agent against wtIDH1 PDAC in preclinical models (8). Additionally, the drug is available for treatment, well-tolerated, and administered orally, making it an attractive experimental treatment option for PDAC patients. To effectively translate our prior mechanistic and preclinical work, it is especially important from a practical standpoint to rigorously test the impact of ivosidenib in the context of conventional and standard-of-care chemotherapies currently in use to treat PDAC. Combination therapy is the most compelling therapeutic context to test drug activity in patients, and preclinical efficacy and toxicity remains unknown. Importantly, there exists a solid mechanistic rationale for the combination. IDH1 plays a key role in redox homeostasis (8–10), and most chemotherapeutics are strong inducers of reactive oxygen species (ROS; refs. 15, 16). Thus, pharmacologic inhibition of IDH1 (an antioxidant enzyme that generates a strong reducing agent, NADPH) may prove to be even more effective in tandem with cytotoxic agents. Consistent with this idea, published phase III results in IDH1-mutant AML have already shown safety and enhanced ivosidenib efficacy when administered in combination with the cytotoxic dose of azacytidine (7, 17–19). Also, a trial is currently accruing elsewhere pairing ivosidenib with gemcitabine and cisplatin to treat IDH1-mutant cholangiocarcinoma (NCT04088188).

Herein, we show that ivosidenib can be safely administered to animals in combination with PDAC standard chemotherapies and that the combinations are very effective therapeutically. Further, we explore the mechanistic underpinnings that drive synergy between ivosidenib and conventional therapeutics. These studies are foundational for current and future clinical trials with chemotherapy targeting wtIDH1 in PDAC.

Human pancreatic cancer cell lines (Mia-Paca2, PANC-1, SW1990, BxPC-3, AsPC-1, Hs766T, and Panc10.05) were obtained from ATCC. Experiments are performed wherever possible in multiple cell lines and preclinical models. The PDAC murine cell line, KPC (K8484: KrasG12D; Trp53R172H/+; Pdx1-Cre), was a gift from Dr. Darren Carpizo. All cell lines were cultured in normal DMEM with 10% fetal bovine serum (Gibco/Invitrogen) and 1% penicillin/streptomycin (Invitrogen). Cell lines were maintained at 37°C in 5% humidified CO₂. Glucose-free DMEM (Life Technologies, 21013-024) low glucose (2.5 mmol/L) and low magnesium (0.08 mmol/L) media were used to simulate aspects of the tumor microenvironment relevant to ivosidenib activity. To do this, glucose-free media was supplemented with varying glucose concentrations for low glucose experiments, and magnesium sulfate–depleted DMEM (Cell Culture Technologies, 964DME-0619) was used for low magnesium experiments. Cell lines were Mycoplasma tested routinely using a Mycoplasma Detection Kit (ATCC; 30-1012K). Before any in vitro experiment, cells were passaged at least twice. For rescue experiments, N-acetylcysteine (NAC; Sigma, no. A9165) and dimethyl-α-ketoglutarate (DMKG; Sigma-Aldrich, no. 349631) were utilized.

IHC was performed on 5-µm-thick formalin-fixed, paraffin-embedded tissue sections on glass slides. Tissues were prepared by first fixing tumors overnight in 10% formalin and then placed in PBS. Tissue sections were incubated for 75 minutes at 60°C, followed by a series of washes for deparaffinization and then rehydrated. Antigen retrieval was performed using a pressure cooker for 10 minutes at 123°C in 10 mmol/L citrate at pH 6.0, then 3% H₂O₂ for 10 minutes, followed by a washing step with water. Tissue sections were then incubated with a blocking solution containing 5% goat serum (Cell Signaling Technology; 5425) at room temperature for 1 hour. Primary antibodies were diluted (SignalStain Antibody Diluent; 8112) and incubated overnight at 4°C. Antibodies included those targeting Ki67 (SP6, BioCare Medical) and cleaved caspase-3 (Cell Signaling Technology; 9579) proteins. Slides were washed thrice with 1× TBS and Polysorbate 20 (TBST) and incubated using IHC detection Signal Stain Boost IHC Detection Reagent (Cell Signaling Technology; 8114) at room temperature for 30 minutes. The slides were washed three times again with 1× TBST and signal was detected using the HRP substrate kit (Vector; SK-4200), according to the manufacture’s recommendations. The primary organs, namely the liver, lung, and bone marrow of the mice, were dissected and subjected to hematoxylin and eosin staining. Quantification of each staining was performed with ImageJ software at ×20 magnification.

Cells were seeded in 96-well plates at 2,000 cells per well, and after 24 hours, cells were treated with drugs, including 5-fluorouracil (5-FU; Sigma, F6627), oxaliplatin (Sigma, O9512), and IDH1 inhibitor ivosidenib (Asta Tech, 40817) at the indicated concentrations. All experiments for cell viability were performed for 6 days unless otherwise detailed. Cell viability was measured using the Quant-iT PicoGreen dsDNA assay kits (Invitrogen; P7589).

Drug combination assays were performed by seeding 1,000 to 2,000 cells per well in 96-well plates. After 24 hours, cells were treated with different concentrations of ivosidenib (dose range, 1.25–20 μmol/L/mL), 5-FU (dose range, 0.15–20 μmol/L/mL), and oxaliplatin (dose range, 0.12–16 μmol/L/mL). Each treatment was done in triplicate. Cells were treated for 6 days, and cell growth relative to vehicle treatments was measured by staining with Quant-iT PicoGreen. For all in vitro experiments using ivosidenib, cells were cultured under low magnesium conditions (<0.4 mmol/L Mg²⁺) to effectively inhibit wtIDH1 enzyme activity with the drug (as a reference, normal culture media and serum contain roughly 1 mmol/L Mg²⁺). Low glucose (2.5 mmol/L glucose) was utilized as indicated to generate conditions of wtIDH1 dependency that simulate glucose levels in the natural tumor microenvironment (20–23).

Drug interaction data were quantified and characterized as synergistic, additive, or antagonistic using Bliss independence modeling, as described previously (24). The effect of the two drugs (X and Y) was calculated as follows: $Effect X+Y=[(Effect X+Effect Y)-(Effect X×Effect Y)]$⁠, where X and Y represent two different drugs.

In six-well plates, cells were initially seeded at 60% confluence. Transient transfections with siRNA (1 µmol/L) were carried out utilizing Lipofectamine 2000 (Invitrogen) and Opti-MEM (Invitrogen), following the manufacturer’s instructions. Typically, experiments were initiated 48 hours after the transfection process. The siRNA oligonucleotides, siIDH1 (s7121), (s7119), and siCTRL (AM4635), were procured from Ambion.

In the upper chamber of a 6.5-mm Transwell with 8.0-μm pore polycarbonate membrane inserts (Corning), cells were seeded at a density of 6 × 10⁴ cells. Subsequently, 100 μL of serum-free DMEM was introduced into the Transwells and incubated for 8 hours at 37°C. In the lower section, complete growth medium served as a chemoattractant. To eliminate nonmigrated cells, the upper surface was gently swabbed using cotton swabs. Cells that had migrated to the lower surface were fixed, stained with 0.5% crystal violet, imaged using a 10× objective on a Nikon TE200 microscope, and quantified using ImageJ analysis software.

Cells were lysed using 1× ice-cold RIPA buffer supplemented with protease and phosphatase inhibitors. Protein concentration was quantified using the Pierce BCA. Protein Assay kit (Thermo Fisher Scientific). Equal amounts of whole protein extracts were loaded and separated by electrophoresis on a 4% to 12% Bis-Tris gel (Life Technologies) and transferred to a PVDF membrane (Thermo Fisher Scientific). Blots were blocked in 5% (wt/vol) non-fat milk and then probed with antibodies against anti-IDH1 (Invitrogen, OTI2H9) and anti-β-actin (Invitrogen, 15739-BTIN). Chemiluminescence (32106, Thermo Fisher Scientific) was measured using a digital imager (Odyssey Infrared Imaging system).

Cells were cultured and seeded in six-well plates at 3,000 to 5,000 per well depending on the growth rate in 2 mL media to assess cell clonogenicity. Media was not changed during experiments unless indicated. Cells were treated with ivosidenib or 5-FU following acclimation of cells to low glucose and low MgSO4 (0.08 mmol/L) for 24 hours. After 8 to 10 days, cells were fixed in a reagent containing 80% methanol and stained with 0.5% (w/v) crystal violet solution for 10 minutes (25). To determine relative proliferation, dye was extracted using 10% glacial acetic acid, and the associated absorbance was measured at 600 nm using a Promega GloMax plate reader.

Cells were plated in 96-well black plates and cellular ROS levels were detected using 2′,7′ dichlorodihydrofluorescein diacetate (H2DCFDA; Invitrogen). Cells were treated in the absence or presence of drugs for 48 hours, and then incubated with 100 μL phenol red–free media containing 10 µmol/L DCFDA for 45 minutes at 37°C in the dark. The fluorescence signal was detected using an excitation wavelength at 485 nm and emission wavelength at 535 nm on a GloMax Promega plate reader. For apoptosis measurements, cells were plated in white 96-well plates and treated for 48 hours in the presence or absence of drugs. The caspase-3/7 (Caspase-Glo Promega G8090) level was measured per the manufacturer’s instructions.

The assessment of lipid peroxidation involved the quantification of malondialdehyde levels, which was performed using a Thiobarbituric Acid Reactive Substances Assay Kit sourced from Cayman Chemical. The assay was carried out following the manufacturer’s instructions.

To harvest intracellular metabolites, 1 × 10⁴ cells were seeded in 10-cm plates using complete growth media, in biological sextuplicates and quintuplicates. After cells were grown to ∼60% confluence, they were treated with the indicated drug for 48 hours. After 48 hours, the media was removed and cells rinsed with ice-cold PBS. Metabolites were extracted using ice-cold 80% high-performance liquid chromatography (HPLC)-grade methanol on dry ice by scraping cells from each plate. The cell lysates were centrifuged at 14,000 g for 10 minutes at 4°C and supernatant was then dried by SpeedVac and frozen at −80°C. For tumors, the samples were flash frozen in liquid nitrogen upon collection. Tumors approximately equal weight of (<100 mg) were collected per sample per experimental group. The tumors were homogenized using ice-cold 80% HPLC-grade methanol on dry ice by sonication and the resulting lysate was centrifuged at 14,000 g for 10 minutes at 4°C and supernatant was then dried by SpeedVac and frozen at −80°C. Samples were resuspended in HPLC-grade water and liquid chromatography-mass spectrometry was performed.

Polar metabolites were quantified by 5500 QTRAP triple quadrupole mass spectrometry (AB/SCIEX) coupled to a Prominence UFLC HPLC system (Shimadzu) using amide HILIC chromatography (Waters) at pH 9.2, as previously described (26). 299 endogenous water-soluble metabolites were measured at a steady state. After the identification of metabolites, differential metabolites were examined in the context of the Kyoto Encyclopedia of Genes and Genomes pathway database (27) for metabolite enrichment pathway analysis. NADPH (NADP/NADPH-Glo Promega G9081) and glutathione (GSH) levels (GSH-Glo Promega V6911) were also measured separately per the manufacturer’s instructions.

Total RNA was extracted using the RNeasy PureLink RNA isolation (Life Technologies; 12183025) according to manufacturers’ instructions. RNA was converted to cDNA using a High-Capacity cDNA Reverse Transcription Kit following the manufacturer’s protocol (Applied Biosystems; 4387406). The qPCR was performed using Taqman Universal Master Mix (Thermo Fisher Scientific; 4440038), along with a probe (IDH1 Thermo Fisher Scientific; 4351372) and cDNA. Analyses were performed using the Bio-Rad CFX96 Maestro Manager 2.0 software.

Mia-Paca2 and PANC-1 cells were seeded at 10,000 cells/well in complete normal DMEM (25 mmol/L glucose and 4 mmol/L glutamine) in an Agilent XFp Cell Culture miniplate (#103025-100) at 37°C in 5% humidified CO₂ incubators, as recommended. Oxygen consumption rates (OCR) were performed using the XFp mini extracellular analyzer (Seahorse Bioscience). For OCR measurements, cells were treated in the absence or presence of 5-FU or oxaliplatin under low glucose concentration for 36 hours. The XFp FluxPak cartridge (#103022-100), was hydrated with 200 μL/well of XF Calibrant (#100840-000) and incubated at 37°C using non-CO₂ incubator overnight. The following day cells were washed and replaced with Seahorse XF base media (at the indicated glucose concentrations), and the plate was incubated in a non-CO₂ incubator at 37°C. The Mito Stress test kit (#103015-100) was reconstituted in an assay medium using the following inhibitor concentrations: 1.5 µmol/L oligomycin, 2.5 µmol/L cyanide-p-trifluoromethoxyphenyl-hydrazone (FCCP), and 0.5 µmol/L rotenone + 0.5 µmol/L antimycin A. OCR was measured using Wave 2.6 software, according to the manufacturer’s instructions, and all data were normalized to cell number with Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen).

All mouse experiments were approved and conducted under the Case Western Reserve University Institutional Animal Care and Use Committee (protocol 2018-0063) and adherent to institutional guidelines. Female athymic nude mice (Nude-Foxn1nu; aged 6–8 weeks) were purchased from Harlan Laboratories (6903M). Mia-Paca2 cells were suspended in a 200 μL solution comprised of 70% Dulbecco’s PBS and 30% Matrigel. 1 × 10⁶ suspended cells were injected subcutaneously into the right flank of mice. Ten-week-old C57BL/6J female mice were purchased from Jackson Laboratories for orthotopic experiments. For orthotopic syngeneic experiments, a suspension of 5 × 10⁴ Luciferase-expressing KPC cells in 30 μL of a 40:60 mixture of PBS and Matrigel solution (Corning, 354234) was injected directly into the mouse pancreas. On postoperative day 10, the presence of pancreatic tumors was confirmed by bioluminescence (BLI) via Spectrum CT (PerkinElmer, 2898979) following intraperitoneal injection of 100 μL D-luciferin (50 mg/mL in PBS). Mice with confirmed tumors were randomized to the indicated treatment groups.

We obtained KP^−/−C tamoxifen-inducible mice (Kras^LSL-G12D;p53^LoxP;Pdx1-CreER) on an C57/BL6 background from the Jackson Laboratory (032429). Tamoxifen was dissolved in corn oil at a concentration of 20 mg mL⁻¹, and the mice were administered tamoxifen solution at a dose of 75 mg kg⁻¹ intraperitoneally once daily for five consecutive days. Both male and female mice were included in the study.

Patient-derived xenografts (PDX) were purchased from Jackson Laboratory (#TM01212) and 2 mm³ tumors were implanted subcutaneously into the right flanks of nude mice. For animal studies, when PDX and Mia-Paca2 tumor volumes reached 100 to 120 mm³ in flanks (approximately 3–5 weeks) or orthotopic tumors were confirmed with BLI in the case of syngeneic PDAC experiments, treatments were initiated as indicated. Mice were administered ivosidenib (AG-120, Asta Tech, 40817; 75 mg/kg in PEG-400, Tween-80, and saline (10:4:86) twice a day orally), 5-FU (Sigma-Aldrich, F6627; 30 mg/kg via intraperitoneal injections two times a week), oxaliplatin (Sigma-Aldrich, O9512; 5 mg/kg, once weekly), FOLFIRINOX [5-FU 12.5 mg/kg, irinotecan 25 mg/kg (Sigma-Aldrich, I1406) and oxaliplatin 2.5 mg/kg, once weekly], or vehicle control.

Body weight and tumor volume were measured weekly. Vernier calipers were used to measure tumor size, using a standard formula (Volume = Length × Width²/2). Upon termination of experiments, mice were euthanized under carbon dioxide inhalation and tumors resected. Tumors were collected in 10% formalin (Richard Allan Scientific, a subsidiary of Thermo Fisher Scientific; 427-098) for IHC analyses and stored at −80°C until processed for further study.

The data are expressed as mean ± SEM (standard error of the mean) of at least three independent experiments unless indicated. Comparisons between groups were determined using unpaired, two-tailed Student t test (^∗, P < 0.05; ^∗∗, P < 0.01; ^∗∗∗, P < 0.001; ^∗∗∗∗, P < 0.0001). One-way or two-way ANOVA test was used to compare more than two groups. The Kaplan–Meier estimate was used for survival experiments, and survival groups were compared using the log-rank test. GraphPad Prism 9.2.3 Software was used for statistical analysis.

The publicly available data from The Cancer Genome Atlas (TCGA) analyzed in this study were obtained from the GDC portal (https://portal.gdc.cancer.gov) and (http://gepia.cancer-pku.cn/). Metabolomics data are available in Datasets S1 and S2. All other raw data are available upon reasonable request from the corresponding authors.

IC₅₀ doses from single-agent dose response experiments (Supplementary Fig. S1A–D) were performed to help select optimal drug doses and ranges for functional studies and drug combination experiments. Metabolomics analyses revealed differences in metabolites and metabolic function after treatment with chemotherapy. For these experiments, liquid chromatography-mass spectrometry (LC/MS) was performed after 36 hours of drug treatment, using a sub-lethal dose (IC₃₀) of chemotherapy to avoid detecting changes that largely reflected cell death. Untargeted metabolomics revealed a large metabolic shift with 5-FU and oxaliplatin treatments (Fig. 1A; Supplementary Fig. S2A). The principal components analysis (PCA) highlights the magnitude of metabolic reprogramming from chemotherapy exposure (Fig. 1B; Supplementary Fig. S2B). Pathway enrichment analysis revealed specific pathways that dominated the metabolic program after 5-FU and oxaliplatin treatment, including the TCA cycle, amino acid, and nucleotide metabolism (Fig. 1C; Supplementary Fig. S2C).

In these same experiments, a closer look at metabolite changes in the enriched pathways after 5-FU and oxaliplatin treatment revealed effects on specific pathways in the TCA cycle, redox metabolism, and nucleotide metabolism (Fig. 1D; Supplementary Fig. S2D). Large changes were observed, including increases in specific TCA metabolites like isocitrate and αKG (Fig. 1E; Supplementary Fig. S2E). Additionally, enhanced reductive power mobilized to overcome the oxidative insult of chemotherapy was reflected by higher levels of NADPH and GSH (Fig. 1F and G; Supplementary Fig. S2F and S2G). Nucleotide-related metabolites were interrogated in 5-FU treated versus nontreated cells and revealed a strong accumulation of all nucleotide intermediates, which likely reflected a compensatory feedback response of treatment, especially in the de novo pyrimidine biosynthetic pathway (Fig. 1H).

Mitochondria play a central role in energy production and respiration (oxidative phosphorylation), and the contribution of this biology toward survival is increased under metabolic stress, including low glucose (8, 10, 28, 29). Cancer cells rely on efficient ATP production under harsh conditions to power prosurvival pathways. This was validated in the aforementioned metabolomic studies after chemotherapy treatment, and further supported by increased mitochondrial respiration after sublethal 5-FU and oxaliplatin treatment, under low glucose conditions (Fig. 1I and J; Supplementary Fig. S2H). In aggregate, these metabolic studies suggest that PDAC cells adapt and respond to chemotherapy stress through increased oxidative phosphorylation, antioxidant defense, and nucleotide synthesis.

TCGA database revealed that wtIDH1 is highly expressed in PDAC samples, as compared to normal pancreatic tissue. Further, higher tumoral expression of IDH1 was associated with worse overall survival and disease-free survival, among patients with PDAC (Fig. 2A–C). Consistent with this notion, examination of both mRNA and protein in a panel of PDAC cell lines demonstrated a moderate to high expression level of IDH1 compared to primary cells (Fig. 2D and E). Despite the assumption that wtIDH1 is a housekeeping enzyme, these clinical and preclinical data, in conjunction with our prior preclinical studies (8), validate wtIDH1 as a promising therapeutic target with a therapeutic window. Of note, gain of function mutations in the protein are exceedingly rare in PDAC (30). We were particularly interested in the possibility that increased wtIDH1 expression and activity impacted PDAC chemoresistance, in part based on our prior observations that wtIDH1 blockade alone induced oxidative DNA damage, as well as in combination with chemotherapy (8, 10). Thus, we investigated a possible relationship between IDH1 expression and chemotherapeutic resistance in the context of 5-FU, which is a backbone ingredient in the multiagent, frontline FOLFIRINOX PDAC regimen (31, 32). Using qPCR and immunoblot, 5-FU-induced expression of IDH1 in both Mia-Paca2 and PANC-1 cells over the course of 48 and 72 hours (Fig. 2F–I). Key reductive metabolites were elevated with 5-FU or oxaliplatin treatment in in vitro experiments (Supplementary Fig. S3A–D), validating the findings in the untargeted metabolomics studies above. Similarly, 5-FU induces the expression of IDH1 and compensatory augmented reductive power in a PDX model of pancreatic cancer (Fig. 2J; Supplementary Fig. S3E). Interestingly, 5-FU did not induce the expression of IDH2 or IDH3 (Supplementary Fig. S3F), underscoring a possible role of the wtIDH1 enzyme under chemotherapeutic stress.

While 5-FU is a common chemotherapeutic used to treat PDAC (33), the drug is only minimally effective as a single agent (34–41). Previously, our group validated ivosidenib as a potent wtIDH1 inhibitor in vitro under low magnesium and low nutrient conditions present in the PDAC microenvironment (8). Findings were validated in in vivo animal studies. We speculated that wtIDH1 inhibition would sensitize PDAC cells to 5-FU because of effects on the enzyme’s protective effects against DNA damage, antioxidant defense, and mitochondrial function. Treatment of Mia-Paca2 and PANC-1 cells were performed with ivosidenib and 5-FU across a range of doses; dose responses of each drug alone were described above (Supplementary Fig. S1A–F) and informed the experimental dosing range used for combination studies. The addition of an IDH1 inhibitor substantially increased 5-FU potency (up to 31-fold and 11-fold in the cell lines) by PicoGreen DNA quantitation (Fig. 3A and B). Similarly, ivosidenib rendered oxaliplatin substantially more potent (up to 18-fold, 23-fold, 27-fold, and 12-fold in Mia-Paca2, PANC-1, AsPC-1, and Hs766Tcells, respectively; Supplementary Fig. S4A and S4B). Bliss index analyses indicated true combination synergy, reflected in the positive Synergy Scores for both cell lines (Fig. 3C and D). As expected, similar to ivosidenib treatment, the knockdown of IDH1 drastically sensitizes cells to 5-FU (Fig. 3E and F).

Next, we performed clonogenic assays to test the impact of combination therapy on PDAC cell survival more directly and over a longer timeframe. Both Mia-Paca2 and PANC-1 cells were treated with increasing doses of chemotherapy, and colonies were stained after 10 days. At lower doses, neither ivosidenib nor 5-FU alone had a marked effect on colony formation of PDAC cells. The combination effect was substantial, however at doses around the IC₅₀ levels (Fig. 3G and H).

Mitochondria are the greatest source of endogenous ROS (42–44). As mitochondrial function is augmented as an adaptive survival response to chemotherapy, we expected chemotherapy to also drive ROS production. Moreover, chemotherapy is known to induce ROS through apoptosis and other mechanisms (16, 45). Indeed, there was an increase in ROS when PDAC cells were treated with chemotherapy for 48 hours (Fig. 4A and B). ROS levels were increased further with the addition of ivosidenib, which directly suppresses NADPH generation. The effect was partially rescued with the glutathione precursor and antioxidant, N- acetyl cysteine, (NAC; 2 mmol/L; Fig. 4A and B). Consistent with in vitro data, combination of ivosidenib and 5-FU increased lipid peroxidation in a PDX model, reflected by increased TBAR as a measure of ROS mediated damage (Supplementary Fig. S5A). Apoptotic studies mirrored this pattern, and the combination of ivosidenib and 5-FU translated to even greater potentiation with this readout (caspase-3/7 activity; Fig. 4C and D). Again, the effect was substantially abrogated with NAC. Similarly, the anapalerotic TCA cycle substrate and IDH1 reaction product, αKG (DMKG) partially rescued cell viability of the combination (Supplementary Fig. S5B and S5C). Increased cell death was also observed with the drug combination, as measured by PicoGreen DNA quantitation (Fig. 4E and F), with a strong rescue effect with NAC. The drug combination even affected cell migration of PDAC cells (Supplementary Fig. S5D and S5E). These data therefore supported a model where chemotherapy induces ROS, and as a part of the survival response, PDAC cells induce IDH1 to bolster mitochondrial and antioxidant activity. IDH1 blockade overcomes this adaptive survival response (Fig. 4G).

Nude mice bearing PDX tumors (TM01212) were treated with either vehicle, low dose ivosidenib, low dose 5-FU, or the combination of both. Xenografts of mice receiving the combination failed to grow at all (Fig. 5A–D). Importantly, mice did not exhibit any signs of treatment-related toxicity, as evidenced by steady body weights in all groups (Fig. 5E). Ki-67 immunolabeling of harvested tumors showed a markedly reduced proliferation rate in tumors exposed to combination treatment (Fig. 5F). In contrast, cleaved caspase-3 labeling of tumors was substantially elevated (Fig. 5G). We then tested ivosidenib efficacy in combination with multiagent chemotherapy, using the standard-of-care FOLFIRINOX regimen (5-FU, irinotecan, and oxaliplatin). Mice treated with the combination of low dose ivosidenib (75 mg/kg oral daily) and FOLFIRINOX (5-FU 12.5 mg/kg, irinotecan 25 mg/kg, and oxaliplatin 2.5 mg/kg once weekly) failed to grow, while low-dose treatments of these therapies alone had only a modest effect (Fig. 5H and I). Low toxicity was reflected in the stable body weights observed during the entire treatment (Fig. 5J).

Chemotherapy and ivosidenib were also tested in a syngeneic, immunocompetent orthotopic PDAC model where tumors were transplanted directly into the mouse pancreas during survival surgery. The combination of treatments using only low doses still resulted in a significant survival benefit, relative to either drug therapy alone, using the same low dose (Fig. 6A–D). The impact of combination treatment was additionally investigated in a more aggressive strain of the KPC mouse model. This particular model, when subjected to tamoxifen treatment, develops PDAC due to conditional mutations in Kras and Trp53 (46). Again, improved survival with the combination was observed (Fig. 6E). This is highly significant since ivosidenib alone did not impact survival either here or in a prior study (8). Histopathological toxicity studies by hematoxylin and eosin staining revealed no morphological changes in major organs (Fig. 6F).

A relatively small number of studies to date have focused on wild-type isoenzyme as a cancer target, although attention is mounting. One of the first was a study by Metallo and colleagues that observed that under hypoxic conditions, melanoma cancer cells rely on wtIDH1 for reductive carboxylation of αKG to generate isocitrate, as critical step toward de novo lipogenesis. Suppression of this enzyme reduced cell growth (1). Similar observations were reported by Jiang and colleagues. The authors showed in various solid tumors that the enzyme was important for spheroid growth in vitro (2). A more recent study by Calvert and colleagues demonstrated that wtIDH1 directs carbon in the opposite biochemical direction. They observed that IDH1 oxidative decarboxylation of isocitrate supports glioblastoma tumor growth, and that the resultant NADPH generated from the reaction supported antioxidant defense and minimized ROS (5).

We recently showed that wtIDH1 is important in PDAC, establishing and validating several key principles in that study. PDAC cells depend on the oxidation of isocitrate by wtIDH1 under nutrient limitation, a key characteristic of tumors, to produce both NADPH and αKG (6, 8–10). These two products support antioxidant defense (through NADPH) and mitochondrial function (through αKG), respectively. Further, we showed that ivosidenib and other allosteric IDH1 inhibitors, originally developed to selectively target mutant IDH1, actually and counterintuitively inhibit wtIDH1 under PDAC-associated conditions. Specifically, low Mg²⁺ levels permit stronger binding of the compound within the allosteric site of wtIDH1, and low nutrient concentrations increase cancer cell dependency on this enzyme (8, 9). These conditions features of the PDAC tumor microenvironment, at least in preclinical models, which renders wtIDH1 cancer cells vulnerable to pharmacologic, allosteric IDH1 inhibition. An independent group recently validated many of these findings in prostate cancer (47).

In the present study, we take the practically important step to show that conventional chemotherapy can potentiate the antitumor effects of IDH1 inhibition and vice versa. Commonly used drugs against PDAC, like 5-FU and oxaliplatin, reprogram metabolism toward greater dependence on oxidative phosphorylation and antioxidant defense. These changes render PDAC cells especially susceptible to IDH1 blockade, because the enzyme is so crucial for these adaptive biologic processes under metabolic stress, and this appears to be much more relevant to cancer cells than to normal tissues. The observation proved true in both in vitro and in vivo PDAC models.

Based on these data, we designed and opened a phase 1b clinical trial (NCT05209074) that is half-way accrued at the time of publication, testing the combination of ivosidenib and FOLFIRINOX in patients with resectable pancreatic adenocarcinoma. The trial provides the first human study to our knowledge testing allosteric IDH1 inhibitors against wtIDH1 cancers. We seek to demonstrate safety of the combination. Additionally, we will perform imaging and metabolic profiling studies as pharmacodynamic assessments of biologically relevant wtIDH1 inhibition in human PDAC.

In conclusion, our results indicate that conventional chemotherapy, including standard-of-care multiagent FOLFIRINOX, potentiates ivosidenib activity and efficacy against PDAC. IDH1 blockade suppresses mitochondrial function and induces ROS, which in turn drives apoptotic-driven PDAC cell death when administered in combination with chemotherapy. Future studies are needed to nominate and test non-chemotherapeutic treatment partners for anti-IDH1 therapy based on biologic insights into the enzyme’s biologic roles in PDAC survival, particularly in the context of austere conditions present in the PDAC microenvironment.

J.M. Winter along with University Hospitals, filed the following patent application on September 24, 2020: Methods for Treating Wild Type Isocitrate Dehydrogenase 1 Cancers. Information regarding this patent application is as follows: PCT/US20/52445 filed 09/24/20, Claiming Priority to U.S. 62/911,717 filed 10/7/19, File Nos: UHOSP-19738 | 2019-014.

M. Zarei: Conceptualization, data curation, software, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. O. Hajihassani: Investigation, methodology. J.J. Hue: Investigation, methodology. A.W. Loftus: Investigation, methodology. H.J. Graor: Investigation, methodology. F. Nakazzi: Investigation, methodology. P. Naji: Investigation, methodology. C.S. Boutros: Investigation, methodology. V. Uppin: Investigation, methodology. A. Vaziri-Gohar: Investigation, methodology. A.S. Shalaby: Investigation, methodology. J.M. Asara: Investigation, methodology. L.D. Rothermel: Investigation, methodology. J.R. Brody: Investigation, methodology. J.M. Winter: Conceptualization, data curation, funding acquisition, writing–original draft.

J.R. Brody is supported by NIH-NCI R01 CA212600; U01CA224012-03; NIH-NCI R21 CA263996. Research supported by the 2015 Pancreatic Cancer Action Network-AACR Research Acceleration Network Grant, grant number 15-90-25-BROD, Lustgarten, and the Hirshberg Foundation. Grant support for this research comes from the American Cancer Society MRSG-14-019-01-CDD, American Cancer Society 134170-MBG-19-174-01-MBG, Gateway for Cancer Research G-22-1100, NCI R37CA227865-01A1, NCI R01 CA281219, the Case Comprehensive Cancer Center GI SPORE 5P50CA150964-08, Case Comprehensive Cancer Center core grant P30CA043703, and University Hospitals research start-up package (J.M. Winter). We are grateful for additional support from numerous donors to the University Hospitals Surgical Oncology Lab, including the John and Peggy Garson Family Research Fund, The Jerome A. and Joy Weinberger Family research fund, the Hieronymous Family, Robin Holmes-Novak in memory of Eugene, Brittan and Fred DiSanto, and Rosi and Saby Behar.