Presence of intratumoral microbiota in BLCAThe flowchart of the present study is shown in Fig. 1. To investigate the presence of microbiota in bladder cancer, we employed fluorescence in situ hybridization (FISH) using the EUB338 probe in both bladder cancer tissues and adjacent normal tissues (Fig. 2A,B). Signal detection via FISH confirmed the presence of bacteria in both bladder cancer and adjacent normal tissues. Importantly, the intratumoral microbiota in bladder cancer tissues was significantly greater than that in normal bladder tissues. Furthermore, we explored the relationship between intratumoral microbiota and immune cells in bladder cancer. The results indicated a lower distribution of CD8+ cells (Fig. 2C) and a greater distribution of macrophages in bladder cancer tissues (Fig. 2D). Interestingly, regions with a greater abundance of macrophages exhibited a significant reduction in intratumoral microbiota.Fig. 1The flowchart of this study. The whole study is divided into three parts. Firstly, we conducted an overall analysis of the intratumoral microbiota in BLCA, including the bacteria composition and abundance. Then, we managed to divide BLCA patients into the high-risk and low-risk groups based on the Immune & antimicrobials related genes in BLCA. Finally, we identified key intratumoral microbiota in BLCA patients that could serve as biomarkers, and based on this, we constructed a novel microbial-based scoring system.Fig. 2Presence of intratumoral microbiota and staining of immune cells in bladder cancer tissues. (A, B) Fluorescence in situ hybridization (FISH) analysis of bacterial 16 S rRNA in normal bladder tissue (A) and bladder cancer tissue (B) stained with the EUB338 probe (green) and DAPI (blue). (C, D) Co-staining of immune cells and bacterial 16 S rRNA in bladder cancer tissue. CD8+ T cells are marked in red (C), while macrophages are marked in pink (D). (E, F) Co-staining of immune cells and bacterial 16 S rRNA in bladder cancer tissues. M1 macrophages are marked in pink (E), while M2 macrophages are marked in red (F). Scale bar = 100 μm.The phenotype and function of macrophages are regulated by the surrounding microenvironment, namely, the phenomenon of macrophage polarization. Polarized macrophages, which can be divided into M1 and M2 types, play a crucial role in regulating tumor progression, metastasis, and clinical outcome. Among them, M2 macrophages have been reported to promote the occurrence and progression of tumors22. We conducted FISH and IF experiments in bladder cancer tissues. The results showed that the infiltrating macrophages in bladder cancer were mainly M2 macrophages (Fig. 2E) rather than M1 macrophages (Fig. 2F).Landscape of the BLCA microbiotaWe analyzed the bacterial diversity at the genus level across cancers in the BIC database. The BLCA Shannon index ranked 11th out of 32 tumors (full name is shown in Supplementary Table 1), indicating high bacterial diversity in BLCA (Fig. 3A). Subsequently, we analyzed the microbiota composition of BLCA. The top ten bacteria at the genus level were Paenibacillus, Pseudomonas, Bacillus, Peptoclostridium, Acinetobacter, Brevibacillus, Corynebacterium, Prevotella, Azotobacter and Actinoplanes (Fig. 3B).Fig. 3Landscape of BLCA microbiota. (A) Bacterial diversity at the genus level across cancers. The red line indicates the mean diversity index of BLCA. (B) Stacked bar plot of the bacterial composition detected in BLCA patients. The bar plot shows the relative abundances of the top ten bacteria at the genus level. (C) The Venn diagram displays the overlap of the top ten genera in terms of relative abundance in BLCA and colorectal carcinomas. (D) The Venn diagram displays the overlap of the top ten genera in terms of relative abundance in BLCA and other urological tumors.The structure of the intratumoral microbiota across cancers is specific. We compared the top 10 intratumoral bacteria in BLCA and colorectal carcinomas, and the Venn diagram showed that only three microorganisms at the genus level overlapped, namely, Pseudomonas, Bacillus and Prevotella (Fig. 3C). We then compared BLCA with other urological tumors, including three different types of kidney cancer. Interestingly, the microbial structure of BLCA is similar to that of other urological cancers. The top ten intratumoral bacteria of BLCA overlapped 7 with KIRC (Paenibacillus, Pseudomonas, Bacillus, Acinetobacter, Brevibacillus, Corynebacterium, and Azotobacter), 7 with KIRP (Paenibacillus, Pseudomonas, Bacillus, Acinetobacter, Brevibacillus, Corynebacterium, and Actinoplanes), 5 with KICH (Pseudomonas, Acinetobacter, Corynebacterium, Azotobacter, and Actinoplanes), and even 8 with PRAD (Paenibacillus, Pseudomonas, Bacillus, Peptoclostridium, Acinetobacter, Brevibacillus, Corynebacterium, amd Azotobacter). A total of three microorganisms, namely, Pseudomonas, Acinetobacter and Corynebacterium, were among the top 10 in bacteria abundance in all five tumors (Fig. 3D).The intratumoral microbiota structure of urogenital system cancers exhibits a high degree of similarity, which may be related to the proximity of anatomical locations and the influence of the urinary microbiome.Construction of a prognostic risk model for BLCA based on differentially expressed immune- and antimicrobial-related genes (DEIARGs)We downloaded a list of antimicrobial-related genes from the ImmPort database and a list of tumor immune-related genes from the TISIDB (Supplementary Table 2). The combination of these two gene lists revealed 1330 genes, which are referred to as immune- and antimicrobial-related genes (IARGs) (Fig. 4A). We identified 4840 differentially expressed genes (DEGs) from the transcriptome data of TCGA bladder cancer and normal bladder tissue samples. By comparing this gene list with those of IARGs, we ultimately obtained 331 DEIARGs (Fig. 4B, Supplementary Table 3).Fig. 4Construction of a prognostic risk model of BLCA based on DEIARGs. (A) A total of 1330 genes were identified as immune- and antimicrobial-related genes (IARGs). (B) Venn diagram shows 331 genes identified as differentially expressed IARGs (DEIARGs) in BLCA. (C, D) LASSO Cox regression analysis of DEIARGs. (E, G) K-M curves of the All dataset, training group and testing group. (H-J) ROC curves of the All dataset, training group and testing group.To construct a DEIARG-based prognostic model for BLCA patients, we first performed univariate Cox regression analysis, and a total of 34 genes were screened (Supplementary Fig. 1). Then, LASSO Cox regression analysis and multivariate Cox regression analysis were conducted (Fig. 4C,D), and the 11 hub genes that were most significantly associated with the prognosis of BLCA patients were ultimately identified (Supplementary Table 4). Patients with bladder cancer were randomly divided into a training group (n = 269) and a testing group (n = 132) to construct and validate prognostic risk models (Supplementary Table 5). Analysis of patient characteristics in the training group and the testing group showed good similarity between the two groups (Supplementary Table 6). All patients were further divided into high-risk and low-risk groups based on the median risk score obtained from the training group. K-M survival curves (Fig. 4E-G) and ROC curves (Fig. 4H-J) show the reliability of the prognostic risk model. Patients with higher risk scores had shorter overall survival and worse prognosis.We further visualized the expression of the 11 hub genes to better visualize the differential expression of these genes between the high-risk and low-risk groups (Supplementary Fig. 2A-C). Overall, our risk model showed that patients with higher risk scores had worse outcomes, while patients with lower risk scores had better outcomes (Supplementary Fig. 2D-I). In addition, using the Metascape database (https://Metascape.org/gp/index.html#/main/step1), a Gene Annotation & Analysis Resource, we conducted an analysis of the 11 hub genes. Pathway and process enrichment analyses were performed using various ontology sources, including KEGG Pathway23, GO Biological Processes, Reactome Gene Sets, Canonical Pathways, CORUM, WikiPathways, and PANTHER Pathway. These hub genes are closely related to various tumor diseases and processes (Supplementary Fig. 2J and Supplementary Tables 7–8).Differential microbiota signatures in the high-risk and low-risk groupsTo investigate the differences in the bladder microbiota between high-risk and low-risk patients, we conducted a series of analyses. First, we analyzed the alpha and beta diversity of the intratumoral microbiota in the high-risk and low-risk groups. No significant difference in the alpha or beta diversity was observed between the two groups at the phylum (Supplementary Fig. 3A, C) or genus levels (Supplementary Fig. 3B, D).To visualize the differences in the composition of the microbiota between the two groups, we identified the top 10 most abundant microbes at different taxonomic levels (Fig. 5A-E). No significant differences were observed between the two groups at the phylum, class and order levels. However, at the family and genus levels, the microbial compositions of the two groups were significantly different. In particular, the abundance of Sanguibacter, Elizabethkingia, Actinobacillus and Spirochaeta in the low-risk group was significantly greater than that in the high-risk group. Whether these bacteria are beneficial to the prognosis of BLCA patients deserves further study.Fig. 5Microbiota composition and STAMP analysis at various taxonomic levels in the high-risk and low-risk groups. (A-E) Stacked bar charts show that the bacterial composition of the two groups is quite different at the level of family and genus. (F-J) Forest plot displaying the results of STAMP analysis at various taxonomic levels in the two groups.Then, we used statistical analysis of metagenomic profiles (STAMP) to specifically analyze the different microorganisms between the high-risk and low-risk groups. Overall, 2, 4, 4, 6, and 15 distinct microorganisms were found between the two groups at the phylum, class, order, family, and genus levels, respectively (Fig. 5F-J).At the phylum level, the abundance of Aquificae was much greater in the high-risk patients than in the low-risk patients, while Chlamydiae showed the opposite trend. At the genus level, Pseudoalteromonas, Thermocrinis, Amycolatopsis, Sanguibacter, and Spirochaeta were the most significantly different bacteria between the two groups.Microbiota co-expression network in the high-risk and low-risk groupsMicrobial communities tend to influence each other; thus, we analyzed the network of co-expression relationships between different microbiomes at the genus level to understand the overall changes between the high-risk and low-risk groups. The results showed that the co-expression networks of the microbiota significantly differed between the two groups. Interestingly, we performed both positive and negative correlation analyses, but only the positive correlation networks showed statistical significance for reasons that require further research.A closely related module appeared in the co-expression network of both the high-risk and low-risk groups; however, the co-expression networks were completely different. The high-risk modules included Oleomonas, Cellulomonas, Rhodopirellula, Cycloclasticus, Candidatus Paracaedibacter, Prosthecobacter, Aphanizomenon, and Thiothrix (Fig. 6A). After reviewing the literature, we found that studies on these bacterial genera are scarce, indicating the potential for further research in this direction.Fig. 6Microbiota co-expression network in the high-risk and low-risk groups. (A) Network showing the microbiota co-expression relationships in the high-risk groups. (B) Network showing the microbiota co-expression relationships in the low-risk group. The size of the dot represents the relative abundance of bacteria, while the color of the dot represents its status in the network. The blue dots represent general points, while the yellow dots represent pivotal points in the network that are co-expressed with at least four other dots.The microorganisms in the module in the low-risk group were Arthrospira, Candidatus Protochlamydia, Rheinheimera, Candidatus Vidania, Halobacteroides, Lebetimonas, Dictyoglomus, Formos, Thermodesulfobium, Anacrofu stis, Dactylococcopsis, Butyrivibrio, Mitsuokella, Nitratiruptor, and Brachyspira (Fig. 6B). Interestingly, in the low-risk group module, Butyrivibrio and Arthrospira have been reported as microorganisms that are beneficial to human health. Previous studies have reported that better diet quality according to multiple healthy dietary patterns is associated with a greater abundance of Butyrivibrio species24. Arthrospira is a free-floating fine filamentous cyanobacteria. It is believed to have potential antioxidant, anti-inflammatory, immunomodulatory, weight loss and antitumor effects25,26,27. In general, the intratumoral microbiota in low-risk patients appears to be composed of beneficial microbial communities.LEfSe analysis and construction of a novel microbial-based scoring (MS) systemTo search for biomarkers of the microbiota in bladder cancer, LEfSe analysis was performed, and 41 vital microorganisms were ultimately identified (Fig. 7B). A detailed cladistic evolution diagram is shown in Fig. 7A.Fig. 7LEfSe analysis and construction of a novel MS system. (A) The cladogram displays bacterial biomarkers that exhibit significant differences between the high-risk and low-risk groups. The colors of the nodes are as follows: yellow for species with no significant differences and red and green for microbial taxa that play a crucial role in the indicated groups. The names of the biomarkers that could not be fully displayed in the cladogram are shown on the right side. (B) Linear discriminant analysis (LDA) score of distinct bacteria in the high-risk and low-risk groups. (C, D) LASSO Cox regression analysis of distinct bacteria based on LEfSe analysis. (E) K-M plot of the three groups based on MS data. (F) The overall survival of BLCA patients in the MS-low group was longer than that in the MS-high group.Based on these 41 biomarkers, we conducted an analysis combining patient survival data and ultimately identified six key prognostic bacteria (Supplementary Table 9). Through LASSO regression analysis, we established a new MS system composed of Syntrophobotulus, Granulicatella, Xanthomonas, Aquificae, Niabella, and Pseudoalteromonas (Fig. 7C-D, Supplementary Table 10). We used this MS system to score the BLCA patients. However, some BLCA patients found none of the key bacteria in the MS system, so they scored 0 and were labeled as the MS-0 group. The rest of the patients were labeled as either the MS- group or the MS+ group, based on their positive or negative scores (Supplementary Table 11). The K-M survival curve showed the prognosis of the three groups (Fig. 7E). The prognosis of patients in the MS-0 group was different from that in the MS + group, but there was no statistical difference compared with that in the MS- group. Therefore, we decided to merge the MS-0 group with the MS- group to form the MS-Low group (patients in this group lack tumor-promoting intratumoral microbiota), while the MS + group was designated as the MS-High group. The K-M survival curve showed the differential prognosis of the two groups (Fig. 7F). We also compared the clinical characteristics of the two groups of patients, including age, gender, grade, stage, race, etc., but there was no statistical difference (Supplementary Table 12). To date, we have identified the hub microbiota within the tumors of BLCA patients, and the novel MS system constructed from this information can subgroup patients.In addition, we analyzed the relative abundance of these six microorganisms and their clinical relevance to patients. The K-M plots revealed that a high abundance of Syntrophobotulus, Granulicatella, Xanthomonas, and Pseudoalteromonas was associated with worse patient outcomes (Supplementary Fig. 4A-F). Although the K-M map of Niabella was not significantly different, a greater abundance of Niabella appeared to be associated with longer OS. Interestingly, we found that Aquificae were significantly more abundant in stage 1 patients, suggesting that intratumoral bacteria may be related to clinical stage (Supplementary Fig. 4G). We analyzed the relative abundance of microorganisms involved in the MS system in the BIC database, and only the Aquificae class showed significant differences between different races (Supplementary Fig. 4H), while the relative abundance of all other microorganisms showed no significant differences between different races.Immune cell infiltration characteristics in distinct MS subgroupsTo further investigate the relationship between microorganisms in bladder cancer and tumor immunity, we calculated the infiltration of 22 types of immune cells in different MS groups using CIBERSORT. The results showed that the expression of B cells memory was significantly greater in the MS-high group than in the MS-low group (Fig. 8A).Fig. 8Analyses of tumor immunity and the MS system. (A) Differences in immune cell infiltration between the MS-High and MS-Low groups. (B, C) Oncoplot showing the top 10 genes with the highest mutation rates in the MS-High and MS-Low groups. (D-J) The differential expression of immune checkpoints in the MS-High and MS-Low groups.In addition, we analyzed the tumor mutation burden (TMB) of patients in different MS groups. Overall, patients in the low MS group had a greater TMB (90.68%) than those in the high MS group (85.06%). We then selected the ten genes with the highest mutation rates for visual analysis (Fig. 8B,C). It is clear that the most frequently mutated genes of patients in different MS groups are significantly different. For example, the mutation rates of TP53 and KDM6A are 6% and 7% greater, respectively, in MS-Low patients than in MS-High patients. Conversely, the mutation rate of SYNE1 is 5% lower in MS-Low patients than in MS-High patients. Whether the different mutations of these genes are related to the different prognosis of bladder cancer patients is worthy of further study.Finally, referring to the previous study28, we compared the differential expression of several important ICI between the high- and low-MS groups. Results can be seen in Supplementary Table 13. The results showed that expression of BTN2A2, BTN3A1, CD96, PD-L1, CEACAM1, HLA-C, and HLA-G was significantly increased in patients with low MS (Fig. 8D-J).
