Intratumoral Fusobacterium nucleatum Recruits Tumor-Associated Neutrophils to Promote Gastric Cancer Progression and Immune Evasion Open Access

Gastric cancer is the fifth most prevalent cancer worldwide, accounting for approximately 5.6% of all cancer cases. It is the fourth leading cause of cancer death, accounting for 7.7% of such mortalities (1). However, strategies for early diagnosis and treatment of gastric cancer are still inadequate. Emerging research indicates that apart from Helicobacter pylori, other microbial communities may contribute to the progression of gastric cancer (2–4). Yu and colleagues (5) identified a novel pathogenic bacterium for gastric cancer, Streptococcus anginosus, which interacts directly with gastric mucosal epithelial cells and fosters gastric cancer development by modulating the TMPC-ANXA2-MAPK axis. Nevertheless, the precise mechanisms by which intratumoral microbiota affect gastric cancer’s malignant progression and tumor immune microenvironment (TIME) remain elusive.

Recent studies have been increasingly focusing on exploring the impact of intratumoral microbiota on malignant tumor progression and TIME. The significance of tumor-associated neutrophils (TAN) in tumor advancement has gained notable attention recently, chiefly because of their distinctive subtypes and influence on neoplastic growth. The two main subtypes of neutrophils, N1 and N2, encompass antitumor and protumor properties, respectively (6–8). Recent research has underscored the significant role of TANs in regulating the TIME and antitumor immune responses (9, 10). Nevertheless, the interplay between the microbiota and TANs in gastric cancer and how TANs can influence the TIME, leading to alterations in malignant progression, remains poorly elucidated. Moreover, their impact on modulating immune checkpoint inhibitor efficacy is frequently undervalued in current literature.

This study primarily aimed to examine the potential influence of intratumoral microbiota on the progression of gastric cancer and its interaction with the TIME. In this study, F. nucleatum was identified as a novel non–H. pylori pathogen that promotes gastric cancer immune evasion and boosts immunotherapy efficacy. This gram-negative, nonspore-forming opportunistic commensal anaerobe is predominantly found in the oral cavity and has been linked to various manifestations in periodontal diseases and cancer (11–15). Recent research has shed light on the role of F. nucleatum in the development of gastrointestinal cancers (16). Particularly, F. nucleatum promotes tumorigenesis (12), inhibits cancer cell apoptosis (13), and induces resistance to chemotherapy (17). However, the precise involvement of F. nucleatum in gastric cancer malignant progression and its specific pathogenic molecular mechanisms remain largely ambiguous. In the current study, F. nucleatum was found to invade gastric cancer cells and activate IL17 signaling, which in turn promotes the recruitment of neutrophils and the differentiation of PD-L1⁺ TANs. This contributes to the exhaustion of CD8⁺ T cells, facilitating immune evasion in gastric cancer and augmenting the effectiveness of PD-L1 immune therapy. Our study has identified a novel non–H. pylori pathogen, F. nucleatum, which plays a crucial role in immune evasion in gastric cancer and modulates the response to immunotherapy.

Tumor tissues were collected from 133 patients [40 patients between 2020 and 2022 (cohort 1) and 93 patients between 2004 and 2007 (cohort 2)] with gastric cancer treated at the First Affiliated Hospital of Sun Yat-sen University (Guangzhou, Guangdong, China). Cohort 1 receives fresh-frozen human tissue samples of patients with gastric cancer primarily for 16S rRNA sequencing and qPCR analysis to investigate microbial community distribution and analyze progression-free survival (PFS). Cohort 2 involves human gastric cancer tissue microarray of formalin-fixed, paraffin-embedded samples, utilized mainly for in situ hybridization to detect F. nucleatum distribution and conduct overall survival (OS) analysis, including data from follow-ups exceeding 5 years. All patients were preoperatively diagnosed with gastric cancer and underwent radical resection. All tissue samples were isolated and frozen at −80°C or formalin-fixed and paraffin-embedded. This study was approved by the Institutional Review Board of the First Affiliated Hospital of Sun Yat-sen University and conducted according to the tenets of the Declaration of Helsinki. Each participant signed a consent form before participating in the study.

The 615, C57BL/6 (RRID: IMSR_JAX:000664), and BALB/c mice (RRID: MGI:2683685) were obtained from the Animal Experiment Center of Sun Yat-sen University and Youda (Guangzhou) Biotechnology Co., Ltd. and were raised in specific pathogen-free animal rooms. The Institutional Ethics Committee of the First Affiliated Hospital for Animal Studies approved the animal study protocols. The ARRIVE reporting guidelines were followed (18). The animal study was carried out in compliance with the guidance of the Animal Care Committee of the First Affiliated Hospital of Sun Yat-sen University [permit number: (2021)737]. The laboratory management personnel allocated the mice and cage placements using a random grouping method, thereby ensuring single blinding for the experiment operators. The research personnel randomly divided the mice into a control group and a treatment group. All mice were raised in a controlled, specific pathogen-free environment. The mice were humanely euthanized via cervical dislocation following anesthesia with isoflurane and subsequently dissected. The sample size was based on estimations by power analysis with a level of significance of 0.05. Animals without tumors were excluded from the study.

The C57BL/6 mice (RRID: IMSR_JAX:000664) were randomly divided into two groups. The model group was orally gavaged with H. pylori SS1 (2 × 10⁹ CFU in 200 μL PBS) 6 times in 2 weeks to establish successful colonization within the stomach. The experimental group was divided into three groups [H.p, Escherichia coli^{H. p+}(E. coli^{H. p+}), and F. n^{H. p+}groups] 12 months after the last feeding of the bacterial solution. The H.p, E. coli^{H. p+}, and F. n^{H. p+} groups were orally gavaged with PBS (200 μL), DH5α (2 × 10⁸ CFU in 200 μL PBS), and F. nucleatum (ATCC 25586, 2 × 10⁸ CFU in 200 μL PBS) strains for 6 months, respectively. In short, C57BL/6 male mice were orally gavaged with H. pylori SS1 (n = 27 for 12 months), F. nucleatum (n = 9 for 6 months), PBS (n = 9 for 12 months, n = 9 for 6 months), or E. coli (n = 9 for 6 months). We examined the color and texture of the stomach and perigastric lymph nodes (LN) following dissection. The perigastric tissues were separated from the stomach and removed. The appetizing tissue along the stomach axis curvature was cut, the gastric contents were washed, the gastric mucosa was exposed, and the mucosa was observed and separated into 4% paraformaldehyde and a cryoprotective solution.

The female C57BL/6 (RRID: IMSR_JAX:000664) conventional mice (6–8 weeks old) were randomly assigned into three groups: PBS (negative control), E. coli (positive control), and F. nucleatum (F.n). The PBS, E. coli, and F.n groups were orally gavaged with the PBS (200 μL), DH5α (2 × 10⁸ CFU in 200 μL PBS), and F. nucleatum (2 × 10⁸ CFU in 200 μL PBS) strains for 3 months.

In this model, bacteria were first retrieved from an anaerobic workstation, whereas F. nucleatum was washed off blood plates and centrifuged at 4,000 g for 3 minutes to remove the supernatant. After washing once with PBS and resuspending, bacterial liquid OD600 absorbance was measured. When the OD600 value reached 1, the quantity of F. nucleatum was determined to be 10⁹ CFU/mL. Subsequently, F. nucleatum was cocultured with gastric cancer cells at a density of multiplicity of infection (MOI) 100, using a culture medium without antibiotics. After coculturing for 4 hours, the cell culture medium was discarded, and the cells were washed twice with PBS before being cultured in complete growth medium containing nonpenetrating antibiotics for 24 to 48 hours. The cells were then retrieved from the 37°C incubator, and the cell pellets were collected at a centrifugal speed of 1,000 rpm/minute for 5 minutes. Bacterial precipitate was resuspended using PBS or serum-free medium. The cell suspension was then injected subcutaneously into BALB/C mice and immunocompetent 615 mice (the 615 mice strain was developed by crossing C57BL/6 and BALB/c mice, resulting in a hybrid strain with genetic characteristics from both parental strains). Around the fourth day of the formation of a subcutaneous tumor, an injection of bacterial suspension containing 5 × 10⁷ CFU in 50 μL PBS was administered into or around the tumor. The 15-cm culture dish was used to cultivate the mouse forestomach carcinoma (MFC) and human MGC803 cell lines to construct a 615 and BALB/c mice subcutaneous tumor xenograft model at a density of 5 × 10⁶ cells per mouse, with six mice per group. The cells were prepared before performing the animal experiments as follows. We injected 100 μL of the cell suspension into the right armpit/groin of nude mice using a 1-mL insulin injection needle, and the subcutaneous tumor formation in the mice was then observed. After 3 to 4 days, the long and short diameters of the subcutaneous tumors were measured twice a week using a Vernier caliper. The tumor volume under the short axis was calculated, and a growth curve for the tumor volume was drawn using the following formula: Mice were euthanized by cervical dislocation approximately 3 to 4 weeks later. The subcutaneous tumors in the axilla/groin of the mice were dissected, weighed, soaked in 4% paraformaldehyde, and fixed for subsequent experiments.

A 615 mice/nude mouse tail vein–lung xenograft tumor model was constructed by injecting 1 × 10⁶ cells per mouse, with 6 to 8 mice per group, and using a 15-cm dish to culture the MFC cell strain (luciferase-labeled) and human MGC803 gastric cancer cell line (luciferase-labeled). In the tail vein–pulmonary metastasis model, the mouse was fixed with clamps or a mortar, 100 μL of cell suspension was injected into the tail vein of the mouse using an insulin injection needle, and the vital signs were observed. The in vivo imaging systems (Lumina II, Xenogen Corp.) were used to capture bioluminescent images after 6 to 8 weeks. Mice were euthanized by cervical dislocation after images were taken, and the lungs were dissected and photographed, fixed with 4% paraformaldehyde, embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E). The number of lung metastatic tumors was counted under a microscope (see below).

In the footpad–popliteal LN metastasis model, cells were prepared as described above and injected into the mouse footpad at 1 × 10⁶ cells. Approximately 1 week after the growth of the footpad tumor, F. nucleatum suspension with 1% PBS solution was injected twice a week or injected with a 1 × 10⁶ cell suspension (cocultured with F. nucleatum) one time, and bioluminescent images were captured after 4 to 6 weeks. Footpad tumors and popliteal LNs were dissected and removed, photographed, and fixed in 4% paraformaldehyde for subsequent experiments.

The gastric cancer tissue samples stored at −80°C were thawed. Subsequently, they were fully ground using a grinding bead homogenizer (BioSpec). The QIAamp PowerFecal DNA Kit (Qiagen) was used to extract microbial DNA from the resulting tissue pellets. Sequencing was performed on the MiSeq platform (Illumina) with the MiSeq Reagent Kit version 3 (Illumina). The off-machine data were split using MiSeq Reporter (Illumina). Using specific primers with barcodes, the V3–V4 region of the bacterial 16S rRNA gene was amplified with the primer pair 338F (5′-ACT​CCT​ACG​GGA​GGC​AGC​AG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). A QuantiFluor fluorometer was used to quantify the PCR amplification product. The purified amplification product was mixed in the same amount, the sequencing connector was connected, the sequencing library was constructed, and the Illumina PE250 was sequenced on the machine (Supplementary Table S1).

Alpha diversity was applied in analyzing the complexity of species diversity for a sample through 14 indices, including richness, chao1, shannon_2, shannon_e, shannon_10, jost, jost1, simpson, dominance, equitability, robbins, berger_parker, reads, and buzas_gibson. All the indices in our samples were calculated with usearch-alpha_div (V10, http://www.drive5.com/usearch/), and rarefaction curve and rank abundance were calculated separately. The differences between groups were analyzed by alpha diversity index using R software. If there were only two groups, the Student t test or Wilcoxon rank-sum test was selected, but if there were more than two groups, the Kruskal–Wallis rank-sum test or one-way ANOVA was used.

Beta diversity analysis was used to evaluate differences in species complexity among samples through nine algorithms, including bray_curtis, Euclidean, abund_jaccard, Canberra, chisq, chord, gower, weighted_unifrac, and unweighted_unifrac, using R software. Principal component analysis was performed to get principal coordinates and visualize complex and multidimensional data using the prcomp function in R software. Principal Coordinates Analysis (PCoA) and non-metric multidimensional scaling (NMDS) were displayed using the vegan package (RRID: SCR_011950) in R software. Sample cluster analysis was performed as an unweighted pair group method with arithmetic means method to interpret the distance matrix using average linkage based on the nine algorithms mentioned above. A distance heatmap was generated based on the results obtained from integrating the nine algorithms, followed by hierarchical clustering using the “hclust” function in the vegan package of R software.

Linear discriminant analysis (LDA) effect size (RRID: SCR_014609) was used to find the biomarker of each group based on a homogeneous OTU table. First, a nonparametric factorial Kruskal–Wallis sum-rank test was used to detect species with significant differences in abundance among different groups. Second, the Wilcoxon rank-sum test was used to judge the difference between the two groups. Finally, LDA was used to evaluate the impact of significant species (LDA score) by setting the LDA score ≥2 and obtaining the biomarkers in different groups. If there were only two groups, Student t test, Wilcox rank sum test, Wilcoxon signed rank test or Welch t test was used to evaluate by R software, and corrected though FDR. But if there were more than two groups, Kruskal Wallis rank sum test or one-way ANOVA was used to evaluate by R software, and corrected though FDR.

RNA sequencing (RNA-seq) was performed on an Illumina HiSeq 3000 platform (Shanghai Genergy BioTech), following the manufacturer’s instructions. Data were analyzed according to the TopHat-HTSeq-DeSeq2 framework. The raw RNA sequence data were deposited in the Gene Expression Omnibus (GEO) database (RRID: SCR_005012). Differentially expressed genes (DEG) between control and F. nucleatum–infected groups in bulk transcriptomic data were identified using DESeq2 (RRID: SCR_015687).

BD Rhapsody system was used to capture the transcriptomic information of the single cells. Single-cell capture was achieved by random distribution of a single-cell suspension across >200,000 microwells through a limited dilution approach. Beads with oligonucleotide barcodes were added to saturation so that a bead was paired with a cell in a microwell. Cell lysis buffer was added to hybridize polyadenylated RNA molecules to the beads. Beads were collected into a single tube for reverse transcription. Upon cDNA synthesis, each cDNA molecule was tagged on the 5′ end (i.e., the 3′ end of an mRNA transcript) with a unique molecular identifier and cell label indicating its cell of origin. Whole-transcriptome libraries were prepared using the BD Rhapsody single-cell whole-transcriptome amplification workflow. In brief, second-strand cDNA was synthesized, followed by ligation of the whole transcriptome analysis (WTA) adaptor for universal amplification. A total of 18 cycles of PCR were used to amplify the adaptor-ligated cDNA products. Sequencing libraries were prepared using random priming PCR of the whole-transcriptome amplification products to enrich the 3′ end of the transcripts linked with the cell label and unique molecular identifier and were quantified using a High Sensitivity DNA chip (Agilent) on a Bioanalyzer 2200 and the Qubit High Sensitivity DNA assay (Thermo Fisher Scientific). All libraries were sequenced by DNBSEQ-T7 Sequencer (MGI) on a 150-bp paired-end run.

To elucidate the biological implications of the DEGs and marker genes, gene ontology (GO) analysis was performed. GO annotations were downloaded from NCBI (http://www.ncbi.nlm.nih.gov/; RRID: SCR_006472), the Gene Ontology database (http://www.geneontology.org/), and UniProt (http://www.UniProt.org/; RRID: SCR_002380). Fisher’s exact test was applied to identify the significant GO categories, and an FDR was used to correct the P values.

Pathway analysis was used to explore the significant pathways of the DEGs and marker genes based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (RRID: SCR_001120). Fisher’s exact test was used to identify significant pathways, and the threshold of significance was defined by the P value and FDR.

To characterize the relative activation of a given gene set, we performed QuSAGE (2.16.1) analysis (RRID: SCR_024255).

To enable a systematic analysis of cell–cell communication molecules, we applied cell communication analysis based on the CellPhoneDB (version 2.0; https://github.com/Teichlab/cellphonedb;RRID: SCR_017054), a public repository of ligands, receptors, and their interactions (19). Membrane, secreted, and peripheral proteins of the cluster of different time points were annotated. Significant mean and cell communication significance (P value < 0.05) were calculated based on the interaction and the normalized cell matrix achieved by Seurat normalization. Cell communication analysis was performed using the R package CellChat (version 1.1.3; RRID: SCR_021946) with default parameters. The CellChat DB human was used for analysis. All two groups of samples were normalized together, and then each group was extracted, analyzed, and compared in parallel. This is with the assumption that they were sharing cell types.

To identify differentially expressed genes among samples, the function FindMarkers with the Wilcoxon rank-sum test algorithm was used under the following criteria: (i) log₂ fold change > 0.25, (ii) P value < 0.05, and (iii) minutes.pct > 0.1, which was more conducive to obtaining different functional genes among samples.

Human gastric cancer cell lines (AGS, RRID: CVCL_0139; MKN1, RRID: CVCL_1415; MKN28, RRID: CVCL_1416; and MGC803, RRID: CVCL_5334) and mouse gastric cancer cell lines (MFC, RRID: CVCL_5J48) were purchased from the Shanghai Branch of the Chinese Academy of Sciences. AGS cells were cultured in DMEM/F-12 (Gibco) supplemented with 10% FBS (Israel Bioindustries) and 1% antibiotics (penicillin/streptomycin, Gibco) at 37°C with 5% CO_2. MKN1/MKN28/MFC cells were cultured in RPMI 1640 (Gibco) supplemented with 10% FBS and 1% antibiotics at 37°C with 5% CO₂. MGC803 cells were cultured in DMEM (Gibco) supplemented with 10% FBS and 1% antibiotics at 37°C with 5% CO₂. All cell lines were confirmed negative for mycoplasma contamination using PCR.

The F. nucleatum strain was purchased from Guangdong Institute of Microbiology and was obtained from ATCC (Fusobacterium nucleatum subsp. nucleatum, ATCC 25586). In brief, F. nucleatum was inoculated, cultured, and subcultured on a blood plate (Huankai Biotechnology); then, it was cultured in the AS-580 chamber of the anaerobic bacteria system at 37°C with 10% CO₂, 10% H₂, and 80% N₂. The E. coli strain, DH5α (Invitrogen), was propagated in Luria-Bertani (BD Biosciences, Difco) medium aerobically at 37°C.

We generally chose a six-well plate as the coculture system. Tumor cells needed to adhere to the wall and reach 70% to 90% density before coculture experiments. We collected F. nucleatum during the logarithmic growth period and adjusted the suspension concentration to 1 × 10⁹ CFU/mL for interaction experiments. MOI was used to examine the proportion of bacteria and cancer cells. A MOI of 100 is often used as a prerequisite for in vivo and in vitro experiments.

Gastric cancer cells and F. nucleatum were cocultivated in a medium containing ampicillin + gentamicin (200 + 200 μg/mL) or doxycycline (10 μg/mL) and cultured at 37°C overnight; then, they were centrifuged to obtain the supernatant and cell pellets. Subsequently, they were streaked on the blood plate and cultured in an anaerobic workstation. The combination of ampicillin and gentamicin was used to remove extracellular bacteria, and doxycycline was used to remove both extra- and intracellular bacteria.

Lymph nodes were dissected and separated from the popliteal fossa of mice and then immediately placed in a PBS-containing well plate for cleaning. Subsequently, they were drained on clean paper, immediately placed into a 2.5% glutaraldehyde electron microscope fixation solution, and fixed at room temperature for 30 minutes (attention was paid to protect the sample from light at this phase). Finally, the samples were transported on ice to a 4°C refrigerator for storage.

Cells were taken out of the incubator (the cell density generally did not exceed 70%), and the medium was discarded. The electron microscope solution at room temperature was added to the samples. The samples were fixed at room temperature and protected from light for 5 minutes. A cell scraper was used to gently scrape the cells in one direction, avoiding back-and-forth movements to prevent cell damage. The samples were centrifuged at 2,000 g for 5 minutes, and the cell pellets after centrifugation were about the size of a mung bean. The original electron microscope fixative was discarded, and a new one was added. The cell mass was blown off and resuspended. For storage, the samples were fixed for 30 minutes at room temperature in the dark and transferred to 4°C on ice.

A 24-well plate was prepared, in which gastric cancer cells were cocultured with F. nucleatum/E. coli for 24 hours. The cell supernatant was discarded, and the cells were washed two to three times with a 1% PBS solution. Endocytosis detection dye (21217, Protonex Green 500 Dextran, working solution prepared by mixing 20 µL stock solution and 980 µL 1% PBS solution) was added to the cells (5–10 µL/well). A fluorescence microscope or confocal microscope was used to take images, and the changes in gastric cancer cell membrane structure in living cells and cell slides were observed.

We cultured 50,000 cells in the top of the transwell chamber (for migration analysis; Corning Costar), with 300 μL of serum-free medium, whereas the chamber was outside the 24-well plate. Then, 600 µL of culture medium containing 10% FBS was added. After 16 to 72 hours of coculturing, the migrated or invaded cells were stained with crystal violet and counted under a microscope (Olympus).

Using the fluorescent carrier Cy3-labeled F. nucleatum 16S rRNA probe (5-/Cy3/CTTGTAGTTCCGCCTACCTC-3), the colonization of bacteria in human and mouse tissues was detected using fluorescence in situ hybridization (FISH). A nonspecific complement probe (5-/Cy3/CTTGTAGTTCCGCCTACCTC-3) served as a control for the hybridization method. A FISH kit (Gimma Gene) was used to detect bacterial localization and enrichment in tissues and cells. A fluorescent confocal microscope and an inverted/upright fluorescence microscope were used to collect images.

Total RNA was extracted using TRIzol reagent (Takara), followed by cDNA synthesis using the Master Mix cDNA Synthesis Kit (Accurate Biotechnology). Quantification of specific RNA was carried out using SYBR Green I (Accurate Biotechnology). The qPCR protocol was executed using a LightCycler 480 II (Roche). Fold change was determined employing the relative quantification method (2^−ΔΔCt). The qPCR primer sequences are detailed in Supplementary Table S2.

The fresh spleen and tumor tissues were extracted from 615 mice, grinding them into tissue fragments. The samples were then digested with trypsin to create a cell suspension, which was filtered through a 70-micron mesh to remove tissue debris. Red blood cells were eliminated using a lysis buffer, followed by cell counting. The samples were treated with 1:1,000 diluted Fixable Viability Stain prepared in 1× PBS and incubated at 4°C for 20 minutes in the dark. Subsequently, the cells were washed 1 to 2 times with staining buffer containing 1% FBS and resuspended. Fc receptor blocking solution (less than 2 μL per test) was added to each tube, followed by a 15-minute incubation in the dark. Antibodies were then mixed proportionally and added to the staining buffer; each antibody was quickly mixed upon addition to prevent accumulation, with a subsequent 15-minute incubation at room temperature. After washing and centrifugation (300 g for 5 minutes), the samples were analyzed using flow cytometry. The samples were divided into “blank,” “single staining,” and “mixed staining” groups (50 µL of the cell suspension was added to each tube, aliquoted into tubes, and incubated). The flow antibody was added according to the antibody instructions and incubated for 1 hour (protected from light). The samples were resuspended in 300 µL PBS and tested on the flow cytometer. Debris were removed by gating for forward and side scatter in the tumor cell suspension. Recognition of neutrophil (CD45⁺ CD11b⁺ Ly6G⁺), CD4⁺ T cell (CD45⁺ CD3⁺ CD4⁺), CD8⁺ T cell (CD45⁺ CD3⁺ CD8⁺), and NK cell (CD45⁺ CD3⁻ NK1.1⁺) was performed. Tumor cell suspensions were gated for forward and side scatter and depleted before recognition of neutrophil (CD45⁺ CD11b⁺ Ly6G⁺) debris. Neutrophils were stained with anti-CD45, CD11b, and anti-CD66b. The CytoFLEX/BD LSRFortessa flow cytometer (BD Biosciences) was used for analysis. FlowJo (RRID: SCR_008520) software was used for data analysis. The antibodies are detailed in Supplementary Table S3.

Protein extraction was conducted using Tissue Extraction Reagent (Thermo Fisher Scientific) supplemented with protease inhibitors (Roche) and PhosSTOP (Roche). Subsequently, the captured image was analyzed using Image Lab software (version 5.2.1, Bio-Rad). The expressed protein levels were normalized to GAPDH (Proteintech, Cat. # CL488-60004, RRID: AB_2919223). The antibodies are detailed in Supplementary Table S3.

The tissue sections were dewaxed, as described above. Subsequently, the sections were washed twice with PBS to remove residual ethanol. The slides were then stained with hematoxylin according to the IHC procedure. When the tap water turned blue, the slides were stained with the eosin solution (Soleibao). After 5 minutes, the staining process was stopped with a tap-water wash. The slides were then left in a fume hood to air-dry naturally. Afterward, the slides were sealed with neutral resin. The slide scanning system was used for archiving.

The IHC immunoreactive score of Remmele and Stegner was calculated as the sum of positive stained tumor cells and staining intensity. Briefly, the percentage of positive staining was scored as 1 (<10%, negative), 2 (11%–55%, sporadic), 3 (56%–80%, focal), or 4 (>81%, diffuse). Staining intensity was scored as 0 (no staining), 1 (weak staining), 2 (moderate staining), or 3 (strong staining). The median immunoreactive score of Remmele and Stegner was used to determine whether the indicated molecule was expressed strongly or weakly immunohistochemically. The slides were subsequently analyzed using a KF-PRO Slide Scanner (KFbio). Five fields from each slide were randomly selected for imaging, and the resulting images were processed using ImageJ (RRID: SCR_003070). The antibodies are detailed in Supplementary Table S3.

Fresh tissues were fixed in a 4% neutral buffered formalin solution and subsequently embedded in paraffin. Multiplex immunofluorescence (MIF) staining was conducted using a kit from Wuhan Servicebio Technology Co., Ltd. Deparaffinization and rehydration followed after melting tissue sections (8 µm) at 60°C for 1 hour. Microwave incubation was used for heat-mediated retrieval of antigens from citrate acid buffer (pH 6.0). After the addition of blocking buffer for 10 minutes, the sections were incubated with primary antibodies against CD8 (GB114196, RRID: AB_3064847), CD16 (GB113963, RRID: AB_3665881), Ly6G (GB11229, RRID: AB_2814689), CD3 (GB111337, RRID: AB_3665880), CD56 (GB112671, RRID: AB_3665882), CD49b (BioLegend, cat. #103501, RRID: AB_313024), NK1.1 (BioLegend, cat. #108714, RRID: AB_389364), and FOXP3 (GB112325, RRID: AB_2922969). The sections were incubated with horseradish peroxidase–conjugated secondary antibodies for 10 minutes at room temperature. Stained signals were further amplified using tyramide signal amplification reagents (G1231, G1223, G1232, and G1222, Wuhan Servicebio Technology Co., Ltd.), and the stained signals were further amplified. Finally, 4′,6-diamidino-2-phenylindole was used to stain the nuclei. A fluorescence imaging system was used to scan the stained slides (Olympus/BX63). The antibodies are detailed in Supplementary Table S3.

Cytokine expression levels, including IL17A, IL1β, IL8, G-CSF, and GM-CSF, were identified through the use of an ELISA kit in both the cell culture supernatant and mouse serum (COIBO).

Between-group differences in quantitative data were analyzed using the unpaired or paired Student t test, Mann–Whitney U test, or Dunnett t test, where appropriate. Among-group differences were analyzed using one-way ANOVA. The effects of clinical parameters were estimated by univariate or multivariate Cox proportional hazards regression analysis. Kaplan–Meier OS curves were generated and compared with the log-rank test. A Venn diagram was used to show flora selection, with F. nucleatum as the dominant flora in the primary gastric cancer and LN tissues simultaneously. Linear discriminant analysis of effect size and phylogenetic tree analysis were performed to screen significant bacteria in primary gastric cancers and LNs. All P values were two-tailed, and P < 0.5 was considered significant. Statistical analyses were conducted using IBM SPSS Statistics 23.0 (RRID: SCR_002865) and GraphPad Prism 9.0 software (RRID: SCR_002798).

The raw sequence data reported in this paper, including the RNA-seq and 16s rRNA-seq data, have been deposited in the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2021) in National Genomics Data Center (Nucleic Acids Res 2022), China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (Project number: PRJCA035816, GSA-Human: HRA010338 and HRA010335, GSA: CRA022889) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human and https://ngdc.cncb.ac.cn/gsa (20, 21). The raw data download can be requested according to the Genome Sequence Archive’s instructions, and all requests for data will be approved after review by its Data Access Committee. The raw data of single-cell RNA-seq (scRNA-seq) can be accessed in the GEO database under the accession number GSE264532. The bulk RNA-seq data and clinical information of The Cancer Genome Atlas (TCGA) cohorts and GEO cohorts were obtained from the Genomic Data Commons Data Portal (https://portal.gdc.cancer.gov/), Timer 2.0 (https://timer.cistrome.org), GEPIA 2.0 (http://gepia2.cancer-pku.cn/#analysis), and Kaplan–Meier Plotter (https://kmplot.com/analysis/). All data generated in this study are available from the corresponding author upon request.

To identify key microbes linked to gastric cancer progression, we analyzed the tissue microbiome of normal gastric mucosa, cancer, and LN of patients with gastric cancer using 16S rRNA amplicon sequencing, qRT-PCR, and FISH (cohort 1, n = 40). All examinations revealed that Fusobacterium, especially F. nucleatum, was highly enriched in gastric cancer lesions compared with normal tissue, particularly in patients with LN metastasis, suggesting a potential association between F. nucleatum and the engraftment process of gastric cancer cells (Fig. 1A–F; Supplementary Fig. S1A and S1B; Supplementary Tables S4–S6). Furthermore, both the 16S rRNA amplicon and FISH results indicated that an even higher F. nucleatum enrichment score was accumulated in the perigastric LNs (Fig. 1G; Supplementary Fig. S1C and S1D). Furthermore, patients with higher F. nucleatum enrichment scores had poorer PFS in cohort 1 and OS in cohort 2 (n = 93). In patients with F. nucleatum positivity, the median survival period was 31.5 months, whereas in patients with F. nucleatum negativity, it was 46 months (Supplementary Table S7). The CT/PET-CT imaging examination indicated that patients with gastric cancer with higher F. nucleatum enrichment scores were more likely to experience recurrence and metastasis after surgery. Moreover, the clinical data correlation analysis in cohort 2 revealed that patients with higher F. nucleatum enrichment scores exhibited increased T, N, and tumor–node–metastasis staging levels (Fig. 1H and I; Supplementary Fig. S1E; Supplementary Tables S8 and S9). These data indicated a close association between F. nucleatum and gastric cancer malignant progression.

To investigate F. nucleatum’s role in gastric cancer malignant progression, we first induced a primary gastric cancer model by orally gavaging H. pylori into C57BL/6 mice for 12 months, followed by gavaging with F. nucleatum to estimate its effect on gastric mucosa dysplasia and LNs metastasis. The F. nucleatum–infected mice displayed exacerbated progression of gastric carcinogenesis compared with those infected with E. coli. H&E staining demonstrated marked atypical hyperplasia, characterized by enlarged nuclei in glandular cells, varying sizes, disrupted polarity, disordered arrangement, and prominent nucleoli. IHC staining revealed higher Ki67 expression in F. nucleatum–infected mice (Fig. 2A–C; Supplementary Fig. S2A). Perigastric LNs in the F. nucleatum–infected mice exhibited significant enlargement, a color change to gray, and a hardened texture; further H&E staining revealed tumor cell infiltration (Fig. 2D and E). Survival times and OS were substantially shorter in F. nucleatum–infected mice (Fig. 2F).

To investigate the colonization ability of F. nucleatum in the gastric mucosal epithelium, gastric mucosal tissues from mice were dissected. Subsequently, microbial culturing was carried out under anaerobic conditions, and the bacterial strains were identified through mass spectrometry analysis. The results indicated the successful cultivation of F. nucleatum strains from the gastric mucosal tissues (Supplementary Fig. S2B and S2C). To clarify F. nucleatum’s impact on normal gastric mucosa, we administered F. nucleatum alone by gavage for 3 months. Compared with E. coli–infected mice, F. nucleatum–infected mice exhibited more severe manifestations of chronic and atrophic gastritis in the gastric mucosa (Supplementary Fig. S2D–S2G). These data illustrate that F. nucleatum colonization occurred in the gastric mucosa and contributed to gastric cancer occurrence and development.

To explore the effect of F. nucleatum on gastric cancer cell proliferation, subcutaneous tumorigenesis models were constructed from immunodeficient BALB/c nude and immunocompetent 615 mice by subcutaneous implantation of MFC cells. Tumor growth in 615 mice infected with F. nucleatum was significantly faster than in the PBS and E. coli groups, with no differences observed between the PBS and E. coli groups. Ki67 expression was notably high in F. nucleatum–infected tumors. Intriguingly, F. nucleatum–infected gastric cancer cells demonstrated accelerated growth in immunocompetent mice compared with control groups but not in immunodeficient mice (Fig. 2G–L; Supplementary Fig. S2H and S2I). Meanwhile, footpad–popliteal LN metastasis and tail vein–lung metastasis models were constructed to examine F. nucleatum’s impact on gastric cancer cell metastasis in vivo (Fig. 2M). The tail vein–pulmonary metastasis model illustrated that F. nucleatum facilitated the distant metastasis of gastric cancer cells (Fig. 2N; Supplementary Fig. S2J). Similarly, the footpad–popliteal LN metastasis model replicates lymphatic metastasis in gastric cancer to some extent, enhancing the precision of in vivo evaluations. The in vivo imaging systems, LN volume, and H&E staining results revealed significantly increased popliteal LN metastasis in the F. nucleatum–infected group compared with control groups. Furthermore, survival curve analysis indicated that F. nucleatum infection in gastric cancer notably decreases survival time (Fig. 2O–S). Additionally, the CCK8 assay and cell monoclonal experiment suggested that F. nucleatum infection does not substantially alter gastric cancer cells’ inherent malignant proliferation capacity (Supplementary Fig. S2K and S2L). Moreover, a tumor model was employed by orally gavaging F. nucleatum into immunocompetent mice, utilizing subcutaneous ectopic transplantation models in the axillary or inguinal regions. The findings demonstrated a considerably heightened malignant proliferative potential of tumors in the group that received oral gavage of F. nucleatum, exhibiting a notable increase in transplanted tumor volume compared with the control group (Supplementary Fig. S3A–S3E). These findings indicated that F. nucleatum might potentially alter the TIME, consequently influencing gastric cancer malignant progression.

To explore how F. nucleatum affects gastric cancer cells, we cocultured F. nucleatum with gastric cancer cells and detected its localization within these cells. Scanning electron microscopy showed that F. nucleatum adhered to the gastric cancer cells’ membrane surface (Fig. 3A). Further confocal fluorescence microscopy revealed F. nucleatum gathering around gastric cancer cells, with three-dimensional (3D) stereoscopic imaging revealing the same 3D plane between F. nucleatum and gastric cancer cells (Fig. 3B). Furthermore, upon infection with F. nucleatum, gastric cancer cells exhibited vesicular protrusions on their membranes (Fig. 3C). The findings indicated that F. nucleatum adhered to the membrane of gastric cancer cells and altered the membrane morphology. Transmission electron microscopy was employed to ascertain the presence of F. nucleatum in gastric cancer cells. Notably, F. nucleatum was found inside the cytoplasm, exhibiting clear structures with intact cell walls (Fig. 3D). When adhered to gastric cancer cell surfaces, F. nucleatum induced a finger-like structure around the cell membranes, similar to endocytosis (Fig. 3E).

To elucidate how F. nucleatum invades gastric cancer cells, fluorescent live cell imaging revealed the presence of numerous vesicle-like structures on the cell membrane surface, through which, F. nucleatum invaded the cells (Fig. 3F). Cell viability experiments showed that F. nucleatum–infected gastric cancer cells remained active (Fig. 3G). Antibiotic experiments revealed that F. nucleatum survives in the presence of cell-impermeable antibiotics (ampicillin and gentamicin) but not in the presence of cell-permeable antibiotics (doxycycline hydrochloride; Fig. 3H). Furthermore, treatment of gastric cancer cells with endocytosis inhibitors [ethylisopropylamiloride (EIPA), 50 µmol/L; genistein, 50 µmol/L; and chlorpromazine, 15 µmol/L] significantly decreased the presence of F. nucleatum in cells (Fig. 3I). To investigate whether intracellular F. nucleatum within gastric cancers affects the malignant progression of gastric cancer, we treated MFC cells with endocytosis inhibitors, followed by coculturing with F. nucleatum. After discarding the supernatant, the cell suspension was injected into the footpads of immunocompetent 615 mice. The results demonstrated that F. nucleatum promotes the malignant progression of gastric cancer, yet inhibiting endocytosis significantly reduced the tumor-promoting effect of F. nucleatum, suggesting that intracellular F. nucleatum mediates the malignant progression of gastric cancer (Fig. 3J–L). To elucidate the impact of intracellular F. nucleatum on the LN metastasis of gastric cancer cells, a mouse popliteal LN metastasis model was employed to track the activity trajectory of F. nucleatum. Intriguingly, FISH and transmission electron microscopy revealed F. nucleatum within the LNs (Supplementary Fig. S4A and S4B). Furthermore, bacterial culture confirmed that live F. nucleatum could translocate from the primary tumor to the LNs through intracellular colonization of gastric cancer cells (Fig. 3M). These results indicate that F. nucleatum invaded gastric cancer cells through endocytosis, establishing intracellular colonization and mediating gastric cancer malignant progression.

To further explore the impact of F. nucleatum on the tumor microenvironment in gastric cancer, the subcutaneous tumorigenesis models of immunocompetent mice were constructed for scRNA-seq. Analysis revealed 17 subclusters and 8 major cell clusters (Fig. 4A and B; Supplementary Fig. S5A and S5B). To observe the temporal heterogeneity of immune cell differentiation during gastric cancer progression, tumor tissues were collected from two periods (time 1 and time 2) and subjected to sequencing analysis. Time 1 corresponds to a 10-day tumor growth period, whereas time 2 represents a 19-day growth period, capturing the acute and chronic stages of tumor growth affected by F. nucleatum.

Remarkably, F. nucleatum group mice had more than three times the higher proportion of neutrophils in their subcutaneous tumors than in the control group. Furthermore, the proportion of neutrophils in late tumors increased distinctly after prolonged F. nucleatum infection. However, the population of other antitumor immune cells, such as B/plasma cells, decreased significantly (Fig. 4C; Supplementary Fig. S5C). Moreover, IHC, MIF, and flow cytometry results in tumor tissue of mice with F. nucleatum infection similarly confirmed this phenomenon (Fig. 4D and E). Additionally, the interaction network between neutrophils and other immune cells in tumor tissue was remarkably enhanced in the F. nucleatum infection group (Fig. 4F). These findings indicate that neutrophils are implicated in the malignant progression of gastric cancer.

Tumor immune responses are increasingly linked to neutrophils, often with poor clinical outcomes (22, 23). Correlation analysis of scRNA-seq revealed nine distinct TAN subclusters within tumor-infiltrating neutrophils, indicating their heterogeneity. Analysis of neutrophil proportions in different subgroups showed that those of Sell⁺ neutrophils, Isg15⁺ neutrophils, G5b and G5c staging neutrophils, and TAN-0 and TAN-3 were substantially declined. Conversely, those of Gbe⁺ neutrophils, Wfdc21⁺ neutrophils, G3 and G5a staging neutrophils, and TAN-1, TAN-2, TAN-4, TAN-5, TAN-6, and TAN-7 were considerably increased in tumors with F. nucleatum infection (Fig. 4G; Supplementary Fig. S5D–S5H). Combining the results of IHC and MIF, we found that intratumoral F. nucleatum infection reshaped the immune microenvironment of gastric cancer and exerted an inhibitory effect on antitumor immunity. Additionally, IHC results also showed significant recruitment of neutrophils in mouse tumor tissues and popliteal LNs after F. nucleatum infection, suggesting a crucial role of TANs in the malignant progression of gastric cancer (Supplementary Fig. S6A–S6D). These data indicate that F. nucleatum–infected gastric cancer cells induced neutrophil recruitment and promoted TANs to protumoral subtype differentiation.

To elucidate the impact of TANs on TIME in response to intratumoral F. nucleatum infection, we integrated a mouse model in which CD4⁺ T cells, CD8⁺ T cells, NK cells, and neutrophils were depleted. The results indicate that depletion of CD4⁺ T cells and NK cells significantly increased tumor size in mice infected with F. nucleatum, whereas there was no significant change in tumor size when CD8⁺ T cells were depleted. Expanding on our prior research discoveries, we hypothesize that within the tumor microenvironment infected with F. nucleatum, the antitumor efficacy facilitated by CD8⁺ T cells is attenuated. Upon depleting neutrophils along with CD4⁺ T cells, CD8⁺ T cells, and NK cells simultaneously, we noted a notable decrease in the tumor’s immunosuppressive effect following neutrophil clearance as opposed to scenarios in which neutrophils remained intact (Supplementary Figs. S7A–S7G and S8A–S8D). This confirms the critical role of neutrophils in reshaping the F. nucleatum–mediated immunosuppressive microenvironment.

Interestingly, we observed a significant upregulation of PD-L1 expression in TANs following F. nucleatum infection based on the single-cell sequencing gene set immune checkpoint expression analysis (Fig. 4H and I; Supplementary Fig. S9A). Upon clarification of the differentiation status of PD-L1⁺ TANs in gastric cancer tissues with intratumoral F. nucleatum infection, a subsequent identification was conducted through flow cytometry analysis. The results demonstrated an elevated proportion of PD-L1⁺ TANs within the F. nucleatum–infected mouse tumor tissue (Fig. 4J). Through Monocle2 (24) pseudotime analysis, it was determined that the F. nucleatum infection group primarily resided in the stage of tumor infiltration. Moreover, GO analysis and KEGG pathway enrichment analysis revealed that the F. nucleatum–infected group was associated with PD-L1 expression and the PD-1 checkpoint pathway in cancer, tumor angiogenesis, cell metastasis, immunosuppression, and inhibition of apoptosis (Fig. 4K–N). Furthermore, Western blotting results indicated that in vitro coculturing with F. nucleatum promoted elevated levels of PD-L1 expression in human neutrophils (Supplementary Fig. S9B and S9C). Additionally, IHC experiments demonstrated a positive correlation between the enrichment level of F. nucleatum and PD-L1 expression in both human and murine tissues (Supplementary Fig. S9D and S9E). These results suggested that intratumoral F. nucleatum infection promoted abnormal differentiation of PD-L1⁺ TANs.

To validate the effect of F. nucleatum on gastric cancer cells and neutrophils, the “F. nucleatum–gastric cancer cells–neutrophils” coculture model was constructed. The supernatant from the coculture of F. nucleatum with gastric cancer cells was centrifuged and placed in the outer chamber, and then induced differentiation of neutrophils dHL60 and aNB4 was added to the inner chamber. Notably, the F. nucleatum group had higher neutrophil migrations than the control group (Fig. 5A and B; Supplementary Fig. S9F), indicating that supernatant from the F. nucleatum–infected gastric cancer cells induced neutrophil recruitment.

A neutrophil depletion model was constructed in immunocompetent mice to investigate the relationship between F. nucleatum–mediated gastric cancer malignant progression and neutrophil recruitment. The tumor and LN volume, tumor weight, H&E staining, and LN metastasis rate results revealed that the impact of F. nucleatum on tumor proliferation and metastasis substantially diminished upon removal of neutrophils (Fig. 5C–J; Supplementary Fig. S9G). These data indicate that neutrophil recruitment plays a crucial role in the process by which F. nucleatum promotes gastric cancer malignant progression. MIF unexpectedly revealed a substantial increase in CD8⁺T cells in regions enriched with TANs in F. nucleatum–infected mice. Following neutrophil depletion, there was a notable decrease in the CD8⁺ T-cell proportion, indicating that the suppressed tumor immune response influenced by F. nucleatum correlated with the recruitment of neutrophils (Fig. 5K and L). To clarify whether intracellular F. nucleatum affects neutrophil recruitment, we also constructed a coculture model involving doxycycline and endocytosis inhibitors. Interestingly, these results revealed a notable decrease in the number of recruited neutrophils when doxycycline and endocytosis inhibitors were used, suggesting further that intracellular F. nucleatum promotes neutrophil recruitment mediated by tumor cells (Supplementary Fig. S10A–S10D).

To explore the molecular mechanism of F. nucleatum–infected gastric cancer cell–mediated TAN recruitment, further scRNA-seq analysis was performed to identify downstream signaling pathways. GO and KEGG enrichment analysis indicated that the promotion of neutrophil activation by F. nucleatum infection was enriched in the IL17 signaling pathway (Supplementary Fig. S11A and S11B). Subsequent gene set activity analysis demonstrated higher scores in protumoral TANs in the IL17 signaling pathway (Fig. 6A). Furthermore, IL17RA (the receptor of IL17) expression was correlated with neutrophil polarization and development, showing higher expression in the F. nucleatum group than in the control group (Supplementary Fig. S11C). Pseudotime analysis revealed that high expression of IL17RA prolonged the lifespan of TANs and promoted the differentiation of the neutrophil-promoting tumor subtype (Supplementary Fig. S11D). These findings indicated a strong correlation between the activation of the IL17 pathway and neutrophil recruitment in gastric cancer.

To examine the molecular mechanisms of F. nucleatum infection in gastric cancer cells, we cocultured MKN1 cells with live (L) or heat-inactivated (D) F. nucleatum and obtained RNA for transcriptome sequencing. Gene expression comparison revealed distinct patterns in the PBS group versus the F.n(L) group (712 genes), the PBS group versus the F.n(D) group (14 genes), and the F.n(L) group versus the F.n(D) group (805 genes; Supplementary Tables S10–S12). Compared with the F.n(D) group, both KEGG pathway enrichment analysis and gene set enrichment analysis revealed significant activation of the IL17 signaling pathway in the F.n(L) group (Fig. 6B–D; Supplementary Fig. S11E–S11H). Combined scRNA-seq of tumor tissues and cellular transcriptome sequencing revealed that F. nucleatum activated the IL17 pathway within gastric cancer cells, thereby promoting neutrophil recruitment. Analysis of TCGA data indicated that IL17A was highly expressed in human gastric cancer tissues and was associated with poor prognosis (Supplementary Fig. S11I and S11J). IHC staining revealed increased expression of IL17A in human and mouse gastric cancer tissue enriched with F. nucleatum. Furthermore, IHC staining demonstrated significantly enhanced expression of IL17A in the gastric mucosa of the primary gastric cancer model with the F.n group. Higher expression of IL17A was significantly associated with an adverse prognosis in gastric cancer (Fig. 6E–G). ELISA and Western blot analysis also demonstrated significantly increased IL17A protein levels in F. nucleatum–infected gastric cancer cells (Fig. 6H–J). Moreover, in the F. nucleatum infection group, cytokines related to the IL17 pathway were secreted more frequently, and neutrophil recruitment of cytokines, such as IL1β, IL8, and G-CSF, was significantly enhanced in vitro (Supplementary Fig. S11K). Additionally, the endocytosis inhibitor and permeabilizing antibiotics (doxycycline) markedly decreased the secretion of IL17A by gastric cancer cells. Compared with the group without endocytosis inhibitors and doxycycline, we observed a significant decrease in the level of IL17A in the cell culture supernatant when endocytosis inhibitors and doxycycline were added during the coculture of F. nucleatum and gastric cancers (Supplementary Fig. S12A–S12D). This indicates that intratumoral F. nucleatum infection promoted the expression and secretion of IL17A within tumor cells.

To further elucidate how F. nucleatum influences the expression of PD-L1 in TANs, we initially performed a correlation analysis utilizing the TCGA database. Our examination of the TCGA database revealed a positive association between the levels of IL17A and PD-L1 expression within human gastric cancer tissues. Furthermore, in vitro cell culture experiments with the addition of IL17A demonstrated a significant upregulation of PD-L1 expression in humanized neutrophils (Supplementary Fig. S12E and S12F). Subsequent flow cytometry analysis revealed a significant decrease in the proportion of PD-L1⁺Ly6G⁺ cells in tumor tissues from F. nucleatum–infected mice following treatment with IL17A neutralizing antibodies (Supplementary Fig. S12G). Recent studies have shown that the innate immune system plays a crucial role in initiating immune responses. Specifically, activation of the cyclic GMP-AMP synthase (cGAS)–stimulator of interferon genes (STING) pathway can generate type I interferon, inducing both innate and adaptive immune responses, thus playing an essential role in tumor immunotherapy (25–30). Our findings demonstrated that the enhanced expression of PD-L1 on neutrophils induced by F. nucleatum was suppressed upon treatment with the cGAS–STING pathway inhibitor (RU.521). Furthermore, the upregulation of neutrophil PD-L1 expression triggered by IL17A could be reversed by the cGAS–STING pathway inhibitor (RU.521; Supplementary Fig. S12H and S12I). These findings suggest that F. nucleatum might influence the aberrant expression of PD-L1 in TANs through the cGAS–STING pathway regulated by IL17A.

An anti–IL17A treatment model was constructed to validate the function of IL17A in TAN recruitment in gastric cancer. Notably, treatment with anti–IL17A remarkably reduced the volume of subcutaneous tumors in the F. nucleatum–infected group. IHC staining further indicated a notable decrease in the expression of Ki67 and Ly6G following anti–IL17A treatment (Fig. 6K–O). To elucidate the roles of IL17A and neutrophils in F. nucleatum–mediated tumor promotion, we developed a combined model involving neutrophil depletion and anti–IL17A treatment. Remarkably, both anti–IL17 treatment and neutrophil depletion separately blocked F. nucleatum–mediated tumor progression. Furthermore, treatment using anti–IL17A combined with neutrophil depletion was more effective in F. nucleatum–infected mice compared with noninfected mice. Conversely, the combined anti–IL17A treatment and neutrophil depletion showed no significant difference in mice without F. nucleatum infection (Fig. 6P–R). These outcomes illuminate the potential of anti–IL17A treatment in inhibiting F. nucleatum–mediated tumor progression, emphasizing the heightened effectiveness of anti–IL17A treatment in combination with neutrophil depletion, specifically within F. nucleatum–enriched gastric cancer.

NF-κB is a pivotal regulator of genes responsible for encoding inflammatory mediators. Recent studies have revealed that F. nucleatum upregulates the canonical NF-κB pathway, augmenting the expression of proinflammatory cytokines, thereby triggering colonic inflammation and carcinogenesis (31, 32). The gene set enrichment analysis revealed that NF-κB signaling pathways were enriched in F. nucleatum–infected gastric cancer cells (Fig. 7A), suggesting the involvement of the NF-κB pathway in neutrophil activation and the malignant progression of gastric cancer. In this study, NF-κB and IL17 signaling pathways were thought to be closely linked in tumor cell–neutrophil interactions.

To determine the interaction between the NF-κB pathway and the IL17 pathway, we conducted transcriptomic differential gene analysis coupled with qRT-PCR assays. There was a considerable upregulation of RelB, TNFATP3, CXCL3, CXCL8, CSF2, and CCL20 genes in both the NF-κB pathway and the IL17 pathway (Fig. 7B; Supplementary Fig. S13A). RelB was widely recognized as a key protein in the noncanonical NF-κB pathway. The Western blot analysis revealed a remarkable increase in p52 and RelB protein levels among NF-κB subunits in gastric cancer cells (Fig. 7C–E). Remarkably, qRT-PCR and Western blot results indicate substantial upregulation of RelB in the F.n(L) group but no significant change in the F.n(D) group. IHC staining demonstrated heightened expression of RelB in human and mouse gastric cancer tissue rich in F. nucleatum. Additionally, RelB was highly expressed in gastric cancer tissues and correlated with poor prognosis of gastric cancer according to the TCGA cohort. Analysis of immune cell infiltration in the TCGA database indicated a positive correlation between RelB expression and neutrophil infiltration in gastric cancer (Fig. 7F–I; Supplementary Fig. S13B–S13F). Furthermore, F. nucleatum infection promoted the phosphorylation and nuclear translocation of RelB proteins (Fig. 7J and K; Supplementary Fig. S13G–S13I).Additionally, the transwell experiment results demonstrated a significant reduction in the chemotactic ability of neutrophils after knocking down RelB in gastric cancer cells (Fig. 7L and M). These results indicated that RelB might be an essential target gene of the NF-κB/IL17 signaling pathway.

Upon further RelB knockdown in gastric cancer cells, secretion levels of IL17A proteins were substantially decreased and significantly inhibited following F. nucleatum infection (Fig. 7N and O). When IL17A was added to the culture medium, RelB protein levels were significantly increased. Immunofluorescence colocalization experiments revealed a significant upregulation of RelB, IL17A, and the neutrophil-associated protein Ly6G in mice infected with F. nucleatum tumor tissues. Moreover, these proteins exhibited a spatially correlated distribution (Fig. 7P; Supplementary Fig. S13J and S13K). Our data indicated that the cross-talk interaction between activation of RelB in NF-κB and IL17 pathways significantly enhances IL17A secretion and further induces neutrophil recruitment and differentiation.

In Fig. 5K and L, we observed a significant enrichment of CD8⁺T cells in the area recruited by neutrophils in tumor tissues following F. nucleatum infection. Furthermore, after depleting neutrophils, CD4⁺ T cells, CD8⁺ T cells, and NK cells, we observed a significant reduction in the tumor’s immunosuppressive influence following neutrophil clearance compared with situations in which neutrophils were preserved. This finding underscores the pivotal role of neutrophils in modifying the F. nucleatum–mediated immunosuppressive microenvironment.

To identify the interaction between neutrophils and other immune cells, we applied CellPhoneDB analysis (19). Upon F. nucleatum infection, there was a noticeable enhancement in the interaction between PD-L1⁺ TANs and PD-1⁺ T cells (Fig. 8A). A substantial upregulation of PD-1 expression in tumor tissue was noted following F. nucleatum infection (Supplementary Fig. S14A). Examination of different CD8⁺ T-cell types showed a significant increase in exhausted CD8⁺ T (CD8⁺ Tex) cells after F. nucleatum infection, suggesting the recruitment and differentiation of TANs further contribute to CD8⁺ T-cell exhaustion (Fig. 8B and C; Supplementary Fig. S14B). Moreover, flow cytometry and Western blotting analysis revealed a significant upregulation of PD-L1 expression in human neutrophils and TANs from mice upon infection with F. nucleatum (Fig. 8D). Based on pseudotime analysis, CD8⁺ T cells differentiate in the direction of PD-1^high in response to F. nucleatum infection, and gene set immune checkpoint expression analysis and KEGG analysis revealed significant enrichment of the PD-L1 pathway in CD8⁺ T cells of the F. nucleatum group (Supplementary Fig. S14C–S14J). Subsequent analysis using flow cytometry unveiled a notable increase in the percentage of PD-1⁺ CD8⁺ T cells in tumor tissues following F. nucleatum infection. Conversely, upon neutrophil depletion, the proportion of PD-1⁺ CD8⁺ T cells conspicuously decreased, suggesting a close association between the recruitment of neutrophils and CD8⁺ T cells as well as their exhaustion within the tumor microenvironment (Fig. 8D; Supplementary Fig. S15A and S15B). Moreover, as described in Fig. 7, after simultaneous neutrophil elimination, the tumor volume in mice noticeably decreased compared with when solely clearing CD8⁺ T cells, indicating a partial reduction in CD8⁺ T-cell exhaustion upon neutrophil clearance, consequently unleashing the antitumor activity of CD8⁺ T cells. These results indicate that F. nucleatum contributes to CD8⁺ T-cell exhaustion by enhancing the expression of PD-L1 in TANs.

Subsequently, an anti–PD-L1 treatment model in mice was established to evaluate the efficacy of PD-L1 immunotherapy in gastric cancer with F. nucleatum infection. Tumor volume considerably decreased in the F. nucleatum group, indicating a superior response to anti–PD-L1 treatment compared with the control group (Fig. 8E–G). By creating a mouse model with neutrophil depletion and anti–PD-L1 treatment, we found that the impact of the anti–PD-L1 treatment lessened when neutrophils were depleted (Fig. 8H–J). This confirmed a strong correlation between anti–PD-L1 treatment in F. nucleatum–infected mice and neutrophil recruitment. Further, MIF staining revealed a significant decrease in the proportion of CD8⁺ T cells and TANs in F. nucleatum–infected tumor tissue with anti–PD-L1 treatment (Fig. 8K and L; Supplementary Fig. S15C). Meanwhile, immunofluorescence staining indicated a notable increase in CD66b⁺ neutrophils in gastric cancer tissue with high F. nucleatum density. Correspondingly, areas rich in CD66b showed an increase in CD8⁺ T cells (Fig. 8M and N; Supplementary Fig. S15D and S15E). In situ hybridization and qRT-PCR results revealed a high abundance of F. nucleatum in gastric cancer tissues of patients responding to anti–PD-L1 treatment (Fig. 8O and P; Supplementary Fig. S15F). PET-CT showed a significant reduction in metastatic lesions in patients with gastric cancer with high F. nucleatum levels following anti–PD-L1 immunotherapy (Supplementary Fig. S15G).

In recent years, an increasing number of studies have revealed a close relationship between microbiota and the malignant progression of tumors, particularly within the gut microbiota (33, 34). Current research gradually confirms that intratumoral bacteria play crucial roles in the malignant progression and treatment outcomes of tumors (35). Recently, although extensive research has focused on the association between F. nucleatum and malignant progression in cancers, notably colorectal cancer, there has been a notable scarcity of investigations about the interplay of F. nucleatum with the TIME. Previous research has predominantly focused on the microbiota’s impact on tumor cells directly (36–38). Nonetheless, the advancement of tumors toward malignancy typically entails the modulation of the microenvironment (39). Furthermore, recent studies emphasize the intratumoral microbiota’s significant role in promoting immune escape and influencing immunotherapy efficacy in cancer (40–44). Our research elucidates that intracellular F. nucleatum promotes the recruitment and differentiation of TANs, reshaping the immunosuppressive microenvironment and leading to CD8⁺ T-cell exhaustion. Mechanistically, F. nucleatum invades gastric cancer cells and activates the IL17/NF-κB/RelB signaling pathway, thereby promoting the recruitment of TANs and their differentiation into a protumoral subtype, which further reshapes the TIME. Moreover, intratumoral F. nucleatum contributes to the exhaustion of CD8⁺ T cells by inducing high expression of PD-L1 in TANs, subsequently facilitating immune evasion in gastric cancer and augmenting the effectiveness of PD-L1 immune therapy. This study elucidated a novel mechanism by which intratumoral microbiota influences immune escape in gastric cancer and mediates the response to immunotherapy, thereby providing new theoretical foundations for the treatment of patients with gastric cancer.

The precise involvement of intracellular microbiota in gastric cancer malignant progression and immune microenvironment remodeling remains largely ambiguous. Prior studies indicate an inverse correlation between F. nucleatum presence in colorectal cancer tissues and the abundance of tumor-infiltrating lymphocytes, affecting antitumor immune response suppression (45, 46). Fu and colleagues (47) confirmed that intracellular bacteria carried by circulating breast cancer cells enhance fluid shear resistance by reorganizing the actin cytoskeleton during tumor cell metastasis and colonization. Our findings indicated that intratumoral F. nucleatum colonization occurred in gastric cancer cells and contributed to gastric cancer malignant progression. Furthermore, scRNA-seq showed that F. nucleatum–infected gastric cancer cells significantly enhanced neutrophil recruitment in the tumor microenvironment. Neutrophils have dual roles in immune responses. Research into their biological activities related to antitumor and tumor-promoting functions has gained increased attention (7). Neutrophil recruitment is primarily initiated by inflammatory mediators such as histamine, cysteine leukotrienes, and cytokines at the endothelial surface when sentinel leukocytes encounter pathogens in tissues (48–50). Jaillon and colleagues and Scapini and colleagues (51, 52) found that neutrophils are a marker of acute inflammation associated with activated, localized, and regulated immune responses. Moreover, neutrophils can suppress T-cell proliferation and hamper their activity (53). In our study, TANs in F. nucleatum–infected tumor tissues mainly tended toward protumoral subtype differentiation. Intratumoral F. nucleatum induced neutrophil-mediated CD8⁺ T-cell exhaustion to further promote gastric cancer immune evasion. Our findings support that F. nucleatum promotes gastric cancer development by altering the TIME through neutrophil recruitment.

Neutrophils promote cancer development both directly and indirectly (54). For instance, TANs remain alive while releasing neutrophil extracellular traps, which promotes the proliferation of cancer cells (55). Kong and colleagues (56) discovered that F. nucleatum induces the formation of neutrophil extracellular traps by activating the Toll-like receptor 4–reactive oxygen species pathway and the NOD1/NOD2 signaling pathways in neutrophils, thereby facilitating the progression of colorectal cancer. Similarly, neutrophils induce the transition from gastric mucosal epithelial cells to mesenchymal cells, promoting the migration and invasion of gastric cancer cells (57). Our results identified a close relationship between IL17 signaling and TAN recruitment, and the data showed that intracellular F. nucleatum activated the NF-κB pathway to enhance IL17A secretion within gastric cancer cells. Moreover, there is cross-talk between NF-κB and IL17 signaling pathways, further promoting neutrophil recruitment through a cascade amplification mechanism. Our data indicated a novel mechanism by which F. nucleatum promotes IL17A–mediated neutrophil recruitment and protumoral TAN differentiation to promote gastric cancer malignant progression. Remarkably, aberrant PD-L1 expression has been noted in these protumoral TANs, suggesting a potential link between F. nucleatum–mediated neutrophil recruitment and PD-L1 therapy.

Anti–PD-1/anti–PD-L1 treatment is progressively reshaping the overarching treatment strategy for patients with gastric cancers (58, 59). However, gastric cancer represents a highly heterogeneous malignant tumor, wherein most tumor tissues demonstrate an immunosuppressive state, significantly affecting the effectiveness of immune checkpoint inhibitor therapy (60–62). Recent research has revealed that the influence of microbiota on tumor immunotherapy is multifaceted. Jiang and colleagues (41) discovered that a high abundance of F. nucleatum limits the infiltration and effector function of CD8⁺ T cells within the tumor through its metabolite succinate, thereby promoting resistance to anti–PD-1 monoclonal antibody therapy. Conversely, Gao and colleagues (42) demonstrated that F. nucleatum may heighten tumor sensitivity to PD-L1 blockade therapy by activating the STING signaling pathway. Our scRNA-seq data revealed a significant interaction between PD-1 in the CD8⁺ T-cell subpopulation and PD-L1 in protumoral TANs. Furthermore, the F. nucleatum–infected gastric cancer showed a more favorable response to anti–PD-L1 treatment in vivo, attributed to the recruitment of PD-L1⁺ TANs. Our findings demonstrate that anti–PD-1/PD-L1 treatment can be enhanced by F. nucleatum infection, which mainly contributes to TAN recruitment and differentiation. Therefore, this provides a theoretical basis and treatment strategy for the clinical application of anti–PD-1/anti–PD-L1 immunotherapy in patients with gastric cancer.

Overall, F. nucleatum mediates neutrophil recruitment to alter the TIME in gastric cancer. Mechanistically, intratumoral F. nucleatum invading gastric cancer cells activates NF-κB/IL17 signaling, promoting TAN recruitment and protumoral TAN differentiation in TIME. This process exhausts CD8⁺ T cells, promotes immune evasion, and enhances the effectiveness of anti–PD-1/anti–PD-L1 immunotherapy in gastric cancer. These findings reveal novel mechanisms through which F. nucleatum affects gastric cancer malignant progression, providing new insights for developing effective treatment strategies.

T. Zhang reports grants from the Guangdong Provincial Natural Science Foundation, the Guangdong Provincial Health Commission, and the Fundamental Research Funds for the Central Universities, Sun Yat-sen University during the conduct of the study. Y. Qian reports grants from the Guangzhou Science and Technology Programme and the Guangdong Provincial Natural Science Foundation project during the conduct of the study. J. Chen reports grants from the Guangdong Provincial Natural Science Foundation project and the National Natural Science Foundation of China during the conduct of the study. S. Cai reports grants from the National Natural Science Foundation of China and the Guangdong Provincial Natural Science Foundation during the conduct of the study. No disclosures were reported by the other authors.

T. Zhang: Conceptualization, data curation, funding acquisition, validation, methodology, writing–original draft. Y. Li: Conceptualization, data curation, funding acquisition, writing–original draft. E. Zhai: Conceptualization, data curation, methodology, writing–original draft. R. Zhao: Software, formal analysis, funding acquisition, writing–review and editing. Y. Qian: Data curation, software, formal analysis, funding acquisition, writing–review and editing. Z. Huang: Validation, methodology. Y. Liu: Validation, methodology. Z. Zhao: Validation, methodology. X. Xu: Software, visualization. J. Liu: Visualization, methodology. Z. Li: Formal analysis, methodology. Z. Liang: Formal analysis, visualization. R. Wei: Software, formal analysis. L. Ye: Formal analysis, visualization. J. Ma: Formal analysis, visualization. Q. Wu: Conceptualization, resources, funding acquisition, project administration, writing–review and editing. J. Chen: Conceptualization, resources, supervision, funding acquisition, writing–review and editing. S. Cai: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

This study was supported by the National Natural Science Foundation of China grant 82173239 (S. Cai), 82472904 (J. Chen), and 82100626 (R. Zhao); Guangdong Provincial Natural Science Foundation project grant 2023A1515111152 (T. Zhang), 2021A1515110759 (Y. Qian), 2023A1515011187 (Y. Qian), 2022A1515011534 (S. Cai), and 2024A1515010739 (J. Chen); Key-Area Research and Development Program of Guangdong Province grant 2022B1111070006 (Q. Wu); GDAS’ Project of Science and Technology Development 2022GDASZH-2022010101 (Q. Wu); Guangdong Provincial Health Commission grant A2024052 (T. Zhang); The Fundamental Research Funds for the Central Universities, Sun Yat-sen University 24qnpy329 (T. Zhang); and Guangzhou Nansha Technology Plan Project (2024MS006, 2023MS003; J. Ma). This study was also supported by the Guangdong Province Excellent Youth Team Project (2024B1515040029), led by Professor Junchao Cai at Sun Yat-sen University. Additionally, we extend our thanks to Novel Bio for their support in single-cell sequencing.