PD-L1 Expression on Tumor Cell Membrane or Tumor-Associated Macrophages (TAMs)
28 male patients with NMIBC had tumor samples and 16S rRNA gene sequence data to be evaluated for PD-L1 expression on tumor cells or TAMs. Among them, the tumor stage of 13 patients was TaN0M0, and that of 15 patients was T1N0M0. Overall, PD-L1 expression was negative in 19 patients (67.9%) and positive in 9 patients (32.1%, including 6 patients with PD-L1 expression on tumor cells and 3 patients with PD-L1 expression on TAMs). And PD-L1 did not appear in tumor cells and TAMs simultaneously in PD-L1 positive specimens. Out of the 28 male patients with NMIBC, the intensity of PD-L1 was assessed and recorded as absent (0) in 19 patients (67.9%), mild (1+) in 6 patients (21.4%), moderate (2+) in 1 patients (3.6%), and severe (3+) in 2 patients (7.1%). Representative immunohistochemical staining results are shown in Figure 1.
Demographic Characteristics of Subjects
A total of 31 mid-stream urine specimens were analyzed, including 11 from male NMIBC patients with positive PD-L1 expression and 20 from male NMIBC patients with negative PD-L1 expression, while 2 specimens in PD-L1-positive group and 1 specimen in PD-L1-negative group were excluded for specimens with too little sequencing reads. Finally, 28 male patients with NMIBC based on their PD-L1 immunohistochemical results were divided into PD-L1-positive group (P group) and PD-L1-negative group (N group) respectively. No statistically significant difference was found in the demographic characteristics between P and N group (Table 1).
TABLE 1. Comparisons of demographic characteristics between PD-L1-positive(P group) and PD-L1-negative groups(N group)
Percentage of smokers
Percentage of multiple tumors
Percentage of hypergrading
Percentage of HER
Percentage of HEP
Percentage of high invasive risk
Percentage of T1N0M0
Data are presented as mean (SD) or median (first quartile to the third quartile) for continuous variables or n (%) for counting data. a indicates that the datum is subject to normal distribution and homogeneity of variance test and is presented as mean (SD). Multiple tumors refer to the number of bladder tumor≥2. Abbreviations: BMI, body mass index; CHD, coronary atherosclerotic heart disease; FHC, family history of cancer; EORTC, European Organization Research and Treatment of Cancer Scoring system, _R refer to recurrence and _P refer to progression; HER, recurrence score of EORTC≥5;HEP, progression score of EORTC≥7.
Sequencing Data, Alpha, Beta Diversity and Alpha Rarefaction Curves
Reads that contain 50% of base whose quality value is less than 20 or that contain 65% of base whose quality value is less than 30 were removed. A total of 962452 reads were obtained from 28 samples after the removal of primer and chimera. The median number of reads in PD-L1 positive group was 26,335, and in the PD-L1 negative group was 34952 (P=0.676). And the reads were classified into 2,124 OTUs which were used for downstream analysis. More OTUs were identified in urine from P group, with an average of 215 OTUs in P group and 141 OTUs in N group (P = 0.029). The number of reads and OTUs for each sample are shown in Supplementary Table S1. The Ace index and Observed species were significantly higher in P group than in N group (P <0.05; Figures 2A, B).While no differences were found in the Chao1 index (Figure 2C), Shannon index (Figure 2D) and Simpson index (Figure 2E) between P group and N group. Rarefaction curves were constructed for different groups showing P group was much more diverse than N group (Figure 2F).
In order to further clarify the relationship between PD-L1 expression and urinary microbiota, we divided P group into C group (with mild PD-L1 expression) and D group (with moderate or severe PD- L1 expression) based on the expression intensity of PD-L1. Interestingly, we found that as the expression intensity of PD-L1 increased, the bacterial richness showed an increasing trend (Supplementary Figure S1), though no statistical difference was found between the two groups. Among PD-L1 positive group, urinary microbiota of patients with ≥5% of tumor cells membrane staining had higher diversity than those of patients with ＜5% of tumor cells membrane staining (Supplementary Figure S2).
To measure the similarity of urine microbial communities between P group and N group, we performed the unconstrained principal coordinate analysis (PCoA) based on Bray Curtis, weighted UniFrac, and unweighted UniFrac distance metrics. We found that the composition of urinary microbiota in P group differed from that in N group. The PCoA plots based on weighted Unifrac and unweighted Unifrac distances revealed that the urinary microbiota of P group and N group formed two distinct clusters (Figures 3A, B). Besides, we detected 643 OTUs were enriched in P group, 934 OTUs were enriched in N group and with 547 OTUs were shared between two groups (Figure 3C).
Relative Abundance of Urinary Bacteria in P group and N group
At phylum level, the urinary microbiota was dominated by Firmicutes (22.8% P group, 34.4% N group) and Proteobacteria (21.1% P group, 25.8% N group), followed by Actinobacteria (6.6% P group, 4.0% N group) and Bacteroidetes (1.1% P group, 9.7% N group) (Table 2, Figure 4A). The urinary microbial composition of different groups at class, order, family and genus level were exhibited in Figures 4B–E. The genera compositions of all samples were exhibited in Figure 4F. Though the relative abundance of so many microbes at various taxonomic level differed between P group and N group, only five taxa including Bacteroidetes (P=0.017), Bacteroidia (P=0.022), Bacteroidales (P=0.025), Prevotellaceae (P=0.028) and Prevotella (P=0.04) were present at higher relative abundance in N group and one genus called Leptotrichia had higher abundance in P group(P=0.001), when the Metastats algorithm was used for comparison of the relative abundance of bacteria and the relative abundance threshold of microbes was set at 0.1% (Table 2).
TABLE 2. Comparison of relative abundance of urinary microbiota between PD-L1 positive(P group) and PD-L1 negative groups(N group).
Data were displayed as mean percentage; For taxa with P-value< 0.05, the taxa shown in the table are those with relative abundance greater than 0.1%. For taxa without significant difference, only taxa with relative abundance greater than 1% are shown in the table; Ns, not significant (based on P < 0.05).
For the OTUs, we defined the core microbes as the OTUs with relative abundance higher than 0.002 resulting in 52 OTUs altogether. The core microbes of each group were displayed by taxonomic dendrogram (Figure 5). The percentages of the total community covered by the core microbes ranged from 85.9% (N group) and 87.1% (P group). And the distribution of core microbes revealed marked changes between the bacterial composition of the different groups. Enrichment of some core microbes (e.g., Ureaplasma, Enhydrobacter, Mucispirillum) and decrease of some core microbes (e.g., Prevotella, Haemophilus, Lactobacillus) were observed in male NMIBC patients with positive PD-L1 expression when compared to PD-L1 negative group.
Specific Taxa Associated With PD-L1 Expression
The specific microbial taxa associated with PD-L1 expression were identified by using LEfSe algorithm. The results showed that Corynebacteriale, Corynebacteriaceae, Corynebacterium_1, uncultured_Corynebacterium_sp, Propionibacterium, Dermabacteraceae, Brachybacterium, Roseomonas, Rhodospirillaceae, Pigmentiphaga were present at significantly higher compositional abundances in P group compared with N group, whereas the relative abundance of Massilia, Oxalobacteraceae, Firmicutes was higher in N group (Figures 6A, B). Heat tree was used to illustrate the taxonomic differences between N group and P group (Figure 6C). Similarly, in P group, Roseomonas was enriched in the group with higher PD-L1 expression (group F, PD-L1≥5%) and Prevotella had higher abundance in the group with lower PD-L1 expression (group O, 1%≤PD-L1<5%) (Supplementary Figure S3).
Potential Functional Pathways Associated With PD-L1 Expression
The functional pathways of different groups were predicted by PICRUSt based on the microbial community profiles. Overall, the functional profiles present in patients with positive PD-L1 expression and negative PD-L1 expression could not be distinguished clearly (Figure 7A). The ATP-binding cassette transporters associated pathway was significantly enriched in NMIBC patients with positive PD-L1 expression, while the predicted KEGG pathways significantly enriched in NMIBC patients with negative PD-L1 expression included Systemic lupus erythematosus, Transcription machinery, Metabolism of Terpenoids and Polyketides, Selenocompound metabolism , Base excision repair and Nucleotide excision repair (Figure 7B).