Inhibiting Fatty Acid Oxidation Reverses Autophagy-Mediated Acquired Chemotherapy Resistance in Pancreatic Ductal Adenocarcinoma

Combined treatment with irinotecan and fatty acid oxidation inhibition reverses acquired drug resistance driven by autophagy in pancreatic cancer.

Abstract

In pancreatic ductal adenocarcinoma (PDAC), irinotecan (1μmol/L) chemotherapy triggers a dual-phase autophagy process that drives drug resistance. Although mTOR and autophagy exert suppressive effects on each other, coactivation of mTOR and autophagy has been observed when PDAC cells begin to regrow after treatment. Therefore, we hypothesized that the distinct temporal phases of autophagy are governed by independent upstream pathways. Initially, DNA damage activated AMPK, inducing early autophagy within 24 hours that fueled fatty acid oxidation (FAO), boosting ATP production. After 48 hours, elevated ATP levels inactivated AMPK and activated mTOR, which typically suppresses autophagy. However, autophagy and FAO activity persisted beyond 72 hours of irinotecan treatment via the JNK1–Beclin-1 pathway. This created a paradoxical state in which mTOR and autophagy were coactivated, promoting cell survival under irinotecan treatment. Irinotecan combined with FAO inhibition using KN510713 (a combination of KN510 targeting the carnitine–acylcarnitine transporter and KN713 targeting acetyl-CoA acyltransferase1/2) or FAO gene knockdown blocked autophagy flux and cell growth. FAO inhibition–induced fatty acid accumulation impaired autophagy flux and induced cytotoxicity, leading to cancer cell death. In xenograft models, combining irinotecan with KN510713 significantly prevented tumor regrowth compared with irinotecan alone. These findings suggest that targeting FAO induced by autophagy activation may overcome acquired drug resistance in PDAC while minimizing the toxic side effects associated with systemic inhibition of autophagy in healthy cells.

Significance:

Combined treatment with irinotecan and fatty acid oxidation inhibition reverses acquired drug resistance driven by autophagy in pancreatic cancer.

Graphical Abstract

Introduction

The main issue with cancer treatment is that tumors often grow back and become resistant to therapy in patients who initially respond well to chemotherapy. This phenomenon is known clinically as cancer relapse through acquired drug resistance (1). To overcome cancer resistance, various strategies have been proposed, including pharmacologic concepts of targeting drug inactivation, reducing intracellular drug accumulation, and altering drug targets (2), as well as biological concepts of regulating DNA repair and cell death (3), cancer stemness (4), and activation of compensatory pathways for cell survival (5). It has also been shown that cancer relapse is closely linked to the activation of autophagy (6, 7). Furthermore, in vitro studies have demonstrated that inhibiting autophagy, when combined with chemotherapy, completely suppresses acquired resistance (8–10). However, the therapeutic benefits remain controversial owing to the severe adverse effects observed in clinical trials (11). Autophagy–associated acquired resistance occurs in two stages. The first stage is the response to anticancer drug treatment, during which early autophagy supports survival (12). The second stage is the regrowth stage, during which cancer cells that have survived anticancer drug treatment resume growth, accompanied by late autophagy (10). The phenomenon of plasticity in cancer cells is referred to as an adaptive stress response (13). Despite extensive research, the controversy surrounding autophagy seems to rest on a simple observation fallacy: The mechanism of autophagy in acquired drug resistance has not been observed in a time-dependent manner. In a recent study, dabrafenib was administered to RAF-mutated melanoma cells, whereas dacarbazine was administered to Raf wild-type cells (14). In a melanoma xenograft model, monotherapy with either dabrafenib or dacarbazine resulted in recurrent growth due to acquired resistance. Both cell lines exhibited concomitant autophagy and mTOR activation during anticancer drug treatment (14). Although mTOR and autophagy are known to exert mutually suppressive effects, paradoxical coactivation has been observed during tumor promotion under anticancer therapy. Recently, we found that autophagy and mTOR signaling pathways are also activated concurrently when PDAC cells begin to regrow after anticancer drug treatment. To reconcile this, we hypothesized that the distinct temporal phases of autophagy are governed by independent upstream pathways. We found that irinotecan treatment induces early autophagy (12) through AMPK activation by DNA damage (15) and that autophagy increases fatty acid oxidation (FAO) in pancreatic ductal adenocarcinoma (PDAC). The increase in ATP by FAO (16, 17) directly induces mTOR activation (18), consistent with increased OxPhos in the mitochondria resulting from irinotecan treatment (18, 19). Therefore, FAO may play a key role in regulating mTOR activation and autophagy during acquired drug resistance. Consequently, we examined the mechanisms underlying mTOR activation and late autophagy induction in response to anticancer drugs, with a focus on FAO-related energy metabolism. Additionally, we assessed whether the anticancer effects were enhanced through the suppression of acquired drug resistance when an anticancer drug was combined with FAO inhibition, either by knocking down FAO genes, such as carnitine–acylcarnitine carrier (CAC) and acetyl-CoA acyltransferase 1 (ACAA1), or using FAO inhibitors, including KN510713 (a combination of KN510 targeting CAC and KN713 targeting ACAA1/2).

Materials and Methods

Institutional Review Board statement

All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of the National Cancer Center Research Institute (protocol nos. for xenograft model: NCC-24-1047, approved on August 12, 2024; NCC-25-1096, approved on February 11, 2025; protocol no. for Kras^G12D; Trp53^R172H; and Pdx1-Cre, KPC, model: NCC-22-767, approved on June 29, 2022; NCC-24-1014, approved on March 28, 2024; NCC-25-1141, approved on June 25, 2025), which is an Association for Assessment and Accreditation of Laboratory Animal Care International–accredited facility that abides by the Institute of Laboratory Animal Resources guide.

Genetically engineered mouse model

The genetically engineered KPC mouse model of spontaneous pancreatic cancer, previously generated by us (20), Kras^G12D; Trp53^R172H; Pdx1-Cre (KPC), was used to evaluate the efficacy of anticancer agents for pancreatic cancer. KPC mice spontaneously develop pancreatic cancer by expressing Kras and p53 mutations in the pancreas. (21). A significant advantage is that it represents a naturally occurring pancreatic cancer that develops into tumors containing various mutations (22). However, although general pancreatic cancer mainly occurs in the pancreatic duct, KPC develops throughout the pancreas and progresses to cancer, making it irreversible and fatal, with a median survival (MS) of 5 months (21). Four groups of KPC mice were analyzed: control (n = 27), irinotecan (n = 20), KN510713 (n = 27), and combination therapy (n = 21).

Xenograft model with irinotecan treatment

BALB/c nu/nu mice (Orient Bio; age 6–8 weeks) were used to establish the xenograft model. MIA-PaCa-2 cells (ATCC, cat. #CRM-CRL-1420, RRID: CVCL_0428, 1 × 10⁷) were suspended in a 1:1 mixture of PBS and Matrigel (354230; BD Biosciences). A total of 200 μL of the suspension was injected subcutaneously into the right flank of each mouse using a 1-mL syringe. To evaluate the antitumor efficacy of the treatments at different stages of tumor progression, experiments were conducted at two distinct time points: an early stage model, in which treatment was initiated when tumors reached approximately 100 mm³ (n = 6 per group), and a mid-stage model, in which treatment began at 300 mm³ (n = 7 per group) to reflect more advanced tumor growth. KN510713 (KN510 25 mg/kg and KN713 25 mg/kg) was administered orally once daily, whereas irinotecan was administered intraperitoneally at 40 mg/kg once per week. The tumor volume and body weight were measured weekly. Overall survival was analyzed using Kaplan–Meier analysis.

Cell culture studies

hTERT-HPNE cells (ATCC CRL-4023, RRID: CVCL_C466) were cultured in 75% glucose-free DMEM (D5030; Sigma-Aldrich) supplemented with 2 mmol/L of L-glutamine, 1.5 g/L of sodium bicarbonate, and 25% Medium M3 Base (Incell Corp.) containing 5% fetal bovine serum (FBS), 5.5 mmol/L of D-glucose, and recombinant human EGF. MIA PaCa-2 cells (ATCC CRL-1420, RRID: CVCL_0428), PANC-1 cells (ATCC CRL-1469, RRID: CVCL_0480), SW1990 cells (ATCC CRL-2172, RRID: CVCL_1723), and HPAC cells (ATCC CRL-2119, RRID: CVCL_3517) were maintained in high-glucose DMEM supplemented with 10% FBS and penicillin/streptomycin. SU.86.86 cells (ATCC CRL-1837, RRID: CVCL_3881) were cultured in RPMI 1640 medium with 10% FBS and penicillin/streptomycin. PA-TU-8988S cells (DSMZ ACC 204, RRID: CVCL_1846) and PA-TU-8988T cells (DSMZ ACC 162, RRID: CVCL_1847) were maintained in DMEM supplemented with 5% heat-inactivated FBS, 5% horse serum, and 2 mmol/L of L-glutamine. KPC cells (Ximbio, 153474) were maintained in DMEM supplemented with 10% FBS, 1% L-glutamine, and 1% penicillin/streptomycin. All cells were cultured at 37°C in a humidified incubator with 5% CO₂. The cells were purchased directly from the ATCC and stored as stock cultures; if the stocks were resuscitated, they were discarded after the fourth passage. Before in vivo experiments, human cell lines were authenticated by short tandem repeat profiling by the Genomics Analysis Team at the National Cancer Center, Korea, and Mycoplasma contamination was also checked in all cell lines using the e-Myco VALiD Mycoplasma PCR Detection Kit (iNtRON Biotechnology, cat. #25239).

KN510 and KN713 and treatment doses

KN510 is an omeprazole that inhibits CAC translocase (SLC25A20; ref. 23). Among compounds with these chemical properties, omeprazole, which has a benzimidazole scaffold, effectively targets the cysteine residue of the active site based on its pharmacologic mechanism (24). Based on this mechanism, mitochondrial CAC is an ideal target for omeprazole (23). The fact that omeprazole efficiently inhibits the CAC transport system is demonstrated by the low IC₅₀ value of 5.6 μmol/L (23). The structural and functional relationships, as well as the interaction mechanism between omeprazole and CAC, have been well elucidated through site-directed mutagenesis combined with molecular docking (23). In addition, omeprazole is a proton pump inhibitor that reduces gastric acid secretion and is used to treat gastroesophageal reflux disease (25). KN713 is a trimetazidine that inhibits ACAA1/2 (3-ketoacyl-CoA thiolase 1/2; ref. 26). Peroxisomal ACAA is known as ACAA1, whereas mitochondrial ACAA is known as ACAA2. KN713 inhibits ACAA1 and ACAA2 expression. KN713 competitively inhibits the catalytic activity of ACAAs, with an IC₅₀ value of 75 nmol/L for long-chain fatty acids and 100 μmol/L for short-chain fatty acids (26). KN713 reduces FAO by blocking ACAA and promotes glucose oxidation in normal cells. Glucose oxidation consumes less oxygen than fatty acid β-oxidation in normal cells (26). Therefore, by increasing glucose oxidation and reducing reliance on fatty acid metabolism, trimetazidine helps optimize cellular energy production under conditions of limited oxygen availability, such as in angina. KN713 was the first cytoprotective anti-ischemic agent developed on the basis of this principle (26). It is used to treat angina pectoris (27). In this study, the experimental design required a 72-hour observation period without changing the culture medium, making it essential to treat cells at an initial concentration sufficiently high to achieve saturation without inducing toxicity. To achieve this, we used relatively high in vitro concentrations: 200 μmol/L of KN510 and 2 mmol/L of KN713, termed the KN510713 combination. Regarding in vivo dosages, the KN510713 dose was tested. The doses of KN510713 were intended to be 25 mg/kg, with a 1:1 ratio for the in vivo study.

Immunohistochemical staining

Immunohistochemistry (IHC) was performed on formalin-fixed, paraffin-embedded pancreatic tissues from mice. Sections were deparaffinized in xylene, dehydrated in ethanol, and rehydrated with water. IHC was performed using standardized methods. The primary antibodies used in this experiment, including their sources and dilution ratios, were carnitine palmitoyltransferase 1A (CPT1A; Abcam, cat. #ab234111, RRID: AB_2864319, 1:500), CK19 (Abcam, cat. #ab52625, RRID: AB_2281020, 1:600), Ki-67 (Abcam, cat. #ab15580, RRID: AB_443209, 1:1,000), phospho-AMPKα (Thr172; Cell Signaling Technology, cat. #2535, RRID: AB_331250, 1:100), phospho-mTOR (Ser2448; D9C2; Cell Signaling Technology, cat. #5536, RRID: AB_10691552, 1:25), and γ-H2AX (Bethyl Laboratories, cat. #A300-081A, RRID: AB_203288, 1:1,000).

Image acquisition and analysis

The stained tissue sections were scanned in high-resolution mode using a Vectra Polaris multispectral imaging system (Akoya Biosciences, RRID: SCR_025508). Protein expression was analyzed based on pathologic evaluation of tissue morphology and staining patterns in pancreatic cancer xenograft tumors. The inForm Image Analysis Software (Akoya Biosciences, RRID: SCR_019155) was used to quantitatively assess the IHC results. H-scores were calculated based on the percentage of positively stained cells and staining intensity. The significance of protein expression across different groups was analyzed using one-way analysis of variance (ANOVA) in GraphPad Prism version 10.3.1 (GraphPad Software, RRID: SCR_002798). Pancreatic samples for IHC were obtained at 16 to 23 weeks of age, corresponding to the MS window when pancreatic lesions are consistently present. For the KPC mouse pancreatic tissues, whole-slide image analysis was initially attempted using inForm software. However, accurate automated segmentation of normal pancreatic components (acinar cells, islet cells, normal ductal cells, and stromal cells) and lesion areas (PanIN and PDAC) was not feasible. Therefore, a semi-quantitative manual evaluation approach was used. Five representative lesion areas were randomly selected from each tissue, and staining intensity was scored as 0 (negative), 1 (weak), 2 (moderate), or 3 (strong) using an H-score–like method. The scores were analyzed using one-way ANOVA and the Kruskal–Wallis test.

Quantitation of metabolites of acylcarnitine using liquid chromatography–tandem mass spectrometry

Metabolites were analyzed using liquid chromatography–tandem mass spectrometry (LC–MS/MS) with a 1290 HPLC system (Agilent), QTRAP 5500 (AB Sciex), and LC column, as previously described (16).

Immunoblotting

To investigate the changes in intracellular signaling pathways following modulation of FAO targets, immunoblotting was performed using an established protocol (16). The primary antibodies used in this experiment, their sources, and dilution ratios were CAC (Abcam, cat. #ab244436, RRID: AB_3740919, 1:250), Total OXPHOS Human WB Antibody Cocktail (Abcam, cat. #ab110411, RRID: AB_2756818, 1:1,000), CPT1A (Abcam, cat. #ab128568, RRID: AB_11141632, 1:1,000), ACAA1 (Thermo Fisher Scientific, cat. #PA5-29956, RRID: AB_2547430, 1:1,000), Bcl-2 (Thermo Fisher Scientific, cat. #MA1-26233, RRID: AB_779361, 1:1,000), β-actin (Santa Cruz Biotechnology, cat. #sc-47778, RRID: AB_626632, 1:1,000), JNK1 (Cell Signaling Technology, cat. #9252, RRID: AB_2250373, 1:1,000), phospho-JNK (Cell Signaling Technology, cat. #9251, RRID: AB_331659, 1:1,000), Beclin-1 (Cell Signaling Technology, cat. #3738, RRID: AB_490837, 1:1,000), phospho-Beclin-1 (Cell Signaling Technology, cat. #84966, RRID: AB_2800045, 1:1,000), ATG14 (Cell Signaling Technology, cat. #5504, RRID: AB_10695397, 1:1,000), phospho-ATG14 (Cell Signaling Technology, cat. #92340, RRID: AB_2800182, 1:1,000), Bax (Cell Signaling Technology, cat. #2772, RRID: AB_10695870, 1:1,000), PARP (Cell Signaling Technology, cat. #9542, RRID: AB_2160739, 1:1,000), cyclin D1 (Cell Signaling Technology, cat. #55506, RRID: AB_2827374, 1:1,000), mTOR (Cell Signaling Technology, cat. #2983, RRID: AB_2105622, 1:1,000), phospho-mTOR (Ser2448; Cell Signaling Technology, cat. #5536, RRID: AB_10691552, 1:1,000), phospho-Bcl-2 (Cell Signaling Technology, cat. #2827, RRID: AB_659950, 1:1,000), AMPKα (Cell Signaling Technology, cat. #2532, RRID: AB_330331, 1:1,000), phospho-AMPKα (Thr172; Cell Signaling Technology, cat. #2535, RRID: AB_331250, 1:1,000), LC3-I/II (Cell Signaling Technology, cat. #12741, RRID: AB_2617131, 1:1,000), and γ-H2AX (Bethyl Laboratories, cat. #A300-081A, RRID: AB_203288, 1:10,000).

Immunoprecipitation

To examine the interaction of Bcl-2 with Beclin-1 or Bax during late autophagy, human pancreatic cancer cell lines MIA PaCa-2 (ATCC, cat. #CRM-CRL-1420, RRID: CVCL_0428) and SU.86.86 (ATCC, RRID: CVCL_3881) were treated with irinotecan (1 μmol/L) for 0, 24, 48, or 72 hours. At each point, the cells were harvested and lysed on ice in immunoprecipitation (IP) buffer (50 mmol/L of Tris–HCl, 150 mmol/L of NaCl, 1 mmol/L of EDTA, 0.5% Triton X-100, pH 7.4) supplemented with protease and phosphatase inhibitor cocktails. The lysates were clarified by centrifugation (14,000 × g, 10 minutes, 4°C), and the supernatants were incubated overnight at 4°C with anti–Bcl-2 antibody (1 μg/mL) under gentle rotation. The immune complexes were captured by adding 10 μL of Protein A/G UltraLink resin (50:50 resin-to-buffer slurry; #35133, Thermo Fisher Scientific) and rotating at room temperature for 2 hours. The beads were collected by centrifugation at 3,000 rpm for 3 minutes, washed five times with 500 μL of ice-cold IP buffer, and gently tapped between washes. Bound proteins were eluted by boiling the beads in 2 × Laemmli sample buffer, resolved by SDS–PAGE, and transferred to polyvinylidene difluoride membranes. The membranes were probed with antibodies against Beclin-1, Bax, and Bcl-2, as indicated, followed by detection with appropriate horseradish peroxidase–conjugated secondary antibodies and chemiluminescence.

Inhibitor treatment and Earle’s Balanced Salt Solution starvation

MIA PaCa-2 cells were treated with irinotecan (1 μmol/L) for 48 hours, followed by treatment with the ULK1 inhibitor XST-14 (5 μmol/L; MedChemExpress, Monmouth Junction; ref. 28) or the JNK1 inhibitor SP600125 (20 μmol/L; MedChemExpress, Monmouth Junction; refs. 29, 30) for an additional 24 hours. For starvation experiments, cells were pretreated with XST-14 (5 μmol/L) for 24 hours and subsequently incubated in Earle’s Balanced Salt Solution (EBSS; Sigma-Aldrich) for 2 hours. For inhibitor-only experiments, cells were treated with XST-14 (5 μmol/L) or SP600125 (20 μmol/L) for 24 hours. XST-14 and SP600125 were dissolved in DMSO and then diluted to the indicated concentrations in culture medium.

Measurement of ATP, AMP, and acetyl-CoA levels

ATP and AMP levels were measured using ATP Colorimetric/Fluorometric Assay Kit (ab83355, Abcam) and AMP Assay Kit (ab273275, Abcam). This assay was performed as previously described (31). Acetyl-CoA levels were measured using an Acetyl-CoA Assay Kit (ab87546, Abcam) following the manufacturer’s instructions.

Apoptosis analysis

Apoptotic and necrotic cell populations were quantified using an Annexin V–AbFluor 488/Propidium Iodide Apoptosis Detection Kit (Abbkine) following the manufacturer’s instructions. Fluorescence was measured using a BD FACSCanto II flow cytometer (BD Biosciences). Data were analyzed using FlowJo software (version 10.10.0, BD Biosciences, RRID: SCR_008520) to determine the percentage of viable (Annexin V⁻/PI⁻), early apoptotic (Annexin V⁺/PI⁻), late apoptotic (Annexin V⁺/PI⁺), and necrotic (Annexin V⁻/PI⁺) cells.

Reactive oxygen species measurement

Reactive oxygen species (ROS) levels were determined using the 2,7-dichlorofluorescin diacetate (DCFDA)/H2DCFDA Cellular ROS Assay Kit (ab113851, Abcam) according to the manufacturer’s protocol. Trypsinized cells (2 × 10⁵) were washed with cold PBS and incubated with 20 μmol/L of the ROS indicator DCFDA in 1× buffer at room temperature for 15 minutes. The levels of cellular ROS were analyzed using the BD FACSVerse flow cytometer. Fluorescence intensity was quantified by FlowJo software (version 10.10.0, BD Biosciences, RRID: SCR_008520).

In vitro colony formation assay (cEC₅₀)

To assess the long-term growth of pancreatic cancer cells following FAO inhibition, a colony formation assay was performed using an established method. The cells were treated with KN510 (control, 12.5, 50, and 200 μmol/L) or KN713 (control, 125, 500, and 2,000 μmol/L). Each experiment was conducted in triplicate, and the growth of pancreatic cancer cell lines under various conditions was analyzed using one-way ANOVA in GraphPad Prism version 10.3.1 (GraphPad Software, RRID: SCR_002798).

Live-cell imaging of autophagic flux using an mCherry–GFP–LC3 stable cell line

Lentiviral particles were generated and used to infect MIA PaCa-2 (ATCC, cat. #CRM-CRL-1420, RRID: CVCL_0428) cells according to a previously described protocol (32) using the FUW mCherry–GFP–LC3 plasmid (Addgene plasmid #110060, RRID: Addgene_110060). MIA PaCa-2 cells stably expressing mCherry–GFP–LC3 were seeded into PhenoPlate 96-well plates (PerkinElmer) at a density of 5 × 10³ cells per well in complete growth medium and incubated at 37°C. After 24 hours of cell seeding, the culture medium was replaced with medium containing one of the following treatments: vehicle, irinotecan (1 μmol/L), KN510 (200 μmol/L) and KN713 (2 mmol/L), irinotecan (1 μmol/L) combination with KN510 (200 μmol/L) and KN713 (2 mmol/L), rapamycin (100 nmol/L; ref. 33), or chloroquine (10 μmol/L; ref. 34). Hoechst 33342 (0.1 μg/mL; Thermo Fisher Scientific) was added to all treatment conditions, followed by a 10-minute incubation at 37°C. This point was defined as 0 hours, and live-cell images were acquired at 24, 48, and 72 hours after treatment. Images were acquired using an Operetta High-Content Imaging System (PerkinElmer, RRID: SCR_018810) with a 40× water-immersion objective and appropriate filters for GFP, mCherry, and Hoechst. For each point, 81 fields per well were captured, and a minimum of 1.5 × 10³ cells per well were analyzed based on the 0-hour baseline. Quantification of cytoplasmic autophagosomes and autolysosomes was performed using the Harmony version 4.5 software (PerkinElmer, RRID: SCR_018809), with analysis parameters adapted from the method described by Zhang and colleagues (35). Spots positive for both mCherry (red) and GFP (green) were counted as autophagosomes. Spots positive for mCherry were counted as autolysosomes.

Fatty acid uptake and colocalization with LC3-II

MIA PaCa-2 cells were seeded at a density of 5 × 10³ cells per well in PhenoPlate 96-well plates (PerkinElmer). On the following day, the culture medium was replaced with medium containing 1 μmol/L of boron–dipyromethene (BODIPY) FL-C16 (D3821, Thermo Fisher Scientific) and the indicated treatments: irinotecan (1 μmol/L), irinotecan (1 μmol/L) combined with KN510 (200 μmol/L), or KN713 (2 mmol/L). Hoechst 33342 (0.1 μg/mL) was added to all treatment conditions, and the mixture was incubated for 10 minutes at 37°C. After 48 hours of treatment, cells were washed twice with cold PBS, fixed with 4% paraformaldehyde, and permeabilized with 0.25% Triton X-100. The cells were blocked with 5% BSA in PBS for 1 hour at room temperature and incubated with an anti-LC3B antibody (Cell Signaling Technology, cat. #3868, RRID: AB_2137707, 1:1,600) overnight at 4°C, followed by incubation with appropriate fluorophore-conjugated secondary antibodies for 1 hour at room temperature. Images were acquired using the Operetta High-Content Imaging System (PerkinElmer, RRID: SCR_018810) equipped with a 40× water-immersion objective and appropriate filter sets for EGFP, Alexa Fluor 594, and Hoechst. Quantification of cytoplasmic lipid droplets and autophagosome-associated spots was performed using Harmony version 4.5 software (PerkinElmer, RRID: SCR_018809). The mean fluorescence intensity of BODIPY FL-C16 was quantified within EGFP-positive (green) spots. Colocalization was determined by calculating the number of Alexa Fluor 594–positive spots overlapping with EGFP-positive spots.

Cell viability measurement

Cells were seeded at a density of 1 × 10⁴ cells/well in 48-well plates. The next day, the culture medium was replaced, and the indicated compounds were added. The cells were then incubated for 72 hours at 37°C in a humidified incubator with 5% CO₂. After the treatment, an equal volume of CellTiter-Glo reagent was added to each well and incubated for 20 minutes on a shaker with gentle agitation. Luminescence was measured using a microplate reader, and cell viability was determined based on luminescence intensity.

Protein identification and quantification

All raw data files were searched in Proteome Discoverer version 3.1 software (Thermo Fisher Scientific, RRID: SCR_014477) using SEQUEST-HT against the SWISS-PROT-Human database, following the established protocol (36).

Statistical analysis

All statistical analyses were performed using GraphPad Prism version 10 (GraphPad Software Inc., RRID: SCR_002798). The data are presented as the mean ± SD unless otherwise indicated; tumor growth curves are presented as the mean ± SEM. The number of independent experiments (n) or animals is stated in each figure legend.

Biological replicates were defined as independent experiments performed using separately prepared samples on different days, whereas technical replicates were defined as repeated measurements within the same biological sample. Unless otherwise specified, n refers to biological replicates for in vitro experiments and to the number of animals for in vivo studies. The variance measure used (SD or SEM) is explicitly stated in the corresponding figure legends.

For comparisons among three or more groups, one-way ANOVA or Kruskal–Wallis tests were used as specified in the figure legends (e.g., ANOVA for apoptosis quantification; Kruskal–Wallis for nonnormally distributed IHC H-scores and xenograft tumor growth comparisons). For two-group comparisons, two-tailed unpaired t tests were used (e.g., autophagic flux quantification). For selected metabolite profiling datasets, the Mann–Whitney U test and multiple unpaired two-tailed t tests were performed as indicated in the corresponding figure.

Overall survival in the KPC model was analyzed using the Kaplan–Meier method and compared using the log-rank (Mantel–Cox) test. The sample sizes for animal studies were determined based on prior experience with these models and feasibility (no statistical methods were used to predetermine sample size, if applicable). Animals/samples were excluded only according to prespecified criteria (e.g., technical failure or humane endpoint unrelated to tumor burden), and all remaining data were included in the analyses. Statistical significance was defined as P < 0.05. The following notation was used: P < 0.05 (* or ^#), P < 0.01 (** or ^##), P < 0.001 (*** or ^###), and P < 0.0001 (**** or ^####); ns, not significant.

Results

Tumors treated with irinotecan exhibited increased autophagy associated with FAO over time

As noted in the Introduction, autophagy is activated during regrowth, even after treatment with anticancer drugs, thereby enabling cancer cells to survive. To determine whether FAO increased under these conditions, we collected and analyzed tumors from xenograft mice that continued to grow well following anticancer drug treatment. At week 7, tumor volumes differed between the vehicle control group and the average values for both doses, although the difference was not statistically significant. At week 7, the expression of p-mTOR in tumor sections increased in a dose-dependent manner with irinotecan treatment (control = 13.7, 20 mg/kg = 36, 40 mg/kg = 66.9; Fig. 1A; Supplementary Fig. S1A). We found that the expression of FAO markers CAC (SLC25A20), CPT1A, and ACAA1 (3-ketoacyl-CoA thiolase) increased in proportion to the irinotecan concentration in the tumor tissue (Supplementary Fig. S1B–S1D). Moreover, p-mTOR activation significantly increased, whereas AMP-activated protein kinase (p-AMPK) showed inactivation (Supplementary Fig. S1B and S1C). To investigate the resistance observed in xenograft models in vitro, pancreatic cancer cells treated with irinotecan were harvested at 0 to 72 hours. Using MS/MS and Western blotting, time-dependent increase in the expression of mTOR, RPS6KB1 (S6K), EIF4EBP1, and LC3-II was observed in both MIA PaCa-2 and SU.86.86, confirming the induction of autophagy alongside the simultaneous activation of cell proliferation signals following irinotecan treatment (Fig. 1B and C). However, p-AMPK activation switched to inactivation after 72 hours of treatment with irinotecan in MIA PaCa-2 and SU.86.86 (Fig. 1C). Conversely, p-AMPK and p-ULK1 decreased sharply 72 hours after irinotecan treatment. When ATP and AMP concentrations were measured in SU.86.86 cells under the same conditions, ATP increased with irinotecan treatment, reaching a 50% increase at 72 hours relative to 0 hours. The difference in AMP levels after irinotecan treatment decreased by 50% at 72 hours compared with 0 hours. The intracellular AMP/ATP ratio fell to 0.33 at 72 hours compared with 0 hours, representing a 67% reduction, indicating that the increase in ATP production led to the inactivation of p-AMPK (Fig. 1D). We found that the expression of FAO-related proteins increased in cells that survived irinotecan treatment, as determined by MS and immunoblot analyses (Fig. 1B and E). As an FAO marker, CAC increased 9.1- and 13-fold in MIA PaCa-2 and SU.86.86 after 72 hours of irinotecan treatment, compared with the control group, respectively (Fig. 1E). This increase in the FAO marker was consistent with the increase in the expression of OxPhos complexes (Supplementary Fig. S1E). These results suggest that PDAC cells may depend on FAO as a key energy source for survival under irinotecan treatment. This is supported by the experimental results above and by previous studies showing that pancreatic cancer cells exhibit a metabolic characteristic in which ATP production relies entirely on FAO (16, 17). We examined whether basal autophagy activity influences responsiveness to autophagy induction by irinotecan (Fig. 1F). The results showed that PaTu-8988T (higher basal autophagy activity) activated pAMPK and autophagy more rapidly than PaTu-8988S (relatively lower basal autophagy activity; Fig. 1F). However, in both cell lines, pAMPK inactivated after 72 hours, and autophagy was activated 4.4-fold in PaTu-8988S and 3.3-fold in PaTu-8988T (Fig. 1F).

Figure 1.

IR treatment simultaneously increased autophagy and FAO. A, MIA PaCa-2 cell–derived xenograft tumors were established, and mice were randomized into Ct (white, n = 5), IR 20 mg/kg (light orange, n = 5), and IR 40 mg/kg (orange, n = 5) groups when tumor volumes reached ∼300 mm³. Tumor growth curves are presented as the mean ± SEM. At the end of the study, tumors were harvested and processed for IHC to evaluate p-mTOR expression. The staining intensity was quantified using H-scores generated using inForm software; H-score data are presented as the mean ± SD. Statistical significance for H-scores was assessed using the Kruskal–Wallis test. Scale bars, 300 μm (black) and 50 μm (blue). B, Heatmap displaying the relative abundance of intracellular phosphorylation or total protein in MIA PaCa-2 and SU.86.86. Cells treated with IR (1 μmol/L) and phosphorylation or total protein levels were measured using LC-MS/MS and normalized as Z-scores. Red indicates increased abundance, and blue indicates decreased abundance relative to the mean across all samples. C, MIA PaCa-2 and SU.86.86 cells were treated with IR (1 μmol/L) and harvested for immunoblotting. Densitometric values shown below each band were normalized to β-actin and expressed relative to the 0-hour Ct. D, SU.86.86 cells were treated with IR (1 μmol/L), and intracellular ATP (orange) and AMP (white) levels were measured. Data are shown as the mean ± SD from n = 3 independent experiments. E, MIA PaCa-2 and SU.86.86 cells were treated as in (C), and FAO-related proteins CAC and ACAA1 were analyzed by immunoblotting. Band intensities were normalized to β-actin and expressed relative to the 0-hour Ct. F, Pancreatic cancer cell lines PaTu-8988-S and PaTu-8988-T were treated with IR (1 μmol/L) and harvested for immunoblotting. Ct, control; IR, irinotecan; p-mTOR, phosphorylated mTOR. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; ns, not significant.

Late autophagy depended on JNK1-induced Beclin-1 activation

Interestingly, we found that the JNK1 pathway was activated 2.4- and 5.8-fold (p-JNK1/JNK1) in MIA PaCa-2 and SU.86.86 cells after 72 hours of irinotecan treatment, respectively (Fig. 2A). JNK1 activity increases even in the presence of DNA damage, leading to cell death. However, JNK1 has also been shown to induce autophagy during serum starvation (30). If JNK1 mediates the activation of late autophagy, it can also mediate mTOR activation by late FAO. To assess this possibility, after treating MIA PaCa-2 cells with irinotecan for 48 hours, ULK1 or JNK1 inhibitors were coadministered with irinotecan for an additional 24 hours (Fig. 2B). The inhibitory activity of the ULK1 inhibitor XST-14 (5 μmol/L) was validated by examining Beclin-1 Ser15 phosphorylation under EBSS-induced starvation conditions (Supplementary Fig. S2A). As a result, p-mTOR increased by 60% compared with the untreated control after 72 hours of irinotecan treatment. However, when an ULK1 inhibitor or a JNK1 inhibitor was coadministered for 24 hours after the initial 48 hours of irinotecan treatment, p-mTOR activation decreased by 25% and 50% compared with the irinotecan group, respectively (Fig. 2B). Furthermore, in the group treated with irinotecan plus both the ULK1 inhibitor and the JNK1 inhibitor, p-mTOR decreased by 50% relative to the untreated control. These results suggest that ATP supply depends more on JNK1 activation during the late stage and that inhibiting JNK1 triggers p-AMPK activation (Fig. 2B). After 72 hours of irinotecan treatment, LC3-II increased 4.3-fold compared with the control (Fig. 2B). After 48 hours of irinotecan treatment, coadministration of ULK1 or JNK1 inhibitors with irinotecan for 24 hours resulted in p-Beclin-1 levels being decreased by 24% or 89%, respectively, compared with the irinotecan-alone group, whereas an LC3-II increase similar to that in the irinotecan-only group (Fig. 2B). However, after 48 hours of irinotecan treatment, simultaneous cotreatment of both ULK1 and JNK1 inhibitors with irinotecan reduced LC3-II by more than 63% compared with the irinotecan group (Fig. 2B). This demonstrates that early and late autophagy induced by irinotecan are significantly dependent on the ULK1 and JNK1 pathways. We measured whether autophagy remained continuously activated and ATP supply persisted even when p-AMPK was inactivated (Fig. 2B). After 72 hours of irinotecan treatment, ATP levels increased by approximately 30% compared with control. Furthermore, when ULK1 and JNK1 inhibitors were each administered for 24 hours after 48 hours of irinotecan treatment, ATP levels decreased slightly compared with the irinotecan-only group (Fig. 2B). When irinotecan was coadministered with both ULK1 and JNK1 inhibitors, ATP production decreased by 10% compared with the control. These results are consistent with the observation that blocking both autophagy pathways via ULK1 and JNK1 inhibition, alongside irinotecan treatment, resulted in a 3.7-fold increase in p62 and a 31% reduction in ATP synthesis relative to irinotecan-alone treatment (Fig. 2B). To verify whether the JNK1–Beclin-1 pathway mediates late-stage autophagy, we treated cells with irinotecan over a series of time points and detected Beclin-1 and Bax via immunoblotting after IP with Bcl-2 (Fig. 2C). When cells were treated with irinotecan for 72 hours, Bax remained completely bound to Bcl-2, whereas Beclin-1 dissociated from Bcl-2 by 80% and 90% in MiaPaCa-2 and SU 86.86, respectively (Fig. 2C). These results align with a 5.5- and 4.4-fold increase in S70 phosphorylation of Bcl-2 in MiaPaCa-2 and SU 86.86, respectively (Fig. 2C). We examined whether ULK1 or JNK1 inhibitors regulate autophagy in the absence of irinotecan. MIA PaCa-2 pancreatic cancer cells were treated for 24 hours with the ULK1 inhibitor and/or the JNK1 inhibitor, either as single agents or in combination. In MIA PaCa-2 cells without irinotecan treatment, inhibition of ULK1 and/or JNK1 does not regulate autophagy (Supplementary Fig. S2B).

Figure 2.

Autophagy depends on JNK1-induced Beclin-1 activation. A, Pancreatic cancer cell lines MIA PaCa-2 and SU.86.86 were treated with IR (1 μmol/L) for immunoblotting. B, Immunoblot analysis of late-stage autophagy in MIA PaCa-2. ULK1 inhibitor (ULK1i, XST-14, 5 μmol/L; ref. 28) and JNK1 inhibitor (JNK1i, SP600125, 20 μmol/L; refs. 29, 30) were administered individually or in combination 48 hours after IR treatment and were maintained for an additional 24 hours. Cell lysates were collected at the 72-hour time point for immunoblotting and ATP assay. Data are shown as the mean ± SD from n = 3 independent experiments. C, co-IP analysis of the Bcl-2 complex in MIA PaCa-2 and SU.86.86 cells treated with IR (1 μmol/L) for 0, 24, 48, and 72 hours. Cell lysates were immunoprecipitated with an anti–Bcl-2 antibody (IP: Bcl-2), and Bcl-2-bound Beclin-1 and Bax were detected by immunoblotting (IB: Beclin-1, Bax). Quantification represents Beclin-1 coimmunoprecipitated with Bcl-2, normalized to immunoprecipitated Bcl-2 (Beclin-1/Bcl-2), and expressed relative to 0 hours (set to 1). Whole-cell lysates (input) from the same samples were analyzed by immunoblotting to determine p-Bcl-2 levels and total Bcl-2, as indicated. Quantification represents p-Bcl-2/total Bcl-2, expressed relative to 0 hours (set to 1). Data are shown as the mean ± SD from n = 3 independent experiments. Statistical comparisons vs. 0 hours were performed using one-way ANOVA. co-IP, coimmunoprecipitation; Ct, control; IR, irinotecan. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

Irinotecan treatment combined with FAO inhibition by ACAA1 or CAC knockdown caused significant cell death through mTOR inactivation via ATP depletion in PDAC cells

To determine whether FAO upregulation during irinotecan treatment drives mTOR activation through increased ATP production, FAO genes were silenced for 24 hours, and irinotecan was treated for 0 to 72 hours. Previous studies have shown that various fatty acids are broken down into appropriately sized acylcarnitines via peroxisomal FAO, which are then transported to the mitochondria via CAC and used to synthesize ATP (36). Therefore, blocking two FAO targets, ACAA1 in peroxisomes and CAC in mitochondria, was considered an ideal strategy to effectively inhibit FAO in PDAC (Fig. 3A). ATP production in the irinotecan treatment group increased by 30% compared with that in the control group (Fig. 3B). When irinotecan was administered together with 20- and 40-nmol/L siRNAs for the double knockdown of CAC and ACAA1, ATP levels decreased by 37% and 56% in Mia PaCa-2 and by 31% and 51% in SU.86.86, respectively, compared with that of irinotecan alone (Fig. 3B). We examined how the reduction in ATP levels due to the double knockdown of CAC and ACAA1 affected the AMPK and mTOR pathways. When PDAC cells were treated with irinotecan alone for 72 hours, p-mTOR was activated two- to four-fold. In contrast, treatment with irinotecan, combined with the double knockdown of CAC and ACAA1, reduced p-mTOR activity by 75% and 79% in MIA PaCa-2 and SU86.86, respectively, compared with irinotecan alone (Fig. 3C). Meanwhile, 10% to 30% of p-AMPK was inactivated after 72 hours of irinotecan treatment. With combination therapy involving irinotecan and double knockdown of CAC and ACAA1, p-AMPK was activated approximately 2.4- and 2.1-fold compared with irinotecan alone in MIA PaCa-2 and SU.86.86, respectively. However, LC3-II levels in combination therapy involving irinotecan and double knockdown of CAC and ACAA1 remained almost the same as in irinotecan monotherapy (Fig. 3C). Additionally, to evaluate the anticancer effects of FAO inhibition, cell death was assessed by Annexin V/PI staining with double knockdown of CAC and ACAA1. Cells were treated with siCAC and siACAA1 for 72 hours simultaneously with irinotecan. Apoptosis was quantified by flow cytometry. The results showed that irinotecan alone caused little change in cell death, whereas its combination with FAO inhibition increased cell death by 2.5 to 8 times in SU.86.86 and MIA PaCa-2 cells, respectively (Fig. 3D).

Figure 3.

IR treatment with double knockdown of ACAA1 or CAC causes significant cell death. A, Cooperative relationship between mitochondria and peroxisomes in FAO and a schematic diagram of this process. B, MIA PaCa-2 and SU.86.86 cells were treated with vehicle (control, white), IR (1 μmol/L, orange), or IR combined with dual siRNA knockdown of CAC and ACAA1 (siCAC + siACAA1) at a total siRNA concentration of 20 nmol/L (sky blue) or 40 nmol/L (blue). Cells were harvested, and intracellular ATP levels were measured and expressed relative to the control (set to 1). C, Immunoblot analysis of mTOR and AMPK signaling and autophagy marker LC3-I/II (LC3-I/II) and p62 (SQSTM1) following the indicated treatments. Phosphorylated and/or total forms were analyzed as indicated. D, Representative flow cytometry dot plots of Annexin V–FITC/PI staining for cells treated with vehicle (control), IR (1 μmol/L), or IR combined with dual knockdown of CAC and ACAA1 (siCAC + siACAA1, 20 nmol/L). Quadrants were defined as indicated in the plots: Q4, viable (Annexin V⁻/PI⁻); Q3, early apoptotic (Annexin V⁺/PI⁻); Q2, late apoptotic/secondary necrotic (Annexin V⁺/PI⁺); and Q1, necrotic (Annexin V⁻/PI⁺). The total cell death means Q1 + Q2 + Q3. All graphs are shown as the mean ± SD from n = 3 independent experiments. Statistical comparisons with 0 hours were performed using one-way ANOVA. Ct, control; IR, irinotecan; PI, propidium iodide. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

Irinotecan treatment combined with FAO inhibitors of KN510713 induced significant cell death in PDAC cells

Previously, we reported that knocking down CAC (36) or ACAA1 (37) induces p-mTOR inactivation and cyclin D1 depletion by decreasing ATP production, thereby reducing tumor growth. We tested whether the same effect could be achieved using FAO inhibitors, such as KN510713 (a combination of KN510 and KN713), which target CAC and ACAA1, respectively. KN510 is omeprazole, which potently inhibits CAC by targeting the active site cysteine (23, 38). KN713 is trimetazidine, which competitively inhibits ACAA1 with various substrates, including short- to long-chain fatty acids (26). Because KN510 and KN713 are noncytotoxic drugs, the inhibitory effect of KN510713 on in vitro proliferation was redefined in several analyses (Supplementary Fig. S3; Supplementary Tables S1–S4). To confirm whether KN510713 inhibits FAO, we treated MIA PaCa-2 cells with KN510713 for 6 hours and performed metabolic profiling (Fig. 4A). Metabolite levels were measured and compared with those in the control. After treatment with KN510713, SC-, MC-, and LC-acylcarnitines from C0 (carnitine) to C18:1 (oleic acid) increased up to nine-fold compared with the control. However, changes in fatty acid levels showed a slight increase with increasing carbon chain length, but these differences were not statistically significant (Fig. 4A; Supplementary Fig. S4A). To assess whether FAO inhibition using KN510713 with irinotecan treatment increased cell death within 72 hours, we used high concentrations of KN510 and KN713, which did not cause toxicity when normal HPNE cells were treated with the same concentration of KN510713 for 72 hours (Supplementary Fig. S4B). The same KN510713 concentration showed no change in ATP levels in HPNE (Supplementary Fig. S4C). We examined whether the increased cell proliferation observed during irinotecan treatment was reduced by cotreatment with KN510713. We analyzed time-dependent changes in intracellular protein expression induced by irinotecan and KN510713 cotreatment using the same setup as shown in Figs. 1C and 4B. In MIA PaCa-2 and SU.86.86 cells, combination therapy resulted in time-dependent decreases in p-mTOR at 72 hours reduced by 79% and 90%, respectively, compared with the control (Fig. 4B; Supplementary Fig. S4D). In MIA PaCa-2 and SU.86.86 cells, combination therapy resulted in time-dependent decreases in cyclin D1 at 72 hours which reduced by 90% in both compared with the control (Fig. 4B; Supplementary Fig. S4D). We tested whether the FAO inhibitor KN510713 could reverse irinotecan-induced p-AMPK inactivation. The combination of irinotecan and KN510713 increased p-AMPK (p-AMPK Thr172) by three- and two-fold in MIA PaCa-2 cells and SU.86.86, respectively, compared with irinotecan alone (Fig. 4C; Supplementary Fig. S4E). The combination of irinotecan and KN510713 also increased LC3-II by 4.3-fold in both MIA PaCa-2 cells and SU.86.86.86. The combination of irinotecan and KN510713 increased p62 levels by 4.5-fold in both MIA PaCa-2 and SU.86.86, compared with irinotecan alone. This suggests that KN510713 treatment impairs autophagy flux in the irinotecan-treated group. Irinotecan treatment combined with KN510713 resulted in a 4.9- and 4.3-fold increase in cell death in MIA PaCa-2 and SU.86.86, respectively, relative to the control (Supplementary Fig. S4F). Because KN510713 treatment caused the accumulation of both LC3-I and LC3-II, indicating possible changes in autophagic flux, we performed an mCherry–GFP–LC3 reporter assay to evaluate autophagic flux during KN510713 treatment (Fig. 4D). Autophagosomes were identified as positive spots for both mCherry (red) and GFP (green), whereas autolysosomes only showed mCherry positivity. In the irinotecan monotherapy group, the changes in the number of autophagosomes were minimal, but the number of autolysosomes increased 2.9-fold, indicating active autophagy was occurring (Fig. 4D). In the group treated with KN510713 and irinotecan for 72 hours, the number of autophagosomes increased by 6.64-fold, whereas the number of autolysosomes decreased to 28% of the control (Fig. 4D). In MIA PaCa-2 cells, irinotecan treatment resulted in a time-dependent increase in ATP production, reaching a 22% increase at 72 hours. In contrast, treatment with a combination of irinotecan with KN510713 resulted in a 50% reduction in ATP production (Supplementary Fig. S4G). This indicates that inhibition of ATP production by FAO prevents autophagosome–lysosome fusion, impairing autophagolysis. We measured colocalization of LC3-II with BODIPY to test whether irinotecan treatment increases autophagy and lipophagy, accompanied by increased FAO flux. To examine whether impaired autophagosome–lysosome flux is associated with fatty acid accumulation caused by FAO inhibition, experiments were performed after treatment with the fluorescently labeled fatty acid BODIPY–FL-C16 (palmitate analog; green; Fig. 4E; ref. 39). The upper graph shows the average fluorescence intensity of lipid droplets (green in Fig. 4E). Compared with the control, lipid droplet levels increased by approximately 1.4-, 2.1-, and 2.4-fold with irinotecan monotherapy, KN510713, and the combination of irinotecan and KN510713, respectively. The lower graph shows the average fluorescence intensity of the LC3-II signal (red), colocalized with the BODIPY-positive region, relative to the control (yellow in Fig. 4E). The ratio of LC3-II colocalizing with BODIPY-positive lipid droplets increased by 1.4-, 1.9-, and 2.2-fold after irinotecan monotherapy, KN510713, and the combination treatment, respectively (yellow in Fig. 4E). These results suggest that although lipophagy and FAO are actively promoted in response to irinotecan treatment, the combination of irinotecan and KN510713 induces an overall functional impairment in autophagy flux due to fatty acid accumulation caused by FAO inhibition.

Figure 4.

Cotreatment with IR and KN510713 (a combination of KN510 targeting CAC and KN713 targeting ACAA1/2) inactivates mTOR and blocks autophagic flux. A, Target engagement of KN510713 in MIA PaCa-2 cells assessed by acylcarnitine profiling. Cells were incubated for 6 hours with vehicle (control; white bars) or KN510713 (KN510 200 μmol/L; KN713 2 mmol/L), and intracellular free carnitine (C0) and short- to long-chain acylcarnitines (C2–C18:1) were quantified by LC–MS/MS. Acylcarnitine species are shown in orange. FFAs, if measured in parallel, are shown in blue and were annotated as a separate metabolite class in the panel. Data are presented as the mean ± SD from n = 3 independent experiments. Statistical significance was assessed using the Mann–Whitney U test, and multiple unpaired two-tailed t tests were additionally performed as indicated in the graph. B, MIA PaCa-2 and SU.86.86 cells were cotreated with IR (1 μmol/L) and KN510713 (KN510 200 μmol/L; KN713 2 mmol/L) and harvested for immunoblotting. p-mTOR (Ser2448), total mTOR, and the downstream cell-cycle regulator cyclin D1 were analyzed. C, MIA PaCa-2 and SU.86.86 cells were exposed for 72 hours to IR (1 μmol/L) and/or KN510713 (KN510 200 μmol/L; KN713, 2 mmol/L) as indicated (±) and analyzed by immunoblotting for p-AMPK (Thr172), total AMPK, p-mTOR (ser2448), total mTOR, p62 (SQSTM1), and LC3-I/II. D, Fluorescence microscopy of autophagic flux in MIA PaCa-2 cells stably expressing mCherry–GFP–LC3. Cells were treated with IR (1 μmol/L) alone (orange), with KN510713 alone (green), or with the combination of IR and KN510713 (blue). Representative images show autophagosomes (GFP⁺/mCherry⁺; yellow puncta) and autolysosomes (GFP⁻/mCherry⁺; red-only puncta). Autophagic flux was quantified as the number (or fraction) of red-only puncta and/or the red-only/yellow puncta ratio per cell, as specified in the graph. Autophagosomes were identified as positive spots for both mCherry (red) and GFP (green). RAP, which inhibits mTOR–ULK1 signaling, is used as a positive control to induce autophagy (33). Treatment with RAP decreases autophagosomes (green) and increases autolysosomes (red). CQ, which inhibits autophagolysosomal fusion, is used as an autophagy blocker in negative controls (34). Treatment with CQ blocks autolysis, thereby increasing the number of autophagosomes (green) and autolysosomes (red). Statistical significance was assessed using one-way ANOVA (n = 6). Scale bar, 25 μm. E, Lipid droplets and autophagosomes were costained, and their colocalization was analyzed. Lipid droplets were labeled with BODIPY FL C16 (BODIPY). After 48 hours of treatment, cells were fixed and immunostained for LC3-II to visualize autophagosomes, with Hoechst 33342 used for nuclear counterstaining. The top graph shows the mean fluorescence intensity of BODIPY staining alone, expressed as relative values normalized to the control group. The bottom graph depicts the mean fluorescence intensity of LC3-II signals colocalized with BODIPY-positive spots, also expressed as relative values normalized to the control group. Statistical significance was assessed using one-way ANOVA (n = 3). Scale bar, 25 μm. CQ, chloroquine; Ct, control; FFA, free fatty acid; IR, irinotecan; KN, KN510 + KN713; RAP, rapamycin. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

The combination of irinotecan and FAO inhibitors disrupted autophagy, leading to increased ROS and inducing cell death

Based on the results shown in Fig. 4, we hypothesized that the combination of irinotecan and KN510713 induces autophagic cell death by inhibiting autolysosomal formation, thereby leading to the accumulation of autophagosomes and toxicity. To test this, we treated cells with various concentrations of bafilomycin A1 (Baf.A1), an inhibitor that blocks the fusion of autophagosomes with lysosomes, causing autophagosome accumulation (Fig. 5A). We investigated whether irinotecan and KN510713 induce cell death through the same pathway when combined with Baf.A1 treatment. Baf.A1 reduced cell survival in a dose-dependent manner in the MIA PaCa-2 cell line, and this reduction was further enhanced in a dose-dependent manner by the combination of irinotecan and KN510713 (Fig. 5A). This suggests that the combination of irinotecan and KN510713 induces cell death by impairing autophagy and disrupting autophagy-related degradation pathways but through a mechanism distinct from autophagy inhibition by Baf.A1. To evaluate how effectively KN510713 suppresses tumor regrowth and boosts anticancer activity, we performed Annexin V/PI staining (Fig. 5B). In the irinotecan group, MIA PaCa-2 cells showed no significant difference compared with the control, whereas SU.86.86 cells demonstrated 1.5 times more cell death than the control (Fig. 5B). Combination treatment increased cell death by 2.7- and 3.5-fold over the control in MIA PaCa-2 and SU.86.86 cells, respectively (Fig. 5B). The halt in autophagolysis upon FAO inhibition is due to fatty acid toxicity. As fatty acids accumulate and the pH drops below the optimal range (pH 4.5–5), the stability of lysosomal membrane proteins and lipids decreases without any gain in protease efficiency (40). This leads to membrane damage, accelerating lysosomal membrane permeabilization and increasing ROS. Therefore, we added NAC rescue experiments alongside ROS measurements, as described below. As a result, no significant change in ROS levels was observed in either MIA PaCa-2 or SU.86.86 cells treated with irinotecan alone compared with the control group. However, when irinotecan was combined with KN510713, ROS levels increased by approximately 1.34- and 2.1-fold in MIA PaCa-2 and SU.86.86 cells, respectively, compared with irinotecan monotherapy (Fig. 5C). Furthermore, to confirm whether the cell death induced by ROS is accurate, NAC treatment reduced cell death by 28% to 40% in the combination treatment group compared with the irinotecan monotherapy group (Fig. 5D). The mechanism of cell death induced by the combination treatment of irinotecan and KN510713 in PDAC cells is ultimately attributable to increased ROS.

Figure 5.

IR with FAO inhibition reverses tumor promotion caused by IR treatment. A, Representative phase-contrast microscopy images of MIA PaCa-2 cells treated for 72 hours with IR (1 μmol/L), KN510713 (KN510, 200 μmol/L; KN713, 2 mmol/L), and/or Baf.A1 (1, 5, or 10 nmol/L), as indicated. All compounds were added simultaneously at the start of treatment. Baf.A1 was used as a late-stage autophagy inhibitor. B, Representative flow cytometry dot plots of Annexin V–FITC/PI staining from cells treated with vehicle (control), IR, or IR plus KN510713. Total cell death was calculated as Q1 + Q2 + Q3 and is presented as the mean ± SD from n = 3 independent experiments. Statistical significance was assessed using one-way ANOVA. C, ROS levels were analyzed by flow cytometry using DCFDA staining after 72 hours of treatment with vehicle (control), IR, or IR plus KN510713 in MIA PaCa-2 and SU.86.86 cells (n = 3). Fluorescence intensity was quantified as the geometric mean in FlowJo software (version 10.10.0). D, MIA PaCa-2 and SU.86.86 cells were pretreated for 1 hour with or without 5 mmol/L of NAC and then treated with the combination of IR and KN510713 for 72 hours. Cell death (red box) was measured by Annexin V–FITC and PI staining and analyzed by flow cytometry (n = 3). All apoptotic cell death rates (Annexin V–positive, PI-positive, and double-positive) were quantified using FlowJo software (version 10.10.0). Ct, control; IR, irinotecan; KN, KN510 + KN713; NAC, N-acetyl-L-cysteine; PI, propidium iodide. *, P < 0.05; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

Treatment with irinotecan and FAO inhibitors improved survival in the KPC spontaneous PDAC mouse model

To confirm that inhibiting FAO can significantly improve therapeutic efficacy even in spontaneous PDAC mouse models, overall survival was analyzed in KPC mice treated with irinotecan and the FAO inhibitor KN510713, either alone or in combination (Fig. 6A). Optimal doses and combination ratios of KN510 and KN713 were established using xenograft models (Supplementary Fig. S5). Both KN510 and KN713 achieved saturation at 25 mg/kg (Supplementary Fig. S5A and S5B). Additionally, the combination of KN510 and KN713 at a 1:1 ratio of 25 mg/kg demonstrated saturation of the synergistic effects (Supplementary Fig. S5C). Consequently, administration of KN510 and KN713 at a 1:1 ratio (KN510:KN713) at 25 mg/kg was deemed appropriate (Supplementary Fig. S5D and S5E). The irinotecan and KN510713 combination groups received 40 mg/kg of irinotecan once weekly and 25 mg/kg of KN510713, starting at 8 weeks of age. We used a 5-day-on/2-day-off schedule (Fig. 6A). The Kaplan–Meier analysis showed that the MS of the control group was 16.4 weeks. The MS increased by 3 weeks in the irinotecan monotherapy group and by 6.2 weeks in the combination therapy group with irinotecan and KN510713 compared with the control group, respectively (Fig. 6A; Supplementary Table S5). During the combination treatment period, mice exhibited no abnormalities, including changes in body weight. Separating KPC tissue to isolate normal and cancerous tissue for metabolite analysis is technically impossible. Therefore, IHC for p-mTOR, Ki-67, p-AMPK, LC3-II, and γ-H2AX was performed on pancreatic tissue from mice at 20 weeks (control), 21 weeks (irinotecan), 21 weeks (KN510713), and 22 weeks (combination; Fig. 6B; Supplementary Fig. S6). p-mTOR, associated with cell survival and growth, showed changes in its levels consistent with the changes in Ki-67 levels, whereas γ-H2AX, associated with cell death, showed changes in its levels consistent with the changes in p-AMPK and LC3-II (Fig. 6B; Supplementary Fig. S6). γ-H2AX, an indicator of DNA double-strand breaks, increased by 52% in the KN510713 monotherapy group and by 311% in the KN510713 and irinotecan combination therapy group compared with the control group (Fig. 6B). Conversely, Ki-67, a cell proliferation marker, expression decreased by 26% in the irinotecan monotherapy group and by 66% in the KN510713 and irinotecan combination therapy group compared with the control group. In irinotecan monotherapy, cell death increased by 52% relative to the control group, whereas p-mTOR, a marker of cell growth, increased four-fold (Fig. 6B; Supplementary Fig. S6). However, in the KN510713 and irinotecan combination treatment group, cell death increased three-fold, whereas p-mTOR, which is essential for growth, decreased to the level observed in the control group (Fig. 6B; Supplementary Fig. S6). These findings suggest that whereas irinotecan monotreatment leads to tumor regrowth, the combination of irinotecan and KN510713 blocks this regrowth. Previously, we demonstrated reduced tumor growth by crossing KPC mice with Cac-knockout mice (36) or Acaa1a-knockout mice (37). We also confirmed reduced tumor growth by knocking down the FAO genes CAC (36) or ACAA1 in human PDAC cells and using xenograft studies (37). Additionally, in the mouse xenograft model using SU.86.86 with CAC knockdown, we observed reduced acylcarnitine levels and approximately one-third decrease in ATP levels in isolated tumors (36). Therefore, the experimental results from KPC mice align with the metabolite analysis in the xenograft model.

Figure 6.

IR treatment combined with FAO inhibition increased OS in the KPC model. A, KPC mice were randomized into four groups: control (black, n = 27), IR (orange, n = 20), KN510713 (green, n = 27), and KN510713 plus irinotecan (blue, n = 21). Beginning at 8 weeks of age, KN510713 (KN510, 25 mg/kg; KN713, 25 mg/kg) was administered daily by oral gavage, and IR (40 mg/kg) was administered once weekly by intraperitoneal injection. OS was evaluated by Kaplan–Meier analysis, and $P$ values were calculated as indicated in Supplementary Table S5 using the log-rank (Mantel–Cox) test. MS and OS metrics are summarized in the accompanying supplementary table. B, IHC was performed on pancreatic tumor tissues collected from treated KPC mice (n = per group): vehicle, IR, KN510713, or IR plus KN510713. Representative images were obtained from mice at 20 weeks (control), 21 weeks (irinotecan), 21 weeks (KN510713), and 22 weeks (combination). Sections were stained for p-mTOR (1:25), p-AMPK (1:300), LC3-II (1:750), γ-H2AX (1:1,000), and Ki-67 (1:1,000). Staining intensity was quantified using H-scores. Statistical significance was assessed using one-way ANOVA for p-AMPK, γ-H2AX, Ki-67, and the Kruskal–Wallis test for p-mTOR and LC3-II (nonnormally distributed data). C,In vivo treatment schedule for the MIA PaCa-2 xenograft model. Mice were randomized when tumor volumes reached ∼300 mm³. Each treatment group included seven mice (n = 7 per group). KN510713 was administered orally at 25 mg/kg, 5 days per week, and IR was administered intraperitoneally at 40 mg/kg, once per week. Tumor volume was calculated using V = (A × B²)/2, in which A is the longest diameter and B is the shortest diameter (mm). Tumor growth curves in (C) are presented as the mean ± SEM, and statistical significance was assessed using the Kruskal–Wallis test, as indicated in the panel. D and E, At 3 weeks of treatment, tumors were excised, and tumor lysates were analyzed for acetyl-CoA; (D) and ATP (E). Data presentation and group sizes (n) are provided in the graphs. Data are shown as the mean ± SD from n = 3 independent experiments. Statistical comparisons vs. 0 hours were performed using one-way ANOVA. Scale bar, 100 μm. acetyl-CoA, acetyl-coenzyme A; Ct, control; IR, irinotecan; KN, KN510 + KN713; OS, overall survival. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, not significant.

To determine whether combining irinotecan with FAO inhibitors produces anticancer effects in a xenograft model similar to those observed in the KPC model, we developed a xenograft mouse model by implanting MIA PaCa-2 cells into BALB/c nude mice. This study aimed to supplement the limited efficacy of irinotecan alone and test whether FAO inhibition could more effectively inhibit tumor growth. By week 6, the tumor size decreased by 34% in the irinotecan 40-mg/kg monotherapy group compared with controls and by 39% in the 25-mg/kg monotherapy group of KN510713 (Fig. 6C). In contrast, in the irinotecan and KN510713 combination group, the number of mice with residual tumors decreased by 89% compared with the controls, and tumor regrowth was halted entirely. To determine whether the significant tumor growth suppression observed with the combined treatment of irinotecan and KN510713 resulted from FAO inhibition, tumors were excised from mouse models at week 3 of treatment, and acetyl-CoA and ATP levels were measured in the tissues (Fig. 6D and E). In the irinotecan-only group, both acetyl-CoA and ATP levels were approximately 20% higher than those in the controls. However, in the group treated with KN510713 alongside irinotecan, acetyl-CoA and ATP levels were approximately 50% lower than those in the irinotecan-only group. This confirmed that tumor regrowth suppression was due to FAO inhibition by KN510713 (Fig. 6D and E).

Discussion

Irinotecan, a topoisomerase 1 inhibitor, shows clear potential as an anticancer drug. However, similar to other chemotherapeutic drugs, irinotecan resistance can develop. In cases of drug resistance, cancer cells often exhibit coactivation of the mTOR and autophagy pathways. This is paradoxical because the typical mTOR activation pathway usually inhibits autophagy. The three primary pathways can activate mTOR. First, receptor tyrosine kinases such as EGFR, PDGFR, FGFR, and IGF-1R are frequently abnormally activated, initiating various cytoplasmic kinase signaling pathways that trigger mTOR activation (41, 42). Second, mTOR can be triggered by increased nutrient levels. The LAMTOR pathway activates mTOR in response to increased essential amino acid levels, thereby promoting ribosome biogenesis and cell growth (43). Third, the mTOR pathway is regulated by intracellular ATP concentration, regardless of amino acid levels, because mTOR itself acts as an ATP sensor (44). The biochemical rationale for how ATP levels control mTOR can be explained by its high Km value (1 mmol/L for ATP; ref. 45). Because of the high level of ATP in the cytoplasm of cancer cells, even small changes in ATP levels can influence mTOR activity. FAO triggers mTOR activation and inactivates AMPK by lowering the AMP/ATP ratio. We have previously demonstrated that inhibiting FAO significantly reduces ATP production, accounting for more than 50% of the total cellular ATP (16, 17). Additionally, CAC knockdown reduces the total ATP production by 50% (36). Activated mTOR, resulting from increased abnormal signaling or ATP production, enhances translation and ribosome biogenesis, while suppressing autophagy. This indicates that mTOR regulates the rate of protein synthesis in response to translational precursor availability and ATP levels, independent of signal transduction (40). However, when tumors are treated with anticancer drugs that inhibit signaling or anabolism, they develop resistance through mTOR activation. Drug resistance has been observed in cancers (46) with mTOR activation (47), including vincristine resistance (48), trastuzumab resistance (49), cisplatin resistance (50), erlotinib resistance (51), and etoposide and doxorubicin resistance (52). Resistance to anticancer drugs is often attributed to a simple metabolic response in living cells. When anabolic metabolism is blocked, cancer cells detect nutrient deprivation and activate FAO, an alternative metabolic pathway to starvation, to produce ATP and survive. This phenomenon has been observed in resistance to anticancer drugs.

In late autophagy, irinotecan-induced FAO directly influences AMPK inactivation by increasing ATP levels. The regulatory mechanism of the AMPK pathway as a sensor of changes in intracellular AMP and ATP levels is well established (53). The primary upstream kinase that activates AMPK during energy stress is LKB1 (54, 55). In this study, a significant increase in ATP levels, induced by irinotecan treatment, inactivated AMPK in cancer cells by decreasing the AMP/ATP ratio. However, an interesting finding was that autophagy remained active, whereas AMPK was inactivated after long-term irinotecan treatment. Therefore, we refer to this as “the late autophagy.” We need to consider the cytotoxic signaling pathway triggered by irinotecan, which induces late autophagy. The active metabolite of irinotecan, SN-38, is a potent agent that causes DNA damage and cell death (56). The activity of c-Jun N-terminal kinase1 (JNK1) was initially identified as the kinase activity of c-Jun triggered by ultraviolet light and oncogenes. However, it was later shown that JNK1 activity is induced by DNA damage (57). DNA damage induced by irinotecan treatment activated JNK1, which accumulated gradually over time. This increased JNK1 activity phosphorylates Bcl-2 (58), which influences cell fate by regulating autophagy and apoptosis (59). Phosphorylation of Bcl-2 by JNK1 disrupts Bcl-2–Bax protein interaction, leading to Bax translocation to the mitochondria and subsequent apoptosis (58, 59), or interferes with Bcl-2–Beclin-1 protein binding, thereby promoting Beclin-1’s role in autophagy (60).

Based on the results from autophagy flux, BODIPY-C16, and ROS, the following conclusions can be proposed. Cancer cells inherently prefer fatty acids as an energy source (16). Irinotecan treatment increases FAO flux via lipophagy, whereas a combination of irinotecan and KN510713 accumulates fatty acids in cancer cells. Excessive fatty acid accumulation halts the autophagolysis process (40), and membrane stress increases ROS (61). DNA damage from irinotecan and ROS toxicity from FAO inhibition powerfully drive cancer cells to death. This suggests that it could be a practical therapeutic approach for selectively killing cancer cells, differentiating them from normal cells that primarily use glucose as their energy source (62).

In this study, we identified three key findings. First, early autophagy induced by anticancer agents promotes FAO, which activates mTOR and inactivates AMPK (Fig. 7A). Second, late autophagy induced by these agents activates the JNK1–Beclin-1 pathway, shifting the cell death process toward autophagy (Fig. 7B). Ultimately, this concurrent activation seems to occur through parallel signaling pathways: mTOR activation via FAO-induced ATP accumulation and autophagy maintenance via the JNK1–Beclin-1 pathway, bypassing canonical nutrient-sensing pathways. Third, anticancer drugs combined with FAO-inhibiting drugs induce cell death by inactivating mTOR survival signals, promoting autophagy flux but impairing autolysis due to fatty acid deposits (Fig. 7C). Therefore, inhibition of FAO suppresses acquired drug resistance, thereby preserving the inherent toxicity of anticancer drugs. Based on this new understanding, FAO inhibition enhances sensitivity to anticancer drugs such as irinotecan as well as to the FOLFIRINOX regimen (irinotecan, oxaliplatin, and 5-fluorouracil; ref. 63). To translate these findings, it is necessary to determine whether FAO inhibition has any side effects. Therefore, the safety of FAO inhibition using KN510713 was assessed in the phase 1 clinical trial (NCT06012708). No serious adverse events occurred; mild adverse events included flu-like symptoms, constipation, nausea, and dizziness (64). Potential systemic toxicity and effects on the heart and skeletal muscle were not observed. KN510713 showed excellent tolerability and safety up to a maximum dose of 120/120 mg/day in patients with advanced solid tumors and is recommended as the phase 2 dose (64). The reason fatty acid accumulation toxicity did not occur in the normal tissue is that during CAC knockdown, fatty acids are converted into acylcarnitine rather than accumulating. It is expected that this acylcarnitine then accumulates (36) and is subsequently excreted from the cells into blood vessels and from the blood vessels into urine. The kidney readily excretes acylcarnitine through glomerular filtration and, for shorter chains, may also secrete a minor amount via tubular secretion, whereas most free carnitine is reabsorbed (65).

Figure 7.

The development model of acquired drug resistance by drug treatment. A, When cancer cells are treated with cytotoxic anticancer drugs, they experience cellular stress such as DNA damage, which activates AMPK-dependent early autophagy. At this stage, cells with severe damage undergo autophagic cell death, whereas surviving cells supply FAs via autophagy. B, When cancer cells face sustained stress from anticancer drugs, FAO is activated, leading to increased ATP production. This increase in ATP inactivates AMPK-dependent autophagy and activates mTOR. However, cancer cells also activate JNK1-dependent late autophagy in response to cellular damage, maintaining mTOR activity through FAO. This process results in tumor regrowth during treatment with anticancer drugs. A key feature of this stage is that autophagy and mTOR, which normally inhibit each other via feedback mechanisms, are concurrently activated through independent pathways due to damage caused by anticancer drugs. C, Both AMPK-dependent and JNK1-dependent autophagy promote mTOR activation through FAO induction. Therefore, anticancer drugs combined with FAO-inhibiting drugs induce cell death by inactivating mTOR survival signals and promoting autophagy flux but impairing autolysis due to FA deposits. In this study, we demonstrated this concept by double knockdown of CAC and ACAA1, the main FAO pathways, or using KN510713, FAO inhibitors. FA, fatty acid.

Supplementary Material

Supplementary Table S1

IC50 values of KN510 and KN713 for Acetyl-CoA reduction in PDAC cell lines

Supplementary Table S2

IC50 values of KN510 and KN713 for ATP reduction in PDAC cell lines.

Supplementary Table S3

cEC50 values of KN510 and KN713 in PDAC cell lines

Supplementary Table S4

Colby analysis for the synergistic effect of KN510 and KN713 in PDAC cell lines

Supplementary Table S5

MS and overall survival (OS) of KPC mice treated with KN510713 and Irinotecan

Supplementary Figure S1

Irinotecan treatment increased autophagy simultaneously with an increase in FAO

Supplementary Figure S2

Validation of ULK1 inhibition and the effects of ULK1 and JNK1 inhibition on autophagy signaling

Supplementary Figure S3

FAO inhibitors suppress the growth of pancreatic cancer cells

Supplementary Figure S4

Irinotecan treatment with FAO inhibitors KN510713 induced significant cell death through mTOR inactivation

Supplementary Figure S5

Anti-tumor effect of KN510 and KN713 combination against a MIA PaCa-2 xenograft mouse model

Supplementary Figure S6

Immunohistochemical staining

Data Availability

The generated data are available upon request from the corresponding author.

Authors’ Disclosures

S.M. Woo reports grants from New Cancer Cure-Bio Co. outside the submitted work. J.H. Kang reports other support from New Cancer Cure-Bio Co. outside the submitted work. W. Choi reports other support from Bayer AG and personal fees from Boryung Pharmaceutical and Astellas outside the submitted work. H. Jang reports grants from the National Research Foundation during the conduct of the study. S.-Y. Kim reports grants, nonfinancial support, and other support from New Cancer Cure-Bio Co. during the conduct of the study, as well as other support from New Cancer Cure-Bio Co. outside the submitted work and a patent for Pharmaceutical composition for the prevention or treatment of cancer comprising a 3-ketoacyl-CoA thiolase inhibitor and a carnitine-acylcarnitine transporter inhibitor issued and licensed to New Cancer Cure-Bio Co. No disclosures were reported by the other authors.

Authors’ Contributions

S.M. Woo: Conceptualization, supervision, investigation, methodology, writing–review and editing. J.H. Kang: Data curation, formal analysis, validation, investigation, visualization, methodology. W. Choi: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–review and editing. H. Lee: Data curation, formal analysis, investigation, writing–review and editing. H. Jang: Data curation, formal analysis, validation, methodology. S.H. Sim: Supervision, funding acquisition, investigation, methodology, writing–review and editing. J.W. Chun: Supervision, funding acquisition, investigation, methodology, writing–review and editing. E.-B. Koh: Data curation, formal analysis, validation, investigation, visualization, methodology. C. Kim: Data curation, formal analysis, validation, investigation, visualization, methodology. W. Ham: Data curation, formal analysis, validation, investigation, visualization, methodology. W. Hong: Data curation, formal analysis, validation, investigation, visualization, methodology. M. Kang: Data curation, formal analysis, validation, investigation, visualization, methodology. J. Park: Formal analysis, validation, investigation, methodology. S. Han: Data curation, formal analysis, validation, methodology. J.W. Kim: Data curation, formal analysis, validation, methodology. E.-W. Lee: Data curation, formal analysis, validation, methodology. W.J. Lee: Supervision, writing–review and editing. S.-Y. Kim: Conceptualization, formal analysis, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.