TGF-β1 drives neutrophil extracellular traps formation to promote CD8+ T cell exhaustion via the ERK-c-Fos-JunB axis, mediating gastric cancer immunotherapy resistance

Abstract

Background

Resistance to anti-programmed cell death protein-1 (PD-1) treatment in gastric cancer (GC) is closely associated with an immunosuppressive tumor microenvironment. However, the role of neutrophils in resistance to anti-PD-1 therapy remains unclear.

Methods

Single-cell RNA sequencing was performed on tumor samples from patients with advanced GC receiving anti-PD-1 therapy to identify neutrophil subsets associated with neutrophil extracellular traps (NETs). Multilevel experimental validation was conducted using multiomics analysis, flow cytometry, multiplex immunofluorescence, and in vitro co-culture. Therapeutic strategies targeting NETs and CD8⁺ T-cell exhaustion were evaluated in a mouse model of YTN16 tumors.

Results

We identified a NETs-associated neutrophil subset enriched in patients with GC resistant to anti-PD-1 treatment. This subset was marked by CD177, and it exhibited a high potential for NETs release. Peripheral blood NETs levels and CD177⁺ neutrophil ratios in patients with GC act as markers for evaluating the efficacy of PD-1 inhibitors. Furthermore, transforming growth factor-β1 (TGF-β1), which was highly expressed in GC and spatially colocalized with CD177⁺ neutrophils, might induce neutrophils to release NETs via the Smad3-NFE2 axis. NETs promoted CD8⁺ T cell exhaustion by activating the MEK/ERK-c-Fos/JunB axis, as evidenced by increased PD-1/TIM3 expression and reduced interferon-gamma/tumor necrosis factor-alpha secretion. In vivo experiments confirmed that targeted inhibition of NETs formation using DNase I or TGF-β1 inhibitors significantly suppressed tumor growth and CD8⁺ T cell exhaustion. Notably, the MEK inhibitor trametinib reversed the immunosuppressive microenvironment associated with CD8⁺ T cell exhaustion and synergistically enhanced the antitumor efficacy with anti-PD-1 therapy.

Conclusions

TGF-β1 drives CD177⁺ neutrophils to release NETs, which induce CD8⁺ T cell exhaustion via the ERK-c-Fos-JunB pathway, thereby mediating resistance to anti-PD-1 treatment in GC. Furthermore, targeting NETs formation and combining trametinib with PD-1 inhibitors can significantly reverse CD8⁺ T cell exhaustion, exert synergistic antitumor effects, and offer a potential therapeutic strategy for overcoming resistance to anti-PD-1 therapy in GC.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Resistance to anti-programmed cell death protein-1 (PD-1) therapy in gastric cancer is linked to an immunosuppressive tumor microenvironment, but the role of neutrophil extracellular traps (NETs) and their specific neutrophil subsets in driving CD8⁺ T cell exhaustion remains unclear.

WHAT THIS STUDY ADDS

• We identified a CD177⁺ NETs-associated neutrophil subset enriched in resistant patients, and show that transforming growth factor-β1 induces these neutrophils to release NETs, which activate the ERK-c-Fos-JunB axis in CD8⁺ T cells, leading to exhaustion and immunotherapy resistance.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Plasma NETs levels and circulating CD177⁺ neutrophil proportions may serve as predictive markers for immunotherapy response. Targeting NETs formation or the MEK/ERK pathway, particularly with trametinib in combination with anti-PD-1 therapy, represents a promising strategy to overcome resistance in gastric cancer.

Introduction

Gastric cancer (GC) remains a leading cause of cancer-related deaths worldwide, with high incidence and mortality rates in East Asia.¹ The advent of immune checkpoint inhibitors (ICIs), particularly programmed cell death protein-1 (PD-1) inhibitors, has revolutionized the treatment for advanced GC (AGC).² Studies such as CheckMate-649 and ORIENT-16 have established the role of a combination of PD-1 inhibitors with chemotherapy in the initial treatment of AGC.^{3 4} However, a significant proportion of patients develop primary resistance to PD-1 inhibitors, which limits their clinical efficacy.⁵ According to the Society for Immunotherapy of Cancer (SITC), primary resistance is defined as disease progression within 6 months of initiating ICIs-based therapy or a stable disease duration of less than 6 months.^{6 7} Therefore, a deeper understanding of the mechanisms underlying immunotherapy resistance is essential to overcome current treatment bottlenecks and achieve precision in treatment.

The tumor microenvironment (TME) significantly influences the efficacy of immunotherapy. Tumor-associated neutrophils (TANs) exhibit marked heterogeneity, adopting either antitumor (N1) or protumor (N2) phenotypes under the influence of the TME.⁸ However, broadly categorizing neutrophils into two groups fails to accurately identify the subpopulations that play specific roles in treatment resistance. Single-cell RNA sequencing (scRNA-seq) is an effective method for understanding the intricacies of cellular heterogeneity in the TME.⁹ In a recent pancancer study, scRNA-seq analysis was performed on tissue and blood samples from 17 cancer types and a pancancer neutrophil atlas was successfully constructed. Neutrophils were subdivided into 10 distinct subpopulations, revealing high heterogeneity and unique differentiation characteristics.¹⁰ Therefore, using scRNA-seq to analyze neutrophil heterogeneity within the GC microenvironment and investigate how specific subpopulations drive and reshape the TME would represent a critical breakthrough in overcoming immunotherapy resistance in GC.

One mechanism by which neutrophils exert their effector functions is by releasing neutrophil extracellular traps (NETs). NETs have been implicated in key regulatory roles in multiple pathological states, including infection,¹¹ autoimmune diseases,¹² and cancer.¹³ Our previous studies confirmed that NETs levels in patients with GC are closely associated with disease diagnosis, treatment response, and prognosis.¹⁴ However, the dynamic changes in NETs during anti-PD-1 therapy and their clinical value remain unclear. Furthermore, NETs participate in the regulation of antitumor immune responses by physically shielding tumor cells from cytotoxic lymphocyte attacks, suppressing T cell function, and inducing the formation of an immunosuppressive microenvironment.¹⁵ CD8⁺ T cells play a pivotal role in antitumor immune responses. However, prolonged exposure to antigens and inflammatory stimuli within the TME gradually drives CD8⁺ T cell exhaustion.¹⁶ The increasing percentage of exhausted CD8⁺ T cells is a key factor contributing to resistance to immunotherapy.¹⁷ Notably, NETs are considered potential drivers of T cell exhaustion. A previous study has confirmed that NETs highly express programmed death ligand 1 (PD-L1), which binds to PD-1 on T cell surfaces and induces T cell exhaustion, thereby establishing an immunosuppressive microenvironment that promotes tumor growth.¹⁸ However, the intracellular signaling cascade by which NETs induce CD8⁺ T cell exhaustion remains unclear. Emerging evidence suggests that transforming growth factor-β (TGF-β), a key immunosuppressive cytokine in the TME, is believed to polarize TANs into a protumor N2 phenotype.⁸ Among its three isoforms, TGF-β1 is the most abundant in the TME and has been shown to be associated with tumor progression, immune evasion, and resistance to immunotherapy.¹⁹ Notably, recent studies have reported that TGF-β1 may act as an upstream regulator of NETs formation.^{20 21} However, whether TGF-β1 specifically drives the release of NETs from distinct neutrophil subsets and how this process contributes to CD8⁺ T cell exhaustion and anti-PD-1 resistance in GC remain poorly understood.

In this study, we used scRNA-seq to map the single-cell landscape of the GC microenvironment. We successfully identified a NETs-associated neutrophil subset closely linked to immunotherapy resistance. We then investigated the process by which TGF-β1 stimulates neutrophils to release NETs in the GC microenvironment. Subsequently, we elucidated the mechanism underlying the interaction between NETs and CD8⁺ T cells. Finally, using a mouse model of subcutaneous GC, we systematically evaluated the effects of targeting key signaling nodes on tumor growth and CD8⁺ T cell exhaustion, providing insights into reversing CD8⁺ T cell exhaustion and overcoming resistance to PD-1 inhibitor therapy.

Materials and methods

Human subjects and clinical specimens

Endoscopic biopsy specimens from primary tumors were collected from six patients with AGC scheduled for anti-PD-1 therapy after obtaining written informed consent. According to the SITC guidelines’ classification of primary resistance, patients were divided into sensitive and resistant groups.^{6 7} Blood samples were collected from patients with GC and from healthy participants.

Mice

6–8 weeks old female C57BL/6 mice (Jicui Yao Kang Biotechnology) were maintained in a specific pathogen-free environment. Mice were randomized into experimental groups. All animals were housed under identical conditions. No expected or unexpected adverse events were observed. No significant body weight loss (>20%), tumor ulceration, or treatment-related deaths were noted.

Compounds

The following compounds were used in this study: SB431542 (HY-10431, MCE), T-5224 (HY-12270, MCE), trametinib (S2673, Selleck), SCH772984 (S7101, Selleck), BIRB 796 (S1574, Selleck), and SP600125 (S1460, Selleck).

Preparation of single-cell suspension

Fresh gastric biopsy tissue was cut into fragments smaller than 0.5 mm³ and digested for 30 min at 37°C in RPMI 1640 medium (C11875500BT, Gibco) containing 0.04% bovine serum albumin (BSA) (G5001, Servicebio) and 0.2% type II collagenase (17101015, Gibco). The suspension was filtered through a 40 µm filter, centrifuged, and treated with erythrocyte lysis buffer. Cell viability was assessed using trypan blue staining.

scRNA-seq and data processing

The BD Rhapsody platform was used for single-cell capture and library preparation. Sequencing was performed by OE Biotech (Shanghai, China). Quality control was performed using Scanpy (V.1.10.4),²² retaining cells with 200–7,500 genes, with 1,000–50,000 unique molecular identifiers (UMIs), where the proportion of UMIs mapped to mitochondrial genes <50%, and showing <5% hemoglobin gene expression. Doublets were removed using DoubletDetection (V.4.3.0). Data were normalized and dimensionality reduction and clustering were performed using Scanpy. The top 2,000 highly variable genes were subjected to principal component analysis, and batch effects were adjusted using Harmonypy (V.0.0.10). Uniform manifold approximation and projection was used for visualization.

Dimension reduction with cluster analysis and enrichment analysis

Marker genes were identified using the Wilcoxon signed-rank test. Cell types were annotated using CellTypist (V.1.6.3)²³ in conjunction with manual curation based on canonical markers. Cells were classified into eight major groups: epithelial (Epi) cells, T cells, B cells, neutrophils (Neu), macrophages (Mφ), fibroblasts (Fib), endothelial cells (EC), and mast cells. For focused neutrophil and T cell analysis, the subsets were reclustered. Differentially expressed genes were defined as those having q-value ≤0.05 and |log2FC|>1. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted using hypergeometric tests. Gene set variation analysis (GSVA)²⁴ and gene set enrichment analysis (GSEA)²⁵ were conducted using gene sets from the KEGG and GO databases.

Monocle2 pseudotime analysis

Pseudotime trajectories were constructed using the Monocle2 package (V.2.9.0).²⁶ The raw count matrix from the Seurat object was converted into a CellDataSet object using the importCDS function. Genes with q-value <0.01 were selected as ordering genes for trajectory reconstruction. Dimensionality reduction was performed using reduceDimension, followed by pseudotime inference using orderCells with default parameters.

SCENIC analysis

To identify key transcriptional factors (TFs) in CD177⁺ neutrophils, TF activity was inferred using SCENIC (V.1.2.4).^{27 28} Gene regulatory networks were constructed using GRNBoost. TF-binding motif enrichment was assessed using RcisTarget to identify over-represented motifs in gene sets. The regulon activity per cell was scored using AUCell (V.1.8.0). To assess cell type specificity of predicted regulons, the Regulon Specificity Score (RSS) was calculated based on Jensen-Shannon divergence between binary regulon activity profiles and cell type annotations.

The Cancer Genome Atlas data analysis

RNA-seq datafor stomach adenocarcinoma (STAD), including tumor, adjacent normal, and GTEx normal samples, were downloaded from UCSC XENA. Differential expression was analyzed using ggplot2. Kaplan-Meier curves were generated and assessed using the log-rank test. Immune scores were calculated using the ESTIMATE algorithm.²⁹ Immune cell infiltration was quantified using the GSVA package.²⁴

Cell culture

GC cell lines (HGC-27 (ZQ0192), AGS (ZQ0240), and MKN-45 (ZQ0457)) and Jurkat T cells (ZQ0083) were purchased from Shanghai ZQXZ-Bio. Cells were cultured in RPMI 1640 or F-12K medium (21127022, Gibco) supplemented with 10% fetal bovine serum (FBS, A5256701, Gibco), 100 U/mL penicillin, and 100 µg/mL streptomycin (C0222, Beyotime) at 37°C with 5% CO₂. The YTN16 cell line was provided by SN. It was cultured in DMEM medium (C11995500BT, Gibco) containing 10% FBS at 37°C with 5% CO₂.

Isolation of primary neutrophils

Human neutrophils were separated from EDTA-anticoagulated peripheral blood via density gradient centrifugation in human neutrophil isolation medium (LZS11131, TBD Science). Blood was carefully layered onto the separation medium and centrifuged at 600 g for 30 min, and the neutrophil layer was collected, washed, and resuspended in RPMI 1640 complete medium. Mouse neutrophils were separated from mouse bone using mouse neutrophil isolation medium (TBD2013NM, TBD Science). After centrifugation at 600 g for 30 min, the neutrophil layer was collected and red blood cells were lysed, washed, and resuspended in RPMI 1640 complete medium. The purity was verified using flow cytometry (>90%).

Isolation and activation of human CD3⁺ and CD8⁺ T cells

Human CD3⁺ and CD8⁺ T cells were isolated from peripheral blood using immunomagnetic negative selection (19661, 19663, STEMCELL). Briefly, whole blood was incubated with an antibody cocktail and magnetic beads, followed by magnetic separation to obtain T cells (>95%). The isolated T cells were activated for 72 hours in complete RPMI 1640 medium with anti-CD3/CD28 antibodies (10971, STEMCELL) and interleukin (IL)-2 (78036.1, STEMCELL) before use in subsequent experiments.

NETs isolation and quantification

Human neutrophils (1×10⁶) were stimulated with 100 nM phorbol 12-myristate 13-acetate (PMA) (P1585, Sigma) at 37°C for 4 hours to induce NETs formation. To remove cell debris, the suspension was centrifuged at 450 g, 4°C, for 10 min. To pellet the NETs, the resultant supernatant was centrifuged at 18,000 g, 4°C, for 10 min. The pellet was resuspended in phosphate-buffered saline (PBS), and the DNA concentration was determined by NanoDrop spectrophotometer. NETs were stored at −80°C.

In vitro NETs formation assays

Purified human or mouse neutrophils (4×10⁵ cells) were seeded on coverslips coated with poly-L-lysine. Cells were treated with TGF-β1 (20 ng/mL, 100–21, Peprotech), PMA (100 nM), and DNase I (100 µg/mL, 10104159001, Roche) for 4 hours. After fixing, permeabilizing, and blocking, the cells were incubated with primary antibodies against citrullinated histone H3 (CitH3) and myeloperoxidase (MPO) overnight at 4°C. Following washing, the appropriate fluorophore-conjugated secondary antibodies were applied and the nuclei were stained with 4',6-diamidino-2-phenylindole (DAPI). Additionally, extracellular DNA was detected using SYTOX Green (5 µM, S7020, Invitrogen) staining. NETs were visualized using fluorescence microscopy and quantified as the percentage of NETs-releasing cells and the percentage of CitH3⁺ coverage area using the ImageJ software.

In vitro T cell-NETs co-culture

Activated human T cells were cultured with purified NETs (0–40 ng/mL) for 48 hours after pretreatment with inhibitors for 1 hour as required. Following manufacturer’s instructions, T cell proliferation was assessed by 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester(CFSE) dilution using flow cytometry (423801, BioLegend), and intracellular reactive oxygen species (ROS) levels were measured using 2',7'-dichlorodihydrofluorescein diacetate (DCFH-DA) in flow cytometry (S0033S, Beyotime).

Flow cytometry analysis

For in vivo experiments, tumor tissues were minced into 1–2 mm³ fragments and digested with digestion medium containing collagenase II (1 mg/mL) and DNase I (0.1 mg/mL) for 1 hour at 37 °C. The digested material was filtered through a 70 µm cell strainer and washed, and the red blood cells were lysed. After counting, 2×10⁶ cells per sample were stained as follows: cells were first blocked with anti-Fc receptor antibody for 10 min at 4°C, following which they were stained with surface-phenotyping antibodies in the dark for 30 min at 4°C. Tubes included unstained controls, single stain controls, fluorescence minus one (FMO) controls, and fully stained samples. After staining, cells were washed twice and resuspended in buffer. 7-AAD was added 5 min before acquisition using flow cytometer.

For in vitro experiments, T cells from each group were stained with fluorochrome-conjugated surface antibodies against CD3, CD4, CD8, PD-1, and TIM3 for 20 min at 4°C. For intracellular cytokine detection, after 6 hours of stimulation with PMA/ionomycin (00–4970-93, eBioscience) and brefeldin A/monensin (00–4980-93, eBioscience), T cells were surface-stained, fixed, permeabilized (88–8824-00, eBioscience), and stained with antibodies against tumor necrosis factor-alpha (TNF-α) and interferon-gamma (IFN-γ) and analyzed using flow cytometer.

The data were analyzed using FlowJo (V.10.8.1). The gating strategy was defined as follows: debris was excluded based on forward scatter-area (FSC-A) and side scatter-area (SSC-A), doublets were removed using FSC-A and forward scatter-height (FSC-H), and dead cells were excluded based on 7-AAD or fixable viability stain-positive events. Within the live singlet gate, CD45⁺ cells or CD3⁺ T cells were selected, and CD4⁺ and CD8⁺ subpopulations were identified. PD-1, TIM3, TNF-α, and IFN-γ were subsequently evaluated on these CD4⁺ or CD8⁺ T cell subsets. The gates were set using blank and FMO controls. Compensation was performed using single-stained controls prepared in parallel with each fluorophore. Representative plots illustrating the gating strategies are shown in online supplemental figures S1,S2. Detailed information on the antibodies used in this study is provided in online supplemental table 1.

Cell sorting

The obtained neutrophils were incubated with anti-human CD16/32 antibody at 4°C for 20 min, following which they were incubated with fluorescently tagged antibodies at 4°C for 20 min (CD66b and CD177). 7-AAD was added to stain the dead cells. Cells were sorted into CD177⁺ and CD177⁻ neutrophils. The sorted CD177⁺ neutrophils were divided into control, TGF-β1 treatment (20 ng/mL), and TGF-β1+SB431542 treatment groups, which were pretreated with SB431542 (10 µM) for 1 hour, followed by stimulation with TGF-β1 (20 ng/mL). All groups were cultured for 4 hours at 37°C and 5% CO₂. At the end of the experiment, cells were harvested for protein extraction.

Western blotting

Cells were lysed using RIPA (R0020, Solarbio) containing protease and phosphatase inhibitors (IP0280, P1260, Solarbio). The Bicinchoninic Acid Assay (BCA) was used to quantify protein concentration (P0012S, Beyotime). Following sodium dodecyl sulfate-polyacrylamide gel electrophoresis, proteins were transferred onto a polyvinylidene fluoride and blocked. The membrane was incubated overnight at 4°C with primary antibodies against PD-1, TIM3, c-Fos, JunB, AKT, p-AKT, p65, p-p65, ERK, p-ERK, E-cadherin, N-cadherin, and β-actin. After the membranes were incubated with secondary antibodies conjugated with horseradish peroxidase (HRP), detection was performed using enhanced chemiluminescence.

Transwell migration and invasion assays

The upper chambers of Transwell inserts (3422, Corning) were seeded with 5×10⁴ GC cells in serum-free medium with or without Matrigel (1:8 dilution, 40 183ES08, YEASEN). RPMI 1640 complete medium with or without NETs (5 ng/mL) and DNase I (100 µg/mL) was present in the lower chambers. After 24 hours, cells were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet (G1063, Solarbio), photographed, and counted.

siRNA-mediated knockdown

Small interfering RNA (siRNA) targeting MAP2K1/2 (MEK1/2) or control siRNA (DN001, D-Nano) was transfected into Jurkat T cells using a lipid-based transfection reagent. After 24 hours, cells were treated with NETs (5 ng/mL) for 48 hours, and PD-1 and TIM3 expression was assessed by flow cytometry. Knockdown efficiency was confirmed using western blotting. The siRNA sequences are listed in the online supplemental table 2.

In vivo mouse models and treatment

5×10⁶ YTN16 cells were subcutaneously injected into female C57BL/6 mice. Once tumors reached approximately 5 mm in diameter, mice were randomized into groups (n=6/group). For NETs-targeting studies, mice were administered PBS (control), DNase I (10 mg/kg/day, intraperitoneally (i.p.)), SB431542 (10 mg/kg/day, i.p. S1067, Selleck), or the combination for 14 days. For MEK inhibitor combination studies, mice received PBS, anti-PD-1 antibody (3 mg/kg, two times per week, i.p. 11430, Innovent), trametinib (1 mg/kg/day, i.p. S2673, Selleck), or the combination for 15 days. Tumor dimensions were measured every 3 days, and tumor volume was calculated as length × width² × 0.5. At the experimental endpoint, the mice were euthanized and samples were collected.

Immunohistochemistry and immunofluorescence

Paraffin-embedded and formalin-fixed sections were subjected to antigen retrieval before staining with antibodies. Next, the sections were treated with primary antibodies against TGF-β1 or MPO for immunohistochemistry, followed by the addition of HRP-conjugated secondary antibodies. Diaminobenzidine was used for signal development (ZLI-9018, ZSGB-BIO), and hematoxylin was used as a counterstain. H-scores were computed for semi-quantification. Sections were stained with antibodies against CitH3, CD177, MPO, TGF-β1, CD8, and PD-1 for immunofluorescence using tyramide signal amplification for multiplexing when required. DAPI was used to stain the nuclei (Beyotime, C1005), and a fluorescence microscope was used to capture the images.

ELISA

Patients and healthy volunteers provided fasting blood samples, which were centrifuged at 1,000 g for 10 min, and stored at −80°C. ELISA kits were used to determine serum TGF-β1 concentrations in accordance with the manufacturer’s instructions (EK981, MULTI SCIENCES). NETs levels were measured using MPO–DNA complexes (11774425001, Roche). Anti-MPO antibody was applied to a 96-well plate overnight at 4°C, blocked with 1% BSA for 2 hour, and then incubated with plasma samples for 2 hour. After washing, HRP-labeled anti-DNA antibody was added and incubated for 2 hour. Next, the ABTS substrate was added and incubated for 40 min. Absorbance was measured at 405 nm. NETs content was expressed as the mean absorbance value.

Statistical analysis

R (V.4.1.2), Python (V.3.10.16), and GraphPad Prism (V.8.0) were used for statistical analysis. Data are presented as mean±SD or median with IQR. The Wilcoxon signed-rank test, paired t-test, or Student’s t-test was used for comparisons. One-way analysis of variance and Tukey’s post hoc test were used to compare groups. Correlations were analyzed using the Pearson or Spearman’s rank test. Survival data were analyzed using Kaplan-Meier curves with log-rank test. Statistical significance was defined as p<0.05 (*p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001). Schematic diagrams were created using BioGDP (GDP2026NMET3E).³⁰

Results

A single-cell atlas of patients with AGC receiving anti-PD-1 therapy

To investigate the TME heterogeneity associated with anti-PD-1 response, we performed scRNA-seq on pretreatment biopsies from six patients with AGC (figure 1A). The clinical characteristics of the six patients are presented in online supplemental table 3. After quality control, 62,375 single-cell transcriptomes were obtained. Patients were stratified into anti-PD-1 treatment-resistant (GC1) and sensitive (GC2–6) groups according to the SITC criteria, with corresponding CT images confirming resistance or sensitivity (figure 1B,C). Eight major cell lineages were identified based on canonical markers: epithelial cells (Epi), B cells (B), T cells (T), macrophages (Mφ), neutrophils (Neu), fibroblasts (Fib), endothelial cells (EC), and mast cells (Mast) (figure 1D,E). A heatmap of the top five markers per cluster confirmed the accuracy of the cell-type annotations (figure 1F). The cellular composition varied substantially among patients (figure 1G). The resistant group showed a higher proportion of B cells and lower proportions of neutrophils, T cells, and macrophages than the sensitive group (figure 1H,I), suggesting that these immune populations may contribute to immunotherapy resistance in patients with AGC.

Figure 1. Single-cell transcriptomic atlas of patients with AGC receiving anti-PD-1 therapy. (A) Workflow of scRNA-seq for six patients with AGC receiving anti-PD-1 therapy. (B) UMAP plot showing single-cell classification by patient ID. (C) Contrast-enhanced CT images of target lesions before and after treatment in an immunotherapy-resistant patient (GC1) and an anti-PD-1-sensitive patient (GC4). (D) Bubble plot of signature genes identified by scRNA-seq. (E) UMAP plot of cellular composition in six patients with AGC prior to anti-PD-1 therapy. Color-coded clusters represent cell types: Epi, T, B, Neu, Mφ, Fib, EC, and Mast. (F) Heatmap of the top five genes for each cell type. (G) Percentage of each cell type across the six samples. (H) Percentage of each cell type in the sensitive and resistant groups. (I) UMAP plot of cell types in the sensitive and resistant groups. AGC, advanced gastric cancer; B, B cells; EC, endothelial cells; Epi, epithelial cells; Fib, fibroblasts; GC, gastric cancer; Mφ, macrophages; Mast, mast cells; Neu, neutrophils; PD-1, programmed cell death protein-1; scRNA-seq, single-cell RNA sequencing; T, T cells; UMAP, uniform manifold approximation and projection.

Identification of a NETs-associated neutrophil subset in the GC

Considering the differential abundance of neutrophils among the response groups, we investigated their heterogeneity in detail. Reclustering of 6,968 neutrophils resolved eight distinct subsets (Neu1-8) with considerable interpatient variability (figure 2A,B). Notably, Neu4-6 were enriched in the resistant group, whereas Neu1-3 and Neu7-8 were predominant in the sensitive group (figure 2C). Functional annotation, based on marker genes and GO enrichment, identified these subsets as Neu1 (proinflammatory), Neu2 (metabolically active), Neu3 (effector chemotactic), Neu4 (stress-adapted), Neu5 (granule release, defined by the marker genes ALPL, CST7, S100A12, MMP9, and CD177), Neu6 (anabolic), Neu7 (ion homeostasis), and Neu8 (matrix remodeling) (figure 2D). To further ascertain the prognostic value of the identified neutrophil subsets, we performed survival analysis on the top 10 marker genes for eight neutrophil subsets using the GEPIA2 database.³¹ Among all the subsets, only the Neu5 gene set was significantly associated with the overall survival (OS) of patients (online supplemental figure S3A-H). This finding indicates that Neu5 has prognostic relevance.

Figure 2. Identification and characterization of a NETs-associated neutrophil subset in the GC microenvironment. (A) UMAP plots of neutrophil subpopulations by patient ID (GC1-6). (B) UMAP plot showing eight neutrophil cluster subpopulations (Neu1–8). (C) Bar charts of neutrophil subpopulations in the anti-PD-1-sensitive and anti-PD-1-resistant groups. (D) Bubble plot of marker genes across eight neutrophil subpopulations, with GO enrichment entries on the right. (E) UMAP plot demonstrating expression of NETs formation genes MMP9 and PADI4 in the Neu5. Violin plots of NETs formation scores (F) and NETs-related gene set scores (G) across neutrophil subpopulations. (H) Monocle pseudotime trajectory of neutrophils colored by clustering. (I) Pseudotime trajectory of neutrophils colored by subpopulation (arrow indicates divergence direction). (J) UMAP plot showing high expression of immunosuppressive genes and the marker gene CD177 in the Neu5. (K) GSVA enrichment analysis revealed significantly enriched signaling pathways in the Neu5. GC, gastric cancer; GO, Gene Ontology; GSVA, gene set variation analysis; mRNA, messenger RNA; NETs, neutrophil extracellular traps; Neu, neutrophils; PD-1, programmed cell death protein-1; TGF, transforming growth factor; UMAP, uniform manifold approximation and projection.

Our previous studies also demonstrated that IL-8-CXCR1/2 axis-mediated NETs production promotes the migration and invasion of GC cells, and that NETs level is an independent prognostic factor influencing OS of patients with GC.^{32 33} Therefore, we aimed to identify the specific subset of NETs-associated neutrophils. The Neu5 cluster uniquely displayed high expression of key NETs-related genes, including PADI4 and MMP9 (figure 2E). AddModuleScore analysis confirmed that Neu5 showed the highest NETs-forming potential³⁴ and expression of NETs-related gene signatures³³ (figure 2F,G). Monocle pseudotime analysis positioned Neu1 at the origin of the trajectory, with one branch differentiating into terminally differentiated Neu5, which is characterized by high degranulation and NETs-forming potential, and another branch differentiating into Neu6 (figure 2H,I). Neu5 also uniquely expressed immunosuppressive molecules (ALPL, MME, CST7) and the marker gene CD177 (figure 2J). GSVA further linked Neu5 to NETs formation and signaling pathways such as JAK-STAT, VEGF, and TGF-β signaling pathways (figure 2K). As shown in online supplemental figure S3I, the highest gene expression difference (gene diff) value was observed for CD177 in the Neu5 subset, indicating that it was the most representative and discriminatory marker gene for this cluster. We then correlated CD177 expression in the Neu5 subset with the NETs formation score and the NETs-associated gene set score (online supplemental figure S3J, K). Moreover, CD177⁺ Neu5 cells exhibited significantly higher NETs scores than CD177⁻ cells (online supplemental figure S3L, M). Hence, we selected CD177 as the Neu5 marker for subsequent experiments. In summary, we identified Neu5, marked by CD177, as an immunosuppressive, NETs-forming neutrophil subset potentially involved in mediating anti-PD-1 resistance.

Functional and clinical characterization of CD177⁺ NETs-associated neutrophils

To evaluate the clinical significance of CD177⁺ neutrophils (Neu5) in the resistance of patients with GC to immunotherapy, we collected blood samples from patients with AGC. Flow cytometry did not reveal any difference in total CD66b⁺ neutrophil proportions between the response groups (figure 3A). However, the CD177⁺ subset within CD66b⁺ cells had expanded significantly in the resistant patients, suggesting that the CD177⁺ subpopulation is associated with immunotherapy resistance (figure 3B,C). Notably, sorted CD177⁺ neutrophils spontaneously formed abundant CitH3⁺MPO⁺ NETs structures in vitro, with significantly higher NETs release rates than CD177⁻ cells (figure 3D,E). Immunofluorescence confirmed that CD177⁺ NETs-associated neutrophils and CitH3⁺ NETs colocalized in the GC tissues of therapy-resistant patients and that these colocalizations were spatially proximal to tumor cells (figure 3F). To further validate the clinical relevance of CD177 in GC, we analyzed messenger RNA (mRNA) expression data from the Kaplan-Meier Plotter database.³⁵ Kaplan-Meier survival analysis revealed that high CD177 expression was significantly associated with worse OS in the entire GC cohort and HER-2-negative subgroup, indicating that elevated CD177 levels predict poor prognosis for GC (online supplemental figure S3N, O).

Figure 3. Functional and clinical validation of CD177⁺ NETs-associated neutrophils. (A) Flow cytometry detection of CD66b⁺ neutrophils in peripheral blood from patients with GC. Flow plots (B) and quantitative analysis (C) of the proportion of CD177⁺ cells within the CD66b⁺ neutrophil population. (D) Quantification of NETs release by sorted CD177⁺ and CD177⁻ human neutrophils. (E) Immunofluorescence images of sorted CD177⁺ and CD177⁻ neutrophils. Cells were stained for DAPI (blue), CitH3 (green), and MPO (red). Scale bar: 40 µm. (F) Immunofluorescence staining of tumor tissue from an anti-PD-1 resistant patient, showing infiltration of CD177⁺ (yellow) neutrophils and NETs marker CitH3 (green) in the GC TME. DAPI (blue). Scale bar: 50 µm. (G) Immunofluorescence staining for DAPI (blue), CitH3 (green) and CD177 (red) of YTN16 tumor model. Scale bar: 50 µm. (H–I) Changes in plasma NETs levels in patients with GC receiving anti-PD-1 therapy: (H) PR group. (I) PD group. Kaplan-Meier curves for PFS (J) and OS (K) in patients with GC receiving anti-PD-1 therapy, stratified by baseline plasma NETs levels. CitH3, citrullinated histone H3; DAPI, 4',6-diamidino-2-phenylindole;GC, gastric cancer; MPO, myeloperoxidase; NETs, neutrophil extracellular traps; ns, not significant; OD, optical delnsity; OS, overall survival; PD, progressive disease; PD-1, programmed cell death protein-1; PFS, progression-free survival; PR, partial response; TME, tumor microenvironment. ** p<0.01, *** p<0.001.

Furthermore, a recent study reported that TGF-β1 promotes NETs formation in hepatocellular carcinoma (HCC).²¹ Combined with the results of our GSVA analysis, which showed that the TGF-β signaling pathway is enriched in the Neu5 neutrophil subset associated with NETs (figure 2K), we speculated that TGF-β1 may promote the release of NETs by the CD177⁺ neutrophil subset in the GC microenvironment. Hence, we conducted in vivo experiments using a TGF-β1 receptor inhibitor, SB431542. We established a subcutaneous GC mouse model using YTN16 cells. SB431542 reduced the formation of CD177⁺CitH3⁺ NETs structures in mouse GC tissues, preliminarily suggesting that the TGF-β1 pathway may be a key regulator of NETs formation (figure 3G). Clinically, plasma NETs levels decreased post-treatment in patients with partial response but increased in patients with progressive disease (figure 3H,I). High baseline NETs levels correlated with shorter progression-free survival and OS (figure 3J,K).

TGF-β1 is upregulated in GC and induces CD177⁺ neutrophils to release NETs

To investigate the link between TGF-β1 and NETs, we performed multilevel validation. Immunohistochemical staining revealed significantly higher TGF-β1 expression in GC tissues than in adjacent tissues (figure 4A,B). Serum TGF-β1 levels were significantly higher in patients with GC than in healthy individuals (figure 4C). Furthermore, patients in the resistant group showed significantly higher serum TGF-β1 levels than patients in the anti-PD-1-sensitive group (figure 4D). Expression of MPO, a neutrophil marker, was also elevated in the GC tissues, indicating neutrophil infiltration within the GC microenvironment (figure 4E,F). Data from The Cancer Genome Atlas (TCGA)-STAD indicated that TGFB1 mRNA expression levels in GC were higher than in normal tissues (figure 4G). In addition, high TGFB1 expression correlated with poorer OS in the TCGA-STAD cohort (figure 4H) and was positively associated with ESTIMATE immune scores and neutrophil infiltration scores (figure 4I,J). Immunofluorescence analysis demonstrated spatial proximity between TGF-β1 and MPO⁺ neutrophils, as well as colocalization of CD177⁺ neutrophils and CitH3⁺ NETs (figure 4K,L), suggesting that TGF-β1 may be actively involved in mediating NETs release from CD177⁺ neutrophils.

Figure 4. TGF-β1 is upregulated in GC and directly induces NETs release from CD177⁺ neutrophils. (A) IHC images of TGF-β1 expression in GC tissue and matched adjacent normal tissue. Scale bar: 100 µm. (B) H-score for TGF-β1 expression in GC and adjacent tissues. (C) Serum TGF-β1 levels in patients with GC and healthy controls. (D) Serum TGF-β1 levels in anti-PD-1 therapy sensitive and resistant patients. (E) IHC images of MPO expression in GC and adjacent tissue. Scale bar: 100 µm. (F) H-score for MPO expression in GC and adjacent tissues. (G) Analysis of TGFB1 mRNA expression levels in GC and normal tissue from the TCGA database. (H) Kaplan-Meier curve for OS from the TCGA-STAD cohort, stratified by TGFB1 mRNA expression. Correlation analysis between TGFB1 mRNA expression and the ESTIMATE immune score (I) and the neutrophil infiltration enrichment score (J). (K) Immunofluorescence staining of GC tissue showing TGF-β1 (green) and MPO-positive neutrophils (red). DAPI (blue). Scale bar: 50 µm. (L) Multiplex immunofluorescence staining of GC tissue showing the spatial relationship between TGF-β1 (purple), CD177⁺ neutrophils (red), and CitH3⁺ NETs (green). DAPI (blue). Scale bar: 50 µm. (M) Immunofluorescence staining for NETs formation in human neutrophils. CD177⁻ and CD177⁺ neutrophils were stimulated with TGF-β1 (20 ng/mL) for 4 hours (×20). DAPI (blue), CitH3 (green), MPO (red). Scale bar: 40 µm. (N) Quantification of NETs-releasing cells by immunofluorescence staining. CitH3, citrullinated histone H3; DAPI, 4',6-diamidino-2-phenylindole;GC, gastric cancer; IHC, immunohistochemistry; MPO, myeloperoxidase; mRNA, messenger RNA; NETs, neutrophil extracellular traps; OS, overall survival; PD-1, programmed cell death protein-1; STAD, stomach adenocarcinoma; TCGA, The Cancer Genome Atlas; TGF-β1, transforming growth factor-β1; TPM, transcripts per million. *p<0.05, ** p<0.01, *** p<0.001.

To validate this finding in vitro, we screened the TGF-β family members and identified TGF-β1 as the key isoform driving NETs release. Stimulation of human peripheral blood neutrophils with TGF-β1 (20 ng/mL) for 4 hours significantly induced the formation of CitH3⁺MPO⁺ NETs structures, which was inhibited by DNase I treatment (online supplemental figure S4A, B). TGF-β1 dramatically expanded the NETs-releasing cells and NETs coverage area (online supplemental figure S4C-F). This effect was also observed in mouse bone marrow neutrophils (online supplemental figure S4G-J). Based on our previous work, we hypothesized that TGF-β1 targets CD177⁺ NETs-associated neutrophils to induce NETs release. To validate this hypothesis, we isolated CD177⁻ and CD177⁺ neutrophils via flow cytometry and stimulated them with TGF-β1 in vitro. Interestingly, TGF-β1 induced abundant NETs formation in sorted CD177⁺ neutrophils, while CD177⁻ cells retained their morphological integrity (figure 4M,N). This demonstrates that TGF-β1 targets CD177⁺ neutrophils to induce NETs.

TGF-β1 activates the Smad3–NFE2 axis to drive NETosis in CD177⁺ neutrophils

To investigate how TGF-β1 drives NETosis in CD177⁺ neutrophils, we analyzed scRNA-seq dataset using SCENIC analysis. Interestingly, the transcription factor NFE2 displayed the highest RSS in the CD177⁺ neutrophil cluster (online supplemental figure S5A, B), indicating that NFE2 is a key regulator. Considering that a previous study has reported that TGF-β1 signals canonically through Smad3-dependent pathways in neutrophils,²⁰ we hypothesized that, in the GC microenvironment, TGF-β1 promotes NETs release in CD177⁺ neutrophils via the Smad3-NFE2 pathway. We therefore stimulated CD177⁺ neutrophils with TGF-β1 in the absence or presence of the TGF-β1 receptor inhibitor SB431542 and examined the downstream targets using western blotting. Compared with the control group, TGF-β1 significantly increased the levels of p-Smad3/Smad3, NFE2, CitH3, and MPO. Notably, cotreatment with SB431542 markedly reversed these TGF-β1-induced upregulations (online supplemental figure S5C-G).

Targeting of NETs and TGF-β1 suppresses tumor growth and reverses CD8⁺ T cell exhaustion

In C57BL/6 mice, we created a subcutaneous YTN16 tumor model and evaluated the therapeutic potential of focusing on TGF-β1 and NETs. Tumor-bearing mice were treated with PBS (control), DNase I (10 mg/kg/day, i.p.), SB431542 (10 mg/kg/day, i.p.), or a combination of DNase I and SB431542 for 14 days (figure 5A). Both monotherapies significantly suppressed tumor growth, while the combination demonstrated the most potent antitumor effect (figure 5B,C). HE staining revealed increased tumor necrosis and fibrosis in the treatment groups (figure 5D). Immunofluorescence confirmed that both SB431542 and DNase I reduced intratumoral NETs (figure 5E). These results indicate that targeting the TGF-β1 pathway can effectively reduce NETs levels in the GC microenvironment.

Figure 5. Targeting NETs and TGF-β1 pathway suppresses tumor growth and reverses CD8⁺ T cell exhaustion. (A) Treatment regimen for the YTN16 murine subcutaneous GC model. Tumor-bearing mice (n=6/group) were treated with PBS, DNase I, SB431542, or a combination of DNase I and SB431542. (B) Images of tumors on day 28. (C) Tumor growth curves of subcutaneous tumors. (D) HE-stained sections of tumors. Scale bar: 200 µm. (E) Immunofluorescence staining for NETs in tumor. DAPI (blue), CitH3 (green), MPO (red). Scale bar: 100 µm. (F) UMAP plot showing six T cell subpopulations. (G) Bar plot displaying the proportion of T cell subpopulations in the anti-PD-1-sensitive and anti-PD-1-resistant groups. (H) Heatmap showing the expression levels of marker genes for the six major T cell subpopulations. (I) Flow plots of exhaustion markers PD-1 and TIM3 on gated CD8⁺ T cells. (J) Flow plots of PD-1 and TIM3 on gated CD4⁺ T cells. (K) Immunofluorescence staining in tumor sections showing tumor-infiltrating CD8⁺PD-1⁺ exhausted T cells. DAPI (blue), CD8 (green), PD-1 (red). Scale bar: 50 µm. (L) Multiplex immunofluorescence staining in the control group. DAPI (blue), CitH3 (purple), CD8 (green), PD-1 (red). Scale bar: 50 µm. CitH3, citrullinated histone H3; DAPI, 4',6-diamidino-2-phenylindole; GC, gastric cancer; i.p., intraperitoneally; MPO, myeloperoxidase; NETs, neutrophil extracellular traps; NKT, natural killer T cell; PBS, phosphate-buffered saline; PD-1, programmed cell death protein-1; s.c., subcutaneously; T_Cyt, cytotoxic T cell; T_EX, exhausted T cell; TGF-β1, transforming growth factor-β1; T_MAIT, mucosal-associated invariant T cell; T_Prolif, proliferating T cell; UMAP, uniform manifold approximation and projection. *p<0.05, ** p<0.01, *** p<0.001.

Considering the differences in T cell abundance across groups, we performed a reclustering analysis of 15,191 T cells and identified six distinct T cell subpopulations (figure 5F). Notably, CD8⁺ exhausted T cell accounted for a larger proportion in the immunotherapy-resistant group (figure 5G). The marker genes for the six major T cell lineages are shown in figure 5H. Considering that NETs have immunosuppressive effects and that our scRNA-seq results indicated an increased proportion of CD8⁺ exhausted T cells in treatment-resistant patients,³⁶ we examined tumor-infiltrating T cells. Flow cytometry revealed that both DNase I and SB431542 monotherapies significantly reduced the expression of T cell exhaustion markers, PD-1 and TIM3, with synergistic effects in combination therapy (figure 5I, online supplemental figure S6A, B). However, no significant changes were observed in CD4⁺ T cells (figure 5J, online supplemental figure S6C, D). Similarly, immunofluorescence staining revealed a reduction of CD8⁺PD-1⁺ T cells following treatment (figure 5K), suggesting that targeted degradation of NETs and inhibition of the TGF-β1 signaling pathway effectively reverses CD8⁺ T cell exhaustion. Furthermore, immunofluorescence colocalization revealed that CitH3⁺ NETs structures were spatially adjacent to exhausted CD8⁺PD-1⁺ T cells. This suggests that NETs may induce CD8⁺ T cell exhaustion (figure 5L). Finally, monotherapies reversed the expression of epithelial–mesenchymal transition markers (decreased N-cadherin and increased E-cadherin expression), and a synergistic effect was observed in the combination group (online supplemental figure S6E-G). These data demonstrate that NETs drive CD8⁺ T cell exhaustion and tumor progression in vivo, and that targeting this axis can reverse exhaustion and inhibit tumor growth.

NETs directly induce CD8⁺ T cell exhaustion and dysfunction

We created an in vitro co-culture system to demonstrate the connection between NETs and T cell exhaustion. Purified NETs were isolated from human neutrophils (online supplemental figure S7A). Human T cells were co-cultured with various concentrations of NETs for 48 hours. Results demonstrated that NETs significantly upregulated exhaustion markers, PD-1 and TIM3, with maximum upregulation observed at 5 ng/mL (online supplemental figure S7B-G). Therefore, we selected 5 ng/mL NETs for subsequent experiments. Furthermore, NETs also impaired CD8⁺ T cell function: reduced TNF-α and IFN-γ levels (online supplemental figure S7H-L), suppressed proliferation (online supplemental Figure S7M, N), and elevated intracellular ROS levels (online supplemental figure S7O-Q). The increase in PD-1 and TIM3 protein levels in CD8⁺ T cells after NETs treatment was verified using western blotting (online supplemental figure S7R-T). Additionally, NETs significantly enhanced the migration and invasion of GC cells in transwell assays, which was abrogated by DNase I (online supplemental figure S8A-H). These findings demonstrate that NETs directly induce CD8⁺ T cell exhaustion and dysfunction while increasing tumor aggressiveness.

NETs drive CD8⁺ T cell exhaustion via the ERK-c-Fos-JunB axis

RNA-seq analysis was performed to thoroughly analyze the mechanisms via which NETs induce CD8⁺ T cell exhaustion. NETs treatment significantly upregulated the expression of exhaustion-related genes (PDCD1, CTLA4, JUNB, FOS, ENTPD1, GZMB) and downregulated the expression of memory/stemness genes (TCF7, CXCR3, EOMES) (figure 6A). Significant enrichment of pathways such as PI3K-AKT, NF-κB, and MAPK pathways was found using KEGG analysis (figure 6B). GSEA confirmed the significant activation of the MAPK pathway (figure 6C,D). Western blotting confirmed that NETs significantly increased the phosphorylation of ERK and P65, but not that of AKT (figure 6E–H). However, the NF-κB pathway inhibitor (QNZ) failed to reverse CD8⁺ T cell exhaustion (online supplemental figure S9A-C). Therefore, we focused on the MAPK pathway, which involves ERK, JNK, and p38. CD8⁺ T cells were pretreated with different inhibitors for 1 hour and then co-cultured with NETs for 48 hours. We found that MEK/ERK inhibitors (SCH772984 and trametinib) reversed NETs-induced upregulation of c-Fos, JunB, p-ERK, PD-1, and TIM3 in CD8⁺ T cells, whereas p38 (BIRB 796) and JNK (SP600125) inhibitors were ineffective (figure 6I–N). Additionally, flow cytometry confirmed that MEK/ERK inhibition significantly reduced CD8⁺PD-1⁺ and CD8⁺TIM3⁺ T cell populations induced by NETs. However, BIRB 796 and SP600125 did not show any significant inhibitory effects (figure 6O–Q).

Figure 6. NETs drive CD8⁺ T cell exhaustion via the ERK-c-Fos-JunB pathway. (A) Volcano plot showing differentially expressed genes from RNA-seq of CD8⁺ T cells treated with or without NETs (n=3). (B) Bubble plot showing the top 15 enriched KEGG pathways. GSEA plots of the MAPK signaling pathway (C) and MAP kinase phosphatase activity (D). (E) Western blotting of proteins in the PI3K-AKT, NF-κB, and MAPK pathways. Quantification of the relative phosphorylation of p-AKT/AKT (F), p-P65/P65 (G), and p-ERK/ERK (H). (I) Western blotting of c-Fos, JunB, p-ERK/ERK, TIM3, and PD-1 expression in CD8⁺ T cells pretreated with specific MAPK pathway inhibitors followed by co-culture with NETs. (J–N) Quantification of relative protein expression levels of c-Fos (J), JunB (K), p-ERK/ERK (L), PD-1 (M), and TIM3 (N), normalized to β-actin. (O) Flow plots showing the effect of MAPK pathway inhibitors on NETs-induced PD-1 and TIM3 expression in CD8⁺ T cells. (P–Q) Quantification of the percentage of CD8⁺PD-1⁺ T cells (P) and CD8⁺TIM3⁺ T cells (Q). FC, fold change; FDR, false discovery rate; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; NETs, neutrophil extracellular traps; ns, not significant; PD-1, programmed cell death protein-1. *p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.

To determine whether the c-Fos-JunB pathway directly contributes to NETs-induced CD8⁺ T cell exhaustion, we treated CD8⁺ T cells with NETs (with or without T-5224, a selective c-Fos/AP-1 inhibitor) and assessed the expression of exhaustion markers. Flow cytometry revealed that NETs treatment increased the surface expression of PD-1 and TIM3, whereas T-5224 cotreatment significantly reduced this effect (online supplemental figure S9D-F). In agreement with these findings, Western blot showed NETs-induced upregulation of PD-1 and TIM3 protein levels, as well as c-Fos and JunB, which was substantially reversed by T-5224 treatment (online supplemental figure S9G-K). Collectively, these results demonstrate that the c-Fos-JunB axis is functionally required for NETs-mediated CD8⁺ T cell exhaustion.

We further validated these results using Jurkat T cells to exclude potential individual variations in primary T cells. Flow cytometry revealed that NETs significantly upregulated the expression of exhaustion markers, PD-1 and TIM3, in Jurkat T cells. Treatment with SCH772984 and trametinib significantly blocked this effect, with trametinib exhibiting the most potent inhibitory activity (online supplemental figure S10A-C). Western blot demonstrated that SCH772984 and trametinib treatment (but not BIRB 796 or SP600125) significantly suppressed NETs-induced expression of c-Fos, JunB, p-ERK, TIM3, and PD-1 (online supplemental figure S10D-I). Finally, validation in Jurkat T cells with siRNA knockdown of MAP2K1/2 confirmed the specific role of the ERK-c-Fos-JunB axis in NETs-induced T cell exhaustion (online supplemental figure S10J-L).

Trametinib synergizes with anti-PD-1 treatment to suppress GC by reversing CD8⁺ T cell exhaustion

YTN16 tumor-bearing mice were randomly assigned to four groups: control (PBS), anti-PD-1 (aPD-1) treatment (3 mg/kg, two times per week, i.p.), trametinib treatment (1 mg/kg/day, i.p.), and combination therapy to confirm that NETs cause CD8⁺ T cell exhaustion via the MEK/ERK pathway and to assess the synergistic antitumor effects of trametinib combined with anti-PD-1 therapy (figure 7A). Both monotherapies significantly suppressed tumor growth, with the combination therapy exhibiting the most potent synergistic antitumor effect (figure 7B,C). HE staining revealed that aPD-1 and trametinib treatment markedly loosened the tumor cell arrangement, significantly reduced cell density, increased cytoplasmic eosinophilia, and induced extensive stromal fibrosis. These phenomena were more pronounced with the combination therapy, accompanied by substantial lymphocytic infiltration, cellular vacuolation in some areas, and stromal fibrosis (figure 7D). Tumor volumes at the endpoint were consistent with the inherent growth characteristics of the YTN16 model,^{37 38} and both tumor volume and the results of histological analysis verified the treatment effects. Flow cytometry showed that the control group exhibited high PD-1 and TIM3 expression in tumor-infiltrating CD8⁺ T cells owing to prolonged tumor stimulation. Interestingly, despite tumor growth inhibition in the anti-PD-1 treatment group, tumor-infiltrating CD8⁺ T cells persisted in a highly exhausted state. In contrast, the trametinib treatment significantly reduced PD-1 and TIM3 levels in tumor-infiltrating CD8⁺ T cells, reversing their exhausted state (figure 7E–H). These results confirm that trametinib reverses CD8⁺ T cell exhaustion in the GC microenvironment by blocking the MEK/ERK pathway and synergizing with anti-PD-1 to enhance antitumor immunity. The mechanism diagram for this study is shown in figure 7I.

Figure 7. Trametinib synergizes with anti-PD-1 therapy to suppress GC growth by reversing CD8⁺ T cell exhaustion. (A) Treatment regimen in YTN16 subcutaneous GC model. Tumor-bearing mice (n=6/group) were treated with PBS, aPD-1, trametinib, or a combination of aPD-1 and trametinib for 15 days. (B) Images of tumors on day 29. (C) Tumor growth curves of subcutaneous tumors. (D) HE-stained sections of tumors. The black dashed line delineates the tumor regression margin. Scale bar: 200 µm. (E) Flow plots showing the expression of exhaustion markers PD-1 and TIM3 on gated tumor-infiltrating CD8⁺ T cells. Quantification of the percentage of CD8⁺PD-1⁺ T cells (F) and CD8⁺TIM3⁺ T cells (G). (H) Immunofluorescence staining showing tumor-infiltrating CD8⁺PD-1⁺ exhausted T cells. DAPI (blue), CD8 (green), PD-1 (red). Scale bar: 50 µm. (I) Schematic model of the mechanism by which TGF-β1-driven NETs mediate anti-PD-1 resistance in GC. aPD-1, anti-programmed cell death protein-1; DAPI, 4',6-diamidino-2-phenylindole; GC, gastric cancer; IFN-γ, interferon-gamma; i.p., intraperitoneally; NETs, neutrophil extracellular traps; PBS, phosphate-buffered saline; PD-L1, programmed death ligand 1; s.c., subcutaneously; TCR, T-cell receptor; TGF-β1, transforming growth factor-β1; TNF-α, tumor necrosis factor-alpha. *p<0.05, ** p<0.01, *** p<0.001.

Discussion

This study integrated single-cell transcriptomics, multidimensional clinical sample analysis, and systematic in vivo and in vitro functional experiments to identify the functions and characteristics of NETs-associated neutrophils (Neu5) in the GC microenvironment. We demonstrated that TGF-β1-induced CD177⁺ NETs-associated neutrophil subpopulations in the GC microenvironment release NETs, thereby activating the ERK-c-Fos-JunB signaling axis in CD8⁺ T cells to induce T cell exhaustion, ultimately mediating resistance to anti-PD-1 immunotherapy. We demonstrated that targeting of NETs formation and the MEK/ERK signaling pathway effectively reverses CD8⁺ T cell exhaustion and that MEK inhibitors synergize with anti-PD-1 immunotherapy to enhance antitumor effects. These findings not only elucidate the composition and function of specific neutrophil subsets in the context of GC immune resistance at the single-cell level but also reveal a direct molecular communication network between NETs and CD8⁺ T cells. This study provides a theoretical basis and therapeutic targets for overcoming immunotherapy resistance in GC.

Neutrophils are essential for innate immune responses and exhibit high heterogeneity within the TME. Traditionally, neutrophils have been categorized into antitumor N1 and protumor N2 types. However, this binary classification fails to adequately capture their complex functional spectrum within the TME.³⁹ In contrast, scRNA-seq technology enables detailed analysis of the high heterogeneity of TANs, not only identifying novel TAN subpopulations but also reconstructing their functional evolution from quiescent to activated states,⁴⁰ providing crucial insights into the complex role of TANs in immunotherapy resistance. To date, no studies have used scRNA-seq to reveal the characteristics of neutrophils in patients with AGC undergoing anti-PD-1 therapy. In the present study, we successfully identified eight functionally distinct neutrophil subpopulations within the GC microenvironment using scRNA-seq. The Neu5 subpopulation was specifically enriched in the immune-resistant group. This subpopulation exhibited specific expression of NETs-related genes, such as CD177, PADI4, and MMP9, along with a pronounced neutrophil degranulation phenotype and NETs formation potential. These findings are consistent with the results of a recent study indicating that CD177 acts as a marker of activated neutrophil subpopulations and correlated with enhanced migration and NETs release potential.⁴¹ Another study using a myocardial infarction model identified MMP9 overexpression as a marker of NETs-associated neutrophil subpopulations. These MMP9-high neutrophils demonstrated enhanced NETs formation capacity and tissue injury potential in myocardial infarction,³⁴ indicating the diversity of neutrophil functional regulation under different pathological conditions. In summary, our study further links the CD177⁺ NETs-associated neutrophil subset to primary resistance to anti-PD-1 therapy in GC.

Using high-throughput sequencing and experimental validation, we identified TGF-β1 as a key upstream driver of NETs release. Traditionally recognized for its immunosuppressive properties, TGF-β1 has been shown to induce the establishment of an immunosuppressive TME by promoting neutrophil polarization toward the N2 phenotype.⁴² Another study showed that TGF-β1 upregulates LIF expression by activating the Smad2/3 complex, ultimately facilitating NETs formation and accelerating peritoneal metastasis in GC.⁴³ Similarly, our data demonstrate that TGF-β1 directly induces human and mouse neutrophils, particularly CD177⁺ NETs-associated neutrophils, to form NETs. This effect was effectively blocked by the TGF-β1 receptor inhibitor, SB431542. In addition, in patients resistant to therapy, the spatial colocalization of TGF-β1 with CD177⁺ neutrophils and CitH3⁺ NETs structures provides compelling in situ evidence for this regulatory axis. A recent study demonstrated that TGF-β1 secreted by HCC induces NETs formation, which further promotes HCC progression by impairing CD8⁺ T cell immunity.²¹ This outcome was consistent with our findings. Our study further validated this regulatory relationship within the GC microenvironment, establishing TGF-β1 as the primary regulator of protumor neutrophil function in the GC microenvironment. It also reveals that the TGF-β1-NFE2-NETs axis represents a potential therapeutic target for overcoming immune therapy resistance. Furthermore, our analysis of clinical relevance revealed that dynamic changes in plasma NETs levels correlated closely with immunotherapy efficacy. Elevated NETs levels act as a risk factor for predicting the prognosis of patients with GC, suggesting the clinical utility of NETs as liquid biopsy markers in GC immunotherapy.

The direct impact of NETs on T cell function has been an area of active investigation. Early studies focused on the physical barrier effects, showing that NETs can encapsulate tumor cells and shield them from cytotoxic lymphocyte attack.¹⁵ Recent studies have revealed direct immunomodulatory effects. Kaltenmeier et al demonstrated that NETs-associated PD-L1 can engage PD-1 in T cells to induce functional impairment.¹⁸ However, the intracellular signaling mechanisms by which NETs induce T cell exhaustion remain poorly understood. In this study, we demonstrated that NETs may directly cause CD8⁺ T cell exhaustion by establishing an in vitro co-culture system using highly pure NETs and CD8⁺ T cells. This is characterized by higher ROS levels, decreased proliferative capability, decreased production of effector cytokines, including IFN-γ and TNF-α, and overexpression of exhaustion markers, PD-1 and TIM3. More importantly, transcriptomic sequencing and signaling pathway analysis revealed an abnormal intracellular kinase cascade triggered by NETs on T cell exposure. Specifically, we observed sustained high-intensity phosphorylation of ERK in CD8⁺ T cells exposed to NETs. Physiological T cell receptor signaling typically involves transient MEK/ERK activation, which promotes clonal expansion.⁴⁴ In contrast, NETs exposure produced sustained hyperactive ERK phosphorylation, which drove the aberrant upregulation of the AP-1 family TFs, c-Fos and JunB, leading to terminal exhaustion. This mechanism aligns closely with the findings of Wang et al,⁴⁵ where sustained activation of the T cell receptor or chimeric antigen receptor caused excessive activation of the MAPK pathway, which increased the expression of c-Fos and JunB. These upregulated TFs further initiate a series of exhaustion-related gene programs, ultimately causing the functional loss of T cells. This study establishes the first direct link between this mechanism and NETs, revealing that NETs promote the exhaustion of CD8⁺ T cells by triggering the ERK-c-Fos-JunB regulatory axis.

To validate this mechanism, we established a subcutaneous tumor model using the YTN16 mouse gastric adenocarcinoma cell line. The selection of the YTN16 cell line enables precise simulation of the key immune-resistant characteristics of human GC while preserving the immune system.⁴⁶ Compared with the MFC mouse gastric squamous cell carcinoma cell line, this cell line authentically replicates the TME of human gastric adenocarcinoma and is now widely used in GC immune microenvironment research and evaluation of immunotherapy strategies.^{37 47} Specifically, this study systematically evaluated the synergistic antitumor effects of targeting NETs and TGF-β1 signaling pathways. We demonstrated that both DNase I-mediated degradation of NETs and SB431542-mediated inhibition of TGF-β1 signaling significantly suppressed tumor growth and reduced NETs levels in the TME, thereby reversing CD8⁺ T cell exhaustion. This aligns with the findings of Snoderly et al that targeting of NETs enhances antitumor immunity.⁴⁸ Building on this, we explored the synergistic antitumor efficacy of targeting PD-1 and MEK/ERK signaling pathway. Results revealed that trametinib effectively blocked NETs-induced c-Fos and JunB upregulation and CD8⁺ T cell exhaustion. This result is consistent with the observation of Dushyanthen et al that MEK inhibitors improve anti-PD-1 effectiveness and reverse T cell exhaustion in breast cancer models.⁴⁹ To our knowledge, this is the first study to link this combined therapeutic strategy to NETs and the GC microenvironment. Notably, tumor-infiltrating CD8⁺ T cells highly expressed exhaustion markers, PD-1 and TIM3, suggesting that the TME persistently drives T cell exhaustion. Interestingly, although anti-PD-1 monotherapy effectively suppressed tumor growth, exhaustion of tumor-infiltrating CD8⁺ T cells was not alleviated. These findings suggest that although PD-1 blockade temporarily inhibits tumor progression, it fails to eliminate the upstream signals that drive T cell exhaustion. Consequently, activated T cells rapidly re-enter an exhausted state with a sustained high expression of inhibitory receptors. In contrast, exhaustion marker levels in CD8⁺ T cells were considerably lowered by trametinib, and its combination with anti-PD-1 showed considerably improved antitumor activity. These findings demonstrate that abnormal activation of the MEK/ERK pathway is a crucial downstream event that mediates NETs-induced T cell exhaustion. Targeting this pathway reverses T cell exhaustion and restores antitumor effector function. This represents not only a significant breakthrough in addressing the efficacy limitations of monotherapy with anti-PD-1 but also provides a biological rationale for the targeted therapy-immunotherapy combination strategy.

Our work provides new insights into the mechanisms of immunological resistance by clarifying the critical function of the TGF-β1-NETs-ERK-c-Fos-JunB axis in GC resistance to anti-PD-1 treatment. Moreover, the findings demonstrated significant clinical translational potential. Plasma NETs levels and peripheral blood CD177⁺ neutrophil ratios serve as predictive biomarkers of immunotherapy efficacy, aiding in the early identification of patients at risk of developing immune resistance. In contrast, combined strategies targeting NETs formation (DNase I, SB431542) or the MEK/ERK signaling pathway (trametinib) with anti-PD-1 therapy offer novel approaches for overcoming immune resistance in GC. Notably, trametinib is already widely used clinically, and its combination with PD-1 inhibitors holds promise for clinical translation.

This study had several limitations. First, the small sample size of the scRNA-seq cohort, particularly among immunotherapy-resistant cases, may not fully reflect the heterogeneity of neutrophils in GC. Future studies with larger scRNA-seq cohorts are warranted to validate these findings. Second, this study primarily focused on CD8⁺ T cells. However, whether NETs influence the functions of other immune cell populations remains unclear. Third, although we identified the ERK-c-Fos-JunB pathway to be a critical modulator of NETs-induced CD8⁺ T cell exhaustion, NETs are complex structures containing DNA, histones, and diverse granular proteins. Determining the specific components that mediate T cell activation and the receptors involved represents a future research direction. Fourth, although our subcutaneous tumor model effectively evaluated therapeutic responses, it failed to fully replicate the complex microenvironment of in situ GC. Future studies using in situ models or humanized mouse systems will provide a more robust validation.

Conclusions

We identified a subset of CD177⁺ neutrophils associated with NETs in the GC microenvironment. This subset has a specific NETs-forming capacity. NETs levels and the proportion of CD177⁺ neutrophils in the peripheral blood of patients with AGC act as biomarkers for predicting the efficacy of PD-1 inhibitors. TGF-β1 acts as an upstream driver of NETs formation. Targeting of the TGF-β1 signaling pathway effectively suppressed NETs production, reversed CD8⁺ T cell exhaustion, and enhanced antitumor efficacy. Furthermore, NETs activated the MEK/ERK pathway within CD8⁺ T cells, upregulating the downstream c-Fos and JunB, thereby inducing CD8⁺ T cell exhaustion. Trametinib blocked the ERK-c-Fos-JunB signaling axis and reversed CD8⁺ T cell exhaustion. When combined with PD-1 inhibitors, trametinib synergistically enhanced antitumor efficacy, offering a novel strategy for overcoming immunotherapy resistance in GC.

Data availability statement

Data are available upon reasonable request.