DGAT1 Inhibition Induces Ferroptosis and Enhances Cancer Immunotherapy Efficacy

DGAT1 inhibition promotes ferroptosis by reducing lipid droplet accumulation, increasing lipid peroxidation, and inducing mitochondrial dysfunction, ultimately enhancing sensitivity to immune checkpoint blockade and offering a promising strategy for improving cancer treatment.

Abstract

Ferroptosis, a form of regulated cell death driven by lipid peroxidation, has emerged as a promising mechanism in cancer therapy. However, the lack of clinically viable ferroptosis inducers has precluded its therapeutic evaluation in patients. In this study, we demonstrated that inhibition of diacylglycerol O-acyltransferase 1 (DGAT1) induces a ferroptosis-like phenotype in cancer cells and enhances the efficacy of immune checkpoint blockade (ICB) therapy. In human cancer cohorts, low DGAT1 expression correlated with improved prognosis and elevated ferroptosis-associated gene signatures. In murine models, both genetic knockout and pharmacologic inhibition of DGAT1 enhanced ICB therapy efficacy by promoting increased infiltration of cytotoxic T lymphocytes. Mechanistically, DGAT1 inhibition reduced lipid droplet accumulation, triggering elevated lipid peroxidation, mitochondrial dysfunction, and reactive oxygen species production. These events culminated in glutathione peroxidase 4 depletion and ferroptosis. Given the availability of clinical-stage DGAT1 inhibitors, these findings provide a strong rationale for repurposing these agents as ferroptosis inducers to improve responses to cancer immunotherapy.

Significance:

DGAT1 inhibition promotes ferroptosis by reducing lipid droplet accumulation, increasing lipid peroxidation, and inducing mitochondrial dysfunction, ultimately enhancing sensitivity to immune checkpoint blockade and offering a promising strategy for improving cancer treatment.

Graphical Abstract

Introduction

Diacylglycerol O-acyltransferase 1 (DGAT1) catalyzes the final step of triglyceride (TG) synthesis by transferring an acyl group from acyl-CoA to diacylglycerol (DAG). This process facilitates the formation of lipid droplets (LD), intracellular organelles critical for energy storage and protection against the cytotoxic effects of free fatty acids and reactive oxygen species (ROS; ref. 1). By preserving mitochondrial integrity and cellular energy homeostasis (2), DGAT1 plays a central role in metabolic regulation.

In the gastrointestinal (GI) system, DGAT1 converts dietary lipids into TGs for incorporation into chylomicrons. Consequently, DGAT1 inhibition has been explored as a therapeutic approach for metabolic disorders such as obesity (3). Recent evidence highlights a role for DGAT1 in cancer metabolism, in which LD biogenesis helps buffer oxidative stress, maintain energy reserves, and support membrane synthesis, all of which are crucial for tumor survival and therapeutic resistance (4–7). These findings position DGAT1 as a potential therapeutic target in cancer.

Ferroptosis is a regulated form of cell death triggered by lipid peroxidation and iron accumulation (8). This process, governed by glutathione peroxidase 4 (GPX4), prevents excessive lipid peroxidation (9). Although some cancer cells are highly susceptible to ferroptosis due to their metabolic profiles, others develop mechanisms to evade it, promoting therapy resistance (10). Notably, ferroptosis is an immunogenic form of cell death that releases damage-associated molecular patterns, enhancing antitumor immunity. However, ferroptosis in immune cells, such as T cells, can suppress antitumor responses, underscoring the complex interplay between ferroptosis and cancer immunotherapy (11).

Despite its potential, ferroptosis has not been clinically evaluated due to the lack of viable inducers. Here, we demonstrate that DGAT1 inhibition induces ferroptosis and enhances ICB efficacy, establishing a foundation for the clinical evaluation of DGAT1 inhibitors as ferroptosis-inducing agents to boost cancer immunotherapy.

Materials and Methods

Patient clinical and whole-genome transcriptional data

We obtained clinical and whole-genome transcriptional data of patients with cancer from The Cancer Genome Atlas (TCGA) PanCancer cohort, accessible through the cBioPortal database (https://www.cbioportal.org; ref. 12). This extensive resource includes data from 32 studies covering approximately 11,000 patients with cancer across 33 cancer types within TCGA (13). Our analysis specifically focused on TCGA data related to skin cutaneous melanoma (SKCM), breast invasive carcinoma (BRCA), lung squamous cell carcinoma (LUSC), and colon adenocarcinoma within the study time frame (April 15, 2024–May 20, 2024). Additionally, we acquired transcriptome profiling data and clinical information for patients with SKCM who received ICB therapy from a previously published study (14).

Cell culture

We acquired HEK 293T (human embryonic kidney cells, ATCC, cat. #CRL-11268, RRID:CVCL_0063), B16F10 (mouse melanoma cells, ATCC, cat. #CRL-6475, RRID:CVCL_0159), MC38 (mouse colon adenocarcinoma cells, Kerafast, cat. #ENH204-FP, RRID:CVCL_B288), 4T1 (mouse breast carcinoma cells, ATCC, cat. #CRL-2539, RRID:CVCL_0125), LLC (mouse lung carcinoma cells, ATCC, cat. #CRL-1642, RRID:CVCL_4358), and A375 (human melanoma cells, ATCC, cat. #1619IG-2, RRID:CVCL_0132) from the Cell Culture Facility (CCF) at Duke University School of Medicine. The identity of the cells was validated by the CCF for A375 by PCR microsatellite assay but not for mouse tumor lines. All cells were cultured in high-glucose Dulbecco’s Modified Eagle Medium (Sigma-Aldrich), supplemented with 10% (v/v) FBS and 1% penicillin/streptomycin (Thermo Fisher Scientific). The cells were maintained at 37°C in a humidified atmosphere with 5% CO₂. Regular mycoplasma testing was conducted on all cell lines using the Universal Mycoplasma Detection Kit (ATCC).

CRISPR/Cas9-mediated gene knockout

We generated knockout (KO) cells using lentivirus-mediated CRISPR/Cas9 technology. Single-guide RNAs (sgRNA) were designed using the CHOPCHOP public CRISPR sgRNA design tool (https://chopchop.cbu.uib.no). Supplementary Table S1 lists the sgRNA sequences targeting both human and mouse DGAT1. Double-stranded oligos encoding the sgRNA sequences were initially cloned into the BsmBI-digested plasmid LentiCRISPRv2 (RRID: Addgene_52961), a vector enabling the coexpression of Cas9 and sgRNA. CRISPR lentivirus vectors were then produced in HEK 293T cells by cotransfecting the psPAX2 (RRID: Addgene_12660) and pMD2.G (RRID: Addgene_12259) plasmids with the sgRNA-encoding plasmid, following the established protocol of the Zhang lab (15). KO cell lines were generated by infecting target cells with the lentivirus and cultivating them in media containing 1 µg/mL puromycin for A375 and B16F10 cells and 5 µg/mL for MC38 and 4T1 cells. After a selection period of 10 to 14 days, cells were seeded into 96-well plates and screened for pure KO clones, confirmed by immunoblot analysis.

Western blot

Cell lysis was initially performed in RIPA buffer supplemented with a protease inhibitor cocktail (Sigma-Aldrich). Total protein concentrations in the lysates were then determined using a protein assay kit (Bio-Rad). Equal amounts of protein were denatured in sodium dodecyl sulfate (SDS) sample loading buffer (Bio-Rad) at 100°C for 15 minutes and subsequently loaded onto a 10% to 12% SDS-PAGE gel for electrophoresis. After electrophoresis, proteins were transferred to a methanol-activated polyvinylidene fluoride membrane (Millipore). The membrane was blocked in TBS with Tween 20 [TBST; 10 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 0.1% Tween 20] containing 5% BSA (MP Biomedicals) for 1 hour at room temperature. The membrane was then incubated with primary antibodies overnight at 4°C. Primary antibodies included those against DGAT1 (Proteintech, cat. #11561-1-AP, RRID:AB_2877779), GPX4 (Proteintech, cat. #67763-1-Ig, RRID:AB_2909469), and GAPDH (Proteintech, cat. #60004-1-Ig, RRID:AB_2107436). Following two washes with TBST, the membrane was incubated with the appropriate horseradish peroxidase (HRP)–labeled secondary antibody for 1 hour at room temperature. HRP-conjugated secondary antibodies were used, including goat anti-rabbit (Proteintech, cat. #SA00001-2, RRID:AB_2722564) or anti-mouse IgG (Proteintech, cat. #SA00001-1, RRID:AB_2722565; 1:5,000). Finally, immunoblots were visualized using the enhanced chemiluminescence detection system according to the manufacturer’s protocol.

Tumor growth delay studies

All animal experiments in this study were approved by the Duke University Institutional Animal Care and Use Committee. Six-week-old female C57BL/6J (RRID:IMSR_JAX:000664) and BALB/c mice (RRID:IMSR_JAX:000651) were obtained from The Jackson Laboratory. Prior to tumor cell injection, the mice’s right hind limbs were shaved. Approximately 1 × 10⁵ B16F10, 2 × 10⁵ LLC, 4T1, or 5 × 10⁵ MC38 cells were resuspended in 50 µL phosphate-buffered saline (PBS) and subcutaneously injected into the shaved hind limbs. Intraperitoneal injections of antibodies were administered with either 100 µg of IgG2 isotype control (Bio X Cell, cat. #BE0089, RRID:AB_1107769) or 100 µg of anti-mouse PD-1 (Bio X Cell, cat. #BP0146, RRID:AB_10949053) antibody in 100 µL PBS per mouse on days 7, 10, and 13 after tumor inoculation. For in vivo DGAT1 or DGAT2 inhibition, 8 mg/kg PF-04620110 (MCE, cat. #HY-13009), 2 mg/kg pradigastat (MCE, cat. #HY-16278), or 10 mg/kg PF-06424439 (MCE, cat. #HY-108341) in 100 µL PBS per mouse was injected intraperitoneally from day 6 to day 15 after tumor inoculation. In vivo alteration of GPX4 function was achieved by i.p. injection of either 100 mg/kg RSL3 (MCE, cat. #HY-100218A) or 5 mg/kg ferrostatin-1 (Fer-1; MCE, cat. #HY-100579) in 100 µL PBS per mouse on days 6, 9, 12, and 15 after tumor inoculation. Tumor volumes were measured every other day and calculated using the following formula: (Length) × (Width)²/2. Mice were euthanized when tumor volumes reached 2,000 mm³. Survival analysis among tumor-bearing mouse groups was conducted using the Kaplan–Meier method and log-rank (Mantel–Cox) tests. In the 4T1 mouse model, lung surface metastases were detected by staining specimens with Fekete’s solution, which enabled the counting of tumor nodules appearing white. Surface metastases were quantified, and differences between study groups were analyzed using unpaired t-test analysis. For the 4T1 orthotopic tumor model, harvested 4T1 cells were resuspended in cold serum-free RPMI 1640 and mixed with Matrigel matrix (Corning) at a 3:1 volume ratio (final density: 5 × 10⁵ cells/100 μL mixture). The cell–Matrigel suspension was maintained on ice throughout the injection procedure to prevent polymerization. Animals were anesthetized using inhaled isoflurane (3% for induction and 1.5%–2% for maintenance) delivered via a precision vaporizer, with medical oxygen as the carrier gas. Orthotopic injections into the mammary fat pad were performed within 15 minutes of mixture preparation using prechilled syringes with 27-gauge needles.

Single-cell RNA sequencing

Tumor tissues were dissected from B16F10 mouse models 14 days after injection (n = 3 per group). Tissues were digested with agitation in 2 mg/mL collagenase D and 30 μg/mL DNase I for 45 minutes at 37°C. Single-cell suspensions containing tumor cells were filtered through 70 μm filters and stained with LIVE/DEAD Aqua (Invitrogen; cat. #L34957) for viability and APC anti-mouse CD45 (Clone 30F11; BioLegend, cat. #103112, RRID: AB_312977). Viable CD45⁺ single cells were sorted using a BD FACSAria III flow cytometer. These cells were then sent to 10x Genomics for library construction and sequencing. Libraries were pooled and sequenced with the Illumina NovaSeq 6000, generating more than 50,000 unique reads per cell with a read length of 50 base pairs and paired-end sequencing.

Analysis of single-cell RNA sequencing

Raw sequencing data were processed using the 10x Genomics single-cell platform, with reads aligned to the mouse reference transcriptome (mm10-2020-A). Downstream analysis was conducted in R version 4.3.2 (Seurat version 4.4.0 package; ref. 16). All Seurat objects were merged and subjected to consistent quality filtering. Gene expression measurements were normalized to the total expression in each respective cell, multiplied by a scale factor of 10,000, and log-transformed using the NormalizeData function. Principal component analysis was performed on the 2,000 most variable features using the RunPCA function for clustering and cell type identification. Optimal dimensions for each dataset were determined using the ElbowPlot and JackStrawPlot functions. Cells were then clustered using the FindNeighbors and FindClusters functions, followed by nonlinear dimensionality reduction with the RunUMAP function under default settings. Clusters were classified and annotated based on specific cell type marker expression. Differential gene expression analysis, gene set enrichment, and scoring calculations were conducted using the FindMarkers function in Seurat, with parameters set to test.use = “wilcox” and logic.threshold = 0.25. Genes were classified as upregulated or downregulated based on log₂ fold change > 0.5 or < −0.5, respectively. Gene Ontology Biological Process (GO-BP) pathway enrichment analysis was performed using the R package clusterProfiler version 4.2.2 and org.Mm.eg.db, with results filtered by Bonferroni-corrected P values <0.05 and q values >0.2. Data visualization was done using enrichPlot (RRID:SCR_026996) and ggplot2 (RRID:SCR_014601; refs. 17, 18). Pathway activity scoring in individual cells was performed with the R package AUCell version 1.24.0 (19). Gene expression rankings were calculated for each cell using the expression matrix with the AUCell_buildRankings function (default parameters). Pathways of interest were obtained via msigdbr (https://github.com/cran/msigdbr, RRID:SCR_016053), and each cell was scored. The area under the curve (AUC) value represented the proportion of genes from the pathway gene set among the top-ranked genes in each cell.

Quantifying tumor-infiltrating lymphocytes, PD-L1, and MHC class I by flow cytometry

We subcutaneously established vector control and DGAT1-KO B16F10 tumors to quantify tumor-infiltrating lymphocytes. Tumor-bearing mice were euthanized 14 days after inoculation, after which, tumors were excised, weighed, minced, and incubated in PBS containing DNase I (50 µg/mL, Sigma-Aldrich, SKU #10104159001) and Collagenase Type IV (2 mg/mL, Sigma-Aldrich, cat. #C5138) for 30 minutes at 37°C. The dissociated cells were filtered through a 70 μm cell strainer (BD), collected as pellets, lysed in red blood cell lysis buffer (Roche, cat. #11814389001) on ice for 5 minutes, and rinsed with 1× wash buffer (2% FBS in 1× PBS). The filtered cells were then blocked with an anti-CD16/32 antibody (BioLegend, cat. #101319, RRID:AB_1574973) and stained with specific surface antibodies (Pacific Blue anti-mouse CD3, BioLegend, cat. #100333, RRID:AB_2028473; APC anti-mouse CD4, cat. #100424, RRID:AB_389324; APC-Cy7 anti-mouse CD8a, cat. #100766, RRID:AB_2572113; FITC anti-mouse CD45, cat. #103107, RRID:AB_312972; APC anti-mouse CD279, cat. #135209, RRID:AB_2251944) for 20 minutes on ice. Dead cells were excluded using LIVE/DEAD Fixable Aqua dye (Thermo Fisher Scientific, cat. #L23105). The cells were fixed with 2% paraformaldehyde (PFA) for 20 minutes and permeabilized with 0.1% Triton X-100 in PBS for 10 minutes on ice. Subsequently, the cells were incubated with antibodies for intracellular targets (PE anti-mouse NK1.1, cat. #108707, RRID:AB_313394; PE anti-mouse Foxp3, cat. #126403, RRID:AB_1089118; APC anti-mouse IFNγ, cat. #505816, RRID:AB_493315; PE anti-mouse GZMB, cat. #372207, RRID:AB_2687031; APC anti-mouse γ/δTCR, cat. #118116, RRID:AB_1731813; PE anti-mouse F4/80, cat. #123110, RRID:AB_893486; BioLegend) in the dark for 30 minutes on ice. Finally, the stained cells were analyzed using a BD FACS Canto II Flow Cytometer.

To evaluate the surface levels of PD-L1 and MHC class I (MHC I) in vitro, cells were treated with 1 ng/mL mouse recombinant interferon γ (IFNγ) for 12 hours before detachment using PBS containing 2% EDTA. The cells were blocked on ice with an anti-mouse CD16/32 antibody for 10 minutes. The cells were stained with LIVE/DEAD Fixable Aqua dye and PE anti-mouse CD274 antibody (BioLegend, cat. #153612, RRID:AB_2894673) or PE anti-mouse H2K^b/H2D^b antibody (BioLegend, cat. #114607, RRID:AB_313598) on ice for 20 minutes in the dark. After staining, the cells were washed twice with FACS buffer (2% FBS in PBS with 2 mmol/L EDTA) and fixed with 1% PFA at room temperature for 15 minutes in the dark. Finally, the stained cells were analyzed using a BD FACS Canto II Flow Cytometer.

Methods for Immunohistochemistry

Subcutaneous tumor tissues from mice were harvested, fixed in 4% PFA, and embedded in paraffin. Sections (4 μm thick) were deparaffinized, rehydrated, and subjected to antigen retrieval using citrate buffer (pH 6.0) via microwave heating. After blocking endogenous peroxidase activity with 3% H₂O₂ and nonspecific binding with 10% normal goat serum, sections were incubated overnight at 4°C with primary antibodies against CD3 (Abcam, cat. #ab16669, RRID:AB_443425, 1:200), CD20 (Abcam, cat. #ab64088, RRID:AB_1139386, 1:400), CD68 (Abcam, cat. #ab283654, RRID:AB_292294, 1:400), or Ki67 (Abcam, cat. #ab16667, RRID:AB_302459,1:500). Subsequently, sections were treated with HRP-conjugated secondary antibodies for 1 hour at room temperature. Finally, sections were counterstained with hematoxylin, dehydrated, cleared, and mounted. Stained images were captured using a light microscope (Nikon Eclipse Ni).

T-cell proliferation assay in tumor tissue

To assess T-cell proliferation within the tumor microenvironment, mice received i.p. injection of bromodeoxyuridine (BrdU; 50 mg/kg, Sigma-Aldrich) 2 hours prior to euthanasia. Tumors were harvested and dissociated into single-cell suspensions; then, cells were surface-stained with anti-CD45 and anti-CD3 antibodies for 30 minutes at 4°C. Cells were then fixed, permeabilized, and incubated with DNase I (37°C, 1 hour) to expose BrdU epitopes, followed by intracellular staining with anti–BrdU-APC (BioLegend, cat. #364113, RRID:AB_2814314) for 20 minutes at room temperature; then, flow cytometry was performed.

Lymphocyte depletion

We depleted CD4⁺ T cells, CD8⁺ T cells, and natural killer (NK) cells via i.p. injection of anti-CD4 (Bio X Cell, cat. #BE0003-1, RRID:AB_1107636), anti-CD8β (Bio X Cell, cat. #BE0223, RRID:AB_2687706), or anti-NK1.1 (Bio X Cell, cat. #BE0036, RRID:AB_1107737) at 100 µg intraperitoneally per mouse, respectively, on days 1, 4, and 7 after tumor cell injection. Equal IgG isotype antibodies were used as controls.

Gene set enrichment analysis in human tumors

We carried out GO analysis by using the R software package “org.Hs.eg.db” (version 3.1.0) for genome annotation and the “ClusterProfiler” (RRID:SCR_016884) package for enrichment analysis in RNA sequence datasets (20). A P value <0.05 was considered statistically significant.

CIBERSORT analysis of the intratumoral lymphocyte infiltration in human tumors

CIBERSORT is a bioinformatics tool used to characterize the status of anticancer immunity and estimate the proportion of tumor-infiltrating immune cells (21). Based on their RNA expression data, we performed a CIBERSORT analysis on patients with SKCM from the TCGA database. The analysis was conducted using the online CIBERSORT tool, TIP (http://biocc.hrbmu.edu.cn/TIP/), to visualize the immune cell subsets within the tumor samples (22).

OT-1 T-cell culture

The culture of OT-1 T cells was conducted following a protocol established in our previous research (23). In brief, CD8⁺ T cells from OT-1 C57BL/6 mice (RRID:IMSR_JAX:003831) were harvested from spleens, expressing a transgene encoding a T-cell receptor (TCR) specific for the SIINFEKL peptide bound to H-2K^b. The spleens were aseptically harvested and homogenized, followed by collagenase P (100 U/mL, Sigma-Aldrich) treatment for 30 minutes at 37°C. Dissociated cells were then passed through a 70 μm cell strainer (BD), pelleted, and resuspended in ACK Lysing Buffer (Thermo Fisher Scientific, cat. #A1049201) to lyse red blood cells for 10 minutes at room temperature. The resulting splenocytes were pelleted, washed, and resuspended in a complete growth medium (RPMI 1640, Sigma-Aldrich) containing 10% FBS, penicillin–streptomycin (Thermo Fisher Scientific), sodium pyruvate (Thermo Fisher Scientific), and 0.1% 2-mercaptoethanol (Thermo Fisher Scientific), supplemented with 0.75 μg/mL SIINFEKL peptide (GenScript, cat. #RP10611). This cell suspension was incubated at 37°C in a humidified environment with 95% air and 5% CO₂. Mouse recombinant IL2 (Thermo Fisher Scientific, cat. #RP-8605) was added on days 3 and 5 at 30 U/mL along with fresh complete growth medium. On day 7, cells were harvested for subsequent assays.

In vitro T-cell killing analysis

All B16F10 cells were engineered to express OVA, luciferase, and GFP fluorescent protein via genetically modified lentivirus infection. The control, DGAT1-KO, control–OVA [expressing ovalbumin (OVA)], and DGAT1-KO–OVA cells were incubated with mouse recombinant IFNγ at 1 ng/mL for 12 hours. The stimulated tumor cells were cultured either with OVA-specific T cells at a 1:5 ratio or without OVA-specific T cells in a T-cell complete growth medium supplemented with mouse recombinant IL2 (30 U/mL) for 24 to 72 hours. GFP fluorescence was captured using the ZEISS Axio Observer.Z1 fluorescence microscope imaging station, and the luciferase intensity was measured using a microplate reader.

LD staining

Cells were seeded at 1 × 10⁴ per dish into 35 mm glass-bottom poly-D-lysine–coated dishes (MatTek). After treatments, the cells were fixed with 4% PFA for 20 minutes and permeabilized with 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 10 minutes. The cells were then incubated with LipidSpot Lipid Droplet Stains (1:1,000, Biotium, cat. #70065) at room temperature for 15 minutes, protected from light. Following incubation, the cells were washed 3 times with PBS. Cellular nuclei were counterstained with DAPI dole (Thermo Fisher Scientific). Digital images were analyzed using a TCS SP5 inverted confocal microscope (Leica) at 40× magnification. Relative integrated fluorescence intensity per cell was obtained using ImageJ (RRID:SCR_003070) software, with quantification performed on images of 30 cells for each group.

Lipid peroxidation assay

The malondialdehyde (MDA) content in tumor cells was measured using the Lipid Peroxidation MDA Assay Kit (Abcam, cat. #ab118970) to monitor lipid peroxidation. Following the manufacturer’s instructions, cells were collected by centrifugation, lysed, and homogenized in MDA lysis buffer. The protein concentration was determined using a BCA protein assay. The reaction mix was then transferred to a 96-well plate, and absorbance was measured immediately on a microplate reader at optical density 532 nm for the colorimetric assay.

Lipid metabolism analysis

Lipid metabolism profiling in tumor cells was performed using the MxP Quant 500 Kit (Biocrates Life Sciences). Cells were harvested, and metabolites were extracted from 1 to 10⁷ cells using methanol-based solutions provided in the kit. The extracted samples, along with internal standards, were analyzed by flow injection analysis tandem mass spectrometry for absolute quantification of lipids. The kit simultaneously measured key lipid classes, including TGs, free fatty acids [e.g., polyunsaturated fatty acid (PUFA) species], phospholipids (phosphatidylcholine, lysophosphatidylcholine), and cholesterol esters. Data were normalized to cell counts and expressed as pmol/10⁶ cells to evaluate lipid metabolic alterations.

Mitochondrial membrane potential assay

Mitochondrial membrane potential (MMP) was detected using the JC-1 (tetraethyl benzimidazolyl carbocyanine iodide) MMP assay kit (Abcam; cat. #113850). FCCP (2-[2-[4-(trifluoromethoxy)phenyl]hydrazinylidene]-propanedinitrile), provided by the kit, was used as a positive control for membrane depolarization at a concentration of 50 ng/mL after 4 hours of treatment. Following the manufacturer’s instructions, immunofluorescence and flow cytometry assessed red fluorescence (excitation 535 nm/emission 590 nm, aggregates) and green fluorescence (excitation 475 nm/emission 530 nm, monomers). Digital image analysis for immunofluorescence was performed using a TCS SP5 inverted confocal microscope (Leica) at 40× magnification. Relative integrated fluorescence intensity per cell was obtained using ImageJ software, with quantification performed on images of 50 cells for each time point. Experiments were generally repeated 3 times.

Cellular ROS assay

We used flow cytometry to assess cellular ROS levels using the DCFDA/H2DCFDA Cellular ROS Assay Kit (Abcam, cat. #ab113851). 2′,7′-Dichlorodihydrofluorescein diacetate (DCFDA) is a cell-permeable fluorogenic probe that directly measures intracellular redox status. Upon introduction to the cells, DCFDA permeates the cell membrane, in which cellular esterases convert it into a nonfluorescent product. This product is subsequently oxidized by ROS into the highly fluorescent compound 2′,7′-dichlorofluorescein, which can be detected at an excitation/emission wavelength of 485 nm/535 nm. For this experiment, cells were treated with 20 µmol/L DCFDA in 1× ROS buffer and incubated in the dark at 37°C for 30 minutes before undergoing flow cytometry using the FITC channel.

Reduced glutathione assay

Intracellular glutathione (GSH) levels were measured using the GSH+GSSG/GSH Assay Kit (Abcam, cat. #ab239709). Following the manufacturer’s instructions, cells were collected by centrifugation and lysed in ice-cold GSH buffer. After incubating on ice for 10 minutes, 5% 5-sulfosalicylic acid was added, mixed well, and centrifuged at 8,000 × g for 10 minutes. The supernatant was then transferred to a 96-well plate to prepare the reaction mix, which was measured at an absorbance of 415 nm using a microplate reader. The concentrations of reduced GSH were calculated by subtracting the oxidized GSH (GSSG) levels from the total GSH (GSH = total GSH −2 × GSSG).

Transmission electron microscopy

Cells were fixed for 30 minutes in 2.5% glutaraldehyde in PBS (pH 7.4) containing 0.1 mol/L sucrose at 4°C and postfixed in 1% osmium tetroxide in PBS for 30 minutes at room temperature. The cells were then stained en bloc with 1% uranyl acetate for 30 minutes, followed by dehydration in a graded ethanol series (30%, 50%, 70%, 90%, and 100%). The cells were embedded in resin after rinsing with propylene oxide. Samples were placed in molds and transferred to an oven for polymerization at 50°C to 60°C for 48 hours. Sections of 70 nm thickness were cut using a Leica EM UC6 ultramicrotome and stained with 2% uranyl acetate and Reynolds lead citrate. Transmission electron microscopy (TEM) was performed using an FEI Tecnai G² Twin at 80 kV at the Duke Center for Electron Microscopy and Nanoscale Technology.

In vitro drug treatment

For in vitro DGAT1 or DGAT2 inhibition, cells were treated with pradigastat (2 µg/mL) or PF-06424439 (10 µg/mL) for 48 hours before harvesting for experiments. To modulate GPX4 function, tumor cell cultures were treated with 1 µmol/L RSL3 or 2 µmol/L Fer-1 for 48 hours prior to analysis. Triacsin C (2 µmol/L; MCE, cat. #HY-N6707) was administered to cells for 48 hours to induce the LDs.

Cell proliferation assay

Cell proliferation was measured using an MTT Cell Proliferation Assay Kit (ATCC). Briefly, cells were seeded in a 96-well microplate at a density of 2 × 10³ cells/well. About 10 mL MTT reagent was added to each well at different time points. After incubation at 37°C for 4 hours, 100 mL of detergent reagent was added, and the cells were left at room temperature in the dark for 2 hours. We then recorded the absorbance at 570 nm with a microplate reader. Each experiment was repeated 3 times.

Statistical analysis and reproducibility

Patients’ prognoses were analyzed using the log-rank (Mantel–Cox) test for clinical data. The Spearman correlation test was conducted for correlation analyses. The remaining statistical analysis methods have been described in the relevant sections.

For in vitro and in vivo experiments, unless otherwise stated in figure legends, data are represented as individual values or mean ± SEM. Group sizes (n) and the statistical tests applied are indicated in figure legends. Statistical significance was assessed by using unpaired t test analysis. All experiments were performed at least 3 times with biologically independent samples. In the murine model, mice were randomly assigned to different treatment groups stratified by tumor sizes at the time of treatment. Two-way ANOVA determined the significance of differences between tumor growth curves. Additionally, differences in survival curves between respective groups were calculated using the Mantel–Cox log-rank test. All statistical calculations were performed using GraphPad Prism 10.1.2 (RRID:SCR_002798).

Results

Low DGAT1 expression correlates with favorable prognosis and enhanced immunotherapy response

To explore the prognostic significance of DGAT1 in cancer, we analyzed its expression across a range of malignancies using publicly available clinical datasets (12). In TCGA cohorts, including SKCM, BRCA, LUSC, and colon adenocarcinoma, low DGAT1 expression was significantly associated with improved overall survival compared with high DGAT1 expression (Fig. 1A–D).

Figure 1.

DGAT1 expression is associated with poor prognosis in human malignancies. A–D, Kaplan–Meier analysis of the overall survival (OS) in patients with TCGA SKCM (A), BRCA (B), LUSC (C), and colon adenocarcinoma (COAD; D), stratified by high or low DGAT1 expression. Cutoff values distinguishing high and low groups were set at the ≥80th vs. ≤20th percentile in SKCM, LUSC, and COAD and at ≥85th vs. ≤15th percentile in BRCA. E–G, Kaplan–Meier analysis of OS in three cohorts of patients with SKCM treated with anti–PD-1 (E), anti–PD-1 (F), and anti–CTLA4 (G) antibodies, stratified by high and low DGAT1 expression. Error bars indicate ± SEM throughout the figure. P values were determined by unpaired t test (E) and log-rank test for the rest.

To further evaluate the clinical relevance of DGAT1 in the context of immunotherapy, we analyzed three independent cohorts of patients with melanoma undergoing immune checkpoint blockade (ICB) therapy (12, 14). In all three cohorts, low DGAT1 expression consistently correlated with superior survival outcomes compared with high DGAT1 expression (Fig. 1E–G). These results suggest that DGAT1 may suppress antitumor immune responses, thereby diminishing the effectiveness of ICB therapy in patients with cancer.

Additionally, tumors with low DGAT1 mRNA expression displayed elevated levels of PD-L1 mRNA in TCGA cohorts, a critical biomarker of response to anti–PD-1 immunotherapy (Supplementary Fig. S1A; ref. 12). On the other hand, in murine tumor cells with DGAT1KO, flow cytometry analysis showed that DGAT1 deficiency did not significantly increase PD-L1 levels (Supplementary Fig. S1B–S1E). We do not know the exact mechanism for this discrepancy. We speculate it could be caused by the microenvironmental differences between human and mouse tumors.

DGAT1 inhibition enhances tumor response to ICB in murine models

To assess the role of DGAT1 in tumor progression and its impact on the efficacy of ICB therapy, we generated DGAT1-KO murine cancer cell lines, including B16F10 melanoma, MC38 colon cancer, and 4T1 breast cancer (Supplementary Fig. S2A, B). Subcutaneous tumors were established in mice using either DGAT1-KO or vector control cells. After tumor establishment, mice were treated with an anti–PD-1 antibody following the protocol outlined in Supplementary Fig. S2C. In the B16F10 melanoma model, DGAT1 deficiency resulted in a small but significant delay in tumor growth and improved survival compared with controls (Fig. 2A and B). Tumor regression was observed in four out of six mice bearing DGAT1-deficient tumors. By contrast, DGAT1 KO alone did not affect tumor growth in the 4T1 breast cancer model (Fig. 2C and D) and produced only a modest, statistically insignificant delay in the MC38 colon cancer model (Fig. 2E and F). Notably, DGAT1 deficiency significantly enhanced the response to ICB therapy in both the 4T1 and MC38 models, which are typically resistant to ICB, with tumor regressions observed in both cases (Fig. 2A–F). In the metastatic 4T1 breast cancer model, DGAT1 deletion also significantly reduced lung metastatic burden. This effect was particularly pronounced in the combination treatment group of DGAT1-KO tumors and anti–PD-1 therapy (Supplementary Fig. S2D and S2E).

Figure 2.

DGAT1 deficiency enhances tumor response to ICB therapy. A and B, Tumor growth delay (A) and Kaplan–Meier survival curves (B) of C57BL/6 mice inoculated subcutaneously with 1 × 10⁵ vector control (VC) or DGAT1-KO B16F10 cells and treated with an anti–PD-1 antibody. C and D, Tumor growth delay (C) and Kaplan–Meier survival curves (D) of BALB/c mice inoculated subcutaneously with 2 × 10⁵ VC or DGAT1-KO 4T1 cells and treated with an anti–PD-1 antibody. E and F, Tumor growth delay (E) and Kaplan–Meier survival curves (F) of C57BL/6 mice inoculated subcutaneously with 5 × 10⁵ vector control or DGAT1-KO MC38 cells and treated with an anti–PD-1 antibody. G and H, Tumor growth delay (G) and Kaplan–Meier survival curves (H) of C57BL/6 mice inoculated subcutaneously with 1 × 10⁵ wild-type B16F10 cells and treated with PF-04620110 and/or an anti–PD-1 antibody. Error bars indicate ± SEM. P values were calculated using two-way ANOVA (A, C, E, and G) and log-rank test (B, D, F, and H). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

To determine whether pharmacologic inhibition of DGAT1 could replicate the effects of genetic deletion, we tested two clinical-stage DGAT1 inhibitors, PF-04620110 and pradigastat (Supplementary Fig. S2F). In the B16F10 model, PF-04620110 alone had no significant impact on tumor growth, but its combination with anti–PD-1 therapy slowed tumor progression and extended survival (Fig. 2G and H). Similarly, pradigastat, when combined with anti–PD-1, significantly delayed tumor growth and prolonged survival in both the B16F10 and MC38 models (Supplementary Fig. S2G–S2J). In the LLC lung cancer model, an extremely immunologically “cold” tumor, the combination of PF-04620110 and anti–PD-1 therapy also showed significant synergy (Supplementary Fig. S2K and S2L). In the 4T1 orthotopic tumor model, we also found that DGAT1 deficiency can directly inhibit tumor proliferation and, when combined with anti–PD-1, significantly delays tumor growth (Supplementary Fig. S2M and S2N). To determine if DGAT2, a member of the same enzyme family as DGAT1, was also involved in a similar manner, we used PF-06424439 (a DGAT2 inhibitor) in combination with immunotherapy. Our studies demonstrated that DGAT2 inhibition did not enhance the efficacy of immunotherapy (Supplementary Fig. S2O and S2P).

Together, these findings demonstrate that DGAT1 inhibition, whether through genetic deletion or pharmacologic targeting, enhances tumor control and potentiates the efficacy of ICB therapy. These preclinical results align with clinical observations of improved outcomes in patients with low DGAT1 expression (Fig. 1).

DGAT1 deficiency promotes a proinflammatory tumor immune microenvironment

To investigate the mechanisms underlying the enhanced antitumor effects of DGAT1 deficiency in combination with ICB therapy, we examined the tumor immune microenvironment (TIME) in DGAT1-deficient tumors. CD45⁺ immune cells were isolated from subcutaneously implanted control and DGAT1-KO B16F10 tumors and subjected to single-cell RNA sequencing (scRNA-seq). Analysis of immune cell populations revealed key tumor-infiltrating immune subsets, including CD4⁺ T cells, CD8⁺ T cells, NK cells, B cells, macrophages, monocytes, and dendritic cells, identified using cell type–specific markers (Fig. 3A and B; Supplementary Fig. S3A). Notably, DGAT1-KO tumors exhibited a significant increase in the proportions of T cells, particularly CD4⁺ T cells, CD8⁺ T cells, and NK cells, compared with controls (Fig. 3C).

Figure 3.

scRNA-seq analysis of DGAT1 deficiency-induced alterations in the TIME. A, Uniform Manifold Approximation and Projection (UMAP) visualization of transcriptional profiles of intratumoral immune cells from B16F10 tumors. Each dot represents a single cell, with colors indicating distinct clusters corresponding to inferred cell types. Tumor-infiltrating live CD45⁺ cells were isolated on day 12 after intracranial tumor implantation (n = 3) for scRNA-seq analysis. B, UMAP representation of intratumoral immune cell populations in tumors from vector control (VC) and DGAT1-KO B16F10 cells. C, Proportions of different immune cell types within vector control and DGAT1-KO tumors, based on scRNA-seq data. D, GO-BP analysis of the top five upregulated and downregulated pathways in CD45⁺ cells from DGAT1-KO B16F10 tumors compared with controls. E, UMAP visualization of intratumoral lymphocytes in vector control and DGAT1-KO B16F10 tumors, with expression levels of lymphocyte activation markers. F, Proportions of different T-cell subtypes in vector control and DGAT1-KO B16F10 tumors, as determined by scRNA-seq analysis. G, UMAP visualization of major macrophage clusters in B16F10 tumors. H, Proportions of distinct macrophage clusters in vector control and DGAT1-KO B16F10 tumors, based on scRNA-seq data. Prolif-TAM, proliferative TAM.

GO-BP analysis of immune cells from DGAT1-KO and control tumors highlighted significant upregulation of pathways associated with lymphocyte activation, T-cell activation, and proliferation in the DGAT1-KO group (Fig. 3D). Further assessment of pathway activity, based on AUC scores, confirmed enhanced lymphocyte activation (Supplementary Fig. S3B and S3C) and T-cell activation (Supplementary Fig. S3D and S3E) in DGAT1-deficient tumors. Detailed analysis of T-cell subsets revealed increased proportions of CD4⁺ memory T cells, as well as CD8⁺ naïve, memory, and effector T cells in the DGAT1-KO group (Fig. 3E and F; Supplementary Fig. S3F).

In addition to T-cell enrichment, we observed a higher proportion of macrophages in DGAT1-KO tumors (Fig. 3C). Subclassification of tumor-associated macrophages (TAM) using gene markers (Supplementary Fig. S3G) identified three dominant subsets: M1-like macrophages, M2-like macrophages, and proliferative TAMs (Fig. 3G). Notably, M1-like macrophages, which are associated with antitumor immunity and suppression of tumor progression and metastasis, were significantly enriched in DGAT1-KO tumors (Fig. 3H; ref. 24). This polarization shift in macrophages suggests the development of a TIME that is more favorable for ICB efficacy.

To validate these observations in human cancers, we analyzed RNA-seq data from the TCGA cohort of patients with SKCM, comparing DGAT1-high and DGAT1-low tumors. Consistent with the findings in murine models, GO analysis revealed that DGAT1-low tumors were enriched for genes involved in stress response, immune activation, and cell death pathways (Supplementary Fig. S3H), further supporting the immune remodeling observed in DGAT1-KO B16F10 tumors.

Validation of intratumoral lymphocyte infiltration and functional relevance

To validate the findings from scRNA-seq, we quantified tumor-infiltrating lymphocytes in DGAT1-KO and control B16F10 tumors using flow cytometry (gating strategy shown in Supplementary Fig. S4A). DGAT1-deficient tumors exhibited a significant increase in the infiltration of CD3⁺, CD4⁺, and CD8⁺ T cells compared with controls (Fig. 4A–C). Additionally, we observed an elevation in NK cells, granzyme B–positive CD8⁺ T cells, IFNγ–positive (IFNγ⁺) CD8⁺ T cells, and macrophages (Fig. 4D–G). There was no significant change in the proportions of γδTCR⁺ T cells, CD4⁺ FoxP3⁺ regulatory T cells (Treg), or PD-1⁺ T cells (Fig. 4H–J). We further investigated the impact of DGAT1 deletion on tumor immune infiltration using immunohistochemistry. The results demonstrated that DGAT1 deletion increased the proportions of CD3⁺ (T cells) and CD68⁺ (macrophages) cells in B16F10, 4T1, and MC38 models, whereas CD20⁺ (B cells) were rare and showed no difference compared with the control group. Additionally, Ki67 (proliferation marker) was significantly reduced in the DGAT1-KO B16F10 model but showed no significant difference in the 4T1 and MC38 models, consistent with our observations in tumor proliferation assays (Supplementary Fig. S4B). We also examined the T-cell proliferation capacity within tumors following DGAT1 inhibition, and the results demonstrated that DGAT1 inhibition significantly enhanced T-cell proliferation, as indicated by increased BrdU staining (Supplementary Fig. S4C and S4D).

Figure 4.

DGAT1 deficiency-induced intratumoral lymphocyte infiltration and immune response. A–J, Quantification of tumor-infiltrating immune cells per milligram of tissue. Flow cytometry was performed to measure the numbers of CD3⁺ T cells (A), CD4⁺ T cells (B), CD8⁺ T cells (C), NK cells (D), CD8⁺ granzyme B–positive (GzmB+) T cells (E), CD8⁺ IFNγ⁺ T cells (F), macrophages (Mφ+; G), γδTCR + T cells (H), CD4+FoxP3+ Treg cells (I), and CD3+PD1+ T cells (J) in vector control (VC) or DGAT1-KO B16F10 tumors grown in C57BL/6 mice (n = 5) on day 14 after subcutaneous inoculation of 1 × 10⁵ tumor cells. K–M, Functional evaluation of immune effector subsets. Tumor growth delay in C57BL/6 mice inoculated with 1 × 10⁵DGAT1-KO B16F10 cells and treated with anti-CD4 (K), anti-CD8 (L), or anti-NK1.1 (M) and anti–PD-1 antibodies or isotype controls. N and O, Flow cytometry analysis of MHC I expression (N) and mean fluorescence intensity (MFI; O) in VC and DGAT1-KO B16F10 cells. P and Q, Flow cytometry analysis of MHC I expression (O) and MFI (P) in nontreated control (NC) and pradigastat-treated B16F10 cells. R, Fluorescence images of GFP-labeled B16F10 cells, with or without OVA antigen, cocultured with either nonspecific T cells or OT-1 OVA-specific T cells after 72 hours. Scale bars, 200 μm for all image panels. S, Quantification of the remaining fraction of control or DGAT1-KO B16F10-GFP-Luc tumor cells after 48 hours of incubation with either activated T cells or OVA-specific T cells (n = 3 OT-I mice per group). Error bars represent ±SEM throughout the figure. P values were calculated using a two-way ANOVA test (K–M) and unpaired t tests for the remaining data. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

To extend these observations to human cancers, we performed CIBERSORT analysis using RNA-seq data from patients with melanoma in the TCGA SKCM cohort. DGAT1-low tumors showed significantly higher infiltration of CD8⁺ T cells, neutrophils, and NK cells, whereas Treg recruitment was reduced compared with DGAT1-high tumors (Supplementary Fig. S4E). These findings suggest a conserved role for DGAT1 in shaping the immune microenvironment across species.

We further assessed the functional relevance of these lymphocyte subsets in the enhanced response to ICB therapy observed in DGAT1-KO tumors through depletion experiments. Anti-CD4, anti-CD8, and anti-NK1.1 antibodies were used to selectively deplete CD4⁺ T cells, CD8⁺ T cells, and NK cells, respectively, in B16F10 tumor-bearing mice (treatment schedule shown in Supplementary Fig. S4F). Depletion of any of these immune cell populations abolished the heightened sensitivity to anti–PD-1 therapy conferred by DGAT1 deficiency, resulting in rapid tumor growth and significantly reduced survival (Fig. 4K–M; Supplementary Fig. S4G–S4I). These findings demonstrate that CD4⁺ T cells, CD8⁺ T cells, and NK cells are all critical mediators of the enhanced ICB response in DGAT1-deficient tumors.

Given the increased infiltration of IFNγ⁺ CD8⁺ T cells in DGAT1-deficient tumors, which is associated with the upregulation of antigen presentation machinery and elevated MHC I surface expression (25), we next examined MHC I expression in both DGAT1-KO and pradigastat-treated B16F10 tumors. Both genetic deletion and pharmacologic inhibition of DGAT1 significantly upregulated MHC I expression on the surface of B16F10 tumor cells (Fig. 4N–Q).

To assess the impact of DGAT1 deficiency on cytotoxic T lymphocyte (CTL)–mediated tumor cell killing, we utilized the OT-1 transgenic mouse model, in which T cells express a TCR specific for the chicken OVA peptide SIINFEKL (26). OVA-transduced, GFP-Luc–labeled DGAT1-KO and control B16F10 cells were cocultured with OT-1–derived CTLs, and tumor cell killing was evaluated over 24 to 72 hours. DGAT1-deficient tumor cells exhibited significantly increased susceptibility to CTL-mediated killing over time compared with controls (Fig. 4R and S).

DGAT1 deficiency–induced LD depletion promotes lipid peroxidation and mitochondrial dysfunction

To investigate the mechanism by which DGAT1 deficiency remodels the TIME and enhances ICB therapy, we first examined the role of DGAT1 in LD formation in cancer cells. DGAT1 is a key enzyme involved in TG synthesis, the primary form of intracellular energy storage, which is sequestered in LDs (27). Genetic KO of DGAT1 and pharmacologic inhibition with pradigastat both led to a significant reduction in LD accumulation in B16F10 melanoma cells. This reduction was similarly observed in DGAT1-KO A375 melanoma cells (Fig. 5A–C; Supplementary Fig. S5A–S5D).

Figure 5.

LD reduction induces lipid peroxidation and mitochondrial dysfunction in DGAT1-deficient cells. A–C, Representative immunofluorescence images (A) and relative integrated intensities of LD staining (B and C) in control and DGAT1-KO or pradigastat-treated B16F10 cells. Scale bars, 25 µm. D and E, Quantification of MDA levels in control and DGAT1-KO (D) or pradigastat-treated (E) B16F10 cells. F–J, Levels of TG (F), DHA-cholesteryl ester (DHA-CE) 22:6 (G), HexCer d18:2/20:0 (H), HexCer d18:2/22:0 (I), and HexCer d18:2/23:0 (J) in control and DGAT1-KO B16F10 cells. K–M, JC-1 dye staining showing representative immunofluorescence images of MMP (K) and monomer/aggregate ratios in control (L), DGAT1-KO (M), or pradigastat-treated B16F10 cells. Scale bars, 25 µm. N and O, Flow cytometry analysis of ROS levels (N) and mean fluorescence intensity (MFI; O) in control and DGAT1-KO B16F10 cells. P and Q, Flow cytometry analysis of ROS levels (P) and MFI (Q) in control and pradigastat-treated B16F10 cells. R and S, Representative immunofluorescence images (R) and relative integrated intensities of LD staining (S) in control and triacsin C–treated B16F10 cells. Scale bars, 25 µm. T, Quantification of MDA levels in control and triacsin C–treated B16F10 cells. U and V, Flow cytometry analysis of ROS levels (U) and mean fluorescence intensity (V) in control and triacsin C–treated B16F10 cells. Error bars represent ±SEM throughout the figure. P values were calculated using an unpaired t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. VC, vector control.

The depletion of LDs impairs the cell’s ability to store lipids, making PUFAs more susceptible to oxidative stress and increasing lipid peroxidation. We observed elevated levels of MDA, a byproduct of lipid peroxidation and a marker of oxidative stress, in both DGAT1-KO and pradigastat-treated B16F10 and A375 cells (Fig. 5D and E; Supplementary Fig. S5E and S5F). Moreover, treatment with a DGAT2 inhibitor did not increase the levels of MDA (Supplementary Fig. S5G). These findings are consistent with previous studies linking LD depletion to increased MDA levels (28). We next conducted lipid metabolism analysis in DGAT1-KO B16F10 cells, and DGAT1 deficiency induced a significant decrease in TG (Fig. 5F), whereas the other four PUFAs—linoleic acid, arachidonic acid, eicosapentaenoic acid, and docosahexaenoic acid (DHA)—were markedly elevated (Supplementary Fig. S5H–S5K; Supplementary Table S2). Especially, we found that the levels of four PUFA-containing oxidized complex lipids, including DHA-cholesteryl ester 22:6, hexosylceramide (HexCer) d18:2/20:0, HexCer d18:2/22:0, and HexCer d18:2/23:0, were increased in DGAT1-KO cells (Fig. 5G–J). This likely occurs because PUFAs that cannot be esterified into triacylglycerol (TAG) are shunted into other metabolic pathways. These increases strongly support the interpretation that DGAT1 loss enhances PUFA oxidation due to the loss of TAG/LD buffering.

MMP (ΔΨm) serves as a critical indicator of mitochondrial integrity and function. Disruption of lipid homeostasis through LD depletion can lead to increased lipid peroxidation, which damages mitochondrial membranes and impairs mitochondrial function (29). Using JC-1 staining to assess ΔΨm, we observed a significant shift toward monomeric (green) fluorescence, with reduced aggregate (red) fluorescence in DGAT1-KO and pradigastat-treated cells. This shift indicates diminished ΔΨm and mitochondrial dysfunction (Fig. 5K–M; Supplementary Fig. S5L–S5N). Analysis of data from patients with melanoma (SKCM) further revealed that DGAT1-low tumors exhibit reduced expression of genes involved in mitochondrial function and energy metabolism, aligning with these observations (Supplementary Fig. S5O). Additionally, elevated levels of ROS were detected following DGAT1 inhibition in both cell lines (Fig. 5N–Q; Supplementary Fig. S5P–S5S).

To confirm that LD depletion drives these effects, we treated tumor cells with triacsin C, a pharmacologic inhibitor of LD formation. Triacsin C significantly reduced LD formation, elevated MDA and ROS levels, and decreased ΔΨm in both B16F10 and A375 cells (Fig. 5R–V; Supplementary Fig. S5T–S5X). These findings confirm that LD depletion is causally linked to increased lipid peroxidation, mitochondrial dysfunction, and ROS generation.

Collectively, our results demonstrate that DGAT1 inhibition disrupts LD formation, leading to lipid peroxidation, mitochondrial dysfunction, and oxidative stress. These changes may contribute to the observed remodeling of the TIME and the enhanced efficacy of ICB therapy.

DGAT1 inhibition depletes GPX4 and induces a ferroptosis phenotype

To examine the downstream consequences of LD depletion, mitochondrial dysfunction, and oxidative stress caused by DGAT1 inhibition, we assessed GSH levels in DGAT1-deficient cells. GSH, a key antioxidant, detoxifies ROS. Elevated ROS levels are known to deplete GSH as part of the cellular antioxidant response (30). Consistent with this mechanism, we observed a significant reduction in GSH levels in DGAT1-deficient B16F10 and A375 cells (Fig. 6A–D). Similarly, treatment with the LD inhibitor triacsin C also led to a decrease in GSH levels in B16F10 cells, confirming that GSH depletion is linked to LD loss (Fig. 6E). However, treatment with the DGAT2 inhibitor did not increase the levels of GSH (Supplementary Fig. S6A).

Figure 6.

DGAT1 deficiency induces GPX4 depletion and a ferroptosis phenotype. A–D, Quantification of GSH levels in control and DGAT1-KO or pradigastat-treated B16F10 (A and B) and A375 (C and D) cells. E, Quantification of GSH levels in control and triacsin C–treated B16F10 cells. F, Representative TEM images of mitochondria in control and DGAT1-KO or pradigastat-treated B16F10 cells. Scale bars, 0.5 µm. G and H, Western blot analysis of GPX4 expression in control and DGAT1-KO or pradigastat-treated B16F10 (G) and A375 (H) cells. I and J, Representative TEM images of mitochondria in control and DGAT1-KO (I) or pradigastat-treated (J) B16F10 cells combined with Fer-1 treatment. Scale bars, 0.5 µm. K and L, Western blot analysis of GPX4 levels in control and DGAT1-KO (K) or pradigastat-treated (L) B16F10 cells combined with Fer-1 treatment. M and N, Tumor growth delay (M) and Kaplan–Meier survival curves (N) in C57BL/6 mice inoculated subcutaneously with 1 × 10⁵DGAT1-KO B16F10 cells and treated with Fer-1 and anti–PD-1 antibody or isotype control. O, Quantification of MDA levels in control, DGAT1-KO, RSL3, and DGAT1-KO combined RSL3-treated B16F10 cells. P and Q, Tumor growth delay (P) and Kaplan–Meier survival curves (Q) in C57BL/6 mice inoculated subcutaneously with 1 × 10⁵DGAT1-KO B16F10 cells and treated with RSL3 and anti–PD-1 antibody or isotype control. Error bars represent ±SEM throughout the figure. P values were determined by unpaired t test (A–E and O), two-way ANOVA (M and P), and log-rank test (N and Q), respectively. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant. VC, vector control.

The combination of increased lipid peroxidation, elevated ROS levels, and GSH depletion is a hallmark of ferroptosis, a distinct form of immunogenic cell death (8). To determine whether DGAT1 deficiency induces ferroptosis, we examined mitochondrial morphology using TEM. DGAT1-deficient cells displayed mitochondrial shrinkage and fragmentation (Fig. 6F), a key feature of cellular ferroptosis (8).

GSH depletion impairs the activity of GPX4, a critical enzyme that neutralizes lipid peroxides. The loss of GPX4 leads to the accumulation of lipid peroxides, which react with iron via the Fenton reaction to generate hydroxyl radicals. These radicals cause extensive lipid peroxidation, membrane damage, and, ultimately, ferroptosis (9). To further investigate the relationship between DGAT1 and ferroptosis, we analyzed the TCGA SKCM dataset. We found a strong correlation between DGAT1 expression and ferroptosis-related genes. Specifically, ferroptosis suppressors such as GPX4, FTH1, SCD, AIFM2, and FXN were positively correlated with DGAT1 expression, whereas ferroptosis drivers, including ACSL4, IREB2, NCOA4, and BECN1, were negatively correlated (Supplementary Fig. S6B). Additionally, DGAT1 and GPX4 levels showed a positive correlation across multiple TCGA cancer cohorts (Supplementary Fig. S6C). Western blot analysis corroborated these findings, showing significantly reduced GPX4 expression in DGAT1-KO or inhibitor-treated B16F10 and A375 cells (Fig. 6G and H). These results confirm that DGAT1 inhibition leads to LD depletion and ROS accumulation, culminating in GPX4 depletion and the induction of ferroptosis.

Ferroptosis induced by DGAT1 inhibition underpins enhanced ICB efficacy

To confirm that DGAT1 deficiency–induced ferroptosis contributes to the enhanced immune response and improved efficacy of ICB therapy, we utilized the ferroptosis inhibitor Fer-1, which neutralizes lipid peroxyl radicals. TEM analysis revealed that the mitochondrial structural damage caused by DGAT1 inhibition was partially reversed by Fer-1 treatment (Fig. 6I and J). Additionally, Fer-1 restored GPX4 levels in DGAT1-deficient cells (Fig. 6K and L), further suggesting that ferroptosis is a key mechanism involved in the observed effects. In vivo, treatment with Fer-1 partially reversed the tumor growth delay and survival benefits conferred by ICB therapy in DGAT1-deficient tumors, leading to a significantly worsened prognosis (Fig. 6M and N; Supplementary Fig. S6D). These results provide strong evidence that the enhanced ICB efficacy associated with DGAT1 inhibition is ferroptosis-dependent.

Furthermore, when we used the ferroptosis inhibitor RSL3 to treat DGAT1-KO cells, RSL3 exacerbated the MDA level induced by DGAT1 deficiency (Fig. 6O). The combination with RSL3 in the mouse tumor model further amplified the ICB treatment effects induced by DGAT1 deficiency (Fig. 6P and Q; Supplementary Fig. S6D). In summary, these results confirm that the lipid peroxidation caused by DGAT1 deficiency further leads to ferroptosis, thereby enhancing tumor immunity and the effectiveness of ICB therapy.

To verify that the DGAT1 inhibitor pradigastat specifically acts on the DGAT1 target to elicit a series of biological effects, we treated DGAT1-KO B16F10 cells with pradigastat and observed that the inhibitor failed to further enhance the DGAT1 deletion-induced increases in MDA, ROS, and GSH levels (Supplementary Fig. S6E–S6H). These results confirm that pradigastat specifically targets DGAT1 to produce its biological effects.

In summary, DGAT1 inhibition promotes ferroptosis through LD depletion, oxidative stress, and GPX4 depletion, which enhances tumor immunity and improves the efficacy of ICB therapy. These findings underscore the therapeutic potential of targeting DGAT1 to induce ferroptosis and enhance cancer immunotherapy outcomes.

Discussion

Oxidative stress globally modulates ferroptosis by influencing lipid metabolism, ROS regulation, and the expression of ferroptosis-suppressing genes (30). In our study, both cellular and clinical analyses demonstrated a significant reduction of GPX4 in DGAT1-low samples, a hallmark of ferroptosis induction. These findings indicate that DGAT1 inhibition promotes ferroptosis by inducing oxidative stress. Consistent with this, samples from patients with DGAT1-low SKCM exhibited downregulation of ferroptosis-suppressing genes and upregulation of ferroptosis-promoting genes. Furthermore, DGAT1 deficiency reduces LD formation and elevates ROS levels through increased lipid oxidation, creating a cellular environment that compromises antioxidant defenses and promotes ferroptosis. This cascade culminates in a “hot” tumor microenvironment conducive to enhanced immune activation and improved response to ICB.

Although there are recent reports of the association of DGAT1 inhibition with ferroptosis, the use of DGAT1 inhibitors to enhance ICB therapy has not been explored (31, 32). Many previous studies have demonstrated that inducing ferroptosis as a strategy to enhance ICB therapy holds significant promise for cancer treatment. However, existing ferroptosis-inducing agents such as RSL3, erastin, and FIN56 are primarily research tools and lack clinical applicability. In this context, the discovery that DGAT1 inhibitors can promote ferroptosis is particularly compelling, as several DGAT1 inhibitors have already reached clinical development. For example, PF-04620110, developed by Pfizer, demonstrated efficacy in reducing TG levels and improving glucose metabolism in early clinical trials. However, its phase I trials were discontinued due to dose-limiting GI side effects, including diarrhea and nausea (33). Similarly, pradigastat (LCQ908), another DGAT1 inhibitor, effectively lowered TG levels but was also associated with GI side effects, particularly diarrhea (34).

Despite these setbacks in noncancer applications, clinical-stage DGAT1 inhibitors present a unique opportunity to evaluate ferroptosis induction as a means of enhancing cancer immunotherapy. The side effects observed in patients with noncancer conditions are likely more acceptable and manageable in oncology settings, in which the therapeutic benefits may outweigh these limitations.

In conclusion, our study identifies DGAT1 as a critical regulator of the TIME and establishes a novel link between DGAT1 inhibition and ferroptosis induction. These findings provide a strong rationale for the clinical evaluation of DGAT1 inhibitors as safe and effective ferroptosis-inducing agents to enhance ICB therapy.

Supplementary Material

Supplementary Table S1

SgRNA sequences used to knockout the DGAT1 gene.

Supplementary Table S2

Mass spectrometry analysis of intracellular lipids in control and DGAT1-KO B16F10 cells.

Supplementary Figure S1

Analysis of PD-L1 levels in TCGA and DGAT1 inhibited samples

Supplementary Figure S2

Additional data on DGAT1 deficiency enhancing tumor response to ICB therapy.

Supplementary Figure S3

Additional data from ScRNA-seq analysis of intratumoral CD45+ cells in DGAT1-deficient B16F10 tumors.

Supplementary Figure S4

Additional data on DGAT1 deficiency-induced intratumoral lymphocyte infiltration and immune response.

Supplementary Figure S5

Additional data on lipid droplet reduction-induced lipid peroxidation and mitochondrial dysfunction in DGAT1-deficient cells.

Supplementary Figure S6

Additional data on DGAT1 deficiency-induced tumor ferroptosis.

Data Availability

The patient data analyzed in this study were obtained from the TCGA PanCancer cohort, accessible through the cBioPortal database (https://www.cbioportal.org). The scRNA-seq raw data and metadata have been submitted to the Gene Expression Omnibus database (GSE292109). All other raw data generated in this study are available from the corresponding author upon request.

Authors’ Disclosures

C.-Y. Li reports grants from the NIH during the conduct of the study. No disclosures were reported by the other authors.

Authors’ Contributions

D. Pan: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Jiao: Resources, validation, investigation, methodology. Y. Zhu: Data curation, software, visualization. M. Hu: Investigation, methodology. F. Li: Resources, supervision, methodology, project administration. J. Yu: Resources, supervision. C.-Y. Li: Conceptualization, resources, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing.