Multi-omics study on tumor-associated macrophages remodeling the tumor microenvironment via the CXCL5-CXCR2 axis to drive immune escape in bladder cancer

Abstract

Background

PD-1/L1 inhibitors improve the prognosis of patients with advanced bladder cancer, but the clinical remission rate remains below 25%. Tumor-associated macrophages (TAMs) and chemokines are critical in the tumor microenvironment (TME), affecting tumor progression, immunotherapeutic efficacy, and patient prognosis; however, their underlying mechanisms remain unclear. This study exhibits innovation by adopting a tumor microenvironment perspective to investigate the interaction mechanism between bladder cancer cells and tumor-associated macrophages, as well as factors affecting the efficacy of immunotherapy.

Methods

Single-cell sequencing, bulk sequencing, and in vivo experiments identified TAM infiltration characteristics, their impacts on prognosis and immunotherapy. In vitro, we established a co-culture model and performed targeted metabolomic sequencing on TAMs. Xenograft and tail vein metastasis models were used to investigate the function of CXCL5-CXCR2 axis in bladder TME.

Results

M2 macrophages were positively correlated with the clinical staging of bladder cancer and resistance to immunotherapy. Single-cell sequencing data revealed that CXCL5⁺ tumor-associated macrophages (TAMs) were associated with poor overall survival but a favorable response to immunotherapy, whereas FOLR2⁺ TAMs were linked to both poor overall survival and immunotherapy resistance. The CXCL5-CXCR2-NF-κB axis was upregulated in the co-culture system, which promoted PD-L1 expression in both tumor cells and TAMs, the formation of an immunosuppressive tumor microenvironment (TME), as well as the migration, proliferation, and lung metastatic potential of bladder cancer cells. Additionally, this axis enhanced IDO1 expression in macrophages and improved the efficacy of immunotherapy for bladder cancer.

Conclusion

The CXCL5-CXCR2 axis mediates bladder cancer cell-macrophage crosstalk: macrophages promote tumor growth, immune escape, and cisplatin tolerance; tumor cells induce macrophage polarization and reshape immunosuppressive TME. Additionally, this axis drives bladder cancer malignant progression and enhances immunotherapy efficacy.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00262-026-04353-8.

Introduction

The incidence rate of bladder cancer is 9.5 per 100,000 person-years for men and 2.4 for women [1]. Non-muscle-invasive and muscle-invasive bladder cancer are the two primary pathological types. Muscle-invasive bladder cancer is characterized by a high recurrence rate and poor prognosis [2]. PD-1/L1 inhibitors have been demonstrated to improve the prognosis of patients with advanced bladder cancer. Targeting PD-1/L1 via immunotherapy undoubtedly offers new hope to these patients [3]. However, the clinical remission rate with immunotherapy is less than 25% [4], and the mechanism of resistance to immunotherapy remains poorly understood. The expression of PD-L1 is a key factor influencing the efficacy of immunotherapy. Nonetheless, studies have indicated that patients with low PD-L1 expression can also respond to immunotherapy, making it controversial whether PD-L1 expression levels influence the efficacy of immunotherapy [5, 6].

An accumulating body of studies has reported that tumor-associated macrophages (TAMs) play a significant role in the tumor microenvironment (TME) [7–9]. Specifically, it has been demonstrated that TAMs can reshape the immunosuppressive microenvironment, thereby promoting the malignant progression of tumors [10]. TAMs are traditionally classified into two types: M1-type macrophages, which secrete pro-inflammatory cytokines and inhibit tumor progression, and M2-type macrophages, which secrete anti-inflammatory cytokines and facilitate tumor progression [11]. However, the traditional classification of macrophages is insufficient to account for the complex physiological functions and their diversity [12]. With the development of single-cell sequencing technology [13], various types of TAMs have been discovered. Many studies indicate that the expression of PD-L1 on immune cells is positively correlated with the response to immunotherapy [14, 15].

Chemokines are a group of low molecular weight (8–12 kDa) cytokines. These molecules are categorized into C, CC, CXC, and CX3C types based on the number and spatial arrangement of amino-terminal cysteine residues. Corresponding receptors are also divided into four classes: CR, CCR, CXCR, and CX3CR [16]. An accumulating body of studies has shown that chemokines play a significant role in tumor progression, serving as a "bridge" connecting tumor cells with immune cells [17]. Chemokines of the CXC family are currently the focus of investigations, but it remains unclear how CXC chemokines in the TME regulate the immune microenvironment [18].

Our study aimed to investigate the potential mechanisms underlying the interaction between bladder cancer cells and tumor-associated macrophages (TAMs), integrating traditional macrophage classification methods with those derived from single-cell sequencing analyses. Concurrently, our research focused on elucidating the role of chemokines as mediators that influence bladder cancer progression and the effectiveness of immunotherapeutic interventions.

Methods

Bioinformatic analysis

Data collection and processing

In order to explore the tumor microenvironment of bladder cancer, single-cell sequencing data containing 11 bladder cancer samples were obtained from a study conducted by Chen et al. [19]. We downloaded the raw data (accession number: HRA000212) from the Genome Sequence Archive for Human (https://ngdc.cncb.ac.cn/gsa-human). Furthermore, we collected four bladder cancer samples that had undergone tislelizumab treatment to conduct single-cell sequencing. Two of these samples were from Shanghai Renji Hospital, while the other two were from the First Affiliated Hospital of Xi’an Jiaotong University. Spatial transcriptome data were downloaded from the Gene Expression Omnibus (accession number: GSE285715), which included three pairs of para-carcinoma and carcinoma tissues.

Four bladder cancer gene expression datasets and clinical data were downloaded from the Gene Expression Omnibus website (https://www.ncbi.nlm.nih.gov/geo/), The Cancer Genome Atlas (TCGA) website (https://www.portal.gdc.cancer.gov), and IMvigor210 cohort website (http://research-pub.gene.com/IMvigor210CoreBiologies/), including GSE13507, GSE128959, TCGA-BLCA, and IMvigor210 data [20]. The expression of CXCL5 in various cancers was visualized using the Timer database (http://timer.cistrome.org/) [21]. To explore the impact of immunotherapy on bladder cancer patients, RNA-sequencing data from 65 bladder cancer patients who received gemcitabine-cisplatin and tislelizumab treatment were obtained from the Second Affiliated Hospital of Sun Yat-sen University [22]. Furthermore, to explore the mechanism underlying the interaction between bladder cancer cells and macrophages, we performed RNA sequencing on the control and co-culture groups. We also performed targeted metabolomics sequencing on co-cultured macrophages. We also performed RNA sequencing on C57 mice treated with immunotherapy.

First, scRNA-seq raw data containing “fastq” files were input into Cell Ranger v7.0 software to obtain three key files, including the single-cell gene expression matrix. The Harmony R package was used to merge the single-cell sequencing data [23, 24]. To control data quality, we filtered out cells with fewer than 1000 or more than 6000 expressed genes, as well as cells with more than 10% mitochondrial genes. Simultaneously, we performed dimensionality reduction via principal component analysis (PCA) and clustered the cells using the Clustree R package [25]. Next, we generated a UMAP plot using 40 PCs, with the resolution set to 0.6. Cell annotation was conducted using the CellMarker tool (http://117.50.127.228/CellMarker/) [26]. To explore TAMs in the bladder cancer tumor microenvironment (TME), we applied the aforementioned method to analyze myeloid cells and identified three distinct TAM subtypes. For spatial transcriptome data, we used the "Seurat" R package to read and analyze the data. We filtered out cells with fewer than 1000 or more than 8000 expressed genes, as well as cells with more than 10% mitochondrial genes to control data quality. Second, RNA-sequencing and metabolomics sequencing data were standardized for subsequent analysis.

Bioinformatic analysis based on RNA-sequence

To estimate the proportion of infiltrating immune cells, especially M2 macrophages, using the datasets previously described herein, we employed the CIBERSORT algorithm [27]. Results with a p-value > 0.05 were excluded. The “ggplot2” R package was used to visualize the results of immune cell infiltration [28]. In addition, the “DESeq2” R package was applied to identify differentially expressed genes (DEGs) between the control and co-culture groups, using the criteria of padj < 0.05 and |log2fold-change|> 1.0 [29]. Then, the top 50 upregulated DEGs were defined as the TAM gene signature. The “GSVA” R package was used to assess the gene signature enrichment of co-cultured macrophages within the immunotherapy cohort. The “clusterProfiler” R package was used to identify activated pathways in co-cultured macrophages, and the top 20 significant pathways were visualized as a bubble plot [30].

Bioinformatic analysis based on scRNA-sequence and spatial transcriptome data

The “Seurat” R package was employed to analyze single-cell sequencing data. Different cell types were annotated within the bladder cancer tumor microenvironment (TME). Myeloid cells were further subdivided into tumor-associated macrophages (TAMs), dendritic cells, and monocytes. We identified three major TAM subtypes in the bladder cancer TME and five major TAM subtypes in the TME following immunotherapy. We explored the biological functions of these three TAM subtypes. The “ClusterProfiler” R package was used to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses, based on the top 50 expressed genes in different TAM types. In addition, we estimated the similarities between M1/M2 macrophages and the three TAM subtypes, and explored their roles in angiogenesis and phagocytosis using the “Addmodulescore” function in the Seurat R package [31]. The “Addmodulescore” function was also utilized to investigate the spatial distribution of FOLR2⁺ TAMs in bladder cancer and adjacent normal tissues. The “Monocle3” R package was applied to perform single-cell trajectory analysis for the different TAM subtypes [32]. To elucidate the mechanism underlying TAM activation, we used the pySCENIC method in Python 3.8 to identify activated transcription factors [33]. To explore the clinical relevance of the three TAM subtypes, the “Scissor” algorithm and “Survival” R package were used to predict patients’ prognosis [34]. We integrated single-cell data with bulk RNA-seq data and applied gene set variation analysis (GSVA) to estimate scores of different TAM-related gene sets in bladder cancer patients [35]. Patients were stratified into high- and low-TAM groups based on the median GSVA scores. The “scTenifoldKnk” R package was used to perform virtual gene knockout in CXCL5⁺ TAMs.

Experimental validation

Chemicals and reagents

Phorbol 12-myristate 13-acetate (PMA; catalog no: ab120297) was purchased from Abcam (Cambridge, UK). PMA powder was dissolved in dimethyl sulfoxide (DMSO), and the tetrazolium compound 3-[4,5-dimethyl-2-thiazolyl]-2,5-diphenyl-2H-tetrazolium bromide (MTT) was obtained from Sigma–Aldrich (St. Louis, MO, USA). Clodronate liposomes (catalog no.: 40337ES) were purchased from Yeason (Shanghai, China). CD3/CD28 T cell activator was purchased from STEMCELL Technologies (Vancouver, BC, Canada). Transwell chambers (8 μm pore size) and co-culture chambers (0.4 μm pore size) were purchased from Millipore (Burlington, MA, USA). SB225002 (CXCR2 inhibitor) and PDTC (NF-κB inhibitor) were purchased from Selleck (Texas, USA).

Cell lines and cell culture

Human bladder cancer cell lines (T24, 5637, 253 J, TCCSUP, J82, RT4), MB49 (a mouse bladder cancer cell line), 293 T cells, SV-HUC-1 (a normal urinary epithelial cell line), THP-1 (CL-0233) human macrophage cell line, and Jurkat T cells (CL-0129) were purchased from Procell Life Science and Technology (Wuhan, China). T24L cells (stably expressing luciferase) were obtained from previous studies. T24 cells, 293 T cells, and MB49 cells were cultured in DMEM, while the other cell lines were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS; Biological Industries, USA). All cell lines were cultured at 37 °C in a humidified cell incubator with 5% CO₂. PMA (150 nM) was used to differentiate THP-1 cells into M0 macrophages for 48 h [36]. The CD3/CD28 T cell activator was used to activate Jurkat T cells for 48 h.

Protein extraction and western blotting

After 48 h of co-culture, M0 macrophages, M0–T24 macrophages, and other bladder cancer cell lines were washed three times with ice-cold PBS. Then, radioimmunoprecipitation assay (RIPA) buffer was added to the 6-well plates. After incubation on ice for 10 min, cells were scraped with a cell scraper, and the cell lysates were ultracentrifuged at 15,000 rpm for 15 min at 4℃. Finally, proteins in the supernatant were extracted into new 1.5 mL tubes. Then, the BCA protein quantification assay was used to determine the concentration of the protein extracts. A microplate reader (Bio-Rad, Hercules, CA, USA) was used to detect the optical density at 562 nm. A volume containing 20 μg of protein was determined as the appropriate amount for loading. Next, 12.5% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) was performed. The proteins were subsequently transferred from the gel onto a PVDF membrane for 2 h. Then, 5% skim milk was used to block non-specific protein-binding sites for 1 h at room temperature. The PVDF membranes were cut according to the molecular weights of the target proteins. Primary antibodies (Table 1) were then incubated with the PVDF membranes for 24 h at 4℃. Subsequently, the PVDF membranes were washed three times with TBST. Secondary antibodies were then incubated with the PVDF membranes for 1 h at RT. Finally, protein bands were visualized using an electrogenerated chemiluminescent (ECL) detection system (Bio-Rad Laboratories, CA, USA).

Table 1.

Primary antibodies used in western blot

Antibody	Species	Company	Dilution rate
CXCL5	Rabbit polyclonal	Proteintech	1:1000
CXCR2	Rabbit monoclonal	Proteintech	1:1000
PD-L1	Rabbit polyclonal	Proteintech	1:1000
P-P65	Rabbit monoclonal	Proteintech	1:1000
P65	Rabbit polyclonal	Proteintech	1:1000
GAPDH	Rabbit monoclonal	CST	1:1000
β-actin	Mouse monoclonal	CST	1:1000
Vinculin	Rabbit monoclonal	CST	1:1000
CD163	Rabbit monoclonal	Abcam	1:1000
CD68	Rabbit monoclonal	Servicebio	1:1000
SPP1	Rabbit monoclonal	Servicebio	1:200
IL1B	Rabbit monoclonal	Servicebio	1:200
FOLR2	Rabbit monoclonal	Abcam	1:1000
CD206	Rabbit monoclonal	HUABIO	1:1000
F4/80	Rabbit monoclonal	Servicebio	1:200
PD-1	Rabbit monoclonal	CST	1:1000
Secondary antibodies	Goat anti-rabbit IgG	Proteintech SA00001-2	1:2000
Secondary antibodies	Goat anti-mouse IgG	Proteintech SA00001-1	1:2000

RNA extraction and RT-qPCR

We used RNAfast 200 Reagent (Feijie Biotechnology, Shanghai, China) to extract total RNA from the cells in our research. Briefly, cells in 6-well plates were washed three times with PBS and then treated with lysis buffer. Total RNA was extracted into 1.5 mL tubes through multiple centrifugation steps, and its concentration was measured. PrimeScript RT-PCR Kit (Takara Biotechnology Co., Ltd., Dalian, China) was used to generate complementary DNA (cDNA). Thereafter, SYBR Green PCR Master Mix (Genestar, Beijing, China) was added to amplify cDNA using a CFX96 Real-Time PCR system (Bio-Rad). The primer sequences used were shown in Table 2.

Table 2.

Primer sequences used in real-time PCR

GENE	Forward primer	Reverse primer
18S	GCAATTATTCCCCATGAACG	GGCCTCACTAAACCATCCAA
CXCL1	TGCTGCCACTAATGCTGATGT	CTCAGGAACCAATCTTTGCACT
CXCL2	TCCAAGAAAGGGCGAAATAAGG	TGCAGCTCTATCTGAATGTCTGT
CXCL3	CGCCCAAACCGAAGTCATAG	GCTCCCCTTGTTCAGTATCTTTT
CXCL4	CTGAAGAAGATGGGGACCTG	AGAGCCACTAACACGTAGCCT
CXCL5	AGCTGCGTTGCGTTTGTTTAC	TGGCGAACACTTGCAGATTAC
CXCL6	AGAGCTGCGTTGCACTTGTT	GCAGTTTACCAATCGTTTTGGGG
CXCL7	GTAACAGTGCGAGACCACTTC	CTTTGCCTTTCGCCAAGTTTC
CXCL8	ACTGAGAGTGATTGAGAGTGGAC	AACCCTCTGCACCCAGTTTTC
CXCL9	CCAGTAGTGAGAAAGGGTCGC	AGGGCTTGGGGCAAATTGTT
CXCL10	GTGGCATTCAAGGAGTACCTC	TGATGGCCTTCGATTCTGGATT
CXCL11	GACGCTGTCTTTGCATAGGC	GGATTTAGGCATCGTTGTCCTTT
CXCL12	ATTCTCAACACTCCAAACTGTGC	ACTTTAGCTTCGGGTCAATGC
CXCL13	GCTTGAGGTGTAGATGTGTCC	CCCACGGGGCAAGATTTGAA
CXCL14	CGCTACAGCGACGTGAAGAA	GTTCCAGGCGTTGTACCAC
CXCR2	CCTGTCTTACTTTTCCGAAGGAC	TTGCTGTATTGTTGCCCATGT
PD-L1	TGGCATTTGCTGAACGCATTT	TGCAGCCAGGTCTAATTGTTTT

Immunohistochemistry and immunofluorescence

In order to validate the results of the bioinformatics analysis, we obtained 12 bladder cancer tissues with different clinical stages from the First Affiliated Hospital of Xi’an Jiaotong University. Primary antibodies against CD68, CD163, and FOLR2 were used for immunohistochemistry and immunofluorescence. Briefly, all tissues were deparaffinized, dehydrated through a graded ethanol series, subjected to antigen retrieval, and blocked with BSA (bovine serum albumin). Subsequently, the samples were incubated with the primary antibodies overnight. For immunohistochemistry, the samples were incubated with secondary antibodies for 30 min, followed by microscopic observation of the images. For immunofluorescence, the samples were treated with fluorescent secondary antibodies for 50 min and then with DAPI for 5 min in the dark. Image capture was performed using a fluorescence microscope.

Flow cytometric analysis

To observe the drug resistance capacity of 5637 cells co-cultured with M0 macrophages. We performed flow cytometry to detect the apoptosis rate in cisplatin-treated tumor cells. Briefly, cisplatin (5 μg/mL) was added to 5637 cells co-cultured with M0 macrophages. After 24 h, we collected the tumor cells and culture media. Then, the cell suspension was centrifuged at 1500 rpm for 10 min. We washed the cells twice with PBS. 7AAD and PI were added to the cell suspension for 30 min in the dark. Finally, we analyzed the samples using a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA, USA). The apoptosis data were inputted into FlowJo software. All experiments were performed in triplicate.

Establishment of co-culture model

Transwell chambers (PET, 0.4 μm, 6-well, Millipore) were employed for the co-culture assay. To simulate the interaction between bladder cancer cells and macrophages, THP-1 cells were differentiated into M0 macrophages and cultured in the lower chamber for 48 h. The tumor cell line was pre-seeded in the upper chamber. Subsequently, the two cell lines were co-incubated for non-contact co-culture. Following 48 h of co-culture, morphological changes in the two cell lines were observed using a light microscope. Culture of the two cell groups was then continued for an additional 24 h, and the conditioned medium was collected.

To investigate whether the macrophages could promote tumor immune escape, Jurkat T cells were mixed with tumor cells at a ratio of 6:1 for contact co-culture. At 24 h, 48 h, and 72 h post-co-culture, morphological changes in the two cell lines were observed using a light microscope.

Lasmid transfection and lentiviral infection

The shRNA constructs with the PLKO.1 vector were generated to target hCXCL5 (human) and mCXCL5 (mouse). Following the transformation and amplification of the plasmid, 293 T cells were used for viral packaging. The collected viral solution was then applied to transfect the tumor cells, and puromycin was used to screen for cells that had been successfully transfected.

MTT assays

To determine whether CXCL5 expression affects bladder cancer cell growth, we performed MTT assays, in which we compared the 5637 and 5637 shn-CXCL5 cell lines. Furthermore, we mixed tumor cells with activated Jurkat T cells to simulate the interaction between bladder cancer cells and Jurkat T cells. Briefly, both cell lines were seeded in 96-well plates at a density of 4000 cells per well in 200 μL of RPMI 1640 medium. Then, we employed a microscope to observe the morphological characteristics of the two cell lines every 24 h. MTT (5 mg/mL) was added to the 96-well plates and incubated for 24 h. Following a 4 h reaction period, the MTT solution was aspirated, and DMSO was added to each well to solubilize the formazan precipitates. Finally, a microplate reader (Bio-Rad, Hercules, CA, USA) was employed to measure the absorbance at 490 nm.

Cell colony formation assay

The 5637 and 5637 shn-CXCL5 cell lines were seeded in 6-well plates at a density of 1000 cells per well. Morphological characteristics were observed using a microscope every 3 days. Subsequently, a gradual increase in the number of cancer cell clones was observed. After 10 days, cell colonies were fixed with 4% paraformaldehyde. Ten minutes later, 1 × crystal violet solution was added to the 6-well plates for staining of the cell colonies. Washing was performed with PBS solution until the cell colonies were visible. The experiment was conducted in triplicate.

Wound healing assay

The 5637 and 5637 shn CXCL5 cell lines were seeded in 6-well plates at a density of 30 × 10⁴ cells per well. After 48 h, we observed that the cells reached approximately 100% confluency. Then, a 200 μL pipette tip was used to make scratches. Next, we observed wound healing using an inverted microscope at ×200magnification. The experiment was performed in triplicate.

Transwell assays

To assess migration of the two cell lines, we performed a transwell assay. Briefly, 4 × 10⁴ 5637 and 5637 shn-CXCL5 cells were seeded into the upper chamber (PET, 8 μm pore size, Millipore) with 200 μL of serum-free medium. Subsequently, 800 μL of complete medium was added to the lower chamber. Following a 24 h incubation, the lower chamber was fixed with 4% paraformaldehyde for 15 min. Next, 1× crystal violet solution was used to stain the migrated cells in the lower chamber. The experiment was performed in triplicate.

ELSIA assay

The CXCL5 ELISA reagent kit was purchased from Proteintech (Wuhan, China). We used this kit to detect the concentration changes of CXCL5 in the co-culture system. The concentrations of IL2, IFN-γ, IL4, and IL6 were determined following the standard protocols of the ELISA kit from Abmart. These kits were used to investigate whether co-cultured tumor cells and macrophages inhibited T cell function.

Xenograft tumor model and tail vein cancer metastasis model

Nude mice and C57 mice used in this study were all approved by the First Affiliated Hospital of Xi’an Jiaotong University. To explore whether macrophages promote tumor cell growth, 5637 cells and M0 macrophages were mixed in a ratio of 2:1, and then inoculated subcutaneously in nude mice. Furthermore, 1 × 10⁶ MB49 bladder cancer cells were inoculated subcutaneously in nude mice. Clodronate liposomes were injected into C57 mice to eliminate macrophages. To investigate whether CXCL5 expression affects tumor cell growth, 5 × 10⁶ 5637shnNC and 5637shnCXCL5 cells were also inoculated subcutaneously in nude mice. Additionally, 1 × 10⁶ MB49 bladder cancer cells were inoculated subcutaneously in C57 mice. A mouse lung metastasis model was established by tail vein injection of T24LL and T24LL-shnCXCL5 at a dose of 1 × 10⁶ cells per mouse. After 30 days, mice were anesthetized with isoflurane and intraperitoneally injected with D-luciferin (150 mg/kg). Bioluminescence imaging (BLI) was then employed to monitor lung metastasis progression. A subcutaneous tumor model in C57 mice was used to compare the influence of CXCL5 on immunotherapy. A PD-1 monoclonal antibody was administered intraperitoneally to mice at a dose of 100 µg per mouse, and a CXCL5 recombinant protein was also administered intraperitoneally at a dose of 0.02 µg/g. Immunohistochemical experiments were conducted using subcutaneous tumors and lung metastases after mouse euthanasia.

Statistical analysis

R software (Version 4.0) and GraphPad Prism (Version 8.0) were employed for statistical analyses. The Wilcoxon test was used to compare continuous variables between two groups, while the Spearman’s rank correlation coefficient was applied for correlation analysis. A Kaplan–Meier curve was utilized to depict the survival outcomes of bladder cancer patients. Statistical significance was defined as P < 0.05.

Results

The infiltration characteristics, molecular functions and clinical significance of TAM in bladder cancer patients based on bulk-seq and ScRNA-seq

CIBERSORT method was used to estimate the proportion of infiltrated immune cells in the TCGA-BLCA database (Fig. S1A). According to the results of CIBERSORT analysis, we found that a large number of M0 and M2 macrophages infiltrated the bladder cancer microenvironment (Fig. 1A). We detected the molecular markers of M0 (CD68) and M2 (CD163) macrophages via immunohistochemical experiments. The results suggested that M0 and M2 macrophage infiltration increased significantly as the T stage and pathological grade of bladder cancer advanced (Fig. 1B-D). However, the above analysis results were based on the traditional classification of macrophages, which is insufficient to explain the heterogeneity and diversity of macrophages. Furthermore, we explored the bladder microenvironment using single-cell data from bladder cancer samples containing eight bladder cancer and three paracancerous tissue samples. The baseline characteristics and tumor-related information of all patients are shown in Supplementary Fig. 1B. After quality control, the single-cell dataset contained 77,926 cells and 33,813 genes (Fig. S1C, D). Next, single cells were divided into nine cell clusters (epithelial, T, endothelial, myofibroblastic cancer-associated fibroblast, inflammatory cancer-associated fibroblast, myeloid, B, mast, and schwann cells) using common cell markers (EPCAM, CD3D, PECAM1, RGS5, PDGFRA, LYZ, CD79A, TPSAB1, and MPZ) (Fig. S1E–G). We focused on myeloid cells: 3,899 myeloid cells were divided into 13 cell clusters based on the common cell markers CD68, MRC1, CD1C, TOP2A, LAMP3, CLEC9A, and CELC4C (Fig. 1E). Among myeloid cells, TAMs were defined as cell clusters expressing the markers APOE, CD68, and CD163. Furthermore, we identified three TAM subtypes with distinct gene expression profiles: TAM1 cells were characterized by high expression of SPP1 and TREM2; TAM2 cells by high expression of FOLR2 and SELENOP; and TAM3 cells by high expression of IL1B and NFKB1 (Fig. 1F).

Fig. 1.

Identification the characteristics of TAM in bladder cancer microenvironment based on bulk-sequence and scRNA-sequence. A The proportion of three type of macrophage infiltrations in the bladder cancer microenvironment. B The characteristics of TAM in bladder cancer TME according to pathological grade. C The TAM marker of immunohistochemical staining in bladder cancer tissue samples according to different clinical stages. D The IHC-score of CD68, CD163 in different stage of bladder cancer tissue. E The UMAP plot of myeloids based on singe cell sequence data in HRA000212. F Six genes (SPP1, TREM2, FOLR2, SELENOP, IL1B, NFKB1) expression in three subtype of TAMs. G The score of M1, M2, phagocytosis, and angiogenesis in three TAMs. H Cell trajectory and pseudo-temporal analysis of myeloid in bladder cancer microenvironment. I Detection of macrophage subpopulations with clinical prognostic significance based on Scissor algorithm. J Survival analysis of FOLR2 TAMs in TCGA-BLCA. K Multiple immunofluorescence of FOLR2 TAM in bladder cancer TME. L The infiltration differences of FOLR2 TAMs in the public bladder cancer spatial transcriptome data GSE285715. **** P < 0.0001. Scale bars, 30 μm

To identify the biological functions of the three TAM subtypes, we compared their angiogenic potential, phagocytic activity, and secreted chemokines. Furthermore, SPP1⁺ TAMs had stronger angiogenic function than the other two TAM subtypes. FOLR2⁺ TAMs exhibited the strongest phagocytic activity when compared to the other two TAM subtypes (Fig. 1G). Additionally, we found that among the three subtypes, IL1B⁺ TAMs and FOLR2⁺ TAMs had the highest M1 and M2 macrophage scores, respectively (Fig. 1G). We performed single-cell trajectory analysis to explore the differentiation trajectory of the three TAM subtypes. The results showed that all TAM subtypes originated from monocytes in the bladder cancer microenvironment. Furthermore, we speculated that IL1B⁺ TAMs appeared earlier than the other two TAM subtypes, which seemed to arise in a more gradual manner (Fig. 1H).

GSVA results based on TCGA, GSE13507, and IMvigor210 datasets showed that SPP1⁺ TAMs had a higher propensity for infiltration into bladder tumor tissues than the other two TAM subtypes (Fig. S2A-C). In addition, KEGG pathway analysis suggested a positive correlation between IL1B⁺ TAMs and the NF-κB signaling pathway (Fig. S2D). Next, we analyzed the specific transcription factors expressed in the three TAM subtypes, which showed that ZNF850 and CEBPG were activated in SPP1⁺ TAMs, while SMAD5, NFBK1, and STAT3 were activated in FOLR2⁺ TAMs and IL1B⁺ TAMs (Fig. S2E).

Interestingly, we found that the FOLR2⁺ TAM score was positively correlated with the overall survival and cancer-specific survival of bladder cancer patients, which was also validated using the Scissors algorithm (Fig. 1I, J). We also discovered that the score of the FOLR2⁺ TAM subtype increased with tumor pathological grade and clinical stage (Fig. S2F). To validate the presence of FOLR2⁺ TAMs in bladder cancer tissues, we performed immunofluorescence staining, spatial transcriptomic data analysis, and immunohistochemical assay. We concluded that CD68 was co-localized with FOLR2 (Fig. 1K). Spatial transcriptome data indicated that FOLR2⁺ TAMs were mainly distributed in paracancerous tissues (Fig. 1L). Immunohistochemical results showed that FOLR2 might be mainly expressed in the tumor stroma (Fig. S2G). The proportion of CD68⁺FOLR2⁺ cells increased with tumor clinical stage, which was consistent with the results of bioinformatics analysis (Fig. S2H).

The infiltration characteristics, efficacy prediction of TAM in bladder cancer patients who have received immunotherapy based on bulk-seq and ScRNA-seq

Based on the IMvigor210 cohort (comprising patients with metastatic urothelial cancer treated with immunotherapy), the proportion of infiltrating immune cells was further analyzed (Fig. S2I). Patients with a complete response (CR) or partial response (PR) were found to have a lower proportion of M2 macrophages and a higher proportion of M1 macrophages (Fig. 2A, B). To validate the above results, RNA sequencing was performed on C57 mice treated with immunotherapy, and CIBERSORT was then used to estimate the proportion of infiltrated immune cells in C57 subcutaneous tumors (Fig. S2J). These data suggested that immunotherapy could suppress tumor growth, with a decreased proportion of M2 macrophages observed after immunotherapy administration (Fig. 2C–F). RNA sequencing data from the Second Affiliated Hospital of Sun Yat-sen University (containing 65 patients with bladder cancer who had received gemcitabine-cisplatin and tislelizumab treatment) were also utilized (Fig. S2K). CIBERSORT results indicated an increased proportion of M1 macrophages in patients with pathological complete response (PCR; Fig. 2G). Collectively, these results suggested that tumor-associated macrophages (TAMs) may exert a potential effect on immunotherapy.​

Fig. 2.

Identification the characteristics of TAM in bladder cancer patients who were received immunotherapy based on bulk-sequence and scRNA-sequence. A The proportion of M1 macrophage in patients with different response of immunotherapy based on IMvigor210 database. B The proportion of M2 macrophage in patients with different response of immunotherapy based on IMvigor210 database. C The flowchart of building C57 immunotherapy model. D The image of subcutaneous tumor. E The tumor weight of different group. F The proportion of M2 macrophage infiltrations in subcutaneous tumor. G The proportion of M1 macrophage in patients with different response of immunotherapy based on immunotherapy cohort from the Second Affiliated Hospital of Sun Yat-sen University. H The UMAP plot of single cell data from bladder cancer patients who were received immunotherapy. I The volcano plot of differential genes in IMvigor210 cohort. J The top 50 differential genes expressed in single cell data. K The UMAP plot of myeloids in bladder cancer ptients who were received immunotherapy. L The marker genes expressed in different type of myeloids cells. M Different type of myeloids cells infiltrations in the bladder cancer patients who were received immunotherapy. N The immunofluorescence image of CXCL5,CD68 in bladder cancer tissue. **** P < 0.0001. Scale bars, 50 μm

Changes in the bladder cancer microenvironment following immunotherapy were further explored. Bladder cancer samples were collected, including 3 cases from patients who had received immunotherapy and 1 case from a patient with no treatment, for single-cell sequencing (Fig. S3A, B). Baseline and tumor information of the four patients is shown in Table 3. Subsequently, single cells were divided into ten cell clusters, including epithelial cells, T cells, endothelial cells, myofibroblastic cells, cancer-associated fibroblasts (CAFs), inflammatory CAFs, myeloid cells, B cells, mast cells, and neutrophils (Fig. 2H, Fig. S3C). Next, the “DESeq2” R package was used to identify differentially expressed genes (DEGs) between the Response and Non-Response groups based on the IMvigor210 cohort (Fig. 2I). The top 50 upregulated DEGs were then defined as a gene signature. The AUCell R package was employed to investigate the expression of this gene signature in single-cell data. Results showed that, compared with non-responder patients, the DEGs in responder patients were mainly concentrated in macrophages and T cells (Fig. 2J). Therefore, the focus was primarily on myeloid cells, which were divided into 8 cell clusters (Fig. 2K). Five TAM subsets were defined based on specific gene expression: TOP2A⁺ TAMs, SPP1⁺ TAMs, CXCL5⁺ TAMs, SERPINH1⁺ TAMs, and FOLR2⁺ TAMs (Fig. 2L). CXCL5⁺ TAMs were identified in the bladder TME, characterized by high expression of CXCL5 and INHBA. Additionally, CXCL5⁺ TAMs were mainly infiltrated in patients with a good response to immunotherapy, whereas FOLR2⁺ TAMs were primarily infiltrated in non-responder patients (Fig. 2M). Immunofluorescence results showed that CD68 was co-localized with CXCL5 (Fig. 2N). Furthermore, we utilized spatial transcriptomic data to demonstrate that CXCL5⁺ TAMs were predominantly infiltrated in bladder cancer tissues (Fig. S3D, E). Moreover, using the transcriptomic data, we found that the infiltration level of CXCL5⁺ TAMs increased with the progression of bladder cancer stage and grade (Fig. S3F). Single-cell trajectory analysis suggested that SPP1⁺ TAMs could be differentiated into CXCL5⁺ TAMs (Fig. S3G).

Table 3.

The clinical data on 4 bladder cancer paitents who received immunotherapy

ID	Sex	Age	Treatment	Grade	T stage	Source	Response
Untreated	M	64	NC	High	T4	Shanghai	Control
Sample1	F	74	Tislelizumab	High	T3	Shanghai	CR
Sample2	M	77	Tislelizumab	High	T3	Xi’an	SD
Sample3	M	76	Tislelizumab	High	T3	Xi’an	PR

Macrophages promote the proliferation of bladder cancer cells, resistance to cisplatin and immune escape.

To explore whether macrophages could affect tumor progression, we mixed THP-1-derived macrophages and bladder tumor cells to construct a subcutaneous tumor model in nude mice (Fig. S4A). After 1 month, the nude mice were sacrificed, we found that subcutaneous tumors in the experimental group in which macrophages were mixed with tumor cells were larger than those in the control group (Fig. S4B, C). Meanwhile, we also used clodronate liposomes to eliminate macrophages in C57 mice (Fig. S4D). We observed that tumors in the control group were larger than those in the macrophage elimination group (Fig. S4E–G). Next, we constructed a co-culture model to simulate the interaction between tumor cells and macrophages (Fig. 3A). After 48 h of co-culture, we found that tumor cells co-cultured with macrophages became resistant to cisplatin (Fig. S4H). The results of flow cytometry for apoptosis showed that co-culture diminished the response of tumor cells to cisplatin (Fig. S4I). We used CD3/CD28 antibodies to activate Jurkat T cells in vitro (Fig. S4J). Subsequently, activated Jurkat T cells were mixed with both tumor cells co-cultured with macrophages and macrophages co-cultured with bladder cancer cells (Fig. 3B). Interestingly, the activated Jurkat T cells failed to suppress tumor cell growth. The results of the MTT assay showed that tumor cells co-cultured with macrophages proliferated faster than regular tumor cells when subsequently mixed with activated Jurkat T cells. (Fig. S4K and S4L).

Fig. 3.

Tumor associated macrophages promote immune escape of bladder cancer cell through the CXCL5-CXCR2-NF-κB-PD-L1 pathway. A The establishment process of co-culture model by using macrophage and tumor cell. B The establishment process of co-culture model by using macrophage, tumor cell and activated Jurkat T cell. C The RT-qPCR results of chemokine changes in M0 macrophage cocultured with T24. D The RT-qPCR results of chemokine changes in T24 co-cultured with M0 macrophage. E CXCL5-CXCR2-NF-κB-PD-L1 pathway related protein and IDO1 protein were increased in macrophage co-cultured with bladder cancer cell. F The heatmap of tryptophan metabolomics. G The concentration of INF-γ, IL4, IL2, IL6 and TNF-α in conditioned medium derived from the co-culture system of Jurkat T cell and macrophage. H CXCL5-CXCR2-NF-κB-PD-L1 pathway related protein were increased in bladder cancer cell co-cultured with macrophage. I The concentration of INF-γ, IL4 and IL2 in conditioned medium derived from the co-culture system of Jurkat T cell and T24 cell. J The concentration of INF-γ, IL4 and IL2 in conditioned medium derived from the co-culture system of Jurkat T cell and 5637 cell. ** P < 0.01, *** P < 0.001, **** P < 0.0001

CXCL5-CXCR2 mediates the interaction between macrophages and bladder cancer cell to reshape the immunosuppressive microenvironment.

To explore the potential mechanism underlying the observed changes in the co-culture system, we performed RT-qPCR assays on tumor cells and macrophages within this system. Interestingly, we found that CXCL5 expression was significantly upregulated in both macrophages and bladder cancer cells in the co-culture system (Fig. 3C, D). The results of ELISA further confirmed that CXCL5 concentration was elevated in the co-culture system (Fig. S4M). Subsequently, we demonstrated via western blot analysis that the protein levels of CXCL5, CXCR2, p-NF-κB, and PD-L1 were increased in macrophages co-cultured with bladder cancer cells (Fig. 3E). We also observed via western blot that the protein levels of immune checkpoint molecules, including PD-L1 and IDO1, were upregulated in macrophages co-cultured with tumor cells (Fig. 3H). Furthermore, we discovered that the NF-κB inhibitor PDTC could diminish the co-culture-induced elevation of PD-L1 (Fig. S4N). Next, we performed targeted metabolomics sequencing on macrophages co-cultured with tumor cells. The results revealed that levels of kynurenine (a metabolite of tryptophan) were elevated in macrophages co-cultured with tumor cells (Fig. 3F). To investigate whether tumor cell-co-cultured macrophages could suppress Jurkat T cell function, we collected conditioned medium from Jurkat T cells co-cultured with macrophages for ELISA analysis. The results showed that the secretion of pro-inflammatory cytokines (IL-4, IFN-γ, IL-2, IL-6, TNF-α) by T cells was inhibited (Fig. 3G). Similarly, western blot analysis showed that bladder cancer cells co-cultured with macrophages also exhibited increased protein levels of CXCL5, CXCR2, p-NF-κB, and PD-L1 (Fig. 3H). We also collected conditioned medium from Jurkat T cells co-cultured with tumor cells for ELISA analysis, which showed that the secretion of pro-inflammatory cytokines (IL-4, IFN-γ, IL-2) by Jurkat T cells was inhibited (Fig. 3I, J).

CXCL5-positive macrophages exhibit high expression of PD-L1

To explore the status of macrophages after co-culture, we performed transcriptomic sequencing using M0 macrophages co-cultured with bladder cancer cells (Fig. 4A). The top 50 upregulated differentially expressed genes in macrophages co-cultured with bladder cancer were designated as the TAM gene signature. Notably, the expression profile of signature genes in macrophages co-cultured with tumor cells indicated enrichment in CXCL5⁺ TAMs (Fig. 4B). Subsequent investigation revealed that patients with higher infiltration levels of CXCL5⁺ TAMs experienced inferior overall survival relative to those with lower infiltration levels (Fig. 4C). Additionally, macrophages expressing CXCL5 exhibited increased PD-L1 expression, as evidenced by the UMAP plot (Fig. 4D). Pathway activity analysis further demonstrated that the NF-κB pathway was activated in CXCL5⁺ TAMs (Fig. 4E). Moreover, Gene Set Variation Analysis (GSVA) scores for CXCL5⁺ TAMs were higher in patients with favorable responses to immunotherapy, based on an immunotherapy cohort from the Second Affiliated Hospital of Sun Yat-sen University (Fig. 4F). Collectively, these findings suggest a potential positive correlation between CXCL5 and PD-L1 expression in macrophages. To substantiate this hypothesis, virtual CXCL5 gene knockout was performed in macrophages, resulting in a significant change in PD-L1 expression (Fig. 4G). KEGG pathway analysis showed that the PD-L1/PD-1 pathway exhibited significant changes in macrophages subjected to CXCL5 knockout (Fig. 4H).

Fig. 4.

Macrophages that are positive for CXCL5 demonstrate a pronounced expression of PD-L1. A The heatmap of differential genes in M0 and M0 coculture with T24. B The similarity score of M0 co-cultured macrophage in single cell data. C Survival analysis of CXCL5 TAMs in TCGA-BLCA. D The expression of CXCL5, CD274 in single cell data. E Activity scores of pathways in various macrophage subtypes. F The GSVA score of CXCL5-TAMs in immunotherapy cohort from the Second Affiliated Hospital of Sun Yat-sen University. G The top 20 differentially expressed genes after virtual knockout of CXCL5 in macrophages. H The KEGG pathway analysis of differentially expressed genes after virtual knockout of CXCL5 in macrophages. **** P < 0.0001

CXCL5-CXCR2-NF-κB mediates the increase of PD-L1 in bladder cancer cells and promotes the proliferation, migration, lung metastasis of bladder cancer cells and the efficacy of immunotherapy.

According to the results of a pan-cancer analysis, CXCL5 was significantly over-expressed in various cancers (Fig. S5A). In addition, we analyzed the relationship between CXCL5 expression and clinical data in bladder cancer patients. We found that CXCL5 expression was elevated with higher tumor grade and stage (Fig. S5B–D). Furthermore, we detected CXCL5 and PD-L1 expression in bladder cancer cell lines, which indicated that CXCL5 and PD-L1 expression were highly correlated (Fig. S5E). Meanwhile, we also used the Spearman method to analyze the correlation between CXCL5 and PD-L1 in the TCGA-BLCA database (Fig. 5A). In the IMvigor210 cohort, we found that CXCL5 expression increased with PD-L1 expression on tumor cells (Fig. 5B). Next, we used the 5637shCXCL5 and MB49 shCXCL5 cell lines to perform Western blot and RT-qPCR assays. The results suggested that reducing CXCL5 expression could decrease the transcriptional and protein levels of PD-L1. NF-κB pathway-related proteins were also downregulated in 5637shCXCL5 and MB49 shCXCL5 cell lines (Fig. 5C–E). Furthermore, we observed that CXCR2 inhibitor SB225002 attenuated the CXCL5-induced upregulation of PD-L1 in bladder cancer cell (Fig. S5F, G).

Fig. 5.

The expression level of CXCL5 is positively correlated with the expression of PD-L1 and enhancing the efficacy of immunotherapy in bladder cancer. A The correlation between CXCL5 and PD-L1 in TCGA-BLAC. B The expression of CXCL5 in bladder cancer patients with different expression of PD-L1 based on IMvigor210 cohort. C The change of PD-L1 RNA level in CXCL5 knockdown T24 cell line. D The change of PD-L1 RNA level in CXCL5 knockdown MB49 cell line E CXCL5-CXCR2-NF-κB-PD-L1 pathway related protein expression in CXCL5 knockdown cell line. F The establishment process of C57 immunotherapy model. G The image of C57 subcutaneous tumors according to different group. H The tumor weight of subcutaneous tumors according to different group. ** P < 0.01, **** P < 0.0001

In order to explore the function of CXCL5 in bladder cancer cells, we performed the MTT assay, transwell assay, cell colony formation assay, and wound healing assay. We found that the proliferative capacity of bladder cancer cells was not affected when we knocked down CXCL5 expression in the 5637 and MB49 bladder cancer cell lines, as indicated by the MTT and cell colony formation assays (Fig. S5H). Furthermore, transwell and wound healing assays confirmed that CXCL5 could promote bladder cancer cell migration (Fig. S5I, J). We established a lung metastasis model via tail vein injection. The results showed that reduced CXCL5 expression could inhibit lung metastasis (Fig. S5K–M). We also found that CXCL5 expression had no effect on subcutaneous tumor growth in nude mice. However, CXCL5 downregulation significantly suppressed subcutaneous tumor growth in C57 mice (Fig. S6A–H). Immunohistochemical results further indicated that CXCL5 downregulation could activate T cells and reduce macrophage recruitment (Fig. S6I, J). Furthermore, SB225002 was used to demonstrate that CXCL5 could attenuate the inhibitory effect of CXCR2 inhibitors on tumor growth (Fig. S6K–M). Finally, we established a subcutaneous tumor model in C57 mice and performed immunotherapy experiments on these mice. The results suggested that CXCL5 could enhance the efficacy of immunotherapy (Fig. 5F–H). Collectively, we confirmed through in vitro and in vivo experiments that CXCL5 plays a critical role in bladder cancer cell growth, migration, lung metastasis, and the efficacy of immunotherapy against bladder cancer.

Discussion

The urothelial layer, lamina propria, muscle layer, fat layer, and urine together constitute a complex bladder cancer microenvironment containing a variety of solid tumor cells, immune cells, and other cell types. As validated by numerous studies [37, 38], the proportion of immunosuppressive cells increases with tumor progression. Tumor-associated macrophages (TAMs) serve as a key component of the TME [39]. Accumulating evidence suggests that TAMs especially M2-type macrophages, exert a role in promoting tumor progression [40–42]. In contrast, tumor cells can secrete cytokines such as IL-4, IL-10, and CSF-1 to induce macrophage polarization [43]. However, numerous studies have demonstrated the presence of distinct TAM subtypes within the TME, which exert significant impacts on tumor progression, immunotherapeutic efficacy, and patient prognosis [44, 45]. Zhang et al. identified over ten TAM subtypes via single-cell sequencing analysis of tumor-infiltrating myeloid cells across multiple cancer types [31]. Peng et al. identified three TAM subtypes within the colorectal cancer TME [46]. While distinct TAM subtypes may exert divergent roles within the TME, the underlying mechanisms through which TAMs influence cancer progression remain incompletely understood. In our study, we found that macrophages in the bladder cancer microenvironment are predominantly categorized as M0 and M2 subtypes according to traditional classification criteria. Furthermore, we utilized single-cell sequencing data to identify six TAM subtypes in the bladder cancer TME, among which FOLR2⁺ macrophages exhibited a significant predictive value for bladder cancer patients’ overall survival (OS) and cancer-specific survival (CSS). Meanwhile, we further concluded that TAMs can promote tumor cell proliferation, induce cisplatin resistance, and facilitate immune escape.

Recently, SPP1⁺ TAMs have been identified across diverse cancer TME. Wu et al. demonstrated that SPP1⁺ TAMs can promote tumor progression and metastasis in head and neck squamous cell carcinoma (HNSCC) [47]. Gao et al. documented that SPP1⁺ macrophages interact with cancer-associated fibroblasts (CAFs) through binding to ITGF1, thereby modulating the TME in hepatocellular carcinoma (HCC) [48]. Matsubara et al. concluded that SPP1⁺ TAMs serve as a predictor of poor prognosis in patients with lung adenocarcinoma [49]. Zhang et al. revealed that SPP1⁺ TAMs may interact with regulatory T cells (Tregs) within CD44-enriched regions in colorectal cancer [50]. Ramos et al. identified specific TAM subsets with high SPP1 and TREM2 expression, and observed that SPP1 was predominantly enriched in breast cancer tumor regions—an expression pattern that correlated negatively with patient prognosis [51]. In our study, we identified SPP1⁺ TAMs via single-cell sequencing data; these cells are characterized by high SPP1 and TREM2 expression.

FOLR2⁺ TAMs constitute an important TAM subtype in the tumor microenvironment (TME). Numerous studies have demonstrated that FOLR2⁺ TAMs exert divergent roles across different cancer TMEs. Ramos et al. observed that FOLR2⁺ macrophages were predominantly enriched in the tumor stroma [51]. Moreover, they proposed that FOLR2 can activate CD8⁺ T cells to inhibit breast cancer progression and correlates positively with patient prognosis. Xiang et al. demonstrated that FOLR2⁺ macrophages modulate Tregs to establish an immunosuppressive TME in lung adenocarcinoma [52]. In a previous study, we established a novel predictive model for tumor-associated macrophages (TAMs) in bladder cancer. We found that FOLR2 functioned as a hub gene in the prognostic signature and was specifically expressed in bladder cancer-associated macrophages [53]. We also identified FOLR2⁺ TAMs in our current study; these cells were enriched in the tumor stroma. Moreover, we found that the TAM2 subset correlated negatively with OS and CSS in bladder cancer patients. Furthermore, we investigated the role of FOLR2⁺ TAMs in bladder cancer immunotherapy. We found that FOLR2⁺ TAMs were enriched in patients who attained stable disease (SD) following immunotherapy.

IL1B⁺ TAMs have also been identified in a multitude of studies. IL1B⁺ TAMs exert crucial roles across diverse tumor microenvironments (TME). Caronni et al. demonstrated that IL1B⁺ macrophages are associated with poor prognosis and modulate the immune TME in early-stage pancreatic cancer, and their infiltration may potentially promote pancreatic cancer invasion and metastasis [54]. They proposed that targeting the PGE2–IL1B axis may possess therapeutic potential for pancreatic cancer. Wang et al. observed that IL1B is capable of inducing TAM accumulation and suppressing the function of cytotoxic T cells within the prostate cancer TME [55]. They concluded that IL1B represents a key therapeutic target for advanced prostate cancer. In the current study, we identified IL1B⁺ TAMs within the bladder cancer TME, and these cells were primarily enriched in biological processes related to chemokine receptor binding and cytokine activity. Furthermore, we identified CXCL5-expressing TAMs (CXCL5⁺ TAMs) in the present study, and these cells exhibit high PD-L1 expression. Meanwhile, based on the results of single-cell trajectory analysis, CXCL5⁺ TAMs may originate from the differentiation of SPP1⁺ macrophages. Notably, we observed that macrophages co-cultured with bladder cancer cells shared a high degree of similarity with CXCL5⁺ TAMs in terms of transcriptional profiles.

CTLA-4 and PD-1 were discovered by American immunologist James Allison and Japanese immunologist Tasuku Honjo, respectively, and hold significant importance in cancer immunotherapy [56]. Undoubtedly, PD-1/PD-L1 inhibitors have driven advancements in cancer immunotherapies. However, a substantial proportion of cancer patients fail to derive benefits from these agents. A previous retrospective study, which analyzed PD-1 blockade therapy in patients with melanoma, confirmed that only 48% of patients achieved significant tumor shrinkage or stable disease (SD) following PD-1 blockade [57]. The IMvigor210 cohort study evaluated the efficacy of atezolizumab in cisplatin-intolerant patients with locally advanced or metastatic uroepithelial carcinoma, showing that only 22.8% of patients responded to PD-1 inhibition therapy [20]. TAMs exert a complex role in mediating resistance to immunotherapy. It has been demonstrated that PD-L1 is highly expressed on TAMs, a finding that may undermine the efficacy of PD-1/PD-L1 inhibitors [58]. During our study, we observed substantial infiltration of M2 macrophages in tissue samples from immunotherapy non-responders, whereas significant infiltration of M1 macrophages was observed in samples from responders. Additionally, single-cell data analysis revealed that differential gene expression in immunotherapy responders was predominantly attributed to macrophages and T cells. One study concluded that PD-L1 and CTLA-4 expressed on TAMs can directly bind to PD-1 and CTLA-4 receptors on T cell surfaces, respectively, thereby inhibiting T cells from exerting their antitumor effects [59]. Matton et al. performed immunofluorescence staining on oral squamous cell carcinoma tissues and found that approximately 14–32% of PD-L1 expression was derived from CD68⁺ TAMs. Furthermore, CD8⁺ T cells exhibited greater co-localization with PD-L1⁺ CD68⁺ TAMs than with CD4⁺ T cells, suggesting that TAMs may modulate CD8⁺ T cell function within tumor tissues via PD-L1 [60]. Additionally, Sun et al. demonstrated that the CXCL3–CXCR2 axis mediates pancreatic cancer metastasis through interactions between TAMs and cancer-associated fibroblasts (CAFs) [61]. Moreover, Liu et al. confirmed that macrophage-derived CCL5 promotes immune evasion in colorectal cancer cells through the p65/STAT3–CSN5–PD-L1 pathway [62]. In our study, we used THP-1-derived macrophages and bladder cancer cells to conduct co-culture experiments. We found that the CXCL5–CXCR2 axis promotes PD-L1 expression in both bladder cancer cells and macrophages by activating the NF-κB pathway. Additionally, we further demonstrated that macrophages co-cultured with bladder cancer cells can inhibit T cell function and remodel the immunosuppressive TME. The CXCL5-CXCR2 axis may play an important role in bladder cancer progression. Therefore, patients with higher CXCL5⁺ TAMs infiltration may had a poor overall survival. However, patients with higher CXCL5⁺ TAMs tumor infiltration exhibited a more favorable response to immunotherapy. We propose three potential underlying mechanisms as follows: Firstly, The CXCL5-CXCR2-NF-κB axis was upregulated in the co-culture system, which promoted PD-L1 expression in both tumor cells and TAMs. Although this process facilitates immune escape of bladder cancer cells and induces T cell dysfunction via PD-L1 upregulation, it also renders the tumor microenvironment more “hot” by enhancing the inflammatory response. Consequently, some patients achieve improved responses to PD-1-targeted immunotherapy. Secondly, T cell dysfunction or exhaustion induced by elevated PD-L1 levels implies that PD-1 blockade may restore the anti-tumor activity of a subset of T cells, thereby augmenting the efficacy of immunotherapy. Thirdly, extensive research has demonstrated that PD-L1-positive macrophages often belong to the pro-inflammatory subset [63, 64]. Consistent with this, our findings reveal that CXCL5-positive macrophages express high levels of PD-L1. Consequently, patients with higher CXCL5⁺ TAM infiltration tend to exhibit a more robust inflammatory response, which contributes to their improved response to immunotherapy.

Chemokines also play a critical role in the TME. They serve as messengers in intercellular communication [65]. Numerous studies have reported that they mediate the recruitment of various immune cell types within the TME [66, 67]. While some chemokines can transform “cold tumors” into “hot tumors”—thereby inhibiting tumor progression—others can recruit immunosuppressive cells that promote tumor progression. Furthermore, CXCL9 and CXCL10 have been shown to recruit effector T cells and natural killer (NK) cells to tumor sites, thereby enhancing therapeutic responses to both cancer immunotherapy and chemotherapy [68]. In our study, single-cell data from bladder cancer patients who received immunotherapy revealed substantial infiltration of CXCL9 and CXCL10 (secreted by TAMs) in tissue samples from immunotherapy responders. Moreover, CCL5 can bind to CCR1 receptors to recruit multiple types of immunosuppressive myeloid cells within tumors [69]. Therefore, acquiring a comprehensive understanding of chemokine functions is essential. Growing evidence indicates that the CXCL5/CXCR2 axis is aberrantly activated across multiple cancer types. Zhou et al. found that CXCL5 can activate the PI3K/AKT/GSK-3β/Snail–Twist pathway by binding to CXCR2, thereby promoting epithelial–mesenchymal transition (EMT) in lung cancer cells and driving malignant progression [70]. Studies have also shown that tumor-fibroblast-derived CXCL5 can activate the PI3K–AKT pathway to upregulate PD-L1 expression in melanoma [71]. CXCL5 can further recruit neutrophils, which secrete various factors (e.g., integrin α2b) to promote the establishment of an immunosuppressive TME [72]. It has also been reported that CXCL5 can recruit large numbers of bone marrow-derived suppressor cells (MDSCs) to inhibit the function of T cells and NK cells [73]. However, few studies have explored the role of CXCL5 in the TME of bladder cancer.

In our previous studies, we demonstrated the potential role of CXCL5 in bladder cancer progression. Briefly, we established orthotopic bladder carcinoma models and lung metastasis models in laboratory mice. We subsequently isolated orthotopic bladder carcinoma cell lines (T24P) and lung metastatic cell lines (T24L) [74]. RT-qPCR analysis of these two cell lines revealed high CXCL5 expression in T24L cells. Furthermore, we demonstrated that CXCL5 knockdown inhibits the invasive capacity, metastatic potential, and stemness of lung metastatic bladder cancer cells by downregulating CD44 expression [75]. To elucidate the role of CXCL5 in the bladder TME, we performed in vitro and in vivo experiments using 5637shnCXCL5 and MB49shnCXCL5 cell lines. Our findings showed that CXCL5 regulates bladder cancer cell proliferation by modulating PD-L1 expression. Additionally, we found that suppressing CXCL5 expression inhibits lung metastasis of bladder cancer. Moreover, we speculated that CXCL5 may enhance immunotherapy efficacy. In summary, we determined that the CXCL5-CXCR2-NF-κB axis promotes polarization of TAMs and upregulates PD-L1 and IDO1 expression on TAMs, thereby suppressing T cell function. PD-1 blockade restores T cell function, which correlates with a favorable response to immunotherapy (Fig. 6). Meanwhile, our research is the first to demonstrate that infiltration of CXCL5⁺ TAMs is associated with enhanced responsiveness to immunotherapy. Furthermore, CXCR2 inhibitors exert a favorable inhibitory effect on bladder cancer. This not only identifies potential biomarkers for predicting immunotherapy efficacy in bladder cancer, but also provides promising therapeutic targets for its targeted treatment.

Fig. 6.

A schematic diagram of CXCL5-CXCR2 axis mediated the crosstalk between bladder cancer cell and macrophage

It is imperative to acknowledge that our study is subject to several limitations. The CXCL5 knockout models in TAMs were lacking in the present study. We aim to employ TAM-specific CXCL5 knockout model mice to further verify our conclusions in future studies. The role of CXCL5 in bladder cancer immunotherapy and TME biology requires comprehensive experimental validation using a wider range of cell lines and animal models.

Conclusions

Studies have determined that the CXCL5-CXCR2 axis facilitates the interaction between bladder cancer cells and macrophages. On one hand, macrophages contribute to tumor growth, enable tumor cells to evade immune surveillance, and enhance their tolerance to cisplatin. On the other hand, tumor cells can induce the polarization of macrophages and reshape the immunosuppressive microenvironment. Additionally, studies have further observed that the CXCL5–CXCR2 axis can drive the malignant progression of bladder cancer and enhance the effectiveness of immunotherapy.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors contribution

F.J.H. and Z.M.Z.conceived the experiments. J.Y.Z. conducted in vitro cell experiments, western blotting, cell flow cytometry, and co-culture assays. J.Y.Z. and L.J.P. performed the bioinformatics analysis of TCGA and GEO databases, and J.Y.Z. wrote the manuscript; L.J.P., W.L., Z.L., and Q.X.W. analyzed and interpreted the results. Z.Y.D., C.Y.T., L.H., and Y.Z.Z. assisted with the in vitro experiments and prepared the references. M.M.H., J.M.X., and P.Y.C. assisted with the in vitro experiments and interpretation of the results. All the authors reviewed the manuscript. All authors (s) read and approved the final manuscript.

Funding

This work was supported by National Natural Science Foundation of China (No: 82570903) and Key Research and Development Program of Shaanxi Provincial Health Commission (2025YF-29).

Data availability

Single cell sequence data was downloaded on the website (https://ngdc.cncb.ac.cn/gsa-human). TCGA-BLCA data, GSE13507 and GSE285715 were downloaded on the website (https://www.portal.gdc.cancer.gov) and（https://www.ncbi.nlm.nih.gov/geo/）. IMvigor210 data was downloaded on the website (http://research-pub.gene.com/IMvigor210CoreBiologies/). The RNA sequencing based on macrophage and single cell data were available from the corresponding author on reasonable request.

Declarations

Conflicts of interest

The authors declare no competing interests.

Ethics approval and consent to participate

This study was approved and supervised by the Ethical Committee of the First Affiliated Hospital of Xi’an Jiaotong University, Xi’an, China (NO.2022018).