HIF1A+CSF3R+ neutrophils-dominated hypoxic niche induced metabolic reprogramming for neoadjuvant therapy resistance in NSCLC Authors: Yaoyao Zhou, Ziyun Liu, Parhat Kaysar, Yuxi Han, Yijian Ma, Xiaozhou Yu, Bo Ni, Zuoyu Chen, Yongjie Xie, Haoling Zhang, Yueguo Li Published: 2026 ABSTRACT AbstractBackgroundNon-small cell lung cancer (NSCLC) is one of the frequently occurring cancers characterized by molecular heterogeneity and multiple immune cell infiltration patterns, which are associated with treatment sensitivity and resistance. However, the specific microenvironmental cells and their mechanisms that lead to treatment resistance in patients need to be explored in greater depth.MethodsOn the basis of patients receiving neoadjuvant therapy in our center, a multicenter, multicohort NSCLC spatial transcriptome, single-cell transcriptome, T-cell receptor repertoire sequencing, bulk RNA transcriptome, phosphorylated proteome, genome mutation, and clinical data were included for a comprehensive assessment of the therapeutic and prognostic impact of HIF1A+ CSF3R+ neutrophils in NSCLC. In vitro experiments validated the functional phenotype of HIF1A+ CSF3R+ neutrophils and co-localization interactions with other cellular subpopulations. Gradient boosting machine (GBM) constructed region of interest (ROI) models for evaluation. Computer-aided drug design (CADD) was used to predict targeted small molecule drugs, and in vivo mouse models were constructed to assess the effectiveness of the combination treatment regimen.ResultsCentered on HIF1A+ CSF3R+ neutrophils, recruited exhausted T cells and stromal cells form a hypoxic niche within the tumor region, which was enriched in non-response patients. ROI composed of these specific cellular subpopulations, associated with senescence and glycolysis, accurately predicting NSCLC progression, prognosis, and microenvironment composition. CADD analysis identified that platycodin-D2 specifically targeted CSF3R, reducing HIF1A expression and inhibiting neutrophil activity. Combining navitoclax, platycodin-D2 with anti-programmed cell death protein 1 (PD-1) significantly suppressed tumor proliferation and improved the immunosuppressive microenvironment.ConclusionOur study emphasized the role of HIF1A+ CSF3R+ neutrophils in immunotherapeutic resistance of NSCLC, constructed a microenvironmental immune dysregulation network in a hypoxic ecological niche with HIF1A+ CSF3R+ neutrophils as the center. Platycodin-D2 specifically targeted HIF1A+ CSF3R+ neutrophils, enhancing the efficacy of anti-PD-1 therapy in NSCLC. WHAT IS ALREADY KNOWN ON THIS TOPIC Previous studies have demonstrated that the hypoxic tumor microenvironment, metabolic reprogramming, and neutrophil-mediated immunosuppression are critical contributors to neoadjuvant therapy resistance in non-small cell lung cancer (NSCLC). Previous studies have demonstrated that the hypoxic tumor microenvironment, metabolic reprogramming, and neutrophil-mediated immunosuppression are critical contributors to neoadjuvant therapy resistance in non-small cell lung cancer (NSCLC). WHAT THIS STUDY ADDS However, the key neutrophil subpopulations and their regulatory networks remain insufficiently characterized. In this study, we identified HIF1A+CSF3R+ neutrophils as a core cell subset closely associated with treatment non-response and revealed that they promote neoadjuvant therapy resistance by establishing a hypoxic immunosuppressive niche and driving glycolytic and senescence-related metabolic reprogramming. Furthermore, through spatial transcriptomics, multi-omics integration, and drug screening analyses, we found that platycodin-D2 specifically targets CSF3R and enhances the therapeutic efficacy of anti-programmed cell death protein 1 treatment. However, the key neutrophil subpopulations and their regulatory networks remain insufficiently characterized. In this study, we identified HIF1A+CSF3R+ neutrophils as a core cell subset closely associated with treatment non-response and revealed that they promote neoadjuvant therapy resistance by establishing a hypoxic immunosuppressive niche and driving glycolytic and senescence-related metabolic reprogramming. Furthermore, through spatial transcriptomics, multi-omics integration, and drug screening analyses, we found that platycodin-D2 specifically targets CSF3R and enhances the therapeutic efficacy of anti-programmed cell death protein 1 treatment. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Collectively, this study establishes a novel resistance mechanism centered on HIF1A+CSF3R+ neutrophils, providing new insights into treatment response prediction, biomarker development, and precision combination therapies for NSCLC. Collectively, this study establishes a novel resistance mechanism centered on HIF1A+CSF3R+ neutrophils, providing new insights into treatment response prediction, biomarker development, and precision combination therapies for NSCLC. Introduction Lung cancer remains the most common type of cancer worldwide, causing approximately 1.8 million deaths annually.1 Non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancer cases, with lung adenocarcinoma (LUAD) and lung squamous cell carcinoma being the most common histologic subtypes.2 The emergence of immune checkpoint inhibitors (ICIs) has revolutionized the treatment landscape of lung cancer over the past few decades. These drugs significantly improve patient survival by targeting the programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) axis or cytotoxic T-lymphocyte associated protein 4 (CTLA-4). compared with conventional standard chemotherapy, patients treated with ICIs have shown a relative reduction in the risk of death of between 24% and 40% during follow-up of up to 5 years or more in several clinical trials.3 However, the objective remission rate of ICIs in lung cancer is usually less than 20%, and most patients eventually relapse due to acquired resistance. The resistance mechanisms are complex and need to be further explored in depth. Neutrophils, as evolutionarily ancient phagocytes, play a key role in defense against small pathogens. They are abundant in the human bone marrow and peripheral blood and are guided by inflammatory mediators and chemokines to the site of tumorigenesis.4 Tumor-associated neutrophils (TANs) play a complex role in cancer, with N1/N2 TANs differentiating into antitumor or protumor phenotypes depending on the tumor microenvironment.5 N1 neutrophils are usually associated with pro-inflammatory responses, are activated in the early stages of inflammation, and have antitumorigenic activity, whereas N2 neutrophils are associated with anti-inflammatory and immune-modulatory functions, and are more likely to appear in the late stages of inflammation. Neutrophils have been reported to directly drive tumor cell proliferation, as demonstrated in a zebrafish model and supported by data from human melanoma.6 7 Furthermore, neutrophil-assisted tumor cell motility, which encompasses epithelial-mesenchymal transition and migration, has been demonstrated in a variety of tumor cell lines and in vivo models. Due to the complex plasticity and functional diversity of TANs, there are many challenges to overcome as potential therapeutic targets. The interaction of TANs with other microenvironmental cells makes targeting more difficult, so personalized therapeutic strategies are in urgent demand. In the immunotherapy of lung cancer, the mechanism of action of TANs is extremely complex. On the one hand, TANs can promote the formation of antitumor immune microenvironment through the release of cytokines and chemokines, thus enhancing the effect of immunotherapy; on the other hand, they may inhibit the activity and function of T cells and promote the occurrence of immune escape, thus weakening the efficacy of immunotherapy. Some studies have shown that the interaction between TANs and tumor cells in the tumor microenvironment can influence the efficacy of ICIs,8 leading to significant differences in therapeutic responses. Therefore, an in-depth understanding of the functions of TANs and their dual roles in immunotherapy is important for optimizing therapeutic strategies in lung cancer. We conducted an in-depth study of a total of more than 30,000 neutrophils in seven datasets by collecting multicenter single-cell datasets of lung cancer in this study to uncover the underlying molecular mechanisms. Neutrophils produce different phenotypes as they progress from naivety to maturity until senescence,9 and new research suggests that neutrophils persist longer in tumors and contribute to tumor progression and treatment resistance.10 Among other things, senescent neutrophils have decreased function, leading to abnormalities in the ability to release NETs (neutrophil extracellular traps), which have a direct impact on tumor progression and treatment resistance through the secretion of factors associated with SASP (senescence-associated secretory phenotype), inflammation, angiogenesis, and cell migration.11 Relatively few studies have been conducted on the role of senescent neutrophils in the lung cancer tumor microenvironment, and identifying specific immune senescent cell phenotypes can help to understand the specific functions of these cells in particular tissues and diseases. In the current study, we identified the neutrophil subpopulation with senescent phenotype and further explored its role in remodeling the lung cancer tumor microenvironment and explored the mechanism by which it affects lung cancer therapy. In the immune microenvironment of lung cancer, the interaction of neutrophils with other immune cells is crucial. Studies have shown that TANs are able to regulate immune responses through interactions with T cells, macrophages and other immune cells.12–14 For example, TANs can secrete cytokines such as interleukin-8, which affect the activation and function of T cells, and thus promote or inhibit antitumor immune responses.15 In addition, neutrophils can alter the immune status of the tumor microenvironment through interactions with macrophages, affecting tumor growth and metastasis. This complex network of interactions not only plays an important role in tumor progression, but also provides potential targets for the development of new immunotherapeutic strategies. In this study, we collected multi-omics sequencing from NSCLC receiving immunotherapy at our center and found that neutrophils promoted the formation of hypoxic niches in drug-resistant patients. Multicenter cohort data and in vitro experiments revealed that HIF1A+ CSF3R+ neutrophils were highly expressive of malignant factors such as iNOS and Arg-1, increasing glycolysis levels and lactate efflux in the microenvironment and inducing the emergence of a senescent phenotype. T-cell receptor (TCR) sequencing demonstrated that HIF1A+ neutrophils led to the expansion of CD8+ Tex and CD4+ Tex clones, co-localizing with PD-1+ T cells. Co-culturing HIF1A+ CSF3R+ neutrophils and stromal cells revealed HIF1A+ CSF3R+ neutrophils promoted the formation of an immunosuppressive microenvironment. Infiltration of HIF1A+ CSF3R+ neutrophils has been predicted in multicenter immunotherapy datasets rendering the patients ineffective for ICIs, and navitoclax and platycodin-D2 were found to be potentially targeted for the treatment of HIF1A+ CSF3R+ neutrophils highly infiltrated tumors by small molecule drug prediction. In vivo experiments revealed that anti-PD-1 combined with navitoclax and platycodin-D2 significantly increased the sensitivity of non-responders, providing a new treatment option for ICIs resistance in NSCLC. Methods Patient and public datasets collection The study protocol complied with the ethical guidelines and was accepted by the Clinical Research Ethics Committee of Tianjin Medical University Cancer Hospital. Lung cancer specimens were obtained from the Tianjin Medical University Cancer Hospital and were performed spatial transcriptome sequencing, single-cell RNA sequencing (scRNA-seq) and bulk RNA sequencing. Multicenter multicohort single-cell RNA datasets were downloaded from GEO and SRA databases. The bulk transcriptomes of patients who received immunotherapy were collected from the GEO database under a specific GEO accession number. The normalized messenger RNA (mRNA) expression data (TPM-Transcripts Per Million data), somatic mutation, and clinical data of NSCLC were downloaded from The Cancer Genome Atlas (TCGA) database (https://portal.gdc.cancer.gov/). The quantification of phosphorylated proteins of LUAD was obtained from the CPTAC database. Single-cell spatial transcriptome sequencing The raw sequencing data were processed using Space Ranger software to extract the barcode and UMI (Unique Molecular Identifier) of the spot, which was compared with the reference genome by STAR, corrected for barcode, filtered for corrected UMI and counted, followed by distinguishing between the barcode of the containing cell and the background barcode according to the algorithm in order to extract the real single-cell data. Finally, the gene expression matrix of each spot was obtained. Expression matrices and spatial information were imported using Seurat (V.4.4.0) for quality control and normalization, combined with spatial coordinates to analyze gene expression patterns and biological features, as well as applying clustering methods to group cells. Annotation classification of spatial transcriptome data was obtained by deconvolution of annotation information of the single-cell dataset by the RCTD package. On the basis of obtaining subpopulation information, ecological niche analysis was performed using the ISCHIA package to obtain the co-localization of different cellular subpopulations.16 The “SpatialFeaturePlot” function was also used to visualize the spatial distribution of specific pathways and genes, which was combined with ggplot2 for further interactive analysis. Bulk RNA-sequencing analysis Strand-specific libraries were constructed using the TruSeq Stranded mRNA Sample Prep Kit (Illumina) according to the manufacturer’s instructions. Poly-adenylated RNA from intact total RNA was refined using oligo-dT beads. The extracted complementary DNA fragment was 3′ end adenylated and ligated to Illumina paired-end sequencing adapters and amplified by PCR. The libraries were sequenced with 65 base pair single-end reads on the Illumina HiSeq 2500 System in high output mode using V.4 chemistry. Raw reads were aligned to GRCh38 using a STAR RNA-sequencing aligner after which gene expression levels were quantified by Salmon (see URLs) using default parameters for both applications. Single-cell RNA analysis We performed the main process of single cell data analysis by “Seurat” (V.4.0.0), including quality control, filtering and normalization, integration of different samples through the harmony package for de-batching and dimensionality reduction clustering with “RunUMAP” function. After obtaining different cell subpopulations, cellular annotation of different subpopulations was performed with marker genes sourced from the Human Primary Cell Atlas database and typical representative genes to obtain tumor parenchymal, stromal and immune cells. Highly expressed genes from different cellular subpopulations were further visualized by comparing them using the “FindMarkers” function. The “addModuleScore” function was applied to calculate the pathway scores of the gene sets and wilcox.test was used to further compare the differences. Pseudo-time analysis Monocle2,17 as a classical single-cell trajectory analysis tool based on R language, its core working principle integrates the Reversed Graph Embedding framework with machine learning algorithms, aiming at revealing the dynamic developmental trajectories of cells through unsupervised learning strategies. The tool achieves trajectory reconstruction through the following key steps: first, the feature space for state judgment is constructed based on high-variance gene screening, followed by non-linear dimensionality reduction and topology preservation using the DDRTree (Discriminative Dimensionality Reduction via Learning a Tree) algorithm, which projects the cells into a low dimensional manifold space; the shortest path algorithm constructs a biologically interpretable trajectory tree model and assigns pseudo-time coordinates to each cell to reflect the developmental process; ultimately, differential expression analysis identifies marker genes that change dynamically during development and locates branching decision points to resolve key nodes of cell fate differentiation. Cellular communication analysis In this study, we used the CellChat18 package to analyze the interactions between different cell subpopulations. CellChat employs a top-down approach, starting from a macroscopic perspective and then refining it in more detailed details on signaling mechanisms to identify signaling changes at different levels. The likelihood of cellular communication is calculated by ligand-receptor gene expression, and for complexes of receptor-ligands, as long as one of the molecules is not expressed, this is considered to be ineffective for the receptor-ligand. Cellular communication significance was calculated using a permutation test to test the significance of each receptor-ligand, with the number of permutations defaulting to 100, comparing the likelihood of communication of a specific ligand-receptor pair after permutations to the true likelihood of communication and calculating the significance p value. Multiplex fluorescent immunohistochemistry 5 µm lung cancer tissue sections were deparaffinized and rehydrated through a graded ethanol series: 100% ethanol (10 min two times), 95% ethanol (10 min), and 70% ethanol rinse. Antigen retrieval was performed in Citric Acid Buffer (pH 6.0) or EDTA Alkaline Buffer (pH 8.0) via microwave treatment for 15 min. Multiplex fluorescent immunohistochemistry was conducted using the multiplex kit (Akoya Biosciences, Cat#: NEL871001KT). After blocking with a protein block, primary antibodies were applied, followed by HRP-conjugated secondary antibodies. Each HRP-conjugated polymer facilitated the covalent binding of distinct fluorophores for signal amplification. Additional antigen retrieval was performed to remove bound antibodies. Nuclei were counterstained with DAPI (4’, 6-Diamidino-2-phenylindole), and slides were mounted with fluorescence mounting medium. Images were captured at 400×magnification using a Zeiss fluorescence microscope and analyzed with ImageJ software. Western blotting Cells were lysed in SDS (sodium dodecyl sulfate) protein lysis buffer with a protease inhibitor, and protein concentration was determined using the BCA assay. Equal protein amounts (20–50 µg) were separated by 10% SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel electrophoresis) and transferred to PVDF (polyvinylidene difluoride) membranes. Membranes were blocked with 5% non-fat dry milk in TBST (Tris-buffered saline with Tween 20) for 1 hour and incubated with primary antibodies overnight at 4°C. After three TBST washes, membranes were incubated with HRP-conjugated secondary antibodies at room temperature for 2 hours. Protein bands were detected using an ECL (enhanced chemiluminescence) system and captured with a Chemi-Scope exposure machine. Antibody details are provided in online supplemental table 1. Pathway enrichment analysis After collecting representative genes highly expressed in HIF1A+ neutrophils, glycolysis-related, and aging-related genes, GSVA (Gene Set Variation Analysis) enrichment analysis was performed. The GSVA method is a nonparametric, unsupervised algorithm.19 In the process of GSVA score calculation, each sample can be calculated independently and give a score that takes into account the inside of the gene set and the outside of the gene set in an integrated manner. It converts a gene expression matrix into an expression matrix in terms of gene sets. After obtaining differentially expressed genes (DEGs) from different comparison groups, we performed KEGG (Kyoto Encyclopedia of Genes and Genomes) and GO (Gene Ontology) pathway enrichment using the “enrichKEGG” and “enrichGO” functions in the “clusterProfiler” package, and both KEGG/GO enrichment analyses are part of over-representation analysis, which is a widely used analytical method to determine whether a known biological function or process is significantly overexpressed in an experimentally obtained list of DEGs. Seahorse XF analysis Cells were cultured under standard conditions (37°C, 5% CO₂) in appropriate medium and were seeded in a 96-well Seahorse XF microplate at 1×105 cells per well and incubated for 24 hours. The glycolysis stress test was performed using the Seahorse XF Analyzer (Agilent Technologies) to measure extracellular acidification rate (ECAR). After baseline measurements, cells were treated with glucose (10 mM), oligomycin (1 µM), and 2-deoxyglucose (50 mM). ECAR was recorded after each injection. Glycolytic rate was calculated as the change in ECAR after glucose addition, and glycolytic capacity was assessed as the maximum ECAR following oligomycin treatment. Lactate levels in the media were measured using a lactate assay kit and normalized to cell number. Specifically, we comprehensively assessed HIF1A expression in patients’ tumor tissues using multiplex immunofluorescence and western blot, and classified patients into high-HIF1A and low-HIF1A expression groups accordingly. Subsequently, we collected peripheral blood samples from the two patient groups, isolated peripheral blood neutrophils, and performed ECAR assays on these neutrophils to evaluate the glycolytic activity of neutrophils derived from patients with different HIF1A expression profiles. TCR sequencing analysis TCR sequencing was performed concurrently in a single-cell sequencing dataset of 234 patients with lung cancer, and on the basis of annotating different subpopulations of T cells we further combined the information on clonal molecular phenotypes and sequences of TCRs to obtain the clonal indices and diversities of TCRs in different T-cell subpopulations. The filtered contig output file filtered_contig_annotations.csv from the 10x Genomics Cell Ranger pipeline was processed through the scRepertoire20 package. This file contains detailed information about the TCR or immunoglobulin (Ig) chains in the single-cell immunome library, such as cell barcodes, gene information, CDR3 sequences, and more. Integration of different samples using combinetcr and combining clonotype data from scRepertoire with single-cell transcriptome data using the “combineexpression” function. At the same time, we applied the STARTRAC package to analyze the changes in T-cell dynamics, an algorithm that assumes that cells with the same clonotype are of the same origin, and that if two T cells have the same TCR clonotype, they are likely to be derived from the same progenitor cell or have the same developmental origin. Startac21 calculates three indicators based on the differences in the distribution of T cells of different clonal types among different tissues and cell types: expansion, migration, and transition. Immune infiltration analysis We used the xCell and CIBERSORT packages to perform cellular infiltration analyses at the bulk tissue level, respectively. xCell was developed by the dviraran team in 2017,22 and the method allows for the assessment of infiltration levels of up to 64 cell types, including multiple adaptive and innate immune cells, hematopoietic progenitors, epithelial cells, and extracellular stromal cells, based on gene expression data. xCell uses the ssGSEA (single-sample Gene Set Enrichment Analysis) enrichment method to return enrichment scores rather than cell occupancy, so scores for the same cell type can be compared across samples, but comparisons across cell types are not meaningful. CIBERSORT (cell-type identification by estimating relative subsets of RNA transcripts) is a computational method for assessing the relative abundance of different cell types in complex mixed tissue samples.23 CIBERSORT is based on support vector regression, a deconvolutional algorithm for estimating the composition of cell types in complex tissue samples from transcriptome data. A predefined matrix of reference gene expression features is used to infer the proportion of individual cell types. The final value obtained by CIBERSORT is a percentage, which can be compared between the proportions of different cell populations in the same sample, or between different cell populations in different samples. TIMER, quanTIseq and EPIC are all inverse convolution-based algorithms that use gene expression data to estimate the infiltration scores of 6, 10 and 6 immune cells, respectively. The difference is that MCPcounter is based on gene expression markers to estimate the infiltration levels of eight immune cells and two stromal cells Tumor immune dysfunction and exclusion analysis Transcriptome samples from the TCGA dataset were normalized and uploaded to the tumor immune dysfunction and exclusion (TIDE) (http://tide.dfci.harvard.edu/) database for predictive analysis.24 TIDE is an algorithm for predicting patients with cancer’s response to ICIs (PD-1/PD-L1 inhibitors) response algorithm. The algorithm was constructed to characterize T-cell dysfunction in other tumor cohorts by examining the correlation between the expression of each gene in the tumor modeling cohort and the level of effector cytotoxic T-lymphocyte infiltration and its impact on survival. Mutation analysis Somatic mutation data for the NSCLC was obtained from TCGA data portal (https://portal.gdc.cancer.gov/) in the form of Mutation Annotation Format. We used the maftools package to filter and visualize the mutation data. Missense_Mutation, Nonsense_Mutation, Frame_Shift_Del, Frame_Shift_Ins, In_Frame_Del, In_Frame_Ins, Splice_Site, and Translation_Start_Site were considered as effectual mutations. Statistics were compiled for the filtered non-synonymous mutations to determine the total number of mutations in each tumor sample. The total mutation count was divided by the length of the coding regions covered by sequencing to yield the tumor mutation burden. Beyoncell drug sensitivity analysis Beyondcell is a method for identifying drug sensitivities in scRNA-seq data.25 It focuses on analyzing drug-related commonalities between cells and classifying them into different therapeutic clusters (TCs). Inputting an scRNA-seq expression matrix and a collection of drug features, the method calculates a Beyondcell Score (BCS) for each drug-cell pair (pair). The BCS ranges from 0 to 1 and measures the sensitivity of each cell to a specific drug. The resulting BCS matrix can be used to determine the TCs of a sample. Thus, therapeutic differences between cell populations can be identified and a sensitivity-based priority ranking is generated to guide drug selection. cMAP drug prediction cMAP is an expression profiling database based on intervening gene expression developed by the Broad Institute.26 It is widely used in the field of pharmacogenomics research by computational biologists and drug screening researchers to reveal functional linkages between small molecule compounds, genes, and disease states. cMAP provides both gene expression profiling data before and after drug treatment of human cell lines and an online tool based on a pattern matching algorithm. The list of expressed genes is first analyzed experimentally, and then cMAP is used to compare the list of differentially expressed genes with a database reference dataset. Finally, the database gets a correlation score (−1 to 1) based on the enrichment of differentially expressed genes in the reference gene expression profile, where a positive number indicates that the drug has promoted gene expression and a negative number indicates that the drug has inhibited gene expression. Computer-aided drug design This study used molecular docking to identify compounds with potential bioactivity. The approach was centered around three key phases: preparation of a compound library, processing of the receptor protein, and execution of the docking simulation. Compound Library Preparation: A set of drug-like small molecules was obtained from the ZINC database (https://zinc.docking.org/) and combined with molecules from preclinical research, investigational agents, and approved drugs to form a diverse screening library. The LigPrep module (Schrödinger Suite) was used for structural standardization. Using the Epik algorithm (pH 7.0±2.0), protonation states were assigned, tautomers were created, and salt ions were eliminated, yielding a three-dimensional library with optimized geometries. Receptor Protein Preparation: Target protein structures were acquired from UniProt (http://www.uniprot.org) and the RCSB PDB (http://www.rcsb.org), with a preference for X-ray structures having a resolution of 2.5 Å or better. These structures were refined in Maestro 10 (Schrödinger), which included adding missing atoms, correcting bond orders, optimizing hydrogen bonding, and applying energy minimization to achieve a stable local energy conformation. Molecular Docking Execution: Initial virtual screening to rank compounds was conducted with MOE 2020 software, which offers an integrated environment for molecular modeling, protein analysis, and ligand docking. Candidates were filtered by their shape complementarity to the active site, with the top 5% by predicted binding score retained for further analysis. Subsequently, refined docking of these selected compounds was performed using AutoDock Vina. The docking grid was centered on the known binding pocket (dimensions: 30×30×30 Å), and a semi-flexible approach allowed ligand conformation optimization. Critical binding interactions, such as hydrogen bonds, hydrophobic contacts, and π-π stacking, were analyzed and visualized in PyMOL to determine the pharmacophoric features of high-affinity binders. Cell culture The LA-795 cell line (RRID: CVCL_G363), LLC cell line (RRID: CVCL_4358), CMT 167 cell line (RRID: CVCL_2405) were obtained from the Type Culture Collection Committee of the Chinese Academy of Sciences (Shanghai, China). Cells were cultured in RPMI (Roswell Park Memorial Institute) 1640 medium (Gibco, Cat#: 11875093) or DMEM (Dulbecco’s Modified Eagle Medium) medium (Gibco, Cat#: C11995500BT) supplemented with 10% fetal bovine serum (FBS) (Gibco, Cat#: A5670701) and 5% penicillin/streptomycin (Gibco, Cat#: 15140122). The cells were maintained at 37°C in a humidified atmosphere with 95% air and 5% CO2. Before initiating the study, Mycoplasma contamination was excluded in the LA-795 cell line. Cells were passaged regularly to maintain growth, and all experiments were performed using cells within 20 passages. The choice of the LA-795 cell line was primarily based on the following considerations: this cell line consists of well-defined epithelioid cells that grow as adherent cultures in vitro, making them easy to culture and observe, and they have a stable genetic background; the tumors formed by this cell line in syngeneic mice exhibit histopathological features very similar to the alveolar subtype of human LUAD, with pathological characteristics closely resembling those observed clinically; following subcutaneous or orthotopic transplantation into syngeneic mice, this cell line demonstrates a consistently high propensity for lung metastasis, with a reproducible and predictable timeline for metastasis. Isolation of PBMCs and neutrophil enrichment from human peripheral blood Peripheral blood was collected from patients with LUAD with informed consent, following the ethical guidelines of the institution. Blood samples were collected into heparinized tubes to prevent clotting. Peripheral blood mononuclear cells (PBMCs) were isolated using density gradient centrifugation. Briefly, using the Ficoll-Hypaque density gradient centrifugation method, the PBMC layer was collected from the interface between the plasma and Ficoll. The collected PBMCs were washed two times with phosphate-buffered saline (PBS) supplemented with 2% FBS (Gibco, Cat#: A5670701) to remove residual Ficoll and plasma. The cells were then resuspended in RPMI 1640 medium supplemented with 10% FBS and 1% penicillin/streptomycin (Gibco, Cat#: 15140122). Cell viability was assessed using trypan blue exclusion (Gibco, Cat#: 15250061) and counted with a hemocytometer. Neutrophils were isolated from PBMCs using a negative selection method, which allows for the enrichment of neutrophils by removing other cell types. Neutrophils were enriched from PBMCs using the Human Neutrophil Isolation Kit (Miltenyi Biotec, Cat#:130-104-434) according to the manufacturer’s instructions. The isolated neutrophils were evaluated for purity and viability using flow cytometry. Cells were stained with anti-CD15 or anti-Ly6G antibodies to confirm neutrophil identity. Neutrophil purity was typically >90%. Viability was assessed using the viability dye and trypan blue exclusion. The isolated neutrophils were resuspended in RPMI medium for downstream applications. Isolation of CAFs Tissue samples were obtained from patients with LUAD with informed consent, in compliance with ethical guidelines. The tumor tissues were collected under ice-cold conditions and immediately placed in transport medium consisting of 10% FBS (Gibco, Cat#: A5670701), 1000 U/mL penicillin-streptomycin (Pen-Strep) (Gibco, Cat#: 15140122), and 25 µg/mL amphotericin B (Himedica). The samples were washed thoroughly with PBS containing 300 U/mL Pen-Strep and 7.5 µg/mL amphotericin B. The collected tumor tissue was finely minced using sterile surgical blades and transferred to a digestion medium containing Collagenase type I (4 mg/mL) (Gibco, Cat#: 17100017) and hyaluronidase (1 mg/mL) (Sigma-Aldrich). The tissue samples were incubated at 37°C in a hybridization chamber at 70 rpm for 30–60 min for digestion. After digestion, the samples were centrifuged at 1.3×10³ rpm for 3 min. The pellet was resuspended in recovery medium (10% FBS, 1,000 U/mL Pen-Strep, and 7.5 µg/mL amphotericin B) and transferred into a culture dish. Primary cancer-associated fibroblasts (CAFs) were enriched by allowing the digested cells to adhere to the culture dish. Non-adherent cells (including immune cells) were removed by washing with PBS. The adherent CAFs were cultured in high-glucose DMEM (4,500 mg/L D-glucose, Gibco, Cat: # 11965092) supplemented with 10% FBS, 1,000 U/mL Pen-Strep, and 7.5 µg/mL amphotericin B. CAFs were cultured at 37°C with 5% CO₂ for 2 weeks. The cells were then passaged and used for further analysis. CAFs were monitored for morphology, purity, and proliferation rate. The purity of the CAF culture was assessed by flow cytometry or immunofluorescence using fibroblast markers. Co-culture of CAFs and neutrophils For co-culture experiments, 1×10⁵ CAFs were seeded into a 24-well culture plate and allowed to adhere overnight. After 24 hours, 1×10⁵ neutrophils were added to the CAF culture and incubated together for an additional 24–48 hours. The cells were co-cultured in RPMI 1640 medium supplemented with 10% FBS, 1% Pen-Strep, and 7.5 µg/mL Amphotericin B at 37°C in a humidified atmosphere with 5% CO₂. Subcutaneous tumor implantation 4–5 weeks old T739 or C57BL/6 mice were randomized assigned to each treatment group of six. The murine lung cancer cells LLC or LA-795 at a dilution range of 1×106 were suspended in a 40 µL PBS and then subcutaneously transplanted into each mouse’s groin. Related administration was intraperitoneally injected 1 week later. The observers and recorders in the study were blinded to the grouping. Tumor growth was monitored every 3 days using a caliper and tumor volumes were calculated by the following formula: Volume=1/2 L1×(L2)², where L1 is the length of the long axis and L2 is the length of the short axis. For mice: Navitoclax: 50 mg/kg/day p.o. (per os); platycodin-D2: 10 mg/kg/day p.o.; anti-PD-1: 12.5 mg/kg/week i.p.(intraperitoneal) Orthotopic tumor implantation During the surgical procedures, anesthesia was induced and maintained with inhaled isoflurane, and all possible measures were taken to minimize animal discomfort. Following a previously described approach, a transverse skin incision was created at the xiphoid process level along the left lateral axillary line. The subcutaneous fat tissue was then dissected to expose the underlying left lung. A suspension containing 1×10⁵ CMT167 cells in PBS supplemented with 1.35 mg/mL Matrigel Basement Membrane Matrix (Corning, Cat#354234) was injected into the left lung using a 30-gage needle. After the injection, the incision was closed with veterinary skin adhesive. 1 week later, the relevant treatment was administered. Following euthanasia, the size of the primary tumor was measured using digital calipers. H&E staining Organ tissues such as heart, liver, spleen and were harvested after sacrifice. Following overnight immersion in 4% paraformaldehyde at 4°C, the tissues were embedded in paraffin. Sections of 5 µm in thickness were subsequently deparaffinized, brought to hydration, and stained with H&E. Evaluation of the histological pathology was carried out separately by two pathologists. Biochemical markers detection in mice Mice were anesthetized by inhalation of isoflurane and placed in a supine position. Blood samples were collected from the retro-orbital venous plexus using heparinized capillary tubes. Approximately 500 µL of whole blood was obtained from each animal. A portion (about 200 µL) was immediately transferred into an EDTA-coated tube for routine hematological analysis. Complete blood count parameters, including red blood cell count, white blood cell count, platelet count and hemoglobin concentration, were determined using an automated hematology analyzer (Mindray, Cat#BC-5000 Vet) within 2 hours after collection. For assessment of liver and kidney functions, serum levels of alanine aminotransferase, aspartate aminotransferase, creatinine, blood urea nitrogen and lactate dehydrogenase (LDH), and were measured using a fully automated biochemical analyzer (Beckman Coulter, Cat# AU480). All measurements were performed in duplicate, and the mean values were used for statistical analysis (online supplemental tables 2 and 3). TUNEL staining for apoptosis detection Tumor tissues were collected, fixed in 10% formalin, and embedded in paraffin. Sections (5 µm) were prepared, deparaffinized, and rehydrated through a graded ethanol series. TUNEL (Terminal deoxynucleotidyl transferase dUTP nick-end labeling) staining was performed using the TUNEL detection kit (Beyotime, Cat: # C1091). After antigen retrieval with proteinase K, tissue sections were incubated with TUNEL reaction mixture containing TdT enzyme and fluorescein-labeled dUTP, which labels DNA breaks in apoptotic cells. After incubation, the sections were counterstained with DAPI and mounted with fluorescence mounting medium. Apoptotic cells were detected under a fluorescence microscope by green fluorescence. The percentage of apoptotic cells was quantified by counting TUNEL-positive cells relative to total cells. Statistical analysis was performed to compare apoptotic rates between treatment groups. ZIP synergy scoring The synergistic effect of the combination therapy of navitoclax or anti-PD-1 and platycodin-D2 was analyzed by cell apoptosis assay. Synergy scores were calculated by SynergyFinder V.2 (https://synergyfinder.fimm.fi). The final synergy scores were interpreted as follows: less than −10, the interaction between two drugs is likely to be antagonistic; between −10 and 10, the interaction between two drugs is likely to be additive; and greater than 10, the interaction between the two drugs is likely to be synergistic. Flow cytometry Tumors were harvested from mice, and single-cell suspensions were prepared by mechanical dissociation and enzymatic digestion. The cell suspensions were filtered through a 70 µm mesh to obtain single cells, which were then divided into separate tubes for antibody staining. The cells were incubated with appropriate concentrations of fluorochrome-conjugated antibodies targeting CD45, CD3, CD8, GrzmB, IFN-γ, TNF-α, and PD-1 for 30 min at 4°C, protected from light. After incubation, the cells were washed three times with PBS to remove any unbound antibodies. For the gating strategy, live cells were first selected by excluding dead cells using the viability dye. Following this, CD45+ cells were gated to identify leukocytes. Among the CD45+population, CD3+ T cells were selected, and then CD8+T cells were identified as the target population. The expression levels of GrzmB, IFN-γ, TNF-α, and PD-1 on CD8+T cells were analyzed for each treatment group. Flow cytometry analysis was performed using a Beckman Coulter fluorescence-activated cell sorter. Data were collected and analyzed using soft FlowJo V.10.0 software. The percentages of positive cells for each marker (GrzmB, IFN-γ, TNF-α, and PD-1) within the CD8+T cell population were calculated for each experimental group. Statistical comparisons between treatment groups were performed using one-way analysis of variance (ANOVA) with post hoc comparisons. The antibodies used for flow cytometry analysis are listed in online supplemental table 1. Statistical analysis Statistical analysis was generated by R software (V.4.2.0) and GraphPad Prism Statistics software V.6.0. For non-normally distributed variables, we performed the Wilcoxon rank-sum test. For comparisons of more than two groups, Kruskal-Wallis tests were used as non-parametric methods. The comparison of percentages was by the χ2 test. One-way ANOVA was carried out for mouse tumor growth. Each experiment was conducted in triplicate. The data values were presented as the means±SDs unless otherwise stated. The median survival time was analyzed using Kaplan-Meier curves and the log-rank test was used to analyze differences in the survival time among the different groups. The level of statistical significance was defined as *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 and ns, non-significant. Results Neutrophil-dominated hypoxic niche induced neoadjuvant therapy resistance in NSCLC The heterogeneity and diversity of the tumor microenvironment is often an important cause of neoadjuvant resistance. In order to deeply investigate the key factors affecting the efficacy of lung cancer, we collected samples of patients with lung cancer treated with anti-PD-1 in our cohort for single-cell transcriptome sequencing, spatial transcriptome sequencing and bulk RNA sequencing (figure 1A). The cellular subpopulation information was used for back-convolution thereby obtaining the annotation of spatial transcriptional organization (figure 1B; online supplemental figures S1A–C and S2A). Based on the treatment effect of the patients, we considered pathological response and major pathological response as treatment effective and stable disease and progressive disease as treatment ineffective. By using the method of ISCHIA niches analysis, we performed subpopulation clustering in the spots of four patients and obtained a total of eight niches (figure 1C and online supplemental figure S2B). Among them, CC3 and CC5 were mainly distributed in patients with effective treatment, while CC7 and CC8 were primarily predominant in the treatment-resistant patients (figure 1D). We performed pathway enrichment analysis at the spatial transcriptome level for each spot, and after summing and averaging, it was more obvious that CC7 showed higher hypoxia properties and angiogenesis (figure 1E). This drew us to pay high attention to CC7, so in order to further explore which cells in CC7 contributed to the increase in hypoxic activity, we performed a correlation analysis based on the proportion of cells in the relative spot with the hypoxia score, and it was relatively obvious that neutrophils were significantly positively correlated with the hypoxia score (figure 1F). Co-localization analysis also revealed that neutrophils had a significant positive co-localization with lung cancer epithelial cells, which also suggested that neutrophils in this hypoxic region influence the progression of lung cancer by some mechanism (figure 1G,H). By immunofluorescence staining, we compared the proportions of HIF1A+ and CD15+ cells in the tissues of non-response and response patients, and we could indeed find a significant increase in the proportion of HIF1A+ CD15+ cells in the resistance group (figure 1I,J). At the same time, we used bulk tissue samples from patients treated with anti-PD-1 in our center, which were divided into high and low groups by the proportion of HIF1A and neutrophil infiltration, respectively, according to the median of each group, and it can be clearly seen that the group of high expression of HIF1A and high infiltration of neutrophils was enriched with the drug-resistant group (figure 1K). The scRNA-seq analysis also confirmed that neutrophils highly express HIF1A were proportionally more abundant in non-response patients (online supplemental figure S1D–F). Given that the formation of tertiary lymphatic vessels (TLS) often promotes a favorable response to immunotherapy in patients, and that there were differences in TP53 mutations between responders and non-responders at our cohort (figure 1C), we further analyzed the relationship between TP53 mutations, HIF1A-overexpressing neutrophils, and TLS. Interestingly, the presence of TP53 mutations promoted the infiltration of HIF1A-positive neutrophils in lung cancer and inhibited the formation of TLS (online supplemental figure S3A). Furthermore, we found that HIF1A+ neutrophils and the TLS score were significantly negatively correlated, and their spatial transcriptomic distributions also exhibited opposite trends. Notably, responders exhibited reduced HIF1A+ neutrophil infiltration and marked TLS formation, whereas non-responders showed increased HIF1A+ neutrophil infiltration and disrupted TLS (online supplemental figure S3B–H). The above results illustrated that neutrophils with HIF1A high expression were negatively correlated with TLS formation and highly enriched in the immunotherapy-resistant patients of NSCLC and may be a potential target. Spatial transcriptome niches analysis for anti-PD-1 sensitive and resistant patients of lung cancer. (A) Flowcharts of spatial transcriptomics sequencing, single-cell sequencing and RNA sequencing of patients treated with anti-PD-1 in our center were collected. (B) UMAP plot showing annotated cell types of our cohort. (C) Spatial distribution maps of different niches on the spatial transcriptome of PR patient, MPR patient, PD patient and SD patient. (D) Sankey diagram demonstrating the distribution of relative proportions of different niches in patients with different efficacies. (E) Bubble diagram demonstrating the activity of cancer-related pathways in different niches. (F) Scatterplot calculating the degree of correlation between the proportions of different cellular subpopulations and hypoxic activity in CC7. (G) Schematic diagram of cell infiltration patterns in hypoxic ecological niche. (H) Triangular heat map matrix indicated whether different cell subpopulations had co-localized relationships in spatial locations. (I) Immunofluorescence staining labeling of HIF1A and CD15 expression in groups with different treatment effects. (J) Bar plots comparing the differences in the proportions of HIF1A+ CD15+ cells in the response and non-response to anti-PD-1 groups. (K) Bar graphs comparing the relative proportions of high and low neutrophils infiltrations and high and low HIF1A-expressing samples in the resistant and sensitive groups. MPR, major pathological response; NSCLC, non-small cell lung cancer; PD, progressive disease; PD-1, programmed cell death protein 1; PR, pathological response; SD, stable disease; UMAP, uniform manifold approximation and projection. HIF1A+ CSF3R+ neutrophils were at the terminal end of neutrophil evolution and promoted glycolytic process Given the significant rise in neutrophils with hypoxic activity found in neoadjuvant therapy-resistant patients, we extracted neutrophils from our center’s single-cell dataset for further stratification (figure 2A). Functional annotation of distinct neutrophil subpopulations revealed a significantly higher proportion of HIF1A+ neutrophil enriched in the non-response group (figure 2B,C). Simultaneously, HIF1A and CSF3R expression in neutrophils showed a significant positive correlation. CSF3R encodes the granulocyte colony-stimulating factor receptor, a key regulator of neutrophil proliferation, differentiation, and survival (figure 2D). At the same time, external public single-cell dataset containing the largest number of samples from patients with lung cancer has a total of 1,254,749 immune cells, consistent with the results of our cohort, neutrophils had higher HIF1A expression relative to other immune microenvironment cells (online supplemental figure S4A–C). Further clustering yielded seven neutrophil subpopulations, and interestingly, the neutrophils_1 subpopulation had a higher proportion and significantly higher expression of HIF1A in the neoadjuvant therapy-resistant group (online supplemental figure S4D–F). Both the slingshot and monocle algorithms confirm that HIF1A+ neutrophils being at the end of the evolutionary trajectory (figure 2C, online supplemental figure S4G–J). Single-cell analysis at our center revealed HIF1A+CSF3R+ neutrophils promoted glycolysis. (A) UMAP plots demonstrating different neutrophils subpopulations. (B) The proportion of different neutrophil subsets in the drug-resistant group and the drug-sensitive group. (C) HIF1A expression in different neutrophil subpopulations. (D) The correlation of HIF1A+ neutrophils with CSF3R expression. (E) Pseudo-temporal evolutionary trajectories of different neutrophil subpopulations. (F) Heatmap showing genes that drive the evolution of cell subpopulations. (G) Genes related to HIF1A+ CSF3R+ neutrophil evolution. (H) Schematic of the process of detecting glycolysis levels in the response and non-response groups. (I) Western blot plot comparing the protein levels of key enzyme genes of the glycolytic pathway in the two groups. (J) Line graphs comparing ECAR levels in HIF1A+ neutrophils and HIF1A− neutrophils. (K) Bar graphs comparing glycolytic rate and glycolytic capacity. (L) Bar graph comparing the relative levels of lactate in different neutrophils. ECAR, extracellular acidification rate; LUAD, lung adenocarcinoma; UMAP, uniform manifold approximation and projection. By evaluating genes closely associated with the pseudo-time evolutionary trajectory, we found that LDHA (lactate dehydrogenase A) aligned with the HIF1A+ neutrophil evolutionary direction (figure 2F,G). LDHA is a key enzyme in the glycolytic pathway, responsible for catalyzing the conversion of pyruvate to lactate. On the basis of the results of data analysis we collected peripheral blood-sorted neutrophils from response and non-response patients with lung cancer and assessed that expression of enzyme genes related to the glycolytic pathway was indeed increased in drug-resistant patients (figure 2H,I). We comprehensively assessed HIF1A expression in patients’ tumor tissues using multiplex immunofluorescence and western blot, and classified patients into high-HIF1A and low-HIF1A expression groups accordingly. Subsequently, we collected peripheral blood samples from the two patient groups and found increased glycolytic activity and lactate efflux in high HIF1A expressed neutrophils group was further confirmed by ECAR assay (figure 2J–L). These results confirmed that HIF1A+ neutrophils represent the terminal stage of evolution toward malignant glycolytic metabolism, promoting resistance to neoadjuvant therapy in patients with NSCLC. HIF1A+CSF3R+ neutrophils exhibited senescent characteristics and malignant potential In order to better analyze the function of the HIF1A+ neutrophils subpopulation, we included other multicenter, multicohort lung cancer single-cell datasets, from which neutrophils were extracted for integrative analysis (figure 1B, online supplemental figures S4A and S5). A total of 34,698 neutrophils from eight cohorts were collected, and after comparing HIF1A expression by subpopulation, we categorized neutrophils into HIF1A+ neutrophils and HIF1A− neutrophils (figure 3A–C). Meanwhile, we found that HIF1A+ neutrophils highly expressed CSF3R (figure 3D). Using immunohistochemical staining and statistical analysis of multicenter single-cell datasets, we determined the proportion of HIF1A+CSF3R+ neutrophils among all neutrophils, finding that they accounted for 18–38% of total neutrophils and 3–8% of all immune cells. Concurrently, we observed an overlap in the distribution of HIF1A+CSF3R+ neutrophils and IL1B+ neutrophils. It has been established that IL1B+ neutrophils can upregulate the expression of genes associated with proliferation and anti-apoptosis in tumor cells. This further demonstrated that HIF1A+CSF3R+ neutrophils constitute a significant proportion of the lung cancer microenvironment and play a role in promoting the malignant progression of tumor cells (online supplemental figure S6A–L). Furthermore, by performing CIBERSORT deconvolution on lung cancer samples from our center and external cohorts to calculate the proportion of HIF1A+CSF3R+ neutrophils, we found that the average proportion of HIF1A+CSF3R+ neutrophils relative to total neutrophils was 51.4%, 7.3%, 78.2%, and their average proportion of all immune cells was 5.2%, 1.1% and 6.3%, respectively (online supplemental figure S6M–O). It was more obvious that the indicators of malignant secretory factor Arg-1 and iNOS were significantly elevated in the high HIF1A+ neutrophils group, and these two proteins have been reported to be significantly associated with tumor proliferation, metastasis, and immunosuppression (figure 3E). Meanwhile, we compared the correlation between HIF1A+ neutrophils scores and the prognosis of patients with lung cancer, and significantly patients with higher scores had worse prognosis (figure 3F). Analyzing the genomic mutation profiles of lung cancer and comparing the high and low HIF1A+ neutrophils score groups, it was found that more TP53 mutations occurred and the total mutation load was significantly increased in patients with lung cancer in the high score group (online supplemental figure S7A–F). Association of HIF1A+ neutrophils with glycolysis and cellular senescence. (A) UMAP shows the distribution of neutrophils from our cohort and other cohorts. (B) The distribution of HIF1A+ neutrophils and HIF1A− neutrophils. (C) Bubble plot showing the percentage of HIF1A+ neutrophils and HIF1A− neutrophils in multicohorts. (D) Violin plots comparing the expression levels of CSF3R in HIF1A+ neutrophils and HIF1A− neutrophils. (E) Western blot depicting the protein levels of secreted iNOS and Arg-1 in HIF1A+ neutrophils and HIF1A− neutrophils. (F) Survival curves comparing survival differences between high and low HIF1A+ neutrophils score groups. (G) Volcano plots showing differentially expressed genes in HIF1A+ neutrophils and HIF1A− neutrophils. (H–I) Meteor plots demonstrating differential gene enrichment pathways in KEGG (H) and GO (I). (J) Violin plots comparing the activity of the senescence pathway in HIF1A+ neutrophils and HIF1A− neutrophils. (K) Western blot comparing the differences in senescent protein levels in HIF1A+ neutrophils and HIF1A− neutrophils. (L) The RNA level of senescent-related genes. (M) Box line plots demonstrating HIF1A+ neutrophils scores in different age groups. (N) Bar graph comparing differences in cellular senescence scores in four different groups of neutrophils. (O–P) Correlation scatter plots demonstrate the correlation between HIF1A+ neutrophils scores and phosphorylated p21 and PD-L1 in different groups. FC, fold change; GO, gene ontology; KEGG, kyoto encyclopedia of genes and genomes; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; UMAP, uniform manifold approximation and projection. Given that HIF1A+ neutrophils had a malignant phenotype, we compared the differences between HIF1A+ neutrophils and HIF1A− neutrophils and found that in addition to the high expression of the HIF1A, there was also a high expression of CD274, which was elevated in expression with the direction of neutrophil evolution in the above analysis, as well as VEGFA, which was an important factor of the VEGF signaling pathway, and ENO1 was an enzyme associated with glycolysis (figure 3G). This compelled us to further explore the function of HIF1A+ neutrophils, and through pathway enrichment we found that HIF1A+ neutrophils were indeed significantly associated with highly expressed functional genes, including the glycolytic pathway, the VEGF signaling pathway, the PD-L1 checkpoint immunity pathway, and cellular senescence (figure 3H,I). Considering the previous enrichment to the cellular senescence pathway and previous studies reported that the glycolytic pathway was associated with cellular senescence, we further explored the intrinsic linkage with cellular senescence after upregulation of glycolytic activity in HIFA+ neutrophils (figure 3J). By detecting senescence-related genes, we found that Bcl-2, p16, p21 and PD-L1, which are senescence-related genes, were significantly highly expressed in the high HIF1A+ neutrophils group (figure 3K,L). At the same time, HIF1A+ neutrophils showed higher infiltration rates among older patients with NSCLC (figure 3M). We further categorized neutrophils into four groups based on HIF1A and glycolytic activity to assess the potential role of glycolysis in promoting cellular senescence and found that elevated glycolytic activity in HIF1A+ neutrophils could further induce cellular senescence (figure 3N). Meanwhile, the assessed phosphorylated senescent proteins also confirmed that the increased glycolysis level could further induce the activation of phosphorylated p21 and PD-L1 and promote cellular senescence in a high HIF1A+ neutrophils infiltration environment (figure 3O,P). The above results indicated that HIF1A+ neutrophils induced cellular senescence and were associated with malignant progression of patients with NSCLC. HIF1A+ neutrophil-induced the formation of an immunosuppressive microenvironment in NSCLC The above analysis identified high PD-L1 expression in HIF1A+ CSF3R+ neutrophils. PD-L1, as a member of senescence genes and a typical immune checkpoint, often interacts with PDCD1 to promote immune escape from tumors. Therefore, we further annotated 840,350 T cells into clusters to obtain 12 classes of T-cell subpopulations (online supplemental figure S8A–C). At the same time, we found that PDCD1 expression was highest in CD4+Tex and CD8+Tex, which potentially indicated the strongest interactions between HIF1A+ neutrophils and Tex (online supplemental figure S8D). By comparing the interactions of HIF1A+ neutrophils and HIF1A− neutrophils with each subpopulation of T cells, we found that the difference was more obvious in the existence of a strong PD-1/PD-L1 signaling interactions pathway between HIF1A+ neutrophils and T cells, and this pairwise interactions molecule in the HIF1A− neutrophils interactions with T cells were not detected (online supplemental figure S8E,F). By comparing the differences in TCR cloning between the high and low HIF1A+ neutrophils groups, it was found that CD8+ Tex and CD4+ Tex demonstrated a significant increase in the level of their clonal amplification in patients with high infiltration of HIF1A+ neutrophils, and the calculation of the clonogenic index by STARTRAC further identified a significant increase in the amplification of CD4+ Tex in the high infiltration group (online supplemental figure S8G–I). Simultaneously, HIF1A+ neutrophil-enriched regions were identified on the spatial transcriptomics map of non-response patients and defined as core ROI. Molecular interaction analysis within the ROI revealed that HIF1A+ neutrophils not only engaged in PD-1/PD-L1 interactions with T cells but also exhibited significant molecular interactions with fibroblasts, endothelial cells, and tumor cells (figure 4A–G). Immunofluorescence localization also revealed that high HIF1A+ neutrophils group with more CD8+ Tex and CD4+ Tex cells infiltration (figure 4H,I). Concurrently, we observed that the high infiltration of HIF1A+ neutrophils promoted the infiltration of fibroblasts and endothelial cells, leading to the formation of an immunosuppressive microenvironment. Intratumoral neutrophils and peripheral blood neutrophils were isolated and co-cultured with CAFs, respectively. We confirmed that high infiltration of HIF1A+ neutrophils promoted CAFs infiltration and upregulated the expression of fibroblast-associated proteins such as αSMA, FAP, S100A4 and PDGFβ (figure 4J–M and online supplemental figure 9). The in vitro experiment also confirmed that HIF1A+ neutrophils co-localized with endothelial cells, promoting endothelial cell infiltration in the tumor microenvironment (figure 4N,O). By tissue sequencing of patients with lung cancer from our cohort, we grouped them according to the HIF1A+ neutrophils score and calculated the proportion of other cellular components in the microenvironment through multiple algorithms. The correlation with other microenvironmental cell subpopulations confirmed that HIF1A+ neutrophils positively correlated with stromal cells, fibroblasts, endothelial cells, and M2 macrophage infiltration, while negatively correlating with CD8+ T cells, plasma cells, and natural killer (NK) cells, which play the roles of immune-killing and antigen-presenting cells (figure 4P–R and online supplemental figure 10). The above results suggest that HIF1A+ neutrophils remodel the stromal state of lung cancer, promote fibroblast infiltration and tumor neovascularization, and further form a malignant microenvironment that is immunosuppressive and potentially promotes tumor metastasis. HIF1A+ neutrophils promoted immunosuppressive microenvironment formation. (A) Spatial distribution of HIF1A+ neutrophils in non-response patient. (B) The region of interest in non-response patient. (C–F) Signaling molecule interaction pairs in different subpopulations. (G) Interaction pattern diagram with other cells centered on HIF1A+ neutrophils. (H–I) Immunofluorescence histochemistry demonstrates the proportion of CD8/CD4 and PDCD1 cells in tissues enriched with HIF1A+ neutrophils or HIF1A− neutrophils, respectively. (J) Schematic diagram of the process of co-culture of neutrophils and fibroblasts after sorting. (K) Staining co-localization of neutrophil and fibroblast-related genes in different subgroups. (L) Bar graph comparing the difference in the proportion of fibroblasts. (M) western blot demonstrating the levels of fibroblast-related proteins in HIF1A+ neutrophils and HIF1A− neutrophils co-cultured with fibroblasts. (N) Immunofluorescence histochemistry demonstrates the localized distribution of neutrophil subpopulations and endothelial cell signature genes. (O) Bar graph comparing the difference in the proportion of endothelial cells. (P) The CIBERSORT algorithm calculated the relative proportions of different cells in the samples of NSCLC from our center. (Q–R) Bubble plots demonstrating the correlation between HIF1A+ neutrophils scores and different cell proportions as well as the difference in the ratio of the proportions of different immune cells in the high and low HIF1A+ neutrophils groups. CAF, cancer-associated fibroblast; NSCLC, non-small cell lung cancer. Gradient boosting machine identifies ROI-specific genes to predict NSCLC tumor progression and microenvironment remodeling Given that HIF1A+ CSF3R+ neutrophils-ROI promoted the immunosuppressive microenvironment in NSCLC among non-responding patients, we explore differences between ROIs and other spatial transcriptomic spots and identified 12 genes exhibiting stepwise expression changes with distance from the ROI, with peak expression at the ROI center (figure 5A–C). Using these 12 genes, we constructed a GBM machine learning model, where FOSB emerged as the key gene distinguishing ROIs from other spatial locations (figure 5D,E). The model’s accuracy was validated in both internal training and validation datasets, enabling spatial prediction of ROI regions (figure 5F). ROI scores calculated using this model showed the highest values within ROI areas (figure 5G). Evaluation across different clinical contexts revealed that ROIs were significantly increased in advanced and metastatic patients (figure 5H,I). Notably, high ROIs independently predicted poor patient prognosis, independent of other clinical factors (figure 5J,K). We further assessed the connection between ROIs and the NSCLC microenvironment and functional pathways. High ROIs were found to induce tumor hypoxia, glycolytic pathway activation, and increased angiogenesis (figure 5L). Additionally, ROIs were negatively correlated with effector T, B, and NK cell infiltration, while positively correlated with endothelial cell, fibroblast, M2 macrophage, and regulatory T-cell infiltration (figure 5M). These findings demonstrated that the constructed GBM model accurately predicted ROIs in spatial omics, with ROI-associated genes correlating with patient prognosis, treatment efficacy, and microenvironmental changes. ROI-related gene-building model predicted patient progression and tumor microenvironment. (A) Spatial positional targeting of areas of interest. (B) Changes in expression gradients on ROI-centered spatial transcriptomes. (C) The line graphs showing the altered expression of different genes as the distance from the ROI changes. (D) The Force plot showing the contribution of each feature to the predicted value through the stepwise change in SHAP values. (E) The Waterfall plot shows the contribution of each feature to the final prediction through the SHAP values. (F) ROC curves evaluated the performance of the model in the training and validation sets. (G) ROI scores for gene set construction of spatial presentation training models. (H–I) Differences in scoring of ROIs among patients with different staging and different M-stages. (J) Survival differences in subgroups of high and low ROIs. (K) Multifactorial cox plots comparing different variables with survival outcomes. (L) The violin plot compares differences in hypoxia, glycolysis, and angiogenesis pathway activity between high-ROI and low-ROI groups. (M) The bar chart illustrates the correlation between ROI scores and the degree of infiltration by different cell types. ROC, receiver operating characteristic; ROI, region of interest; SHAP, shapley additive explanations. Navitoclax and platycodin-D2 targeted HIF1A+CSF3R+ neutrophils increased sensitivity to immunotherapy in patients with lung cancer Given that HIF1A+ CSF3R+ neutrophils promoted the formation of an immunosuppressive microenvironment in lung cancer, we further evaluated the association of HIF1A+ CSF3R+ neutrophils infiltration with immunotherapy in multiple datasets. In the bladder cancer dataset, patients treated with anti-PD-1 with nonresponse exhibited higher HIF1A+CSF3R+ neutrophils scores, and patients with high scores had significantly poorer prognosis (figure 6A). Meanwhile, in patients with colorectal cancer treated with anti-PD-1 and anti-CTLA-4 the HIF1A+ CSF3R+ neutrophils scores were significantly higher in the same non-responsive patients and survival was worse in the high-scoring patients (figure 6B). In other multicenter datasets, we grouped patients according to their HIF1A+ CSF3R+ neutrophils scores and found that the high-scoring group was indeed enriched in patients with resistance to different combinations of immunotherapies received (figure 6C–I). These results illustrated that HIF1A+ CSF3R+ neutrophils not only affected the efficacy of immunotherapy in patients with lung cancers, but also have a role in hindering immunotherapy efficacy in pan-cancer. Therefore, in order to improve patients’ drug resistance, we explored potential drugs targeting HIF1A+ CSF3R+ neutrophils by drug sensitivity analysis of single cells. Lung cancer tumor epithelial cells were grouped according to the percentage of HIF1A+CSF3R+ neutrophils infiltration in patients with lung cancer, and it can be found that the tumor cells in this high infiltration group showed potential sensitivity to the navitoclax drug, and higher beyondCell scores represented the more sensitive the cells were to the drug (figure 6J–M). We also applied the cMAP database for small molecule drug analysis and found that navitoclax indeed had a role in negatively regulating HIF1A+CSF3R+ neutrophils (figure 6N). Our previous research revealed that HIF1A+CSF3R+ neutrophils exhibited senescent characteristics, while navitoclax is an inhibitor targeting senescent cells. This suggested that the drug may be a potential effective agent for suppressing the function of HIF1A+ CSF3R+ neutrophils. Prediction of small molecule drugs targeting HIF1A+ neutrophils. (A–C) IMvigor210 is a bladder cancer dataset treated with anti-PD-1 (A), GSE179351 is a colorectal cancer dataset treated with anti-PD-1 and anti-CTLA-4 (B), GSE78220 is a metastatic melanoma dataset treated with anti-PD-1 (C), violin plots comparing the scores of HIF1A+ neutrophils in the response and non-response groups. Survival curves compare survival differences between high and low HIF1A+ neutrophils scores. (D–I) Bar graphs show the differences in the relative proportions of patients with high and low HIF1A+ neutrophils scores in the non-response and response groups, respectively, in the different datasets; PRJEB23709 is metastatic melanoma that received anti-PD-1/anti-CTLA-4 (D); GSE67501 is renal cell carcinoma received anti-CTLA-4 (E); GSE176307 was metastatic urothelial carcinoma that received anti-PD-1 (F); SRP067586 melanoma dataset received anti-CTLA-4 (G); GSE115821 was metastatic melanoma that received anti-PD-1/anti-CTLA-4 (H); GSE145996 is metastatic melanoma treated with anti-PD-1 (I). (J) UMAP plots demonstrating the distribution of epithelial cell subpopulations with high and low HIF1A+ neutrophils ratios. (K) Bar graph comparing the predicted drug sensitivity scores of tumor epithelial cells infiltrated with high HIF1A+ neutrophils to different drugs. (L) UMAP plot demonstrating the distribution of navitoclax scores in epithelial cells. (M) Violin plot comparing the epithelial cells with high and low HIF1A+ neutrophils proportions to the navitoclax drug sensitivity scores. (N) Lollipop plot demonstrating the potential effects of different drugs on HIF1A+ neutrophils. CTLA-4, cytotoxic T-lymphocyte associated protein 4; PD-1, programmed cell death protein 1; UMAP, uniform manifold approximation and projection. CADD prediction identified platycodin-D2 as the small-molecule drug with the highest specificity for binding to the CSF3R protein. This compound, derived from the campanulaceae plant family, exhibits anti-inflammatory, antioxidant, antitumor, and immune-boosting properties (figure 7A–D). As the concentration of platycodin-D2 increases, the surface plasmon resonance signal exhibits a concentration-dependent response, suggesting that platycodin-D2 can directly bind to the CSF3R protein (figure 7E). At the same time, we isolated neutrophils from CSF3R-OE, CSF3R-KO, and corresponding wild-type control mice. Using western blot, quantitative PCR, and pharmacological intervention experiments, we found that CSF3R activation promotes HIF1A expression via the JAK1-STAT3 signaling axis, while inhibition or activation of STAT3 can, respectively, block or restore the CSF3R-mediated regulatory effects on HIF1A, validating the mechanism by which CSF3R regulates HIF1A via STAT3 (online supplemental figure S11A–G). Platycodin-D2 not only targeted CSF3R, specifically inhibited CSF3R+ neutrophils, but also suppressed HIF1A function (online supplemental figure S11H). In vivo therapeutic experiments in mice with subcutaneous tumor transplantation and mice with in situ tumors revealed that compared with anti-PD-1 monotherapy, combination therapy effectively suppressed tumor growth (figure 7F–H, online supplemental figures S13A–C and S14A,B). Combined anti-PD-1 therapy with navitoclax and platycodin-D2 significantly reduced HIF1A+ CSF3R+ neutrophils infiltration in tumors (figure 7I,J, online supplemental figures S13D,E and S14C,D). The proportion of Ki67+ tumor cells, associated with proliferation, was significantly reduced in the combination therapy group, while the proportion of TUNEL+ tumor cells, associated with apoptosis, was markedly increased (figure 7K,L, online supplemental figures S12A–D, S13F,G and S14E,F). Combined anti-PD-1 therapy with navitoclax and platycodin-D2 significantly improved the immunosuppressive microenvironment in lung cancer, which increased the proportion of IFN-γ+ CD8+ T cells and decreased the proportion of PD-1+ CD8+ T cells (figure 7K,L, online supplemental figures S13H,I and S14G,H). Platycodin-D2 and navitoclax exhibited synergistic effects with anti-PD-1 (figure 7M,N). At the same time, we generated a model specifically knocking out CSF3R in neutrophils by crossing csf3rfl/fl and Ly6G-cre mice to evaluate the effect of platycodin-D2 on HIF1A+CSF3R+ neutrophils, confirming that platycodin-D2 primarily inhibits lung cancer progression by targeting HIF1A+CSF3R+ neutrophils (figure 7O–Q). We also assessed the toxicity of navitoclax and platycodin-D2, both as monotherapies and in combination, in multiple models, confirming that they are non-toxic to a variety of organs (online supplemental figure S15). The above findings revealed that HIF1A+ CSF3R+ neutrophils induced anti-PD-1 resistance in patients, but combination therapy with navitoclax and platycodin-D2 could sensitize patients to the efficacy of immune inhibitors. Navitoclax, platycodin-D2 and anti-PD-1 significantly suppressed the malignant function of HIF1A+CSF3R+ neutrophils. (A–D) Computer-aided molecular drug screening visualization depicts the process of platycodin-D2 specifically binding to the CSF3R protein. (E) Surface plasmon resonance analysis of the binding affinity between platycodin-D2 and CSF3R. (F) Schematic diagram of subcutaneous xenografts in mice receiving different treatment regimens. The dosage and timing of administration are indicated on the diagram: Navitoclax: 50 mg/kg/day p.o.; platycodin-D2: 10 mg/kg/day p.o.; anti-PD-1: 12.5 mg/kg/week i.p. (G) The line chart compares tumor volume across different treatment groups. (H) The bar plot compares tumor weight across different treatment groups. (I–J) Immunofluorescence and bar charts displaying the number of HIF1A+ CSF3R+ Neutrophils in different treatment groups. (K–L) Flow cytometry analysis compared the proportions of IFN-γ+ CD8+ T cells across different treatment groups. (M–N) The heatmap illustrates the correlation between anti-PD-1 and navitoclax or platycodin-D2, respectively. (O) The flowchart illustrates the process of generating a mouse model with specifically knockout CSF3R in neutrophils by crossing CSF3Rfl/fl and Ly6G-Cre mice. (P) The survival curve compares the survival times of mice in different groups. (Q) The bar chart compares tumor volume across different combination groups based on the presence or absence of CSF3R knockout and the use of drug platycodin-D2. IFN, interferon; i.p., intraperitoneal; PD-1, programmed cell death protein 1; p.o., per os. Discussion Lung cancer, as one of the frequently occurring cancer types, is characterized by obvious invasiveness, proliferation and metastasis. With the progress of medical research, there are various choices of treatment for lung cancer, but the complexity of the internal environment of the tumor leads to drug resistance and tumor recurrence, thus making the patients suffer from multisite metastasis of the tumor.27 28 Immunotherapy, as an important method of lung cancer treatment, has been widely used in the clinic, but the increasing number of patients with drug resistance has aroused our concern. Therefore, by collecting lung cancer tissue samples from patients treated with anti-PD-1 in our center, we performed spatial transcriptomic sequencing and RNA transcriptomic sequencing, respectively. In-depth analysis of the differences in the tumor microenvironment in treatment-resistant and sensitive groups of patients, thus revealing important factors in immunotherapy resistance. After the niche analysis of spatial, niches with hypoxic properties were clearly present in drug-resistant tissues. Hypoxia is a widespread cancer hallmark in solid tumors, and the interplay of the hypoxia signaling pathway with other cellular pathways is closely related to cancer cell proliferation, migration, invasion, and angiogenesis, and affects cancer treatment outcomes. While previous studies have mostly focused on the exploration of hypoxic tumor cells, we identified highly relevant neutrophil subpopulations in the ecological niche of hypoxia and their co-localization with lung cancer tumor cells. This phenomenon led us to further consider that the role of neutrophils with hypoxic activity in lung cancer treatment needs to be explored in depth. Neutrophils have a dual role in tumor progression, being both antitumor and promoting tumor proliferation. For a long time, their role in diseases such as cancer has been seriously underestimated because they are considered “short-lived cells” and have low transcriptional activity.29 Clinical data have shown that neutrophils are strongly associated with poor prognosis in a variety of cancers.30–32 The CIBERSORT team’s analysis of 18,000 tumor specimens found that neutrophil counts were significantly associated with reduced survival in 25 common cancers.33 To explore the role of neutrophils in lung cancer, we integrated a multicenter, multicohort single-cell dataset to extract neutrophils for more in-depth study. Through neutrophil prototyping, we clarified that HIF1A+ neutrophils were at the end of neutrophil subpopulation evolution and are significantly associated with drug resistance and poor prognosis in patients. And neutrophils highly expressing HIF1A secreted Arg-1 and iNOS, factors that inhibit the immune response and promote tumor metastasis, and this population of cells was significantly enriched in patients with advanced T staging. The hypoxic microenvironment in tumors is often associated with metabolic reprogramming, and through differential analysis and pathway enrichment, we also identified a significant increase in glycolytic activity in HIF1A+ neutrophils. When hypoxia occurs, cells are unable to generate sufficient energy through aerobic respiration, whereas glycolysis can be carried out under anaerobic conditions to provide cells with the necessary ATP. We verified in vitro that both enzyme genes related to the glycolytic pathway and lactate secretion were indeed significantly upregulated in HIF1A+ neutrophils to meet the survival of this neutrophil subpopulation in lung cancer tissues. At the same time, in HIF1A+ neutrophils, we also found an upregulation of cellular senescence-related gene activity, with both western blot and phosphorylated protein dataset analyses confirming a significant increase in the protein levels of p16, p21, Bcl-2 and PD-L1. PD-L1 is normally an immune checkpoint associated with immune escape from tumors, but recent publications have indicated that PD-L1 has also been included as a senescence marker, and several articles have reported increased expression of PD-L1 in senescent cells.34 Meanwhile, previous studies have revealed that a large amount of lactic acid is produced during glycolysis, and the accumulation of lactic acid may alter the acid-base balance in cells, affecting the activity of various enzymes and the structure and function of proteins in cells.35 At the same time, glycolysis also produces reactive oxygen species (ROS), and the excessive production of ROS can trigger oxidative stress, resulting in oxidative damage to intracellular lipids, proteins, DNA and other macromolecules, thereby accelerating cellular aging.36 37 Therefore, we divided HIF1A+ neutrophils and glycolysis scores into high and low groups, respectively, and compared the activity of the senescence pathway and the expression of senescence-related genes between the different groups, and found that cells or tissues highly infiltrated with HIF1A+ neutrophils and accompanied by a high level of glycolytic activity possessed a stronger tendency to senescence. Thus, the HIF1A+ neutrophils we found in the hypoxic niches were characterized by glycolytic activity and senescence. Neutrophils influence the degree of infiltration of other immune cells or stromal cells in the tumor microenvironment, and previous studies have reported that neutrophils of the antitumorigenic type can directly kill the tumor capacity and promote the immune function of CD8+ T cells,38 whereas neutrophils of the protumorigenic type accelerate the progression of the tumor through the secretion of inflammatory factors, the release of NETs, and the inhibition of CD8+ T-cell function.39 Therefore, in order to investigate the role of HIF1A+ neutrophils on the remodeling of the immune microenvironment in patients with lung cancer, we performed an interactions analysis between HIF1A+ neutrophils and other T-cell subpopulations and stromal cell subpopulations, and found that HIF1A+ neutrophil promoted the CD8+ T and CD4+ T cells underwent functional depletion through PD-1/PD-L1 interaction, and the results of further TCR sequencing analyses also confirmed that HIF1A+ neutrophils could promote the clonal expansion of CD8+ Tex and CD4+ Tex. In vitro experiments also verified that the ratio of CD8+Tex and CD4+ Tex increased in tissues infiltrated by HIF1A+ neutrophil, and these phenomena indicated that the HIF1A+ neutrophil in the present study was a group of pro-carcinogenic neutrophils that suppressed immunoreactivity. In interaction and co-culture experiments with stromal cells, HIF1A+ neutrophil promoted both neovascularization and fibroblast infiltration, and remodeled the lung cancer ECM (extracellular matrix), which had a significant role in the formation and progression of the microenvironment of lung cancer. The upregulation of CXCL8 in neutrophils promotes its interaction with the ACKR1 receptor on endothelial cells and the activated VEGF signaling pathway further promotes neovascularization, which potentially facilitates the possibility of metastasis in lung cancer. The VEGF is a highly biologically active glycoprotein, and the most active of the family group is VEGFA. It has the ability to promote neovascularization and fragmentation and also to increase capillary permeability.40 It has been shown that on binding of VEGF factors to the receptor, three important signaling pathways were formed: prompting the neovascularization of blood vessels, inducing vascularization towards tumor cells and increasing vascular permeability.41 Given the tumor-promoting effects of HIF1A+ neutrophil, and its role in promoting patient resistance in both pan-cancer immunotherapy datasets, we unearthed the potentially targeted drug navitoclax by performing drug sensitivity analyses in single cells and further targeted screening with a small molecule drug database. Navitoclax, also known as ABT-263, ABT263, is a potent, orally bioactive inhibitor of the Bcl-2 family of proteins that acts on Bcl-xL and Bcl-2 to clear senescent cells.42 CADD identified platycodin-D2 as a potential effective drug for inhibiting HIF1A expression and activity of neutrophils. Compared with the single-agent anti-PD-1 and no drug intervention groups, the combination therapy group showed a significant reduction in tumor load and an increase in the number of functionally-killed CD8+ T, and the immunosuppressive microenvironment of immunosuppression was improved. We therefore propose a potentially viable treatment option for patients with lung cancer resistant to anti-PD-1 therapy, in which the combination of navitoclax and platycodin-D2 could sensitize the efficacy of conventional ICIs. Although our study highlights that the presence of HIF1A+ neutrophils in lung cancer was critical in causing immunotherapy resistance, there are still some shortcomings. The specific mechanism by which platycodin-D2 targets HIF1A+ neutrophils was not explored in depth, and we should expand the clinical cohort and validate the utility of the combination regimen through a multicenter cohort trial to make up for the shortcomings of the animal model. Platycodin-D2 may exhibit multitarget regulatory characteristics; therefore, the involvement of other molecules in its biological effects cannot be completely ruled out, and further systematic evaluation of its off-target effects is needed in the future. Overall, our study clarifies that HIF1A+ neutrophils undergo glycolytic metabolic reprogramming to promote cellular senescence and increase the activity of PD-L1 and VEGF signaling pathways in lung cancer, which further leads to impaired T-cell immune-activated function and neovascularization in the tumor microenvironment, resulting in therapeutic resistance and tumor progression. Supplementary material