Intratumoral Klebsiella pneumoniae drives metabolic-immune remodeling and immunotherapy resistance in bladder cancer

Ruilin Zhuang; Zhiming Wu; Runhao Zheng; Yuelong Chen; Xingliang Tan; Xinpei Deng; Chichen Zhang; Jiamin Zeng; Yanjun Wang; Haixia He; Yang Liu; Qianghua Zhou; Kai Yao

doi:10.1136/jitc-2026-016574

. 2026 Oct 27;14(10):e016574. doi: 10.1136/jitc-2026-016574

Intratumoral Klebsiella pneumoniae drives metabolic-immune remodeling and immunotherapy resistance in bladder cancer

Ruilin Zhuang ^1,^2,^3,⁰, Zhiming Wu ^1,^2,^3,⁰, Runhao Zheng ^1,^2,^3,⁰, Yuelong Chen ^4,⁰, Xingliang Tan ^1,^2,³, Xinpei Deng ^1,^2,³, Chichen Zhang ^1,^2,³, Jiamin Zeng ^1,^2,³, Yanjun Wang ^1,^2,³, Haixia He ⁵, Yang Liu ^6,^7,^8,^9,^10,^*, Qianghua Zhou ^1,^2,^3,^*, Kai Yao ^1,^2,^3,^*

PMCID: PMC13723928 PMID: 42946394

Abstract

Background: Immune checkpoint blockade has expanded the therapeutic landscape of bladder cancer, yet durable clinical benefit remains limited, and the tissue-level ecological determinants of therapeutic resistance are incompletely understood. Intratumoral microbiota are increasingly recognized as spatially organized components of cancer ecosystems, but whether microbial dysbiosis within bladder cancer tissues functionally shapes antitumor immunity remains unclear.

Methods: We integrated contamination-controlled 2bRAD-M microbiome profiling with fluorescence in situ hybridization, single-cell transcriptomics, metabolomics, clinical annotation, organoids, and mouse models to characterize intratumoral microbial states and investigate their roles in immune remodeling and immunotherapy response.

Results: Bladder tumors harbored compositionally heterogeneous intratumoral microbial states. Enrichment of Klebsiella pneumoniae (K. pneumoniae) was associated with aggressive disease and reduced immunotherapy benefit, with receiver operating characteristic AUCs of 0.77 and 0.72 in the discovery and validation cohorts, respectively. Mechanistically, K. pneumoniae-associated valylphenylalanine (Val-Phe engaged an SLC15A1-linked epithelial response, activated NF-κB/p65 signaling, and induced an immunoregulatory cytokine program that promoted M2-like macrophage polarization, impaired CD8+ T-cell effector function, remodeled the tumor immune microenvironment, and limited anti-PD-1 efficacy. SLC15A1-guided virtual screening identified MK-0557 as a candidate modulator that restrained tumor growth and improved anti-PD-1 response in preclinical models.

Conclusions: These findings identify intratumoral microbial dysbiosis as a functional regulator of metabolic-immune remodeling in bladder cancer and define the K. pneumoniae-associated valylphenylalanine–SLC15A1–NF-κB/p65 axis as a potential therapeutic target for overcoming resistance to immune checkpoint blockade.

Keywords: Bladder Cancer, Immune Checkpoint Inhibitor, Tumor microenvironment - TME

WHAT IS ALREADY KNOWN ON THIS TOPIC

Intratumoral microbiota can influence tumor metabolism and antitumor immunity, but their role in bladder cancer and resistance to immune checkpoint blockade remains poorly defined.

WHAT THIS STUDY ADDS

By integrating contamination-controlled 2bRAD-M microbiome profiling, spatial validation, single-cell transcriptomics, metabolomics, two clinical cohorts, organoids, and mouse models, this study identifies intratumoral Klebsiella pneumoniae enrichment as a marker of aggressive disease and reduced immunotherapy benefit. It delineates a Klebsiella pneumoniae-associated valylphenylalanine-SLC15A1-NF-κB/p65 axis that drives M2-like macrophage polarization and CD8+T cell dysfunction, and identifies MK-0557 as a candidate modulator that suppresses tumor growth and enhances anti-PD-1 efficacy in preclinical models.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

These findings position intratumoral Klebsiella pneumoniae as a candidate biomarker for identifying patients at risk of anti-PD-1 resistance and reveal the valylphenylalanine-SLC15A1-NF-κB/p65 axis as an actionable therapeutic vulnerability. The activity of MK-0557 across organoid, patient-derived xenograft, and immunocompetent models provides a translational basis for targeting microbe-associated metabolic signaling to reverse macrophage-mediated immunosuppression and broaden the benefit of anti-PD-1 therapy in bladder cancer.

Introduction

Bladder cancer (BLCA) remains clinically heterogeneous despite therapeutic advances. Radical cystectomy and platinum-based chemotherapy remain central for muscle-invasive disease, while immune checkpoint inhibitors and other systemic therapies have expanded options for advanced urothelial carcinoma.^1–5 Nevertheless, durable benefit from PD-1/PD-L1 blockade remains limited. Established biomarkers, including PD-L1 expression, tumor mutational burden, molecular subtype, FGFR alteration status, and immune-cell infiltration, improve biological stratification but incompletely predict treatment response.^6
7 This gap suggests that determinants beyond tumor-cell genetics and conventional immune markers may influence the therapeutic state of BLCA.

Growing evidence indicates that microbes can reside within tumor tissues and participate in cancer biology. Pan-cancer profiling identified tumor type-specific bacterial communities,⁸ and single-cell and spatial analyses have shown that tumor-resident bacteria can occupy defined cellular compartments and influence metastatic behavior.⁹ These observations shift the key question from whether microbes are detectable in tumors to where they are located, which host cells they interact with, and whether they alter tumor behavior or therapy response.

Across cancer types, intratumoral microbial heterogeneity and taxon-specific microbial programs have been linked to immune, transcriptomic, metabolic, and clinical differences.^10–13 Together, these studies show that intratumoral microbes are context-dependent and that functional interpretation requires species-level validation and mechanistic follow-up.

BLCA is an appropriate but underexplored setting for this question. The bladder is continuously exposed to urine and recurrent inflammation, and the clinical activity of BCG illustrates how microbial stimuli can reshape antitumor immunity. Yet most BLCA microbiome studies have profiled urine, stool, or reproductive-tract samples rather than tumor tissue.^14–19 Although these studies provide clinically relevant microbial signatures, they cannot determine whether bacteria reside in BLCA tissues, preferentially associate with specific host-cell compartments, or directly contribute to local immune suppression.

Microbial metabolites can regulate epithelial signaling and local antitumor immunity.^20–24 For example, inosine produced by Bifidobacterium pseudolongum enhanced checkpoint blockade in mouse cancer models by supporting antitumor T-cell immunity,²⁰ whereas Fusobacterium nucleatum-derived metabolites have been linked to impaired CD8+T cell trafficking and immunotherapy resistance.^21
25 In BLCA, Parabacteroides distasonis-derived indole-3-acetic acid was reported to suppress tumor progression through the AhR-FASN axis and enhance ferroptosis sensitivity.²⁶ These findings indicate that microbial metabolites can either enhance or restrain antitumor immunity depending on metabolite identity and host context. Whether bacteria residing in BLCA tissues generate local metabolic cues that influence the tumor immune microenvironment or response to immune checkpoint blockade remains unclear.

Here, we profiled the intratumoral microbiota of BLCA tissues and integrated microbial, spatial, single-cell, clinical, metabolomic, and functional analyses to examine whether tumor-associated bacteria are linked to disease progression and immunotherapy response. We focused on Klebsiella pneumoniae as a representative BLCA-enriched bacterium and traced a microbial-metabolic mechanism in which K. pneumoniae-associated valylphenylalanine engages an SLC15A1-related tumor-cell program, activates NF-κB/p65-dependent inflammatory signaling, and promotes macrophage M2-like polarization with impaired CD8⁺ T-cell effector function. We further evaluated whether targeting this axis with MK-0557 could enhance anti-PD-1 efficacy. These findings provide a mechanistic framework for understanding how intratumoral microbes shape BLCA immune escape and suggest a potential strategy for targeting microbe-associated therapeutic resistance.

Methods

Patients and samples

Patients with histologically confirmed BLCA who underwent surgical resection or diagnostic biopsy at Sun Yat-sen University Cancer Center between January 2022 and December 2025 were enrolled. Tumor tissues, paired para-tumor tissues, and normal bladder tissues were collected according to the requirements of each downstream assay. Para-tumor tissues were obtained from macroscopically non-tumor regions at least 2 cm from the tumor margin when sufficient tissue was available. Normal bladder tissues were obtained from matched non-tumor bladder regions or non-malignant bladder specimens and confirmed by pathological review. For microbiome profiling, human tissue metabolomic profiling, and fluorescence in situ hybridization (FISH) validation, formalin-fixed, paraffin-embedded (FFPE) tissue blocks were selected only when sufficient tissue area and pathological quality were available. Tissue materials for microbiome sequencing and metabolomic profiling were aseptically obtained from FFPE sections, and FISH was performed on FFPE tissue sections. Freshly resected samples were used only for organoid culture and patient-derived xenograft (PDX) model establishment and were processed under sterile conditions immediately after surgery.

Patients who had received preoperative antibiotics, chemotherapy, radiotherapy, or intravesical therapy within 4 weeks before tissue collection were excluded from microbiome discovery analyses. Clinical information, including age, sex, T stage, N stage, M stage, clinical stage, treatment history, immunotherapy response, recurrence-free survival (RFS), and overall survival (OS), was extracted from medical records. Immunotherapy response was evaluated according to Response Evaluation Criteria in Solid Tumors version 1.1, which is the standard radiographic response criterion for solid tumors. Complete response (CR) was defined as disappearance of all target lesions and reduction of pathological lymph nodes to a short-axis diameter of <10 mm. Partial response (PR) was defined as at least a 30% decrease in the sum of diameters of target lesions relative to baseline. Progressive disease (PD) was defined as at least a 20% increase in the sum of diameters of target lesions relative to the smallest recorded value, with an absolute increase of at least 5 mm, or the appearance of new lesions. Stable disease (SD) was defined as neither sufficient shrinkage to qualify for PR nor sufficient increase to qualify for PD. In this study, patients with CR or PR were classified as responders, whereas patients with SD or PD were classified as non-responders. RFS was defined as the interval from radical cystectomy to first recurrence, with patients without recurrence censored at last follow-up. OS was defined as the interval from radical cystectomy or treatment initiation to death, with surviving patients censored at last follow-up.

Cell lines and reagents

Human BLCA cell lines T24 and 5637, the mouse BLCA cell line MB49, and RAW264.7 macrophages were obtained from authenticated cell banks. T24 and 5637 cells were cultured in RPMI-1640 medium, and MB49 and RAW264.7 cells were cultured in DMEM. All media were supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin, and cells were maintained at 37°C in a humidified incubator containing 5% CO2. Cells were confirmed to be mycoplasma-free before experiments. Human cell lines were authenticated by short tandem repeat profiling before use. L-Valyl-L-phenylalanine (Val-Phe; HY-107378, MedChemExpress), MK-0557 (AABH93DE0637, Sigma-Aldrich/Aldrich partner), BAY 11–7082 (B5556, Sigma-Aldrich), and anti-mouse PD-1 antibody for in vivo blockade (clone RMP1-14; A2122, Selleck Chemicals) were used in this study.

BLCA-derived organoid culture

Fresh human BLCA tissues were transported in ice-cold Hank’s balanced salt solution (HBSS) supplemented with 1% penicillin-streptomycin (final concentrations: 100 U/mL penicillin and 100 μg/mL streptomycin) within 0.5 hour of resection. After washing, specimens were minced into small fragments and digested in DMEM/F12 containing 0.001% DNase I (Sigma-Aldrich), 1 mg/mL collagenase/dispase (Roche), and 1% penicillin-streptomycin at 37°C for 0.5 hour with intermittent agitation. The digested material was triturated, filtered through 100 μm strainers (BD Falcon), and centrifuged at 1000 rpm for 5 min. The pellet was washed, re-centrifuged, resuspended in Matrigel (356231, Corning), and seeded into 24-well plates (3337, Corning) for gelation. Organoids were maintained in serum-free medium based on DMEM/F12 (Gibco) supplemented with 50 ng/mL EGF (PeproTech), 100 ng/mL FGF10 (PeproTech), 100 ng/mL FGF4 (PeproTech), B27 (Gibco), and 1% penicillin/streptomycin. Medium was changed every 3 days, and organoids were passaged every 10–14 days.

For drug-response assays, organoids derived from treatment-naive BLCA tissues were dissociated or mechanically sheared into comparable fragments and seeded in Matrigel. After establishment, organoids were treated with MK-0557 at the indicated concentrations for 48 hours. Bright-field images were captured using an inverted fluorescence microscope. Viability and cell death were assessed using Calcein AM/propidium iodide (PI) live-dead staining according to the manufacturer’s instructions. Fluorescence intensity or positive areas were quantified using ImageJ or Image-Pro Plus from at least three randomly selected fields per organoid line.

2bRAD-M microbiome sequencing of FFPE tissues

FFPE blocks were sectioned using sterile blades under contamination-controlled conditions. Tumor, para-tumor, and normal tissue regions were reviewed by a pathologist and aseptically obtained from unstained FFPE sections. To reduce environmental and reagent-derived contamination, sample processing was performed in a clean workspace using DNA-free consumables. Paraffin-only sections, blank extraction controls, library-preparation controls, and environmental controls were processed in parallel.

DNA was extracted from FFPE tissue sections using the QIAamp DNA FFPE Tissue Kit (Qiagen, 56404) according to the manufacturer’s protocol. DNA quantity was measured using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific), and DNA fragment distribution was assessed using the Agilent 2100 Bioanalyzer. 2bRAD-M libraries were constructed using type IIB restriction enzyme digestion to generate iso-length microbial genomic tags. Barcoded libraries were amplified, purified, pooled, and sequenced on an Illumina paired-end sequencing platform. Low-quality reads, adaptor sequences, and reads not matching the expected tag structure were removed.

Microbial taxonomic profiling was performed using the 2bRAD-M analytical pipeline. High-quality tags were aligned against a custom species-level 2bRAD marker database (2b-Tag-DB), and species-level relative and absolute abundance matrices were generated after removal of potential false-positive assignments. Taxa detected in negative controls or showing contamination-like prevalence patterns were filtered before downstream analyses. α-diversity indices, including Chao1, richness, Shannon, and Simpson indices, were calculated using the vegan and phyloseq packages in R. β diversity was evaluated by principal coordinate analysis (PCoA), non-metric multidimensional scaling (NMDS), permutational multivariate analysis of variance (PERMANOVA), and analysis of similarities (ANOSIM). Differentially enriched taxa were identified using linear discriminant analysis effect size (LEfSe) with an LDA score threshold of 2.0.

Fluorescence in situ hybridization

FISH was performed to detect total bacterial 16S rRNA signals, gram-negative-associated signals, gram-positive-associated signals, and selected taxa in tissue sections. FFPE sections were deparaffinized, rehydrated, and treated with prehybridization buffer at 37°C for 30 min to block non-specific binding. Hybridization was performed by applying probes diluted in hybridization solution at a ratio of 1:40, followed by denaturation at 73°C for 5–8 min and overnight incubation at 37°C. Sections were washed, counterstained with DAPI, mounted, and scanned using a fluorescence microscope or slide scanner.

Bacterial abundance was quantified using ImageJ. For each section, three non-overlapping fields were randomly selected. The positively stained area fraction or positive signal counts per mm² were measured in each field, and the mean value was used as the representative value for that section. Investigators performing image quantification were blinded to group information.

Bacterial isolation, culture, and scanning electron microscopy

Fresh tumor tissues were processed under sterile conditions for bacterial isolation. Tissue fragments were washed with sterile PBS, homogenized, serially diluted, and plated on blood agar or LB agar under aerobic conditions at 37°C for 18–24 hours. Colonies with morphology consistent with K. pneumoniae were picked and expanded. Species identity was confirmed by 16S rRNA sequencing, and isolates were stored in bacterial freezing medium containing 20% glycerol at −80°C. Bacterial cultures used for functional experiments were grown to logarithmic phase, washed, and standardized by OD600 or colony-forming units.

For scanning electron microscopy, bacterial pellets were fixed in 2.5% glutaraldehyde at 4°C, washed with PBS, post-fixed with 1% osmium tetroxide if applicable, dehydrated through a graded ethanol series, dried by critical-point drying, sputter-coated with gold, and imaged using a scanning electron microscope.

Bacterial coculture, proliferation, colony formation, migration, and EdU assays

For bacterial coculture experiments, BLCA cells were exposed to live K. pneumoniae at a bacteria-to-cell ratio of 10:1 (MOI=10) for 12 hours at 37°C with 5% CO₂. The culture medium used during live K. pneumoniae exposure was antibiotic-free. After transient coculture, cells were washed with PBS and switched to fresh complete medium containing gentamicin (100 μg/mL) for 2 hour to terminate bacterial exposure and prevent bacterial overgrowth before downstream functional assays. For CCK-8 assays, cells were seeded into 96-well plates at 2×103 cells per well, and absorbance at 450 nm was measured after the addition of CCK-8 reagent at the indicated time points. For colony formation, treated cells were seeded into 6-well plates at 500 cells per well and cultured for approximately 2 weeks, fixed with 4% paraformaldehyde, stained with 1% crystal violet, and quantified. For Transwell migration assays, cells in serum-free medium were placed in the upper chamber, and complete medium containing 10% FBS was added to the lower chamber. After 24 hours, migrated cells on the lower membrane surface were fixed, stained with crystal violet, and counted.

For Val-Phe treatment experiments, T24 and 5637 cells were treated with 0, 1, or 2 μM Val-Phe, or with the concentrations indicated in the figures, for 48 hours. For EdU assays, cells seeded in confocal dishes were incubated with 10 μM EdU for 4 hours at 37°C, fixed in 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, and stained using the BeyoClick EdU Cell Proliferation Kit with Alexa Fluor 555 (Beyotime) according to the manufacturer’s protocol. Nuclei were counterstained with DAPI, and the proportion of EdU-positive cells was quantified using Image-Pro Plus 6.0.

Conditioned medium preparation and macrophage polarization assays

K. pneumoniae culture supernatant was collected from logarithmic-phase cultures by centrifugation at 4000× g for 10 min and filtration through a 0.22 µm filter. For tumor cell-conditioned medium, T24 or 5637 cells were cultured to approximately 70%–80% confluence, washed, and incubated in complete medium for 24 hours. For K. pneumoniae-stimulated tumor cell-conditioned medium, BLCA cells were cocultured with K. pneumoniae at MOI=10 for 12 hours, washed with PBS, and then cultured in fresh complete medium for 24 hours. The resulting conditioned medium was collected, centrifuged, filtered, and either used immediately or stored at −80°C.

Bone marrow-derived macrophages (BMDMs) were generated from mouse femurs and tibias in the presence of macrophage colony-stimulating factor (M-CSF) for 7 days. BMDMs and RAW264.7 cells were stimulated for 48 hours with conditioned media derived from control BLCA cells, K. pneumoniae-stimulated BLCA cells, or BLCA cells pretreated with increasing concentrations of Val-Phe. Macrophage polarization was evaluated by flow cytometric detection of CD86 and CD206.

Animal experiments

5-week-old female C57BL/6 mice were used for all MB49-based syngeneic subcutaneous tumor experiments, including the K. pneumoniae, Val-Phe, MK-0557, and anti-PD-1 treatment studies. 5-week-old female NOD-scid IL2Rγnull (NSG) mice were used for PDX experiments. All mice were purchased from Zhuhai Bestest Bio-Tech Co. (Zhuhai, China) and maintained under specific pathogen-free conditions. Mice were randomly assigned to treatment groups. Investigators were blinded to group allocation during tumor measurement and histological quantification.

For K. pneumoniae-associated subcutaneous tumor experiments, MB49 cells were cocultured with K. pneumoniae at MOI=10 for 12 hours, washed with PBS to remove residual culture medium and non-cell-associated bacteria, resuspended in PBS, and injected subcutaneously into the flank of mice at 5×10⁵ cells per mouse. For drug-treatment experiments, mice received intraperitoneal injections once every 3 days. Anti-PD-1 antibody was administered at 10 mg/kg for a total of three injections. Val-Phe and MK-0557 were administered at 20 mg/kg, and tumors were collected 21 days after the first injection. For combination treatment, MK-0557 and anti-PD-1 were administered using the same schedules as the corresponding monotherapy groups. Tumor length and width were measured every 3 days using calipers, and tumor volume was calculated as length×width²/2. On day 21 after the first injection, mice were euthanized, and tumors were excised, photographed, weighed, and processed for histology, flow cytometry, RNA extraction, or metabolomics.

For PDX models, freshly resected BLCA tissues were cut into approximately 2–3 mm³ fragments and implanted subcutaneously into immunodeficient mice. When tumors reached approximately 100–150 mm³, mice were randomized to receive vehicle or MK-0557 by intraperitoneal injection at 20 mg/kg once every 3 days, and tumors were collected 21 days after the first injection. Tumor growth and body weight were monitored throughout treatment.

Flow cytometry

Single-cell suspensions were prepared from subcutaneous tumors, in vitro cultured macrophages, or sorted immune cells. Tumors were minced into 1–2 mm³ fragments and digested in RPMI 1640 supplemented with 1 mg/mL collagenase II, 1 mg/mL collagenase IV (Yeasen), and 200 μg/mL DNase I (Solarbio) at 37°C for 60 min with shaking. The cell suspension was passed through a 100 μm strainer, centrifuged at 450× g for 5 min, and washed with PBS. Erythrocytes and debris were removed using a Percoll density gradient (80%/40%, Biosharp), and dead cells were excluded using a Dead Cell Removal Kit (Miltenyi Biotec) or viability dye.

Cells were incubated with Fc block and stained with fluorophore-conjugated antibodies against CD45, CD3, CD4, CD8, F4/80, CD86, CD206, PD-1, IFN-γ, and GZMB; antibody details are provided in online supplemental table S1. For intracellular IFN-γ or GZMB staining, cells were stimulated with PMA/ionomycin in the presence of brefeldin A, fixed, and permeabilized using the BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences, 554714). Data were acquired on a CytoFLEX flow cytometer (Beckman Coulter) and analyzed using FlowJo v10.

Immunohistochemistry, multiplex immunofluorescence, and TUNEL staining

Tissues were fixed in 10% neutral-buffered formalin, embedded in paraffin, and sectioned at 4 µm. For immunohistochemistry (IHC), sections were deparaffinized, rehydrated, subjected to heat-mediated antigen retrieval in citrate buffer, blocked, and incubated with primary antibody against Ki-67; antibody details are provided in online supplemental table S1. HRP-conjugated secondary antibodies and DAB substrate were used for visualization. Hematoxylin was used for counterstaining.

For multiplex immunofluorescence, sections were stained with antibodies against F4/80, CD86, CD206, CD8, PD-1, and GZMB, followed by fluorophore-conjugated secondary antibodies or tyramide signal amplification according to the manufacturer’s protocol. Nuclei were counterstained with DAPI. TUNEL staining was performed using the In Situ Cell Death Detection Kit, Fluorescein (Roche, 11684795910), according to the manufacturer’s instructions. Images were acquired using a fluorescence microscope or slide scanner, and positive cells or H-scores were quantified using ImageJ and Image-Pro Plus from three randomly selected fields per sample.

RNA extraction, RNA sequencing, and transcriptomic analysis

Total RNA was isolated from cells or tumor tissues using TRIzol reagent (Invitrogen) according to the manufacturer’s protocol.

For RNA sequencing, RNA integrity was assessed using the Agilent 2100 Bioanalyzer. Poly(A)-selected mRNA-seq libraries were constructed and sequenced on an Illumina NovaSeq platform by Annoroad (Beijing, China). Raw reads were quality-filtered using fastp, aligned to the mouse reference genome GRCm39 using STAR, and gene counts were generated using featureCounts. Differentially expressed genes were identified using DESeq2 with thresholds of adjusted p<0.05 and |log2 fold change|>1. GO and KEGG pathway enrichment analyses were performed using clusterProfiler. Immune cell composition was inferred from bulk RNA-seq data of resected subcutaneous tumors using the CIBERSORT deconvolution framework with the mouse-specific ImmuCC signature matrix.^27
28

Cytokine measurement

T24 and 5637 cells were treated with Val-Phe in the presence or absence of BAY 11–7082 for 48 hours. Cell-culture supernatants were collected, centrifuged to remove cellular debris, aliquoted, and stored at −80°C until analysis. Human CCL2, TGF-β, CSF1, IL-6, and IL-10 levels in the culture supernatants were measured using R&D Systems Quantikine ELISA kits according to the manufacturer’s instructions. The kit catalog numbers for CCL2, TGF-β, CSF1, IL-6, and IL-10 were DCP00B, DB100C, DMC00B, D6050B, and D1000B, respectively. For TGF-β detection, samples were acid-activated and neutralized before measurement. Absorbance was measured at 450 nm with wavelength correction at 540 nm using a microplate reader.

Single-cell RNA-seq processing and microbial read analysis

Single-cell RNA-seq data from three paired BLCA and matched normal tissues were processed using Cell Ranger with the GRCh38 reference genome. Cells with fewer than 200 detected genes, more than 6000 detected genes, or more than 20% mitochondrial transcripts were removed, and predicted doublets were excluded before downstream analysis. Gene expression matrices were analyzed using Seurat. Data were normalized, highly variable genes were identified, and batch effects were corrected using Harmony. Principal component analysis, shared-nearest-neighbor graph construction, clustering, and Uniform Manifold Approximation and Projection (UMAP) visualization were performed in Seurat. Cell types were annotated based on canonical marker genes and cluster-specific markers. Cell-cell communication was inferred using CellChat with the CellChatDB human ligand-receptor database. Communication probabilities were calculated among annotated cell types, and signaling pathways with permutation-based p values <0.05 were retained for downstream visualization.

For single-cell-associated microbial read analysis, reads not assigned to the human transcriptome by Cell Ranger were extracted from the alignment files and subjected to microbial taxonomic classification. Low-quality reads, low-complexity reads, ribosomal/repetitive host-derived reads, and reads mapping back to the human genome (GRCh38) were removed before microbial assignment. Taxonomic classification was performed using Kraken2 and Bracken against a custom microbial database constructed from NCBI RefSeq bacterial, archaeal, fungal, and viral genomes. The SILVA 138.1 rRNA database was used to support rRNA-based microbial annotation, and bacterial/archaeal taxonomic labels were further harmonized according to the Genome Taxonomy Database (GTDB). Taxa detected in negative controls, empty-droplet barcodes, or showing contamination-like profiles across samples were excluded. Microbial reads passing quality-control and contamination-filtering steps were mapped back to single-cell barcodes, and microbial abundance was summarized at the phylum, class, family, genus, and species levels to evaluate cell type-specific microbial distribution. Virtual SLC15A1 knockout analysis was performed using scTenifoldKnk in epithelial cells, and pathway changes were evaluated by GO and GSEA analyses.

Untargeted metabolomics and pathway analysis

Human tumor and para-tumor metabolomic profiling was performed using FFPE tissue sections. Pathologist-reviewed FFPE blocks were sectioned with clean blades, and tumor-rich or para-tumor regions were selected from adjacent H&E-stained sections. Unstained sections were deparaffinized using xylene, rehydrated through a graded ethanol series, and air-dried before metabolite extraction. Blank paraffin sections and extraction blanks were processed in parallel to monitor FFPE-derived, solvent-derived, and reagent-derived background signals.

For FFPE-compatible metabolite extraction, deparaffinized tissue material was extracted using pre-cooled methanol/acetonitrile/water extraction solvent containing internal standards. Samples were vortexed, sonicated, incubated at low temperature to promote protein precipitation, and centrifuged at high speed. Supernatants were collected for LC-MS/MS analysis. Mouse tumor tissues and K. pneumoniae supernatant samples were collected separately, stored at −80°C, and extracted using the same organic-solvent-based metabolite extraction strategy. Quality-control samples were prepared by pooling aliquots from all biological samples and were injected at regular intervals throughout the analytical run. Extracts were subjected to untargeted LC-MS/MS analysis in both positive and negative ion modes. Features detected in paraffin blanks, extraction blanks, or with poor reproducibility in quality-control samples were removed before downstream analysis. Raw LC-MS data were processed using MS-DIAL for peak detection, alignment, normalization, and annotation against HMDB, KEGG, mzCloud, and the in-house metabolite database. Differential metabolites were identified using fold-change thresholds, VIP scores from partial least squares discriminant analysis (PLS-DA), and statistical testing with false discovery rate correction. PLS-DA, volcano plots, heatmaps, KEGG enrichment, and pathway topology analyses were performed using MetaboAnalyst and R packages. Shared metabolites across K. pneumoniae supernatant, mouse tumors, and human FFPE tumor samples were identified by overlap analysis.

Western blotting

Cells or tissues were lysed in RIPA buffer supplemented with protease and phosphatase inhibitor cocktails (Selleck.cn). Protein concentrations were determined using a BCA assay. Equal amounts of protein were separated by SDS-PAGE and transferred onto PVDF membranes (Millipore). After blocking with 5% non-fat milk or BSA in TBST, membranes were incubated with primary antibodies against SLC15A1, p65, phospho-p65, IκBα, phospho-IκBα, and GAPDH at 4°C overnight; antibody details are provided in online supplemental table S1. Membranes were then incubated with HRP-conjugated secondary antibodies, and bands were detected using enhanced chemiluminescence (Boster). Band intensity was quantified using ImageJ. SLC15A1, p65, IκBα, and GAPDH were normalized to GAPDH, whereas phospho-p65 and phospho-IκBα were normalized to total p65 and total IκBα, respectively.

Chromatin immunoprecipitation-qPCR

Chromatin immunoprecipitation (ChIP) was performed using a ChIP assay kit (Beyotime, P2078) according to the manufacturer’s protocol. Briefly, T24 and 5637 cells were cross-linked with 1% formaldehyde at 37°C for 10 min, quenched with glycine, and lysed. Chromatin was sheared by sonication to an average size of 200–500 bp. Lysates were incubated overnight at 4°C with anti-p65/RELA antibody or control IgG; antibody details are provided in online supplemental table S1. Protein A/G agarose beads were added and incubated for 1 hour at 4°C. After washing, reverse cross-linking, and DNA purification, enriched DNA was analyzed by qPCR using primers targeting p65-binding regions in the CCL2, TGFB1, CSF1, and IL6 promoters. Primer sequences are provided in online supplemental table S2.

Virtual screening and molecular docking

The AlphaFold-predicted human SLC15A1 protein structure was prepared by removal of water molecules, addition of hydrogens, and charge assignment using AutoDockTools. A compound library containing 12 749 small molecules from ChEMBL was downloaded and converted to docking-compatible formats using Open Babel. Ligands were energy-minimized before docking. The predicted binding pocket of SLC15A1 was used to define the docking grid box. Batch docking was performed using AutoDock Vina, and compounds were ranked according to predicted binding free energy. The top 5% candidates were retained for refined screening, and the highest-affinity compounds were visualized using PyMOL.

Quantification and statistical analysis

Statistical analyses were performed using GraphPad Prism V.9.5.1 and R V.4.2.2. Data are presented as mean±SEM unless otherwise indicated. Normality was assessed before selecting parametric or non-parametric tests. Two-group comparisons were performed using two-sided unpaired Student’s t tests for normally distributed data or Wilcoxon rank-sum tests for non-normally distributed data. Comparisons among more than two groups were performed using one-way ANOVA followed by Tukey’s or Dunnett’s multiple-comparison tests, or Kruskal-Wallis tests followed by Dunn’s multiple-comparison tests for non-parametric data. Tumor growth curves and repeated time-course measurements were analyzed using two-way repeated-measures ANOVA. Survival curves were compared using log-rank tests. ROC curves were evaluated by area under the curve (AUC) analysis, and clinical net benefit was assessed by decision curve analysis.

Microbial β-diversity was assessed using PERMANOVA and ANOSIM. PLS-DA model separation in metabolomic analyses was evaluated by permutation testing. Differential metabolites and differentially expressed genes were analyzed using false discovery rate correction, and adjusted p values were reported for multiple-testing analyses. GO, KEGG, and GSEA analyses were performed using false discovery rate-adjusted statistics. Statistical significance was defined as p<0.05. ns, p≥0.05; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001.

Results

Intratumoral microbiota diversity in patients with BLCA

To characterize the intratumoral microbiota in BLCA, we established a 2bRAD-M workflow for FFPE-derived tumor, para-tumor, and normal tissues, incorporating aseptic tissue acquisition from FFPE sections, DNA extraction, contamination-aware library preparation, and downstream decontaminated microbiome profiling (figure 1A). Given the limited direct histologic evidence for intratumoral microbes in BLCA tissues, we further performed FISH, which revealed increased bacterial signals in tumor tissues relative to normal bladder tissues (figure 1B). Quantitative analysis confirmed that both gram-positive (G+) and gram-negative (G−) bacterial loads were significantly higher in tumor tissues than in normal bladder tissues (online supplemental figure 1A). We next evaluated microbial community structure across the three tissue groups. PCoA and NMDS revealed significant structural differences among normal, para-tumor, and tumor tissues, indicating a distinct microbial community shift in BLCA-associated tissues relative to normal bladder tissue (figure 1C). By contrast, tumor and para-tumor samples showed substantial overlap, consistent with broadly similar community structures in these two compartments. α-diversity analysis based on Chao1, richness, and Shannon indices further showed significantly higher bacterial diversity in both tumor and para-tumor tissues than in normal bladder tissue (figure 1D). This pattern was supported by the Venn diagram, which showed extensive overlap in operational taxonomic units (OTUs) between tumor and para-tumor tissues despite the presence of group-specific OTUs in each group (figure 1E).

BLCA intratumoral microbiota workflow, FISH images, PCoA and NMDS plots, diversity indices, Venn diagrams, LEfSe cladograms, and LDA scores show enrichment and compositional differences among normal, para-tumor, and tumor tissues.

To define the taxa underlying these compositional differences, we examined relative abundance across multiple taxonomic levels. At the phylum level, Pseudomonadota was the dominant phylum across all three groups (online supplemental figure 1B). At the class, order, and family levels, Gammaproteobacteria, Enterobacterales, and Enterobacteriaceae showed higher representation in tumor-associated and para-tumor-associated tissues than in normal bladder tissue (online supplemental figure 1C-E). At lower taxonomic ranks, both Klebsiella and K. pneumoniae were consistently enriched in tumor and para-tumor tissues relative to normal bladder tissue (Online supplemental figure 1G-H). Consistent with these findings, LEfSe analysis, summarized by cladogram and LDA score plots, identified distinct bacterial clades enriched in tumor and para-tumor tissues (figure 1F–G).

We next examined microbial heterogeneity within tumor tissues. PCoA restricted to tumor samples revealed two distinct clusters, subsequently designated G− and G+ according to their microbial signatures and supported by LEfSe analysis (figure 1H–J). The G+ subgroup showed significantly lower Shannon and Simpson diversity than the G− subgroup, and ANOSIM further supported a significant difference in microbial community structure between these two tumor subgroups (R=0.542, p=0.001) (figure 1K). In paired comparisons within the G− subgroup, para-G− tissues showed higher Chao1 and richness indices than tumor-G− tissues, whereas ANOSIM did not detect significant overall community separation between the paired para-G− and tumor-G− groups (R=0.125, p=0.147) (figure 1L). A similar increase in α-diversity indices was observed in para-G+ tissues relative to tumor-G+ tissues, although ANOSIM did not detect a significant difference in overall community structure in this subgroup (online supplemental figure 1H). Finally, FISH analysis further demonstrated the intratumoral presence of representative gram-negative-associated and gram-positive-associated taxa, including Pseudomonadota, Gammaproteobacteria, Alphaproteobacteria, Bacillota, and Bacilli (figure 1M). Collectively, these findings demonstrate marked heterogeneity in the BLCA microbiota and identify K. pneumoniae as a representative enriched species for downstream investigation.

Single-cell transcriptomic analysis reveals the cell type-specific distribution of intratumoral microbes in BLCA

To define the cellular distribution of intratumoral microbes, we performed single-cell RNA sequencing on three paired BLCA and matched normal tissues. Quality-control metrics, including transcript counts, detected gene numbers, and the fractions of mitochondrial and ribosomal transcripts, were comparable across samples, supporting robust downstream integration (online supplemental figure 2A–D). After batch correction and clustering, 28 clusters were resolved and assigned to nine major cell types, including epithelial cells, smooth muscle cells (SMCs), endothelial cells, monocytes, T cells, NK cells, B cells, neutrophils, and dendritic cells (figure 2A,B; online supplemental ffigure 2E–G). Tumor tissues were enriched for epithelial cells, whereas SMCs and B cells were relatively more abundant in normal tissues; T cells, endothelial cells, and monocytes also differed in proportion between the two groups (figure 2C). Marker analysis further delineated each cell compartment, with the epithelial population showing high expression of SCHLAP1, CAPN8, ITGB6, LIPH, KRT19, CDH1, CLDN4, SLC9A2, and PTPRR, consistent with a malignant bladder epithelial identity (figure 2D–F). Functional enrichment linked the epithelial program to cell adhesion, cell-substrate junction organization, protein folding, regulation of protein transport, and phospholipid metabolism, whereas endothelial signatures were enriched in cell adhesion, wound repair, TGF-β response, and actin cytoskeleton remodeling (figure 2G; online supplemental figure 2H).

UMAP, dot plots, violin plots, and box plots show cell type composition, marker gene expression, GO enrichment, and elevated microbial and Klebsiella pneumoniae abundance across BLCA versus normal cell types.

Cell-cell communication analysis positioned epithelial cells at the center of the tumor interaction network, with strengthened signaling toward monocytes, endothelial cells, and neutrophils in tumors relative to normal tissues (figure 2H; online supplemental figure I). Given that bladder and urinary tract infections are predominantly caused by gram-negative bacteria, and that chronic infection-associated inflammation may contribute to BLCA development, we prioritized the G− subgroup for downstream analyses. We therefore focused on representative taxa showing the strongest enrichment and highest abundance within this subgroup, including Pseudomonadota, Gammaproteobacteria, Alphaproteobacteria, Enterobacterales, and K. pneumoniae (figure 1I). At single-cell resolution, overall microbial abundance was increased across most tumor cell populations (figure 2I). Among the representative G− taxa, Pseudomonadota, Gammaproteobacteria, Alphaproteobacteria, and Enterobacteriaceae were broadly elevated across multiple tumor-associated cell types, with the largest increases observed for Gammaproteobacteria and Enterobacteriaceae (online supplemental figure 3A-D). Notably, K. pneumoniae showed the highest abundance among the five representative taxa (online supplemental figure 3E) and was most strongly enriched in the tumor-expanded epithelial compartment, while remaining detectable at higher levels in monocytes, SMCs, B cells, endothelial cells, and T cells (figure 2J; online supplemental figure 3F). Together, these data indicate that intratumoral microbial signals in BLCA are distributed in a cell type-biased rather than random manner, with K. pneumoniae emerging as a dominant epithelial-associated species for downstream investigation.

K. pneumoniae promotes BLCA progression and immunotherapy resistance

Given its epithelial enrichment and its association with microenvironmental remodeling at single-cell resolution, we next examined whether intratumoral K. pneumoniae abundance correlates with clinical behavior and immunotherapy outcome using 2bRAD-based microbial profiles. Correlation analysis revealed that K. pneumoniae abundance positively correlated with T stage, N stage, distant metastasis, and a poor immunotherapy response (figure 3A,B). A model based on K. pneumoniae abundance discriminated immunotherapy response with an AUC of 0.77, and decision curve analysis supported its potential clinical utility (figure 3C). These findings were independently validated by multiplex FISH, which colocalized K. pneumoniae signals with bacterial 16S in tumor tissues (figure 3D). Consistently, higher K. pneumoniae abundance tracked with advanced clinical stage and inferior immunotherapy response, with an AUC of 0.72 in the validation cohort (figure 3E–G). Patients with high intratumoral K. pneumoniae also had significantly worse OS, whereas RFS showed the same direction of association without reaching significance (figure 3H–I).

Box plots, bar charts, ROC curves, Kaplan-Meier curves, and assay images show high K. pneumoniae abundance linked to advanced disease stage, immunotherapy resistance, increased tumor growth, and altered immune cell composition.

To validate the pro-carcinogenic role of K. pneumoniae, we isolated and cultured K. pneumoniae from tumor tissues and confirmed its morphology by colony appearance and scanning electron microscopy (figure 3J). Coculture of BLCA cells with K. pneumoniae increased growth kinetics, migration, EdU incorporation, and colony-forming capacity in both T24 and 5637 cells (figure 3K–M; online supplemental figure 4A–D). Subcutaneous implantation of BLCA cells after K. pneumoniae coculture markedly accelerated tumor growth, which was corroborated by Ki-67 IHC and multiplex fluorescence imaging (figure 3N–P).

To define the underlying molecular program, we performed transcriptomic profiling of control and K. pneumoniae-treated tumors. PCA, inter-sample correlation, and differential-expression heatmaps showed clear group separation and high within-group consistency (online supplemental figure 4E–G). Differentially expressed genes included prominent inflammatory and chemotactic mediators—Cxcl1, Cxcl2, Cxcl3, S100a8, S100a9, Ptgs2, Slfn4, Slc7a11, Acod1, and Mxd1—and were enriched in cytokine, chemokine, NF-κB, JAK-STAT, IL-17, MAPK, and related inflammatory pathways (figure 3Q and online supplemental figure 4H–I). Consistent with these transcriptional changes, CIBERSORT analysis showed increased M2 macrophages together with reduced T-cell-associated subsets in K. pneumoniae-treated tumors (figure 3R). Beyond this immune remodeling, preranked GSEA revealed enrichment of hypoxia and EMT programs in K. pneumoniae-treated tumors, whereas MET and stem cell population maintenance were not enriched (online supplemental figure 4J). Together, these data support a model in which K. pneumoniae promotes BLCA progression while fostering an immunotherapy-resistant microenvironment.

K. pneumoniae promotes immunosuppression by enhancing M2 polarization and impairing CD8 T-cell effector function

To validate these transcriptional findings at the protein level, we performed flow cytometry analysis of tumor-derived immune cells. Consistent with the CIBERSORT predictions, K. pneumoniae-treated tumors showed a reduced proportion of CD86⁺ M1-like macrophages and an increased proportion of CD206⁺ M2-like macrophages relative to PBS controls (figure 4A). Similarly, the proportion of CD4⁺ T cells increased, whereas the proportions of CD8⁺ T cells and IFN-γ⁺CD8⁺ T cells decreased (figure 4B,C), indicating altered immune-cell composition and an impaired CD8 T-cell effector phenotype. Because absolute tumor-infiltrating lymphocyte (TIL) numbers were not quantified, these results reflect relative changes in immune-cell composition and phenotype rather than differences in overall TIL infiltration. Multiplex fluorescence staining provided spatial support for this immune remodeling, showing increased F4/80⁺CD206⁺ and reduced F4/80⁺CD86⁺ signals in K. pneumoniae-treated tumors, together with reduced CD8⁺GZMB⁺ and increased CD8⁺PD-1⁺ signals, consistent with M2-skewed polarization and CD8⁺ T-cell exhaustion (figure 4D,E and online supplemental figure 5A–E).

Flow cytometry, immunofluorescence, and histology show K. pneumoniae increases M2 macrophages and PD-1 expression while reducing CD8 T cells, and blunts anti-PD-1 efficacy, elevating Ki-67 and reducing TUNEL staining.

To test whether K. pneumoniae-mediated M2 polarization is dependent on tumor cells, we exposed BMDM and RAW264.7 cells to conditioned media from different sources. Conditioned medium derived from tumor cells pretreated with K. pneumoniae supernatant induced the strongest increase in CD206⁺ cells, surpassing the effects of either K. pneumoniae supernatant or tumor cell-conditioned medium alone, indicating that K. pneumoniae can promote macrophage polarization by reprogramming tumor cell-derived soluble factors (figure 4F,G). We next assessed how this immunosuppressive state affects anti-PD-1 therapy. Notably, anti-PD-1 monotherapy markedly suppressed tumor growth, whereas its antitumor effect was partially blunted in the presence of K. pneumoniae (figure 4H,I). Increased Ki-67 expression and reduced TUNEL positivity further confirmed that K. pneumoniae counteracted the anti-proliferative and pro-apoptotic actions of anti-PD-1 treatment (figure 4J). Together, these data indicate that K. pneumoniae fosters an immunosuppressive tumor microenvironment by promoting macrophage M2-like polarization, weakening CD8 T-cell effector activity, and reducing anti-PD-1 responsiveness.

K. pneumoniae-associated metabolic reprogramming identifies Val-Phe as a candidate mediator of protumor activity

To identify microbial metabolites from K. pneumoniae involved in tumor progression, we first conducted untargeted metabolomic profiling on tumor tissues from subcutaneous tumor models (figure 5A). Differential-metabolite analysis coupled with unsupervised clustering revealed a coordinated metabolic shift in K. pneumoniae-treated tumors, producing a distinct profile readily distinguishable from that of PBS controls (figure 5B,C). KEGG enrichment identified nucleotide and purine metabolism, glutathione metabolism, the cGMP-PKG pathway, ferroptosis, ABC transporters, and arginine-proline metabolism as the most affected pathways (figure 5D). To more directly capture the soluble components of this metabolic shift, we also profiled the metabolome of K. pneumoniae culture supernatant (figure 5E). Volcano analysis showed extensive metabolite redistribution in response to the K. pneumoniae supernatant (figure 5F). Pathway topology analysis highlighted the most perturbed pathways as alanine/aspartate/glutamate metabolism, tryptophan metabolism, arginine-proline metabolism, pyruvate metabolism, and nucleotide metabolism (figure 5G). Consistently, the altered metabolites were predominantly assigned to amino acid, lipid, cofactor and vitamin, nucleotide, and carbohydrate metabolism (figure 5H); KEGG analysis further implicated ABC transporters, tryptophan metabolism, D-amino acid metabolism, pyrimidine metabolism, and other cancer-associated metabolic pathways (figure 5I).

Multimodal metabolomic analysis comparing K. pneumoniae and PBS conditions via PLS-DA, volcano plots, heatmap, and KEGG pathway analyses converges on Valylphenylalanine as the shared metabolite across mouse and patient tumors.

We next extended metabolomic profiling to human BLCA tissues. Tumor and para-tumor tissues formed distinct metabolic clusters by heatmap analysis and PLS-DA (online supplemental figure 6A, B). The corresponding differential metabolites were enriched in glycerophospholipid metabolism, glutathione metabolism, purine metabolism, arginine biosynthesis, alanine/aspartate/glutamate metabolism, and fatty acid biosynthesis (online supplemental figure 6C). Stratifying patients by our previously defined microbial subgroups maintained this metabolic separation; enriched pathways included vitamin B₆ metabolism, tyrosine metabolism, the pentose phosphate pathway, pyrimidine metabolism, tryptophan metabolism, and fatty acid biosynthesis (online supplemental figure 6D,E). Volcano plots further confirmed broad metabolic differences both between tumor and para-tumor tissues and across patient subgroups (figure 5J,K). Integration of the differential metabolite profiles from K. pneumoniae culture supernatant, K. pneumoniae-treated tumors, and human BLCA tissues identified Val-Phe as the only metabolite consistently detected across all three datasets (figure 5L). These findings nominate Val-Phe as a candidate metabolite linking K. pneumoniae-induced metabolic reprogramming to immune remodeling of the BLCA microenvironment.

Val-Phe promotes BLCA progression and immunosuppressive microenvironment remodeling

Given that metabolomic profiling identified Val-Phe as a key K. pneumoniae-associated candidate metabolite, we next evaluated its functional role in BLCA. In vitro, Val-Phe treatment markedly increased colony formation, migratory capacity, and EdU positivity in both T24 and 5637 cells, indicating that Val-Phe directly enhances proliferation-related and migration-related phenotypes in BLCA cells (figure 6A–C; online supplemental figure 7A–C). Consistently, compared with controls, Val-Phe-treated subcutaneous tumors grew substantially faster, with significantly increased terminal tumor volume and weight (figure 6D). Moreover, Ki-67 immunostaining confirmed enhanced proliferative activity in Val-Phe-treated tumors, supporting a tumor-promoting effect of Val-Phe in vivo (figure 6E).

Val-Phe treatment increases colony formation, migration, proliferation, tumor growth, Ki-67, M2 macrophages, and CD206 expression, while reducing CD8 T cells and IFN-gamma in BLCA models.

Given the link between Val-Phe and K. pneumoniae-associated immune remodeling, we further examined its impact on the tumor immune microenvironment. Multiplex fluorescence staining revealed that Val-Phe treatment increased F4/80⁺CD206⁺ cells while reducing F4/80⁺CD86⁺ cells, indicative of a shift toward an M2-like macrophage phenotype. Concurrently, CD8⁺GZMB⁺ signals declined, whereas CD8⁺PD-1⁺ signals rose, suggesting impaired CD8⁺ T-cell effector function and an exhausted state (figure 6F,G). Flow cytometry further supported these findings: Val-Phe treatment reduced CD86⁺CD206⁻ M1-like macrophages and increased CD206⁺ M2-like macrophages. Furthermore, intratumoral CD4⁺ T cells increased, while CD8⁺ T cells decreased, and IFN-γ⁺ CD8⁺ T cells were markedly reduced, indicating suppression of CD8 T-cell-mediated antitumor immunity (figure 6H–J and online supplemental figure 7D–G). To determine whether Val-Phe enhances the macrophage-polarizing activity of tumor-cell-derived soluble mediators, we stimulated BMDM and RAW264.7 cells with conditioned medium collected from BLCA cells pretreated with increasing concentrations of Val-Phe. Conditioned medium from Val-Phe-pretreated BLCA cells dose-dependently increased the proportion of CD206⁺ macrophages in both BMDMs and RAW264.7 cells (figure 6K,L). Collectively, these findings demonstrate that Val-Phe not only promotes malignant phenotypes in BLCA cells, but also enhances tumor-cell-mediated macrophage M2-like polarization and suppresses CD8 T-cell effector function, thereby contributing to the establishment of an immunosuppressive tumor microenvironment.

Val-Phe activates NF-κB/p65 signaling and induces immunoregulatory programs through a mechanism involving SLC15A1

Given the combined transcriptomic evidence from K. pneumoniae-treated tumors of NF-κB pathway activation and its canonical role as a driver of M2-polarizing factors, we next examined whether Val-Phe exerts its immunoregulatory effects through NF-κB/p65 signaling. In both T24 and 5637 cells, Val-Phe treatment increased the supernatant levels of CCL2, TGF-β, CSF1, IL-6, and IL-10, as measured by cytokine assay OD450 readouts, and this effect was attenuated by BAY 11–7082 (figure 7A–E). We next assessed the activation status of the NF-κB pathway after Val-Phe exposure. Immunoblotting showed a dose-dependent increase in p-p65, accompanied by changes in IκBα and p-IκBα, consistent with activation of canonical NF-κB signaling in both T24 and 5637 cells (figure 7F; online supplemental figure 8A, B). Immunofluorescence further demonstrated increased nuclear p65 signal following Val-Phe treatment, supporting enhanced p65 nuclear translocation (figure 7G). Together, these data indicate that Val-Phe acts upstream of NF-κB/p65 rather than merely altering the secretory phenotype of tumor cells. To determine whether NF-κB/p65 directly regulates the transcription of these factors, we examined p65 binding at their promoter regions. Motif analysis using the JASPAR database identified predicted p65-binding motifs within the promoters of CCL2, TGFB1, CSF1, and IL6 (figure 7H). Consistently, analysis of public p65 ChIP-seq datasets from Cistrome DB showed evident p65 enrichment peaks at these loci (figure 7I). ChIP-qPCR further confirmed that p65 occupancy at the CCL2, TGFB1, CSF1, and IL6 promoters was significantly higher than IgG control in both T24 and 5637 cells (figure 7J). These results support a model in which Val-Phe enhances p65-dependent transcription of immunoregulatory genes and thereby reprograms the tumor cell secretory landscape.

Bar charts, immunoblots, immunofluorescence, ChIP-qPCR, UMAP, volcano plot, and GSEA show Val-Phe activating NF-κB p65 signaling via SLC15A1, increasing cytokines CCL2, TGF-β, CSF1, IL-6, and IL-10.

We then asked how Val-Phe enters BLCA cells. Based on the dipeptide-like structure of Val-Phe, we screened candidate peptide transporters from the SLC15 family (SLC15A1, SLC15A2, SLC15A3, and SLC15A4) and examined their expression in TCGA-BLCA. Among these candidates, only SLC15A1 was significantly upregulated in BLCA tissues (Online supplemental figure 8C). Single-cell analysis further showed that SLC15A1 was primarily expressed in epithelial cells and was elevated in tumor-associated epithelial populations (figure 7K). To evaluate its functional relevance, we applied scTenifoldKnk virtual knockout analysis, which showed broad transcriptional perturbation after SLC15A1 depletion (figure 7L), enrichment of metabolic and transcriptional regulatory programs (figure 7M), and attenuation of NF-κB pathway activity (figure 7N). We next generated SLC15A1-overexpressing T24 and 5637 cells. SLC15A1 overexpression increased intracellular Val-Phe levels and further enhanced p-p65, together with corresponding changes in IκBα and p-IκBα (figure 7O; Online supplemental figure 8D–F). These findings indicate that SLC15A1 functions as a Val-Phe-associated transporter and facilitates downstream NF-κB/p65 activation. Collectively, these results indicate that Val-Phe enters BLCA cells through a mechanism involving SLC15A1, activates NF-κB/p65 signaling, and induces expression of CCL2, TGFB1, CSF1, IL6, and IL10, thereby providing a molecular basis for macrophage M2 polarization and immunosuppressive microenvironment remodeling.

Small-molecule screening targeting SLC15A1 identifies MK-0557, which enhances anti-PD-1 efficacy

Given that SLC15A1 mediated Val-Phe uptake and subsequent NF-κB/p65 activation in BLCA cells, we next explored whether SLC15A1 could be pharmacologically targeted. We assembled a screening set of 12 749 compounds from the ChEMBL small-molecule library and used the SLC15A1 protein structure as the receptor model for docking-based virtual screening. Candidate compound structures were converted, energy-minimized, and imported into AutoDockTools. Batch docking was then performed against the predicted binding pocket of SLC15A1, and compounds were ranked according to the lowest binding free energy. After high-throughput screening, 4202 compounds were retained. Affinity-based virtual screening of the top 5% yielded 211 candidates, and further refined screening identified 10 high-affinity molecules (figure 8A). Among these candidates, MK-0557 showed the strongest predicted binding capacity, with a binding energy of −10.242 kcal/mol (figure 8B). Docking analysis showed that MK-0557 occupied the putative SLC15A1 binding pocket and interacted with several key residues, including SER-174, THR-178, GLN-326, PHE-297, and GLN-300, suggesting that MK-0557 may interfere with SLC15A1-related transport activity through direct binding (figure 8C). We then evaluated the antitumor activity of MK-0557 in organoids generated from tumor tissues of three treatment-naive patients newly diagnosed with BLCA. Bright-field imaging showed that MK-0557 treatment reduced organoid size and disrupted organoid architecture. Consistently, Calcein AM/PI live-dead staining revealed a dose-dependent reduction in viable-cell signals together with increased dead-cell signals, indicating that MK-0557 suppresses viability and induces cell death in patient-derived BLCA organoids (figure 8D).

Virtual screening workflow identifies MK-0557 as top predicted SLC15A1 binder. MK-0557 reduces BLCA organoid viability and PDX tumor growth, and combined with anti-PD-1 further decreases Ki-67 and increases TUNEL staining.

To further validate its activity in vivo, we established BLCA PDX models, which better preserve clinical tumor heterogeneity. Compared with controls, MK-0557 markedly delayed PDX tumor growth and reduced terminal tumor weight, as supported by tumor images and quantitative measurements (figure 8E,F). Histological analysis further showed decreased Ki-67 positivity and increased TUNEL-positive cells in MK-0557-treated tumors, indicating that MK-0557 inhibits tumor-cell proliferation and promotes cell death in the PDX model (figure 8G,H). Because the Val-Phe-SLC15A1 axis was closely linked to immunosuppressive microenvironment remodeling, we next assessed whether MK-0557 could enhance anti-PD-1 efficacy in the MB49 subcutaneous tumor model. Compared with either monotherapy, combined MK-0557 and αPD1 treatment produced a stronger reduction in tumor burden, as shown by tumor images, growth curves, and terminal tumor weights (figure 8I,J). Consistently, tumors from the combination group showed further decreased Ki-67 staining and markedly increased TUNEL positivity, indicating enhanced suppression of proliferation and induction of cell death (figure 8K,L). Together, these findings demonstrate that MK-0557 suppresses BLCA growth and potentiates anti-PD-1 therapy, supporting pharmacological blockade of the Val-Phe-SLC15A1-NF-κB/p65 axis as a potential strategy to improve immunotherapy response.

Discussion

The concept of an intratumoral microbiome has reshaped how the tumor microenvironment is interpreted. Pan-cancer studies have demonstrated that bacteria are detectable in human tumors, exhibit tumor type-specific patterns, and may associate with particular cellular compartments rather than existing as random contaminants.^8
9 More recent studies in hepatocellular carcinoma, cholangiocarcinoma, gastric cancer, and breast cancer further suggest that tumor-associated microbes can influence immune contexture, metabolic state, and clinical outcome.^10–13 However, the field still lacks mechanistic studies that move from microbial detection to a defined bacterium–metabolite–host pathway. This limitation is especially relevant in BLCA, where most microbiome work has focused on urine or gut microbial communities, while direct evidence for tumor-resident microbes and their functional role in treatment resistance remains limited.^{14–19
26
29} The present study therefore extends prior BLCA microbiome research by shifting the focus from microbial association to a tissue-based, species-level mechanism of immune remodeling.

Technical rigor is central to any tumor microbiome study because tumor tissues are low-biomass samples and are vulnerable to reagent, environmental, and processing-derived signals.^8
30–32 Contemporary tumor microbiome studies increasingly rely on orthogonal validation and contamination-aware interpretation rather than sequencing alone.^8
9
30–32 Against this background, the combination of FFPE-compatible 2bRAD-M profiling, aseptic tissue acquisition from FFPE sections, negative-control-aware analysis, and FISH validation strengthens the biological interpretation of the BLCA microbial signal. The partial overlap between tumor and para-tumor microbial communities should not be viewed simply as a lack of tumor specificity. Similar observations in other cancers support the idea that microbial remodeling may involve a broader tissue field shaped by mucosal barrier disruption, inflammation, vascular dissemination, or tumor-conditioned niches.^10
11 Thus, our data support BLCA as a tumor type in which microbial heterogeneity is embedded within the local tissue ecosystem, not merely superimposed on malignant cells.

Single-cell studies have added an important dimension to tumor microbiome research by showing that bacteria can associate with tumor cells, immune cells, and stromal compartments in a cell type-dependent manner. This matters because the biological effect of a microbe is likely determined not only by its abundance, but also by the host cell compartment in which it is detected. In BLCA, epithelial cells represent the malignant compartment and are major sources of inflammatory mediators, chemokines, metabolic signals, and immune-regulatory factors. The epithelial enrichment of K. pneumoniae-associated signals in our study therefore provides a cellular context for microbial-host interaction. Rather than simply confirming that bacteria are present in BLCA tissues, these data suggest that K. pneumoniae may be positioned to influence tumor-cell programs that subsequently shape macrophage behavior and T-cell function. This interpretation is consistent with the broader literature showing that tumor-resident microbes can act through host transcriptional and metabolic remodeling rather than through microbial abundance alone.

K. pneumoniae is best known as an opportunistic gram-negative pathogen, but recent cancer microbiome studies have broadened the relevance of pathobionts beyond classical infection. Tumor-enriched bacteria may contribute to cancer progression by sustaining inflammation, altering nutrient availability, producing bioactive metabolites, and modifying local immunity.^{8–13
30–32} In BLCA, the clinical association between high intratumoral K. pneumoniae abundance and poor immunotherapy response supports the possibility that a tumor-resident pathobiont can participate in immune escape. This finding complements studies linking gut microbiota to checkpoint blockade response, but adds an important local dimension: the microbial feature is detected in tumor tissue rather than inferred from stool or urine. The adverse association with OS and the same-direction trend for RFS further suggest clinical relevance, although prospective cohorts will be required before K. pneumoniae can be considered a predictive biomarker. The main implication is that BLCA immunotherapy resistance may partly reflect local microbial-metabolic states within the tumor bed.

Macrophage polarization is one of the most common routes by which tumors establish an immunosuppressive microenvironment. Tumor-associated macrophages with M2-like features are frequently linked to angiogenesis, tumor growth, extracellular matrix remodeling, T-cell exclusion, and immune checkpoint resistance.^33–35 Cytokines and chemokines such as CCL2, CSF1, TGF-β, IL-6, and IL-10 are well-recognized mediators of monocyte recruitment and macrophage polarization in cancer.^33–35 The immune phenotype associated with K. pneumoniae in this study is consistent with this framework: the bacterium did not merely increase tumor growth, but pushed the microenvironment toward CD206⁺ macrophage enrichment and reduced CD8⁺ T-cell effector features. The conditioned-medium experiments are particularly important because they suggest an indirect mechanism in which bacterial exposure first reprograms BLCA cells, and tumor-cell-derived soluble factors then instruct macrophage polarization. This places tumor cells as intermediaries between a microbial cue and an immunosuppressive immune state, a mechanism that is more nuanced than direct bacterial stimulation of immune cells alone.

Microbial metabolites have emerged as key mediators linking bacteria to cancer biology. Short-chain fatty acids, tryptophan metabolites, bile acid derivatives, and indole compounds have all been implicated in tumor immunity and therapeutic response.^20–24 In BLCA, a recent study reported that gut Parabacteroides distasonis-derived indole-3-acetic acid suppresses tumor progression by regulating the AhR-FASN axis and ferroptosis sensitivity.²⁶ That work supports a broader principle: microbial metabolites can act as host-pathway modulators in BLCA. Our study adds a complementary but opposite example in the intratumoral setting. Val-Phe, a dipeptide-like metabolite, emerged as a shared K. pneumoniae-associated signal across bacterial supernatant, mouse tumors, and human BLCA tissues. Compared with many microbial metabolites that remain correlative, Val-Phe was functionally linked to tumor-cell behavior and macrophage polarization. This finding suggests that dipeptide metabolites, which have received far less attention than indoles or short-chain fatty acids in cancer microbiome research, may also participate in tumor immune remodeling.

SLC15A1, also known as PEPT1, is a proton-coupled oligopeptide transporter classically involved in intestinal uptake of di- and tripeptides and peptide-like drugs.^36
37 In cancer, solute carriers are increasingly recognized as metabolic gatekeepers that determine how tumor cells acquire nutrients, import metabolites, and respond to therapy. The relevance of SLC15A1 to BLCA has not been well defined, and its connection to microbial metabolites has been particularly unclear. By linking Val-Phe accumulation to SLC15A1 expression and NF-κB pathway activity, our study suggests that BLCA epithelial cells may use a peptide transporter-associated mechanism to interpret bacteria-derived dipeptide-like signals. This does not yet prove definitive transport kinetics for Val-Phe, but it provides a testable mechanism: microbial metabolite availability alone is insufficient; host-cell transporter context may determine whether that metabolite becomes immunologically active inside tumor cells.

NF-κB is a central inflammatory transcription factor that couples microbial and cytokine stimuli to chemokine production, survival signaling, and immune modulation. In cancer cells, NF-κB activation can induce factors such as CCL2, CSF1, IL6, IL10, and TGFB1, thereby promoting macrophage recruitment, M2-like polarization, and suppression of cytotoxic T-cell activity.^25
33–35 The Val-Phe-SLC15A1-NF-κB connection in this study therefore provides a plausible transcriptional bridge between microbial metabolism and immune evasion. BAY 11–7082 was used here specifically as an NF-κB pathway inhibitor to test whether Val-Phe-induced cytokine expression depends on this pathway. The reduction of Val-Phe-enhanced cytokine output by BAY 11–7082, together with p65 activation and promoter occupancy analyses, supports a model in which Val-Phe activates a tumor-cell inflammatory program capable of reshaping macrophage polarization.

Therapeutically, the study raises the possibility that microbe-associated immune resistance can be targeted through host metabolic signaling rather than by attempting to eliminate the bacterium itself.³⁸ This distinction is important because intratumoral microbial communities are heterogeneous, and direct antimicrobial approaches risk broad ecological effects and uncertain tumor penetration. MK-0557 was originally developed as an orally available neuropeptide Y Y5 receptor antagonist in metabolic disease research rather than as an oncology drug.³⁹ Its identification here through SLC15A1-oriented virtual screening therefore represents a drug-repurposing style strategy rather than a conventional targeted therapy pipeline. The organoid, PDX, and anti-PD-1 combination data support MK-0557 as a candidate tool compound for interrupting the Val-Phe-SLC15A1 axis. More broadly, these findings suggest that host pathways sensing microbial metabolites may be therapeutically tractable nodes for reversing tumor immune suppression.

This work also helps place BLCA within the broader tumor microbiome literature. Prior studies have shown that intratumoral microbes can be tumor type-specific, compartmentalized, prognostically relevant, and metabolically active.^{8–13
30–32} BLCA studies have separately implicated urinary and gut microbiota in tumor risk, prognosis, and immunotherapy response.^{14–19
26
29} What has been missing is a direct tumor-tissue mechanism explaining how a BLCA-associated bacterium might alter immune state and treatment response. By connecting K. pneumoniae to Val-Phe, SLC15A1, NF-κB-dependent cytokine expression, macrophage polarization, CD8⁺ T-cell dysfunction, and anti-PD-1 resistance, this study provides a mechanistic bridge between these previously separate areas. It therefore supports the view that the BLCA microbiome should be considered not only as a diagnostic or prognostic signature, but also as a functional component of tumor ecology.

Several limitations should be considered. The clinical associations require prospective validation in larger cohorts treated with standardized immunotherapy regimens. Although contamination-aware sequencing and FISH support the presence of BLCA-associated microbial signals, multicenter tissue studies with harmonized negative controls will be needed to confirm reproducibility. Moreover, because the study did not include a matched cross-cancer comparison, we cannot determine whether the enrichment of Gammaproteobacteria, Enterobacterales, and Enterobacteriaceae is specific to BLCA. Related taxa have been reported in other solid tumors, and standardized cross-cancer studies are needed to distinguish shared from tumor type-specific intratumoral microbial features.^8
40 Because patients with antibiotic exposure within 4 weeks before tissue collection were excluded to minimize treatment-related confounding, the effects of antibiotic administration on intratumoral microbial composition, K. pneumoniae abundance, and tumor pathology could not be assessed. The potential impact of antibiotic administration on the intratumoral microbiota warrants further investigation in prospective longitudinal studies. The Val-Phe-SLC15A1 relationship should be strengthened by direct transport assays, isotope tracing, SLC15A1 loss-of-function experiments, and rescue studies. The precise contribution of live bacteria, bacterial structural components, bacterial metabolites, and tumor-cell-derived secondary mediators also remains to be separated. Finally, MK-0557 requires pharmacologic characterization, including target engagement, specificity, pharmacokinetics, toxicity, and optimized scheduling with anti-PD-1 therapy. These limitations do not diminish the central message, but define the next experimental steps needed to move from mechanistic discovery toward therapeutic translation.

Supplementary material

online supplemental file 1

jitc-14-10-s001.docx^{(8MB, docx)}

DOI: 10.1136/jitc-2026-016574

online supplemental file 2

jitc-14-10-s002.tiff^{(9.2MB, tiff)}

DOI: 10.1136/jitc-2026-016574

The funders had no role in the study design, data collection, data analysis, data interpretation, writing of the manuscript, or decision to submit the manuscript for publication.

Footnotes

Funding: This study was supported by the National Natural Science Foundation of China (82472805), the Guangdong Basic and Applied Basic Research Foundation (2024A1515013007), and the Beijing Xisike Clinical Oncology Research Foundation (Y-Gilead2024-ZD-0239) to KY; the Guangdong Medical Science and Technology Research Fund (A2024379) to YW; the National Science Foundation for Young Scholars of China (82503944) to XT; and the Yunnan Fundamental Research Kunming Medical University Projects (202501AY070001-243) to YC.

Provenance and peer review: Not commissioned; externally peer reviewed.

Patient consent for publication: Not applicable.

Ethics approval: All animal experiments were approved by the Experimental Animal Ethics Committee of Sun Yat-sen University Cancer Center (approval no. L102002025000B) and conducted in accordance with institutional guidelines for animal care and use. The use of human tissues was approved by the Ethics Committee of Sun Yat-sen University Cancer Center (approval no. G2024-147-01). Archived clinical information and tissue specimens were used, and the requirement for written informed consent was waived by the ethics committee.

Data availability free text: Sequencing data generated in this study are available from the corresponding author on reasonable request. Public datasets and reference resources used in this study are available from TCGA-BLCA, ChEMBL, AlphaFold, NCBI RefSeq, SILVA, GTDB, HMDB, and KEGG.

Data availability statement

Data are available on reasonable request.

References

1.Powles T, Valderrama BP, Gupta S, et al. Enfortumab Vedotin and Pembrolizumab in Untreated Advanced Urothelial Cancer. N Engl J Med. 2024;390:875–88. doi: 10.1056/NEJMoa2312117. [DOI] [PubMed] [Google Scholar]
2.Roviello G, Santoni M, Sonpavde GP, et al. The evolving treatment landscape of metastatic urothelial cancer. Nat Rev Urol. 2024;21:580–92. doi: 10.1038/s41585-024-00872-0. [DOI] [PubMed] [Google Scholar]
3.van der Heijden MS, Sonpavde G, Powles T, et al. Nivolumab plus Gemcitabine-Cisplatin in Advanced Urothelial Carcinoma. N Engl J Med. 2023;389:1778–89. doi: 10.1056/NEJMoa2309863. [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Loriot Y, Matsubara N, Park SH, et al. Erdafitinib or Chemotherapy in Advanced or Metastatic Urothelial Carcinoma. N Engl J Med. 2023;389:1961–71. doi: 10.1056/NEJMoa2308849. [DOI] [PubMed] [Google Scholar]
5.Patel DM, Mateen R, Qaddour N, et al. A Comprehensive Review of Immunotherapy Clinical Trials for Metastatic Urothelial Carcinoma: Immune Checkpoint Inhibitors Alone or in Combination, Novel Antibodies, Cellular Therapies, and Vaccines. Cancers (Basel) 2024;16:10.:335. doi: 10.3390/cancers16020335. [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Yang T, Luo W, Yu J, et al. Bladder cancer immune-related markers: diagnosis, surveillance, and prognosis. Front Immunol. 2024;15:1481296. doi: 10.3389/fimmu.2024.1481296. [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Lopez-Beltran A, Cookson MS, Guercio BJ, et al. Advances in diagnosis and treatment of bladder cancer. BMJ. 2024;384:e076743. doi: 10.1136/bmj-2023-076743. [DOI] [PubMed] [Google Scholar]
8.Nejman D, Livyatan I, Fuks G, et al. The human tumor microbiome is composed of tumor type-specific intracellular bacteria. Science . 2020;368:973–80. doi: 10.1126/science.aay9189. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Fu A, Yao B, Dong T, et al. Tumor-resident intracellular microbiota promotes metastatic colonization in breast cancer. Cell. 2022;185:1356–72. doi: 10.1016/j.cell.2022.02.027. [DOI] [PubMed] [Google Scholar]
10.Li S, Xia H, Wang Z, et al. Intratumoral microbial heterogeneity affected tumor immune microenvironment and determined clinical outcome of HBV-related HCC. Hepatology. 2023;78:1079–91. doi: 10.1097/HEP.0000000000000427. [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Chai X, Wang J, Li H, et al. Intratumor microbiome features reveal antitumor potentials of intrahepatic cholangiocarcinoma. Gut Microbes. 2023;15:2156255. doi: 10.1080/19490976.2022.2156255. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Gui F, Zhang L, Xiao J, et al. Decoding the role of intratumoral microbiota in gastric cancer. Biochim Biophys Acta Rev Cancer. 2025;1880:189355. doi: 10.1016/j.bbcan.2025.189355. [DOI] [PubMed] [Google Scholar]
13.Liu S, Pan Y, Zheng C, et al. Tumor-colonizing Pseudoalteromonas elyakovii metabolically reprograms the tumor microenvironment and promotes breast ductal carcinoma. MBio. 2025;16:e03873-24. doi: 10.1128/mbio.03873-24. [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Lou K, Chi J, Wu J, et al. Research progress on the microbiota in bladder cancer tumors. Front Cell Infect Microbiol. 2024;14:1374944. doi: 10.3389/fcimb.2024.1374944. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Pallares-Mendez R, Brassetti A, Bove AM, et al. Insights into the Interplay between the Urinary Microbiome and Bladder Cancer: A Comprehensive Review. J Clin Med. 2024;13:10.:4927. doi: 10.3390/jcm13164927. [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Surber J, Semmler M, Enderlin D, et al. Microbiome landscapes of the bladder, intestine, and vagina in bladder cancer: a systematic review. Discov Oncol. 2025;16:525. doi: 10.1007/s12672-025-02165-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Bourgi A, Bruyère F, Rusch E. Microbiome and immunotherapy in bladder cancer: The missing link. Fr J Urol. 2025;35:102922. doi: 10.1016/j.fjurol.2025.102922. [DOI] [PubMed] [Google Scholar]
18.Russo AE, Memon A, Ahmed S. Bladder Cancer and the Urinary Microbiome-New Insights and Future Directions: A Review. Clin Genitourin Cancer. 2024;22:434–44. doi: 10.1016/j.clgc.2023.12.015. [DOI] [PubMed] [Google Scholar]
19.Gherasim RD, Chibelean C, Porav-Hodade D, et al. Microbiome Shifts in Bladder Cancer: A Narrative Review of Urobiome Composition, Progression, and Therapeutic Impact. Medicina (Kaunas) 2025;61:10.:1401. doi: 10.3390/medicina61081401. [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Mager LF, Burkhard R, Pett N, et al. Microbiome-derived inosine modulates response to checkpoint inhibitor immunotherapy. Science . 2020;369:1481–9. doi: 10.1126/science.abc3421. [DOI] [PubMed] [Google Scholar]
21.Jiang S-S, Xie Y-L, Xiao X-Y, et al. Fusobacterium nucleatum-derived succinic acid induces tumor resistance to immunotherapy in colorectal cancer. Cell Host Microbe. 2023;31:781–97. doi: 10.1016/j.chom.2023.04.010. [DOI] [PubMed] [Google Scholar]
22.Kovtonyuk LV, McCoy KD. Microbial metabolites and immunotherapy: Basic rationale and clinical indications. Semin Immunol. 2023;67:101755. doi: 10.1016/j.smim.2023.101755. [DOI] [PubMed] [Google Scholar]
23.Lu S, Wang C, Ma J, et al. Metabolic mediators: microbial-derived metabolites as key regulators of anti-tumor immunity, immunotherapy, and chemotherapy. Front Immunol. 2024;15:1456030. doi: 10.3389/fimmu.2024.1456030. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Zheng D, Liwinski T, Elinav E. Interaction between microbiota and immunity in health and disease. Cell Res. 2020;30:492–506. doi: 10.1038/s41422-020-0332-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Martin-Gallausiaux C, Salesse L, Garcia-Weber D, et al. Fusobacterium nucleatum promotes inflammatory and anti-apoptotic responses in colorectal cancer cells via ADP-heptose release and ALPK1/TIFA axis activation. Gut Microbes. 2024;16:2295384. doi: 10.1080/19490976.2023.2295384. [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Li W, Shangguan W, Huang W, et al. Gut Parabacteroides distasonis ‐derived Indole‐3‐Acetic Acid Promotes Phospholipid Remodeling and Enhances Ferroptosis Sensitivity via the AhR‐FASN Axis in Bladder Cancer. Advanced Science. 2025;12:10. doi: 10.1002/advs.202504688. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Chen Z, Huang A, Sun J, et al. Inference of immune cell composition on the expression profiles of mouse tissue. Sci Rep. 2017;7:40508. doi: 10.1038/srep40508. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12:453–7. doi: 10.1038/nmeth.3337. [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Masone MC. Gut microbiome’s effect on immunotherapy response. Nat Rev Urol. 2024;21:63. doi: 10.1038/s41585-024-00854-2. [DOI] [PubMed] [Google Scholar]
30.Cao Y, Xia H, Tan X, et al. Intratumoural microbiota: a new frontier in cancer development and therapy. Signal Transduct Target Ther. 2024;9:15.:15. doi: 10.1038/s41392-023-01693-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Kyriazi AA, Karaglani M, Agelaki S, et al. Intratumoral Microbiome: Foe or Friend in Reshaping the Tumor Microenvironment Landscape? Cells. 2024;13:10.:1279. doi: 10.3390/cells13151279. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Che S, Yan Z, Feng Y, et al. Unveiling the intratumoral microbiota within cancer landscapes. iScience. 2024;27:109893. doi: 10.1016/j.isci.2024.109893. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Mantovani A, Marchesi F, Malesci A, et al. Tumour-associated macrophages as treatment targets in oncology. Nat Rev Clin Oncol. 2017;14:399–416. doi: 10.1038/nrclinonc.2016.217. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Cassetta L, Pollard JW. Targeting macrophages: therapeutic approaches in cancer. Nat Rev Drug Discov. 2018;17:887–904. doi: 10.1038/nrd.2018.169. [DOI] [PubMed] [Google Scholar]
35.DeNardo DG, Ruffell B. Macrophages as regulators of tumour immunity and immunotherapy. Nat Rev Immunol. 2019;19:369–82. doi: 10.1038/s41577-019-0127-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Smith DE, Clémençon B, Hediger MA. Proton-coupled oligopeptide transporter family SLC15: physiological, pharmacological and pathological implications. Mol Aspects Med. 2013;34:323–36. doi: 10.1016/j.mam.2012.11.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
37.César-Razquin A, Snijder B, Frappier-Brinton T, et al. A Call for Systematic Research on Solute Carriers. Cell. 2015;162:478–87. doi: 10.1016/j.cell.2015.07.022. [DOI] [PubMed] [Google Scholar]
38.Kwon SY, Thi-Thu Ngo H, Son J, et al. Exploiting bacteria for cancer immunotherapy. Nat Rev Clin Oncol. 2024;21:569–89. doi: 10.1038/s41571-024-00908-9. [DOI] [PubMed] [Google Scholar]
39.Erondu N, Gantz I, Musser B, et al. Neuropeptide Y5 receptor antagonism does not induce clinically meaningful weight loss in overweight and obese adults. Cell Metab. 2006;4:275–82. doi: 10.1016/j.cmet.2006.08.002. [DOI] [PubMed] [Google Scholar]
40.Geller LT, Barzily-Rokni M, Danino T, et al. Potential role of intratumor bacteria in mediating tumor resistance to the chemotherapeutic drug gemcitabine. Science . 2017;357:1156–60. doi: 10.1126/science.aah5043. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

online supplemental file 1

jitc-14-10-s001.docx^{(8MB, docx)}

DOI: 10.1136/jitc-2026-016574

online supplemental file 2

jitc-14-10-s002.tiff^{(9.2MB, tiff)}

DOI: 10.1136/jitc-2026-016574

Data Availability Statement

Data are available on reasonable request.

[R1] 1.Powles T, Valderrama BP, Gupta S, et al. Enfortumab Vedotin and Pembrolizumab in Untreated Advanced Urothelial Cancer. N Engl J Med. 2024;390:875–88. doi: 10.1056/NEJMoa2312117. [DOI] [PubMed] [Google Scholar]

[R2] 2.Roviello G, Santoni M, Sonpavde GP, et al. The evolving treatment landscape of metastatic urothelial cancer. Nat Rev Urol. 2024;21:580–92. doi: 10.1038/s41585-024-00872-0. [DOI] [PubMed] [Google Scholar]

[R3] 3.van der Heijden MS, Sonpavde G, Powles T, et al. Nivolumab plus Gemcitabine-Cisplatin in Advanced Urothelial Carcinoma. N Engl J Med. 2023;389:1778–89. doi: 10.1056/NEJMoa2309863. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R4] 4.Loriot Y, Matsubara N, Park SH, et al. Erdafitinib or Chemotherapy in Advanced or Metastatic Urothelial Carcinoma. N Engl J Med. 2023;389:1961–71. doi: 10.1056/NEJMoa2308849. [DOI] [PubMed] [Google Scholar]

[R5] 5.Patel DM, Mateen R, Qaddour N, et al. A Comprehensive Review of Immunotherapy Clinical Trials for Metastatic Urothelial Carcinoma: Immune Checkpoint Inhibitors Alone or in Combination, Novel Antibodies, Cellular Therapies, and Vaccines. Cancers (Basel) 2024;16:10.:335. doi: 10.3390/cancers16020335. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R6] 6.Yang T, Luo W, Yu J, et al. Bladder cancer immune-related markers: diagnosis, surveillance, and prognosis. Front Immunol. 2024;15:1481296. doi: 10.3389/fimmu.2024.1481296. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R7] 7.Lopez-Beltran A, Cookson MS, Guercio BJ, et al. Advances in diagnosis and treatment of bladder cancer. BMJ. 2024;384:e076743. doi: 10.1136/bmj-2023-076743. [DOI] [PubMed] [Google Scholar]

[R8] 8.Nejman D, Livyatan I, Fuks G, et al. The human tumor microbiome is composed of tumor type-specific intracellular bacteria. Science . 2020;368:973–80. doi: 10.1126/science.aay9189. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R9] 9.Fu A, Yao B, Dong T, et al. Tumor-resident intracellular microbiota promotes metastatic colonization in breast cancer. Cell. 2022;185:1356–72. doi: 10.1016/j.cell.2022.02.027. [DOI] [PubMed] [Google Scholar]

[R10] 10.Li S, Xia H, Wang Z, et al. Intratumoral microbial heterogeneity affected tumor immune microenvironment and determined clinical outcome of HBV-related HCC. Hepatology. 2023;78:1079–91. doi: 10.1097/HEP.0000000000000427. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R11] 11.Chai X, Wang J, Li H, et al. Intratumor microbiome features reveal antitumor potentials of intrahepatic cholangiocarcinoma. Gut Microbes. 2023;15:2156255. doi: 10.1080/19490976.2022.2156255. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R12] 12.Gui F, Zhang L, Xiao J, et al. Decoding the role of intratumoral microbiota in gastric cancer. Biochim Biophys Acta Rev Cancer. 2025;1880:189355. doi: 10.1016/j.bbcan.2025.189355. [DOI] [PubMed] [Google Scholar]

[R13] 13.Liu S, Pan Y, Zheng C, et al. Tumor-colonizing Pseudoalteromonas elyakovii metabolically reprograms the tumor microenvironment and promotes breast ductal carcinoma. MBio. 2025;16:e03873-24. doi: 10.1128/mbio.03873-24. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R14] 14.Lou K, Chi J, Wu J, et al. Research progress on the microbiota in bladder cancer tumors. Front Cell Infect Microbiol. 2024;14:1374944. doi: 10.3389/fcimb.2024.1374944. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R15] 15.Pallares-Mendez R, Brassetti A, Bove AM, et al. Insights into the Interplay between the Urinary Microbiome and Bladder Cancer: A Comprehensive Review. J Clin Med. 2024;13:10.:4927. doi: 10.3390/jcm13164927. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R16] 16.Surber J, Semmler M, Enderlin D, et al. Microbiome landscapes of the bladder, intestine, and vagina in bladder cancer: a systematic review. Discov Oncol. 2025;16:525. doi: 10.1007/s12672-025-02165-3. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R17] 17.Bourgi A, Bruyère F, Rusch E. Microbiome and immunotherapy in bladder cancer: The missing link. Fr J Urol. 2025;35:102922. doi: 10.1016/j.fjurol.2025.102922. [DOI] [PubMed] [Google Scholar]

[R18] 18.Russo AE, Memon A, Ahmed S. Bladder Cancer and the Urinary Microbiome-New Insights and Future Directions: A Review. Clin Genitourin Cancer. 2024;22:434–44. doi: 10.1016/j.clgc.2023.12.015. [DOI] [PubMed] [Google Scholar]

[R19] 19.Gherasim RD, Chibelean C, Porav-Hodade D, et al. Microbiome Shifts in Bladder Cancer: A Narrative Review of Urobiome Composition, Progression, and Therapeutic Impact. Medicina (Kaunas) 2025;61:10.:1401. doi: 10.3390/medicina61081401. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R20] 20.Mager LF, Burkhard R, Pett N, et al. Microbiome-derived inosine modulates response to checkpoint inhibitor immunotherapy. Science . 2020;369:1481–9. doi: 10.1126/science.abc3421. [DOI] [PubMed] [Google Scholar]

[R21] 21.Jiang S-S, Xie Y-L, Xiao X-Y, et al. Fusobacterium nucleatum-derived succinic acid induces tumor resistance to immunotherapy in colorectal cancer. Cell Host Microbe. 2023;31:781–97. doi: 10.1016/j.chom.2023.04.010. [DOI] [PubMed] [Google Scholar]

[R22] 22.Kovtonyuk LV, McCoy KD. Microbial metabolites and immunotherapy: Basic rationale and clinical indications. Semin Immunol. 2023;67:101755. doi: 10.1016/j.smim.2023.101755. [DOI] [PubMed] [Google Scholar]

[R23] 23.Lu S, Wang C, Ma J, et al. Metabolic mediators: microbial-derived metabolites as key regulators of anti-tumor immunity, immunotherapy, and chemotherapy. Front Immunol. 2024;15:1456030. doi: 10.3389/fimmu.2024.1456030. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R24] 24.Zheng D, Liwinski T, Elinav E. Interaction between microbiota and immunity in health and disease. Cell Res. 2020;30:492–506. doi: 10.1038/s41422-020-0332-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R25] 25.Martin-Gallausiaux C, Salesse L, Garcia-Weber D, et al. Fusobacterium nucleatum promotes inflammatory and anti-apoptotic responses in colorectal cancer cells via ADP-heptose release and ALPK1/TIFA axis activation. Gut Microbes. 2024;16:2295384. doi: 10.1080/19490976.2023.2295384. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R26] 26.Li W, Shangguan W, Huang W, et al. Gut Parabacteroides distasonis ‐derived Indole‐3‐Acetic Acid Promotes Phospholipid Remodeling and Enhances Ferroptosis Sensitivity via the AhR‐FASN Axis in Bladder Cancer. Advanced Science. 2025;12:10. doi: 10.1002/advs.202504688. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R27] 27.Chen Z, Huang A, Sun J, et al. Inference of immune cell composition on the expression profiles of mouse tissue. Sci Rep. 2017;7:40508. doi: 10.1038/srep40508. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R28] 28.Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12:453–7. doi: 10.1038/nmeth.3337. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] 29.Masone MC. Gut microbiome’s effect on immunotherapy response. Nat Rev Urol. 2024;21:63. doi: 10.1038/s41585-024-00854-2. [DOI] [PubMed] [Google Scholar]

[R30] 30.Cao Y, Xia H, Tan X, et al. Intratumoural microbiota: a new frontier in cancer development and therapy. Signal Transduct Target Ther. 2024;9:15.:15. doi: 10.1038/s41392-023-01693-0. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R31] 31.Kyriazi AA, Karaglani M, Agelaki S, et al. Intratumoral Microbiome: Foe or Friend in Reshaping the Tumor Microenvironment Landscape? Cells. 2024;13:10.:1279. doi: 10.3390/cells13151279. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R32] 32.Che S, Yan Z, Feng Y, et al. Unveiling the intratumoral microbiota within cancer landscapes. iScience. 2024;27:109893. doi: 10.1016/j.isci.2024.109893. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R33] 33.Mantovani A, Marchesi F, Malesci A, et al. Tumour-associated macrophages as treatment targets in oncology. Nat Rev Clin Oncol. 2017;14:399–416. doi: 10.1038/nrclinonc.2016.217. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R34] 34.Cassetta L, Pollard JW. Targeting macrophages: therapeutic approaches in cancer. Nat Rev Drug Discov. 2018;17:887–904. doi: 10.1038/nrd.2018.169. [DOI] [PubMed] [Google Scholar]

[R35] 35.DeNardo DG, Ruffell B. Macrophages as regulators of tumour immunity and immunotherapy. Nat Rev Immunol. 2019;19:369–82. doi: 10.1038/s41577-019-0127-6. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] 36.Smith DE, Clémençon B, Hediger MA. Proton-coupled oligopeptide transporter family SLC15: physiological, pharmacological and pathological implications. Mol Aspects Med. 2013;34:323–36. doi: 10.1016/j.mam.2012.11.003. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R37] 37.César-Razquin A, Snijder B, Frappier-Brinton T, et al. A Call for Systematic Research on Solute Carriers. Cell. 2015;162:478–87. doi: 10.1016/j.cell.2015.07.022. [DOI] [PubMed] [Google Scholar]

[R38] 38.Kwon SY, Thi-Thu Ngo H, Son J, et al. Exploiting bacteria for cancer immunotherapy. Nat Rev Clin Oncol. 2024;21:569–89. doi: 10.1038/s41571-024-00908-9. [DOI] [PubMed] [Google Scholar]

[R39] 39.Erondu N, Gantz I, Musser B, et al. Neuropeptide Y5 receptor antagonism does not induce clinically meaningful weight loss in overweight and obese adults. Cell Metab. 2006;4:275–82. doi: 10.1016/j.cmet.2006.08.002. [DOI] [PubMed] [Google Scholar]

[R40] 40.Geller LT, Barzily-Rokni M, Danino T, et al. Potential role of intratumor bacteria in mediating tumor resistance to the chemotherapeutic drug gemcitabine. Science . 2017;357:1156–60. doi: 10.1126/science.aah5043. [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Intratumoral Klebsiella pneumoniae drives metabolic-immune remodeling and immunotherapy resistance in bladder cancer

Ruilin Zhuang

Zhiming Wu

Runhao Zheng

Yuelong Chen

Xingliang Tan

Xinpei Deng

Chichen Zhang

Jiamin Zeng

Yanjun Wang

Haixia He

Yang Liu

Qianghua Zhou

Kai Yao

Abstract

WHAT IS ALREADY KNOWN ON THIS TOPIC

WHAT THIS STUDY ADDS

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

Introduction

Methods

Patients and samples

Cell lines and reagents

BLCA-derived organoid culture

2bRAD-M microbiome sequencing of FFPE tissues

Fluorescence in situ hybridization

Bacterial isolation, culture, and scanning electron microscopy

Bacterial coculture, proliferation, colony formation, migration, and EdU assays

Conditioned medium preparation and macrophage polarization assays

Animal experiments

Flow cytometry

Immunohistochemistry, multiplex immunofluorescence, and TUNEL staining

RNA extraction, RNA sequencing, and transcriptomic analysis

Cytokine measurement

Single-cell RNA-seq processing and microbial read analysis

Untargeted metabolomics and pathway analysis

Western blotting

Chromatin immunoprecipitation-qPCR

Virtual screening and molecular docking

Quantification and statistical analysis

Results

Intratumoral microbiota diversity in patients with BLCA

Single-cell transcriptomic analysis reveals the cell type-specific distribution of intratumoral microbes in BLCA

K. pneumoniae promotes BLCA progression and immunotherapy resistance

K. pneumoniae promotes immunosuppression by enhancing M2 polarization and impairing CD8 T-cell effector function

K. pneumoniae-associated metabolic reprogramming identifies Val-Phe as a candidate mediator of protumor activity

Val-Phe promotes BLCA progression and immunosuppressive microenvironment remodeling

Val-Phe activates NF-κB/p65 signaling and induces immunoregulatory programs through a mechanism involving SLC15A1

Small-molecule screening targeting SLC15A1 identifies MK-0557, which enhances anti-PD-1 efficacy

Discussion

Supplementary material

Footnotes

Data availability statement

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases