Inflammation Mediated by Gut Microbiome Alterations Promotes Lung Cancer Development and an Immunosuppressed Tumor Microenvironment

Lung adenocarcinoma, the most common histologic subtype of non–small cell lung cancer (NSCLC), is a leading cause of cancer-related mortality globally, representing a substantial public health burden (1). Despite advances in targeted and immune-based therapies, the 5-year survival rate of patients with lung adenocarcinoma remains low, accentuating the urgent need to elucidate external mechanisms beyond the traditionally explored territories of driver genomic alterations and tumor-intrinsic properties that contribute to its progression. Understanding these factors could unveil novel, actionable targets for therapeutic intervention.

The gut microbiome, when in a state of dysbiosis, defined as an alteration in the composition and functionality of the microbiota that is a deviation from the homeostatic state in the healthy host, has been implicated in the pathogenesis and progression of various diseases, including cancer (2). Previous studies have highlighted mechanisms by which the gut microbiome can directly and indirectly modulate tumor progression. For instance, metabolites like gallic acid produced by microbes in the gut have been shown to alter the functional consequences of mutant P53, influencing a switch from tumor-suppressive to oncogenic activities in WNT-driven intestinal cancers (3). Furthermore, impaired epithelial gut barrier in colitis and primary sclerosing cholangitis mouse models can facilitate bacterial translocation and cholangiocarcinoma progression (4). Additionally, distinct clades of Fusobacterium nucleatum have been shown to exacerbate colorectal cancer and metastasis through multiple mechanisms, including attachment to epithelial cells, modulation of the immune microenvironment, and activation of Toll-like receptor 4 signaling (5–8).

Although most of these aforementioned studies, among others, have probed the role of the gut microbiome in the progression of cancers within the gastrointestinal tract, its influence beyond these confines, such as in lung cancer, remains largely unknown. To date, studies have established correlative relationships between specific gut microbiome alterations and lung cancer, highlighting significant microbiome composition changes in patients with lung cancer compared with healthy individuals (9). Distinct microbial signatures have been found to influence responses to immune checkpoint blockade (ICB; refs. 10, 11). Despite these notable associations, the mechanisms that underlie the causal relationships between gut microbiome changes and lung cancer development, progression, and response to therapy remain very poorly defined and warrant further investigation.

Our earlier work showed that mice deficient in G protein–coupled receptor 5A (Gprc5a^−/−), upon exposure to nicotine-specific nitrosamine ketone (NNK), a tobacco carcinogen, develop preneoplastic lesions and lung adenocarcinoma, recapitulating the disease progression in humans (12). We also observed in these mice an upregulation of lipocalin-2 (LCN2), an antimicrobial and acute-phase protein (13, 14), during the early stages of lung oncogenesis (15). Moreover, we showed that knockout of Lcn2 (Gprc5a^−/−;Lcn2^−/−) reduced microbial diversity while enhancing tumor-promoting inflammation (15, 16). These findings suggest that LCN2 mediates a protective mechanism against protumorigenic inflammation and lung adenocarcinoma progression. However, the broader relationship between gut microbial imbalance, such as that incurred by the loss of Lcn2, and lung cancer development remains poorly understood. To address these knowledge voids, we sought to probe how alterations in the gut microbiome influence lung adenocarcinoma progression. Using models of fecal microbiota transfer (FMT) from Lcn2-deficient mice, fecal microbial community profiling, and comprehensive immunologic analyses, we investigated how the gut microbiome modulates the lung tumor microenvironment (TME) and progression. We elucidated a mechanism by which the gut microbiome impacts lung adenocarcinoma pathogenesis as well as response to immunotherapy, thereby providing insights into the systemic effects of microbial alterations and thus potential targets for therapeutic intervention.

All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee at the University of Texas MD Anderson Cancer Center. The mice were housed and maintained in a pathogen-free animal facility at MD Anderson Cancer Center. C57BL/6;129sv Gprc5a^−/− (G) and C57BL/6;129sv Gprc5a^−/−;Lcn2^−/− (GL) mice were generated as previously described (15, 17). Gprc5a^−/−;Il6^−/− (GI) mice were generated by crossing Gprc5a^−/− and Il6^−/− (JAX strain code: 002650) mice. Wild-type 129sv mice were acquired from Charles River Laboratories (Strain code: 476).

Mouse lung adenocarcinoma cell lines used in this study were LKR13 cells (mutant Kras-driven), which we obtained from Dr. Jonathan M. Kurie at the University of Texas MD Anderson Cancer Center in 2021 (18), and MDA-F471 cells (Gprc5a deficient and Kras mutant), which we previously generated and described (12). Mycoplasma testing was performed using the IMPACT III PCR Profile by IDEXX BioAnalytics with all results confirmed negative for contamination. The cells were cultured in DMEM (Gibco, Catalog No. 11995-065) supplemented with 10% FBS (Gibco, Catalog No. 16000-044), 2 mmol/L L-glutamine (Sigma-Aldrich, Catalog No. G7513), and 1% penicillin–streptomycin (Gibco, Catalog No. 15140-122) and maintained at 37°C in a humidified atmosphere with 5% CO₂. Cells were used between passages 5 and 15, with passage occurring every 2 to 3 days to prevent overgrowth. Mycoplasma testing was conducted regularly throughout the study to ensure the cell lines remained uncontaminated. Although the cells have not been reauthenticated within the past year, they were fully tested at the time of acquisition and before use in this study.

Feces were collected from male and female donor mice that were housed in multiple cages. Collected feces were pooled and promptly stored at −80°C in sterile containers. Prior to treatment, fecal pellets were resuspended in anaerobic/reduced PBS supplemented with 0.5 g/L-cysteine (MilliporeSigma, Catalog No. 1028380025). The suspension underwent vigorous grinding and vortexing for 3 minutes. Subsequently, samples were centrifuged at 800 × g for 5 minutes at room temperature, and the final volume of the fecal supernatant was adjusted to achieve a concentration of 100 mg/mL, based on the weight of fecal pellets prior to resuspension.

For experiments with transplant of mouse tumor cells, 8-week-old mice, matched for age with a 1:1 male-to-female ratio (8–10 mice per group), underwent a 2-week regimen of broad-spectrum antibiotics containing 1 g/L ampicillin (Pfizer, NDC 00409-3720-01), 2 g/L streptomycin (XGen Pharmaceuticals, NDC 39822-0706-02), 0.5 g/L vancomycin (Firvanq, Azurity Pharmaceuticals, NDC 65628-205-10), and 1 g/L metronidazole (Pfizer, NDC 0025-1821-50) and flavored with 20 g/L Kool-Aid (Kraft Heinz). Subsequently, 100 μL of the fecal supernatant derived from G or GL mice was administered orally three times a week for 2 weeks via oral gavage prior to tumor inoculation. Two weeks following FMT, mice were subcutaneously injected with mouse Mycoplasma-free lung adenocarcinoma cell lines: either 2 × 10⁶ million LKR13 cells or 5 × 10⁶ million MDA-F471 cells (see “Cell lines”). These cells were embedded in Matrigel (Corning, Catalog No. 354262) at a 1:1 volume ratio, with a total injection volume of 100 μL. Tumor growth was monitored every 3 days by measurement using a caliper; tumor volume was calculated using the formula: tumor volume = ½ × length × width². FMT continued twice per week until endpoint. Throughout the experimental period, mice bearing tumors were regularly monitored for signs of reduced feeding, weight loss, or dehydration to ensure their well-being. Fecal samples were collected at four distinct timepoints: T1 (baseline before antibiotics treatment), T2 (after antibiotics treatment and before FMT), T3 (after 2 weeks of FMT and before tumor inoculation), and T4 (at the experimental endpoint). At endpoint, sera, tumors, and colons were collected.

For tobacco-mediated carcinogenesis experiments, 8-week-old Gprc5a^−/− mice were divided into two groups. Each group consisted of 10 to 12 mice per FMT donor genotype (G < G or G < GL), matched for age with a 1:1 male-to-female ratio. Mice were subjected to intraperitoneal injections of 75 mg/kg body weight of NNK (TragetMol, Catalog No. T20533) three times per week for 8 weeks; the NNK was dissolved in sterile normal saline. Following the 8-week NNK treatment, mice received a 2-week regimen of pharmaceutical-grade broad-spectrum antibiotics (1 g/L ampicillin, 2 g/L streptomycin, 0.5 g/L vancomycin, 1 g/L metronidazole and flavored with 20 g/L Kool-Aid), followed by FMT from G or GL for 3 months, three times per week, and then once weekly thereafter, until they reached the 7-month post-NNK endpoint. At endpoint, mice were humanely euthanized, and various samples were collected, including serum, bronchoalveolar lavage fluid, lungs, and fecal pellets.

Following sacrifice, lungs were inflated with 10% neutral-buffered formalin (Sigma-Aldrich, Catalog No. HT501128) and subsequently excised for histopathologic evaluation. Lungs were fixed in 10% neutral-buffered formalin for 24 hours before being processed and embedded in paraffin. Formalin-fixed, paraffin-embedded sections were cut at a thickness of 5 µm using a microtome (Leica RM2235, Leica Microsystems, Inc.). Hematoxylin and eosin staining was performed using an automated stainer (Leica ST5020, Leica Biosystems) following the manufacturer’s standard protocol. All reagents used during the automated process, unless stated, were sourced from Leica Microsystems, Inc. The staining process included deparaffinization, rehydration, staining with Harris, counterstaining with Eosin Y (Thermo Fisher Scientific, Catalog No. 6766007), dehydration, and clearing. The stained sections were digitally scanned using the Aperio ScanScope Turbo slide scanner (Leica Microsystems Inc.) at 200× magnification. Scanned images were visualized using ImageScope software (Leica Microsystems, Inc., RRID: SCR_014311). Total lung area, as well as the area of each tumor mass and malignant tumor lesion, was delineated and measured by an experienced pathologist utilizing the Aperio Image Toolbox (Leica Microsystems Inc.). Enumeration of lung tumors from each mouse group was conducted through histopathological analysis of hematoxylin and eosin–stained slides, employing previously established criteria for characterizing mouse lung lesions (19). Tumor volume (measured in mm³) was calculated by multiplying the area of malignant tumor lesions (measured in mm²) by the thickness of tissue sections on each slide. Tumor burden was subsequently defined as the percentage of tumor volume relative to the entire lung volume.

Alistipes finegoldii (AF, MC166976) and Alistipes onderdonkii (AO, PCO2303) were obtained from Dr. Robert Jenq at the University of Texas MD Anderson Cancer Center and propagated anaerobically on Columbia agar (Thermo Fisher Scientific, Catalog No. R452954). AF was originally retrieved from a patient with advanced stage metastatic melanoma prior to treatment, whereas AO was recovered from a patient with colitis secondary to immunotherapy treatment for clear cell carcinoma of the kidney. Animals were gavaged with 10⁸ colony-forming units (CFU) AF, AO, or anaerobic/reduced PBS supplemented with 0.5 g/L-cysteine (MilliporeSigma, Catalog No. 1028380025), used as the vector control.

Syngeneic mice bearing MDA-F471 tumors were randomized according to FMT donor groups and were intraperitoneally injected with either anti–PD-1 (Bio X cell Catalog No. BE0273) or an anti-trinitrophenol IgG2a isotype control (InVivoMAb Catalog No. BE0089), once the tumor volume reached 200 mm³. This injection was administered at a dosage of 200 μg per mouse twice per week until reaching the experimental endpoint.

Total DNA from mouse tumors and fecal pellets was isolated using the DNeasy Blood and Tissue kit (Qiagen Catalog No. 69504) and the DNeasy PowerSoil Pro Kit (Qiagen Catalog No. 47014), respectively, following the manufacturer’s instructions. Following extraction, DNA was quantitated using a Qubit fluorometric assay (Thermo Fisher Scientific) and normalized to a concentration of 15 ng/μL as appropriate.

Protocols employed for 16S gene profiling were adapted from the NIH Human Microbiome Project and the Earth Microbiome Project (www.earthmicrobiome.org). Briefly, 2 μL of template DNA used for 16S gene profiling was isolated from mouse tumors and fecal pellet (see “Total DNA isolation from mouse tumor tissues and fecal pellets”). The v4 region of the 16Sv4 rDNA was amplified by PCR using the Thermo DreamTaq DNA Polymerase (Catalog No. EP0701) with the forward primer 515F (GTGCCAGCMGCCGCGGTAA) and the reverse primer 806R (GGACTACHVGGGTWTCTAAT). The primers were engineered to integrate Nextera XT adapters (Illumina, Catalog No. 20015960) for sequencing compatibility, and the reverse primer included a unique single-index barcode to facilitate multiplexing. These barcodes were custom-synthesized by Integrated DNA Technologies. The sequencing was performed using the MiSeq Reagent Kit v2 (Illumina, Catalog No. MS-102-2003) on the Illumina MiSeq platform at the Alkek Center for Metagenomics and Microbiome Research at Baylor College of Medicine, generating 2 × 250 base pair read pairs. The target per-sample yield was approximately 25,000 read pairs.

Preprocessing of the sequencing data started with an initial quality filtering of demultiplexed read pairs using bbduk.sh from BBMap version 38.82. This step was crucial for eliminating Illumina adapters, PhiX reads, and any reads that did not meet the quality threshold of a Phred score above 15 or a length exceeding 100 bp posttrimming. Subsequently, quality-controlled reads were merged, applying amplicon-specific parameters to ensure accurate assembly of the sequences. The parameters set a maximum error rate of 0.05 and length thresholds, with a minimum of 252 bp and a maximum of 254 bp, congruent with the V4 amplicon characteristics.

Following the merging process, reads were consolidated into a single FASTA file, followed by operational taxonomic unit picking using the UPARSE algorithm. This step facilitated the recovery of abundances by aligning the demultiplexed reads against the representative sequences, resulting in a Feature table in biom format while simultaneously discarding any chimeric reads. The representative sequences underwent taxonomic classification against an optimized subset of the SILVA database (RRID: SCR_008452), reflecting only sequences from the V4 region of the 16Sv4 rRNA gene, utilizing the usearch_global function of usearch70 at a 97% similarity threshold. The phylogenetic context was constructed by aligning centroid sequences with MAFFT and generating a phylogenetic tree through FastTree (RRID: SCR_015501). The biom file, enriched with phylogenetic information, summarized the number of reads per sample and when merged with the overall read statistics, produced a final summary document encapsulating both read statistics and taxonomy information.

Data analysis was executed using the Agile Toolkit for Incisive Microbial Analysis, a comprehensive web application combining publicly available R packages to import sample data and elucidate trends in taxa abundance, α-diversity, and β-diversity in relation to the sample metadata. For statistical analyses, categorical variables were assessed through the Mann–Whitney U and Kruskal–Wallis tests for dichotomous or polytomous groups, respectively, whereas correlations with continuous variables were determined using the linear regression models in R’s lm function. Community composition variances were quantified using the vegan:adonis function to approximate permutational multivariate analysis of variance (PERMANOVA) P values. To ensure the reliability of the results, all P values were adjusted for multiple comparisons using the Benjamini and Hochberg approach to control the FDR. Agile Toolkit for Incisive Microbial Analysis’s suite of tools and visualizations is accessible online at https://atima.research.bcm.edu. Linear discriminant analysis (LDA) effect size (RRID: SCR_014609; ref. 20) was used to identify taxa that were enriched in the different treatment groups (G < G and G < GL). An LDA score of 2.0 was used with a 0.05 α-value for the Kruskal–Wallis all-against-all test.

DNA extracted from fecal samples collected from the study animals (see “Total DNA isolation from mouse tumor tissues and fecal pellets”) was used for qPCR analysis to detect AF and AO. A measure of 7.5 ng/µL of DNA per sample was run in triplicate. The qPCR reactions were performed using the PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, Catalog No. A25742), with specific primers curated from the literature (21). For AF, the forward (AfinF; GTA​CTA​ATT​CCC​CAT​AAC​ATT​CGA​G) and reverse (AfinR; CTA​ATA​CAA​CGC​ATG​CCC​ATC​TT) primers were utilized, yielding an amplicon of 83 base pairs. In the case of AO, the forward (AondF; AGA​GGC​ATC​TCT​CCG​GGT​T) and the reverse (AondR; AGT​CTG​GTC​CGT​GTC​TCA​GT) primers were employed to produce a 152-base pair fragment. qPCR was carried out on a QuantStudio 3 Real-Time PCR System (Applied Biosystems, Catalog No. A28137) with an initial denaturation at 95°C for 2 minutes, followed by 45 cycles of 95°C for 10 seconds, annealing at species-specific temperatures (63°C for AF and 61°C for AO) for 15 seconds, and extension at 72°C for 45 seconds. After cycling, a melt curve analysis was performed, beginning with 95°C for 5 seconds, followed by 65°C for 1 minute, and increasing to 98°C at a rate of 0.11°C per second. The amplified products were verified by agarose gel electrophoresis for qualitative assessment based on size.

MDA-F471 tumor tissues freshly collected from syngeneic mice were dissociated into single-cell suspensions using the Tumor Dissociation Kit (Catalog No. 130-096-730, Miltenyi Biotec). Colon tissues were dissected, promptly immersed in ice-cold PBS, and kept on ice for subsequent processing. The colon sections were carefully opened, and any feces present were removed. Colonic tissues were then cut into pieces measuring 1 to 2 cm in length. Lamina propria cells were isolated using the Miltenyi Biotec Lamina Propria Dissociation Kit (Catalog No. 130-097-410), following the guidelines provided by the manufacturer. Tissues were processed in a gentleMACS Octo Dissociator with Heaters (Miltenyi Biotec), operating under the program settings recommended by the manufacturer for optimal tissue dissociation while preserving cell viability. The resulting cell suspensions were filtered through a 70-μm cell strainer to remove debris and aggregates. Subsequently, red blood cells were lysed using ACK Lysis Buffer (Thermo Fisher Scientific, Catalog No. A1049201) for 5 minutes at room temperature. Lysed suspensions were then neutralized and washed, and the cells were resuspended in ice-cold FBS (Catalog No. 16140071, Thermo Fisher Scientific) to protect cell integrity. Cell suspensions were then passed through a 40-μm Flowmi cell strainer (Millipore, Catalog No. BAH136800040) to ensure homogeneity for downstream processing in single-cell RNA sequencing (scRNA-seq) and multiparameter flow cytometry.

Single-cell suspensions were freshly processed and labeled with SYTOX Green (Thermo Fisher Scientific, Catalog No. S7020) to stain dead cells for exclusion. Live cells were sorted using a FACSAria I flow cytometer (BD Biosciences), collecting only SYTOX Green–negative cells. Sorted cells were then washed with PBS containing 2% FBS (Catalog No. 10270106, Thermo Fisher Scientific) to remove any extracellular RNA. Cell viability and number were assessed via trypan blue (Catalog No. T8154, Sigma-Aldrich) exclusion using a hemocytometer. Viable cells were encapsulated into gel beads-in-emulsion (GEM) using the Chromium Next GEM Single Cell 5′ Kit v2 (10X Genomics, Catalog No. PN-1000263) and Chromium Next GEM Chip G Single Cell Kit (10X Genomics, Catalog No. PN-1000120), following the manufacturer’s protocol. Barcoding, reverse transcription, and cDNA amplification were conducted using the Chromium Single Cell 5′ Gel Bead & Library Kit v2 (10X Genomics, Catalog No. PN-1000153). Amplified cDNA was subjected to a quality check using the Bioanalyzer High Sensitivity DNA Analysis Kit (Agilent, Catalog No. 5067-4626). The cDNA libraries were fragmented, size-selected, and underwent end-repair, A-tailing, adapter ligation, and PCR amplification with unique indexing using the Chromium i7 Multiplex Kit (10X Genomics, Catalog No. PN-1000084). The final library was validated for quality and quantified using a Qubit 4 Fluorometer with the dsDNA HS Assay Kit (Thermo Fisher Scientific, Catalog No. Q32851). The libraries were normalized based on molarity, pooled, and sequenced on an NovaSeq 6000 system (Illumina, Catalog No. 20012866) following the manufacturer’s instructions for high-throughput sequencing.

Raw sequencing reads were demultiplexed, and initial quality control was performed using the Cell Ranger software suite (10X Genomics, version 3.1). Read counts were aligned to the mouse reference genome (GRCm38) and quantified to generate feature-barcode matrices. The Seurat R package (v3.1.5; RRID: SCR_007322) was utilized for the preliminary analysis, which included filtering of barcodes based on unique molecular identifier (UMI) counts to identify and retain genuine single-cell transcriptomes.

A stringent quality control pipeline was implemented to ensure high data fidelity in the same manner as our previous reports (22). To remove damaged or dying cells, those with low gene counts (less than 200 genes per cell) and high mitochondrial content (over 15% of total reads) were excluded from downstream analysis. Library size normalization was applied to the remaining dataset using the NormalizeData function in Seurat, followed by the identification and regression of unwanted sources of variation such as cell-cycle effects.

DoubletFinder (version 3.0) was used alongside Seurat to predict and exclude doublets from the dataset. Parameters for doublet identification were determined by assessing the distribution of read counts and the coexpression of genes typically not coexpressed in the same cell. This iterative process involved refining the parameters and reassessing the dataset until the removal of probable doublets was maximized.

Dimensionality reduction was first performed via principal component analysis based on highly variable genes identified using the FindVariableFeatures function in Seurat. The RunUMAP function was used to visualize the data in two dimensions, and the FindClusters function was utilized with a resolution parameter set between 0.3 and 1 to identify distinct cellular clusters. For each identified cluster, the FindAllMarkers function was used to detect differentially expressed genes, allowing for the characterization of cell types based on the expression of known cell type–specific marker genes. Cell-type identification was undertaken by analyzing differentially expressed genes and reviewing canonical markers.

Clustering results and marker gene expression were visualized using the Uniform Manifold Approximation and Projection dimensionality reduction technique. The FeaturePlot, VlnPlot, and DotPlot functions from Seurat were used to visualize gene expression patterns across clusters. Additionally, the ggplot2 package (RRID: SCR_014601) was used for enhanced visualization and the patchwork package for complex plot arrangements.

The composition of specific cell types within each cluster was quantified by computing the percentage of each cell type relative to the total cell count within the cluster. Expression levels of genes encoding key inflammatory markers (Il1a, Il1b, Il6, Nlrp3, and Tlr5) and signaling pathways (inflammasomes, TNFR1, TNFR2, NOD1 and 2, cytokine, Toll, TGFβ, IL17, IL6, and IL1R pathways) in myeloid cells were assessed by computing average expression values of the gene sets from the Molecular Signatures Database for each cell using the AddModuleScore function in Seurat. Average expression values were scaled to normalize the data using the ScaleData function, which adjusted the module scores by standardizing across the dataset. The scaled module scores were then used to create a heatmap, comparing the average expression levels between the G < G and G < GL groups.

Single-cell suspensions were plated in 96-well V-bottom plates at a concentration ranging from 1 × 10⁵ to 3 × 10⁶ cells per sample. These samples were then incubated with a solution containing PBS supplemented with 80× Zombie UV fixable viability dye (1000×; BioLegend Catalog No.: 423108), 1× monensin (1000×; BioLegend, Catalog No. 420701), and 1× brefeldin A (1000×; BioLegend, Catalog No. 420601) for 30 minutes on ice. Following incubation, cells were blocked with rat anti–mouse CD16/32 antibody (diluted 1:100; BioLegend, Catalog No. 156604) at a concentration of 0.25 μg per 1 million cells for 15 minutes at 4°C. Subsequently, cells were incubated with fluorescently conjugated antibodies targeting surface markers for 30 minutes at 4°C. Following this, cells were fixed and permeabilized using commercial reagents (Intracellular Staining Permeabilization Wash Buffer and Fixation Buffer; BioLegend, Catalog No.: 421002; 420801) and stained with antibodies against intracellular proteins for an additional 30 minutes at 4°C. Antibodies used are listed in Supplementary Table S1. Flow cytometry analysis was conducted using Fortessa flow cytometers (BD Biosciences), with data acquisition performed using FACSDiva 8.0.1 (BD Biosciences; RRID: SCR_001456). Flow cytometry files (.fcs) were analyzed using FlowJo 10.10.0 software (BD Biosciences; RRID: SCR_008520) employing supervised gating strategies.

Proteins from the lamina propria were extracted utilizing the Bio-Plex Cell Lysis Kit (Catalog No.: 171304011), following the manufacturer’s instructions. Proteins were quantitated using Pierce BCA Protein Assay Kits (Catalog No.: 23225) and adjusted to a concentration of 400 mg/mL following normalization.

ELISAs were conducted on serum samples and proteins extracted from the lamina propria of MDA-F471 tumor–bearing mice using the Bio-Plex Pro Mouse Chemokine Panel 31-Plex Kit (Catalog No.: 12009159), following the manufacturer’s instructions. The panel included the following chemokines: BCA-1/CXCL13; IL4; MIP-1α/CCL3; CTACK/CCL27; IL6; MIP-1β/CCL4; ENA-78/CXCL5; IL10; MIP-3α/CCL20; eotaxin/CCL11; IL16; RANTES/CCL5; eotaxin-2/CCL24; IP-10/CXCL10; MIP-3β/CCL19; fractalkine/CX3CL1; I-TAC/CXCL11; SCYB16/CXCL16; GM-CSF; KC/CXCL1; SDF-1α/CXCL12; I-309/CCL1; MCP-1/CCL2; TARC/CCL17; IFNγ; MCP-3/CCL7; TNFα; IL1β; MCP-5/CCL12; IL2; MDC/CCL22.

LCN2 and IL6 ELISAs were conducted on mice sera using the Invitrogen Lipocalin-2 (LCN2) Mouse ELISA Kit (Catalog No. EMLCN2) and Invitrogen Mouse IL-6 ELISA Kit (Catalog No. KMC0061), respectively, following the guidelines provided by the manufacturer.

L929 cells were cultured in DMEM (Gibco, Thermo Fisher Scientific, Catalog No. 11995-065) supplemented with 10% FBS (Gibco, Thermo Fisher Scientific, Catalog No. 16000-044) and 1% penicillin–streptomycin (Gibco, Thermo Fisher Scientific, Catalog No. 15140-122) at 37°C with 5% CO2. Cells were maintained at 80% to 90% confluency and passaged every 3 to 4 days. To generate the conditioned medium, L929 cells were seeded at a density of 5 × 10⁶ cells in T175 flasks and allowed to grow for 7 to 10 days. The supernatant was collected, filtered through a 0.22-µm filter (Millipore, Catalog No. SLGP033RS), and stored at −20°C until use. This conditioned medium was used as a source of macrophage colony-stimulating factor in bone marrow–derived macrophages (BMDM) culture.

BMDMs were generated as previously described (23). Briefly, bone marrow cells were harvested from the femurs and tibias of 8-week-old naïve mice, followed by erythrocyte removal via hypotonic lysis. These cells were cultured for 6 days in 100-mm Petri dishes at an initial seeding density of 4 × 10⁵ cells per dish. The culture medium consisted of DMEM (Gibco, Thermo Fisher Scientific, Catalog No. 11995-065), supplemented with 10% FBS (Gibco, Thermo Fisher Scientific, Catalog No. 16000-044), 30% L929 cell–conditioned medium (see “L929-conditioned medium preparation”), and 1% penicillin–streptomycin (Gibco, Thermo Fisher Scientific, Catalog No. 15140-122). The fecal microbiota concentrates, prepared as described earlier (see “FMT preparation and administration”), were diluted to a working concentration of 10 mg/mL. Subsequently, BMDMs were exposed to either G or GL microbiota concentrates for 12 hours in DMEM supplemented with 10% FBS. Following incubation, the medium was centrifuged at 3,000 × g. The supernatant was subsequently filtered through a 0.22-µm pore-size filter to eliminate residual cellular debris. The filtered supernatant was then used in subsequent coculture experiments with bone marrow cells.

Bone marrow cells were obtained from G mice and distributed into six-well plates at a density of 7 × 10⁵ cells per well, with each well containing 2 mL of supernatant from the coculture of BMDMs with G or GL microbiota concentrates. After a 24-hour incubation period, cells were harvested, and the proportion of polymorphonuclear myeloid-derived suppressor cells (PMN-MDSC; CD45⁺CD11b⁺Ly6G⁺I-A/I-E⁻) was assessed using flow cytometry, following the previously described methods.

We analyzed whole-genome shotgun (WGS) metagenomic sequencing data of pretherapy stool samples obtained from 21 patients with operable lung adenocarcinoma, selected from the NEOSTAR trial (NCT03158129; ref. 11). The cohort comprised patients receiving neoadjuvant nivolumab alone or in combination with ipilimumab. Informed consent, inclusion/exclusion criteria, and sample collection protocols were defined by the NEOSTAR trial (11), which was approved by the University of Texas MD Anderson Cancer Center’s Institutional Review Board. From the selected cohort, seven patients achieved major pathologic response (MPR), whereas 14 were classified as no MPR. Stool samples used in this analysis were collected before treatment initiation and stored at −80°C until processing for WGS metagenomic sequencing.

Metagenomic WGS was performed on bacterial genomic DNA extracted from patient stool samples collected before the initiation of treatment. Genomic DNA extraction from stool samples was performed using the QIAamp Fast DNA Stool Mini Kit (Qiagen), with a modification to the standard protocol that included an additional bead-beating lysis step. A 3.2-mm steel bead, approximately 150 mg of zirconium beads, and the lysis buffer were added to each tube containing stool samples. The samples were then subjected to bead-beating for a total of 8 minutes at a speed of 3,800 rpm (BioSpec) to lyse the bacteria for DNA extraction. Fecal microbial DNA was prepared into individual libraries, constructed from each sample using the DNA prep kit (Illumina), pooled, and loaded onto the NovaSeq 600 platform (Illumina) for sequencing. Sequencing was performed using the 2 × 150 bp paired-end read protocol, following the manufacturer’s instructions.

MetaPhlAn4 (24) was used to profile fecal bacterial taxonomies to the species level from the resulting shotgun sequencing libraries. MetaPhlAn4 taxonomic counts at the species level were processed with MaAsLin2 analysis (version 1.16.0; ref. 25) employing TMM normalization to identify species with differential abundance between patients with lung adenocarcinoma using MPR as the fixed effect (reference: No MPR). A minimum abundance threshold of 1 and standardization with no data transformation was used. The relative abundance of microbial taxa in the samples is displayed using centered log-ratio transformation.

Data were analyzed and visualized using GraphPad Prism software (version 10.1.1; RRID: SCR_002798). Normality of data distribution was assessed using the Shapiro–Wilk test. For comparisons between two groups, an unpaired t test was applied if data were normally distributed. For nonnormally distributed data, the Mann–Whitney U test was utilized. When assessing differences among more than two groups, ANOVA or Kruskal–Wallis test was used, followed by post hoc tests for multiple comparisons where applicable. A P value of less than 0.05 was considered statistically significant.

Mouse scRNA-seq data generated in this study have been deposited into the NCBI’s Gene Expression Omnibus with accession number GSE271872. Data for 16S rRNA gene sequencing from our mouse studies have been deposited into the NCBI’s Sequence Read Archive under BioProject accession number PRJNA1136021. Alistipes data pertaining to patients selected from the NEOSTAR trial (NCT03158129) are available upon request by contacting Dr. Tina Cascone ([email protected]) and the corresponding author Dr. Humam Kadara. They have not been deposited in a public repository because of patient confidentiality concerns and the terms of the trial’s data sharing agreements. All other data generated in the study are available in the article and its Supplementary Materials or from the corresponding author upon reasonable request.

We recently showed that gut microbiome changes are closely associated with lung adenocarcinoma development in a human-relevant, tobacco-induced mouse model (G; ref. 12). Knockout of the antimicrobial protein Lcn2 in these mice (GL) further reduced microbial diversity in the gut (16) and augmented tumor-initiating inflammation (15). To better understand how gut microbiome alterations impinge on lung adenocarcinoma development, we functionally investigated the role of gut microbiome changes in lung adenocarcinoma development using FMT. MDA-F471 lung adenocarcinoma cells were transplanted into syngeneic G mice, in which native gut microbiota had been eradicated with broad-spectrum antibiotics and then subjected to FMT to reestablish their gut flora, either with microbiota from GL mice (G < GL) or with control G microbiota (G < G; Fig. 1A; Supplementary Fig. S1A). G < GL mice exhibited significantly increased tumor growth relative to littermates with FMT from G mice, G < G (Fig. 1B). These effects were recapitulated in an independent syngeneic model (Kras^G12D LKR13 cells in wild-type mice; Supplementary Fig. S1B). Of note, performing the reverse FMT, that is, transferring G microbiota into GL hosts, did not significantly alter tumor burden (Supplementary Fig. S1C).

We conducted 16Sv4 rRNA gene sequencing (16S) analysis of fecal pellets in recipient mice from the G < G and G < GL FMT models. We found significant differences in gut β-diversity between syngeneic G < G and G < GL mice (Fig. 1C). G < GL fecal pellets were characterized by increased relative abundance of bacteria such as Alistipes species (spp.), which have been shown to promote the growth of other solid tumors such as colon cancer (26, 27), as well as Blautia and Lactobacillus, whereas the relative abundance of Ruminococcus genus, known to contain species associated with a favorable response to immunotherapy (28), was diminished (Fig. 1D; Supplementary Table S2). LDA effect size (20) highlighted Alistipes spp. as a distinguishing driver of the G < GL group, in contrast to Ruminococcus for G < G mice (Fig. 1E; Supplementary Table S3). Furthermore, 16S analysis on tumors showed no significant differences in β-diversity (Supplementary Fig. S1D), bacterial load (Supplementary Fig. S1E), or composition between the two groups (Supplementary Table S4), suggesting that the increased tumor burden in extraintestinal organs following gut microbiome modulation in our model was not influenced by differential intratumoral bacterial composition.

To better understand the influence of gut microbiome alterations on lung cancer development in its native environment, we explored the effects of FMT on tumorigenesis in our tobacco carcinogen–exposed mouse model, which develops tumors following NNK exposure (Fig. 1F). Tobacco carcinogen–exposed G < GL mice exhibited increased lung tumor burden and multiplicity compared with similarly exposed G < G littermates (Fig. 1G). Concordant with our findings in our syngeneic model, we found a significant increase in the relative abundance of Alistipes (26) in tobacco carcinogen–exposed G < GL mice compared with G < G littermates (Supplementary Fig. S1F). We also found a significant increase in the relative abundance of Odoribacter (29) in tobacco carcinogen–exposed G < GL mice compared with G < G mice (Supplementary Fig. S1F). Moreover, a positive correlation was found between the relative abundance of Alistipes spp. and tumor burden (Fig. 1H), further emphasizing the critical role of the gut microbiome’s composition in lung adenocarcinoma progression.

These findings collectively underscore significant differences in the gut microbiome landscape due to Lcn2 deficiency, culminating in a microbial ecosystem conducive to lung adenocarcinoma development.

Having shown that the loss of Lcn2 leads to microbial compositional differences linked to lung adenocarcinoma progression, we sought to understand how these alterations affect the local milieu of the gut. We found substantial gut remodeling, as evidenced by a significant reduction in colon length, a surrogate marker for inflammation (Fig. 2A and B; ref. 30). We comprehensively probed the colonic lamina propria of syngeneic G < G and G < GL mice (n = 2 each) by scRNA-seq. Unsupervised clustering of 39,556 quality-controlled cells revealed cell clusters representing major cellular lineages, namely, stromal, epithelial, myeloid, T and NK, B, and plasma cells (Fig. 2C; Supplementary Fig. S2A), with a consistent representation of each major cluster across all experimental groups (Supplementary Fig. S2B). To refine our understanding of these cellular landscapes, we further dissected each major lineage through subsequent clustering, revealing subclusters indicative of heterogeneous functions and states within the broader categories (Supplementary Fig. S2C–S2F; Supplementary Table S5). We found a higher relative abundance of enterocytes in the G < GL group, coupled with a decrease in mucin-producing goblet cells and stem cells in G < GL mice compared with G < G littermates (Fig. 2D; Supplementary Fig. S2C). Two populations of crypt-bottom fibroblasts, essential for maintaining intestinal homeostasis and function (31), differed significantly between the two FMT treatment groups (Supplementary Fig. S2D). Specifically, Wnt⁺ crypt-bottom fibroblasts, known to support stem cell proliferation and differentiation (31), were more prevalent in the G < G group (Fig. 2D). Conversely, G < GL littermates displayed a higher abundance of Cd55⁺C3⁺ crypt-bottom fibroblasts, which are marked by the expression of key components of the complement system cascade (Fig. 2D). Next, we focused on delineating immune changes in the colonic lamina propria in response to microbial alterations in the gut. In the lymphoid group, we observed increased abundance of mucosal-associated invariant T cells in the G < G group (Fig. 2D; Supplementary Fig. S2E). Concurrently, lymphoid tissue inducer–like cells, marked by elevated Rorc and Il22 expression (Supplementary Fig. S2E; Supplementary Table S5), were found in greater abundance in G < GL mice (Fig. 2D). In the myeloid compartment, comprising distinct subclusters with varied functions and states (Supplementary Fig. S2F; Supplementary Table S5), we observed a higher relative abundance of Ccr2⁺ M2 macrophages in G < GL mice (Fig. 2D). In addition, myeloid subsets from G < GL animals showed elevated activity scores for several key pathways, including inflammasomes, TNFR1 and TNFR2, Toll, and NOD 1 and 2, and various cytokine modules, such as TGFβ, IL17, IL6, and IL1R (Supplementary Fig. S3A). We also observed elevated expression levels of Nlrp3 and Tlr5, as well as the proinflammatory cytokines Il6, Il1a, and Il1b in myeloid subsets of G < GL mice compared with G < G littermates (Fig. 2E). To corroborate our scRNA-seq findings on colonic immune responses, we performed multiparameter flow cytometry of colonic lamina propria (Supplementary Fig. S3B and S3C; Supplementary Table S1). In G < GL mice, we observed increased infiltration of proinflammatory immune cell subsets, including IL6⁺ monocytes, type 1 conventional dendritic cells (cDC1s), cDC2s, and pro-IL1β⁺ and PDL1⁺ dendritic cells (DC), as well as IL17⁺ CD4⁺ T cells, suggesting a local inflammatory response and aligning with our scRNA-seq data (Fig. 2F).

Given the absence of significant differences in bacterial load and composition in tumor tissues between the two groups (Supplementary Fig. S1D and S1E; Supplementary Table S4), we sought to focus on the gut and hypothesized that systemic propagation of inflammation could serve as a mechanism for communication along the gut–lung axis, thereby influencing lung pathology. To explore this, we quantified serum levels of proinflammatory cytokines using multiplex assays. We found that local inflammation in the gut of G < GL mice was propagated systemically, as evidenced by increased serum levels of IL6 compared with G < G littermates (Fig. 2G). Albeit not reaching statistical significance, we also found increased serum levels of pro-IL1β in G < GL mice (Supplementary Fig. S3D). This pattern was mirrored in our tobacco carcinogen–induced tumor model (Fig. 1F), in which gut microbiome alterations led to elevated serum levels of IL6 and pro-IL1β in the G < GL group (Supplementary Fig. S3E). Furthermore, in our syngeneic model, additional cytokines such as IL2, TNFα, granulocyte-macrophage colony-stimulating factor (GM-CSF), CX3CL1, and CCL22 were increased in the sera of G < GL mice compared with G < G animals (Fig. 2H), with IL2, CXCL10, CCL27, and CX3CL1 also elevated in tobacco carcinogen–exposed G < GL mice (Supplementary Fig. S3F).

Overall, our findings highlight the intricate interplay between microbial alterations and gut homeostasis, demonstrating how local inflammation within the gut’s colonic lamina propria can escalate into a systemic inflammatory state, potentially modulating tumor development beyond the intestinal tract.

To begin to understand how inflammation, instigated in the gut and propagated systemically following microbial alterations, could potentially influence the TME, we charted immune changes in the TME by scRNA-seq. Analysis of 20,572 high-quality cells from the tumors of G < G and G < GL syngeneic mice (n = 2 each group) identified distinct clusters representing major cellular lineages, namely, stromal, epithelial, myeloid, T and NK, and B cells (Fig. 3A; Supplementary Fig. S4A; Supplementary Table S6), with a consistent representation of each major cluster across all experimental groups (Supplementary Fig. S4B). The TME in G < GL mice displayed an overall enhanced immunosuppressive phenotype evidenced by prominently increased fractions of exhausted CD4⁺ T, regulatory T cells, exhausted CD8⁺ T cells (cluster 2), and immunosuppressive PMN-MDSCs and, conversely, reduced levels of tissue-resident memory T cells (Fig. 3B; Supplementary Fig. S4C and S4D, Supplementary Table S6). Furthermore, myeloid cells of G < GL tumors displayed increased expression of the proinflammatory cytokines Il6 and Il1b and the inflammasome component Nlrp3 (Fig. 3C), in accordance with our analysis of the myeloid compartment in the colonic lamina propria (Fig. 2E). Consistent with our scRNA-seq analysis, multiparameter flow cytometry analysis showed that the TME in G < GL mice relative to that in G < G littermates was overall marked by increased immunosuppression and reduced antitumor immunity (Supplementary Fig. S3B and S3C; Supplementary Table S5). The TME in G < GL mice comprised increased regulatory and exhausted CD4⁺ T cells and immunosuppressive myeloid subsets, including neutrophils, PMN-MDSCs, monocytic myeloid-derived suppressor cells, and M2 macrophages, and conversely, fewer M1 macrophages, DC1, and CD27⁺ memory B cells as opposed to their G < G counterparts (Fig. 3D and E). Similarly, in our tobacco carcinogen–induced tumor model, we noted increased immunosuppression within the TME of G < GL mice compared with G < G counterparts characterized by increased regulatory CD4⁺ T cells and PMN-MDSCs, as well as increased proinflammatory pro-IL1β⁺ DCs (Supplementary Fig. S4E).

Previous studies have shown that during chronic inflammation, PMN-MDSCs produce IL6, which in turn contributes to their own expansion and maturation (32). We thus probed IL6 expression in these cell subsets in the TME of our FMT models. Indeed, we observed a significant increase in IL6⁺ PMN-MDSCs in the TME of G < GL mice compared with their G < G littermates (Fig. 3F). Reflecting on increased levels of PMN-MDSCs in the TME of G < GL mice and their established role in inflammation, immunosuppression, and tumor progression (33, 34), we were prompted to probe the differentiation of myeloid-derived suppressor cells (MDSC) in the bone marrow. To this end, we exposed naïve bone marrow cells to supernatants from BMDMs cultured with fecal microbiota concentrates from either G < GL or G < G mice (Fig. 3G). Our findings showed a significant shift toward PMN-MDSC differentiation in cells treated with the supernatants conditioned with G < GL fecal microbiota concentrates compared with the control (Fig. 3H). This suggests that microbial alterations, simulated by G < GL FMT, can enhance the differentiation of these immunosuppressive cells, highlighting a potential pathway through which the gut microbiome influences tumor immune evasion, which is a key feature of cancer progression.

Given the observed upregulation of Il6 in colonic and tumor-associated myeloid cells, its systemic elevation in the serum, and its markedly increased expression by PMN-MDSCs, further compounded by the well-documented roles of IL6 in both inflammation and lung cancer pathogenesis (35), we were prompted to explore lung tumor progression mediated by gut microbiome alterations in Gprc5a^−/− mice with knockout of Il6 (GI; Fig. 4A). In concordance with our previous findings (Fig. 2G), systemic serum levels of IL6 were significantly higher in G < GL mice compared with G < G littermates, and these levels were undetectable in hosts with knockout of Il6 (Fig. 4B). Moreover, hosts with knockout of Il6 displayed a significant decrease in tumor burden following FMT from GL mice (GI < GL), in contrast to their Il6-intact littermates (Fig. 4C). However, knocking out Il6 had no effect on tumor burden when hosts were given a eubiotic FMT (G < G vs. GI < G), highlighting that Il6-mediated inflammation is downstream of microbial alterations (Fig. 4C). Moreover, knocking out Il6 in the host restored colon length following GL FMT (GI < GL) compared with mice with intact Il6 (G < GL; Fig. 4D). Additionally, deletion of Il6 in the host mitigated gut microbiome alteration–associated immunosuppression in the TME, as evidenced by a reduction in exhausted PD-1⁺CD8⁺ T cells, neutrophils, PMN-MDSCs, monocytic myeloid-derived suppressor cells, and cDC2 in GI < GL mice relative to G < GL littermates (Fig. 4E). These findings suggest that gut microbial alterations trigger an IL6-mediated inflammatory cascade, which in turn contributes to tumor progression and immunosuppression.

Next, we sought to identify specific bacterial taxa that could be pivotal in inflammation mediated by gut microbiome alterations and enhanced lung tumor growth. We focused on the genus Alistipes, drawn by its marked increase in the gut with microbial alterations and its association with increased tumor burden (Fig. 1D and H; Supplementary Fig. S1F). Using qPCR, we confirmed the presence of two Alistipes spp. known for their roles in inflammation (36), AF and AO in the guts of G < GL mice (Supplementary Fig. S5A). We found that oral gavage of 10⁸ CFU of AF into G mice (G < AF) following broad-spectrum antibiotics significantly increased lung tumor growth compared with mice receiving anaerobic PBS as the vehicle control (G < control; Fig. 5A) while markedly decreasing colon length (Fig. 5B). Knocking out Il6 in the host (GI < AF) nearly reversed this effect, significantly reducing lung tumor growth (Fig. 5A) and restoring colon length (Fig. 5B), underscoring AF as a pathogenic bacterium that thrives in altered gut microbiota and mediates inflammation and tumor development via an IL6-dependent pathway. To closely replicate physiologic conditions, we challenged the mice with AF without prior antibiotic preconditioning, preserving the native gut microbiota, and found a significant increase in tumor burden in G < AF compared with G < control animals (Supplementary Fig. S5B). Introducing AO similarly augmented lung tumor growth, suggesting a broader pathogenic role for Alistipes spp. in cancer promotion (Supplementary Fig. S5B). Enhanced lung tumor growth induced by AF was accompanied by enhanced IL6 levels in the colonic lamina propria (Fig. 5C). Additionally, we observed increased levels of other proinflammatory mediators in the colonic lamina propria of G < AF mice compared with G < control, including IL12, TNFα, CXCL1, and CCL4 (Fig. 5D). Upon Il6 knockout (GI < AF), the levels of these cytokines and chemokines were nearly reduced back to baseline, illustrating IL6’s central role in mediating the inflammatory response elicited by AF in the gut (Fig. 5C and D). AF administration significantly elevated serum levels of LCN2 in G < AF mice compared with controls (Fig. 5E), pointing to LCN2 induction as a response mechanism against Alistipes. Moreover, elevated LCN2 levels were maintained even after Il6 knockout, indicating that whereas IL6 significantly influences inflammation and tumor progression, LCN2 engages in a distinct, possibly complementary, and upstream pathway to directly regulate microbial populations (Fig. 5E). Furthermore, we found increased levels of the chemoattractant CCL4 in the serum of G < AF compared with the control group. Knocking out Il6 reversed these effects albeit not reaching statistical significance (Fig. 5F). Although systemic levels of IL1β were significantly higher in GI < AF compared with their G < AF counterparts, administration of AF did not affect these levels (Fig. 5F). These findings underscore species of Alistipes as drivers of lung tumor progression induced by gut microbiome alterations through an IL6-dependent mechanism.

Having demonstrated the role of gut microbiome alterations in promoting lung adenocarcinoma development and immunosuppression within the TME, we further explored its impact on ICB. We studied mice with blockade of PD-1, which now constitutes the backbone of ICB-based therapy as the standard of care for many patients with lung adenocarcinoma. G < GL mice exhibited increased resistance to ICB compared with G < G mice (Fig. 6A). Similarly, mice treated with both ICB and AF exhibited significantly larger tumors compared with those treated with ICB and vehicle control (Fig. 6B). These findings indicate that inflammation induced by gut microbiome alterations not only accelerates tumor progression but also reduces the effectiveness of ICB therapy.

To evaluate the translational relevance of our findings, we investigated the association between Alistipes spp.’ relative abundance and the response to neoadjuvant nivolumab, or nivolumab combined with ipilimumab, in 21 patients with operable lung adenocarcinoma participating in the phase II randomized NEOSTAR trial (NCT03158129; ref. 11). We utilized MPR, defined as ≤10% residual viable tumor in the original resected tumor bed after neoadjuvant therapy, as a proxy for treatment efficacy. WGS data from fecal samples collected before treatment initiation were processed with MetaPhlAn4 (24) and MaAsLin2 (25) to profile fecal microbiomes and detect differentially abundant taxa present in patients with MPR (n = 7) versus no MPR (n = 14). Alistipes putredinis (FDR-adjusted P < 0.05, −1.74 log₂ fold change) and AF (FDR-adjusted P < 0.05, −1.47 log₂ fold change) were among a group of bacteria enriched in the no-MPR group (Fig. 6C). Although not significant, we observed a higher abundance of Alistipes spp. in in the no-MPR group (P = 0.37; Fig. 6D). Although constrained by a limited sample size and moderate effect sizes, these results suggest a potential link between the baseline relative abundance of certain Alistipes spp. and reduced ICB treatment responsiveness, suggesting a pathogenic influence for Alistipes spp. and a potential role in mediating resistance to immunotherapy. These findings collectively emphasize the importance of gut microbiota, such as Alistipes spp., in modulating therapeutic outcomes in lung adenocarcinoma.

In this study, we aimed to uncover the mechanisms by which modulation of the gut microbiome affects lung adenocarcinoma development and progression. We found that in a model of gut alterations incurred by the loss of Lcn2, inflammation was triggered and propagated systemically, exerting an immunosuppressive effect within the lung TME. This cascade resulted in enhanced tumor growth and a diminished response to immunotherapy. Crucially, gut microbiome–related lung tumor progression seemed to be mediated, at least in part, by microbes that foster inflammation (Alistipes spp.) and via an IL6-dependent mechanism, highlighting a potential avenue for therapeutic intervention (Fig. 6E).

Our experiments demonstrated that FMT from Lcn2-deficient mice significantly increased tumor burden and accelerated lung adenocarcinoma progression, a consistent outcome across various syngeneic animal models and a tobacco-related lung cancer model. Our observations support the supposition that LCN2 may partly counteract lung adenocarcinoma development by maintaining microbial homeostasis. Additionally, the gut microbial alterations resulting from Lcn2 deficiency are characterized by changes in gut microbiome taxonomic composition that might lead to a compromised integrity and regenerative capacity of the gut, as indicated by shortened colon length and a decrease in mucin-producing goblet cells and stem cells, coupled with inflammation. The changes in the colonic epithelial compartment, while interesting, need to be validated using additional methods to better understand how microbial imbalances affect the local milieu in the gut. Our findings are concordant with a previous report showing that Lcn2 deficiency is associated with alterations in the microbiota that precipitate intestinal inflammation in Il10^−/− mice and subsequent colon cancer formation (26). The antimicrobial roles of LCN2, tied to its ability to bind iron-loaded bacterial siderophores like enterobactin (37), suggest that Lcn2 deficiency may confer a growth advantage to certain pathogenic bacteria by increasing iron availability. Therefore, our study proposes that FMT from Lcn2-deficient mice serves as a viable method to induce microbiome alterations, offering a valuable model to explore the gut microbiome’s impact on systemic diseases, including cancer. Although we and others have demonstrated the important roles of LCN2 in maintaining microbial homeostasis, it is important to highlight that LCN2 is a multifaceted protein (14, 16, 38). Our previous work has shown that augmented expression of Lcn2 is associated with chronic obstructive pulmonary disease and counteracts lung adenocarcinoma development (15). Therefore, it is plausible to surmise that inherent reduced expression of Lcn2 in patients at risk for lung cancer could similarly exacerbate lung adenocarcinoma pathogenesis by inducing dysbiosis and other protumorigenic mechanisms, underscoring the importance of LCN2 in maintaining both microbial and systemic homeostasis. Additionally, it is important to consider that the impact of Lcn2 deficiency may vary depending on its site (gut or lung), as this could differentially influence local microbial communities and systemic iron homeostasis. These variations, along with other factors that could affect iron accessibility (e.g., other iron chelators), may contribute to the predisposition to lung adenocarcinoma in high-risk individuals, potentially serving as biomarkers.

Although previous studies have mechanistically explored the roles of the local lung (39) and intratumoral (33) microbiomes in lung adenocarcinoma development, our study functionally probed the gut–lung tumor axis, delineating a mechanism by which gut microbiome alterations influence extraintestinal organs and tissues, such as the lungs. We identified systemic inflammation as a key conduit of communication along this axis. Central to our findings is the role of IL6, a cytokine well recognized for its involvement in chronic inflammation and cancer (35). We found that gut microbiome alterations were linked to elevated systemic levels of IL6, correlating with increased tumor promotion and immunosuppression within the TME. IL6, transcriptionally regulated by NF-κB, activates the STAT3 signaling pathway (40), which is implicated in lung adenocarcinoma pathogenesis, a disease notably associated with inflammation. Moreover, we observed that gut microbiome changes induce increased expression of Il1b and Nlrp3 in myeloid cells of the colonic lamina propria and TME, alongside increased systemic levels of IL1β. The activation of the NLRP3 inflammasome leads to the cleavage of pro-IL1β into its active form, IL1β, which is known to act upstream of IL6, amplifying the inflammatory cascade (41). Thus, our findings are congruent with earlier reports implicating proinflammatory cytokines such as IL1β in dysbiosis-induced lung and hepatocellular tumorigenesis (33, 42). It is important to note that in our study, animals with gut microbiome alterations and enhanced inflammation also showed increased MDSCs as critical elements in the TME. Indeed, an inflammatory milieu characterized by elevated levels of IL6 and IL1β fosters expansion of immunosuppressive cell types, such as MDSCs, into the TME, thus promoting an immunosuppressive environment conducive to tumor growth and evasion from immune responses (43). Additionally, it is reasonable to hypothesize that systemic inflammation induced by gut dysbiosis could also influence lung bacterial composition, compounding inflammatory effects in the lung microenvironment and exacerbating lung adenocarcinoma development.

Our findings also support the pathogenic roles of bacteria within the genus Alistipes. This observation is congruent with existing literature that implicates Alistipes spp. in colorectal cancer development in mouse models of colitis (26), and in mice maintained on a high-fat diet (27). We also showed that knocking out Il6 mitigates the pathogenic proinflammatory roles of Alistipes spp., echoing and extending upon the findings of Moschen and colleagues, demonstrating that Alistipes spp. promote tumorigenesis through an IL6-dependent pathway (26). The clinical relevance of our findings was further highlighted by assessing response to ICB in our animal model and a limited cohort of patients with early-stage lung adenocarcinoma from the phase II randomized NEOSTAR trial (NCT03158129) receiving neoadjuvant immunotherapy (10, 11). Our data show a possible association between the increased abundance of Alistipes spp. and relatively poor response to ICB, yet other studies report varied outcomes. For example, A. putredinis was found in higher abundance in the gut microbiota of Chinese patients with NSCLC who responded to treatment (44), whereas AO correlated with durable clinical benefit in patients with NSCLC undergoing ICB therapy (45). The variances in these findings could be attributed to several factors: the small sample sizes across studies, the specific focus on NSCLC in the cited literature versus lung adenocarcinoma in our study, and the influence of diverse factors like diet and environmental exposures (e.g., smoking) on the gut microbiome composition, as well as the interactions between different microbial species within the gut microbiome and their collective impact on tumorigenesis (46). Furthermore, in our cohort, we selected patients who received exclusively ICB, avoiding the confounding effects of chemotherapy on gut microbiome composition (47). Moreover, responses to gut microbiome alterations are mediated by contextual inflammatory cues influenced by host factors, which may further modulate the effectiveness of immunotherapies, indicating that a one-size-fits-all approach may not always be applicable. Although our study, along with others, provides compelling mechanistic evidence of the role of Alistipes in enhancing extrinsic drivers of tumorigenesis, such as inflammation, confirming its effect on ICB responses requires further validation in larger patient cohorts. It is also plausible to surmise that IL6-dependent inflammation can distract immune cells from executing their antitumor effects during cancer progression and ICB treatment (48). It is worth noting that although IL6 emerged as a key mediator in our study, the complexity of communication within the gut–lung tumor axis likely involves additional factors, such as other proinflammatory cytokines, immunomodulatory molecules, and metabolic products, which were not exhaustively explored.

Collectively, our findings highlight the pivotal role of inflammatory host factors, particularly IL6, in orchestrating the downstream effects of gut microbiome alterations, resulting in immunosuppression within the TME and lung adenocarcinoma progression. We also underscore the involvement of specific bacterial taxa, such as Alistipes spp., in fostering lung tumor–promoting inflammation. These insights provide a compelling rationale for targeting inflammation triggered by gut microbiome changes as a novel strategy to mitigate lung adenocarcinoma pathogenesis and augment responses to ICB.

B. Sepesi reports personal fees from AstraZeneca and Bristol Myers Squibb outside the submitted work. D.L. Gibbons reports grants from Bristol Myers Squibb and Rexanna’s Foundation for Fighting Lung Cancer during the conduct of the study, as well as grants from Mirati Therapeutics, NGM Biopharmaceuticals, and Boehringer Ingelheim and personal fees from Eli Lilly and Company, Onconova, and Menarini Ricerche outside the submitted work. J.A. Wargo reports receiving compensation for speaker’s bureau and honoraria from PeerView, as well as being a consultant and/or advisory board member for Gustave Roussy Cancer Center, Ose Immunotherapeutics, Bayer Therapeutics, James Cancer Center OSU, and Daiichi Sankyo. R.R. Jenq reports other support from Seres Therapeutics, Postbiotics Plus, MaaT Pharma, LISCure, and Prolacta BioScience outside the submitted work, as well as a patent for US-20220080001-A1 issued, a patent for US-20170258854-A1 issued and licensed, a patent for US-11597979-B2 issued and licensed, a patent for US-20240084403-A1 issued, a patent for US-20240241107-A1 issued, and a patent for US-20240216410-A1 issued. S.J. Moghaddam reports grants from NIH/NCI during the conduct of the study. T. Cascone reports grants, personal fees, and other support from Bristol Myers Squibb during the conduct of the study, as well as personal fees from ASCO Post, Bio Ascend, Clinical Care Options, Mark Foundation for Cancer Research, Medscape, Targeted Oncology, oNKO-innate, Pfizer, RAPT Therapeutics, and Regeneron, personal fees and other support from AstraZeneca, IDEOlogy Health, Medical Educator Consortium, OncLive, PEAK Medical, PeerView, Physicians’ Education Resource, Genentech, and Merck, and other support from European Society for Medical Oncology, International Association for the Study of Lung Cancer, and Society for Immunotherapy of Cancer outside the submitted work. K. Hoffman reports grants from NCI during the conduct of the study. H. Kadara reports grants from Johnson & Johnson outside the submitted work. No disclosures were reported by the other authors.

Z. Rahal: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. Y. Liu: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. F. Peng: Data curation, formal analysis, visualization. S. Yang: Data curation, formal analysis, validation, visualization. M.A. Jamal: Resources, data curation, methodology. M. Sharma: Investigation, methodology. H. Moreno: Data curation, formal analysis. A.V. Damania: Data curation, formal analysis. M.C. Wong: Data curation, formal analysis. M.C. Ross: Data curation, formal analysis. A. Sinjab: Data curation, supervision. T. Zhou: Data curation, supervision. M. Chen: Data curation. I. Tarifa Reischle: Data curation. J. Feng: Data curation. C. Chukwuocha: Data curation. E. Tang: Data curation. C. Abaya: Data curation. J.K. Lim: Data curation. C.H. Leung: Formal analysis, validation, writing–review and editing. H.Y. Lin: Formal analysis. N. Deboever: Formal analysis. J.J. Lee: Formal analysis, supervision, validation. B. Sepesi: Resources, supervision. D.L. Gibbons: Resources, supervision. J.A. Wargo: Resources, supervision. J. Fujimoto: Data curation, formal analysis, visualization. L. Wang: Data curation, software, formal analysis, supervision, methodology. J.F. Petrosino: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, methodology. N.J. Ajami: Data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–review and editing. R.R. Jenq: Resources, supervision. S.J. Moghaddam: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–review and editing. T. Cascone: Resources, data curation, supervision, funding acquisition, validation, investigation, writing–review and editing. K. Hoffman: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. H. Kadara: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

This work was supported in part by the NCI grants R01CA248731 (to H. Kadara) and P30-CA016672 Cancer Center Support Grant to the MD Anderson South Flow Cytometry and Cellular Imaging Core Facility and to the Biostatistics Resource Group and the NIH grant 1S10OD024977 to the Sequencing Core Facility at the MD Anderson Cancer Center, as well as the MD Anderson Cancer Center Office of Strategic Research Programs (to H. Kadara) and the University Grant Foundation at MD Anderson Cancer Center. Partial support was also provided by Bristol Myers Squibb and the philanthropic contributions to the University of Texas MD Anderson Cancer Center Lung Cancer Moon Shots Program and the Rexanna’s Foundation for Fighting Lung Cancer. L. Wang, R.R. Jenq, T. Cascone, and H. Kadara are Andrew Sabin Family Foundation Fellows of the University of Texas MD Anderson Cancer Center.