Description of TMB in BC
The research process was showed in Figure 1. Full transcriptome data belonging 433 individual BC tumor samples were retrieved from TCGA database with outrightly open acquisition access, among which 398 were included into study after processing. Then we analyzed the whole-exome sequencing data using the “maftools” package to visualize an exhaustive presentation of mutational information. The waterfall plots of mutational information from TCGA and ICGC were separately established in Figure 2A and Figure 2B. The color category beside the plot represents different genomic mutation types. Selectively, we focused on the missense mutations marked with green.
An intersection of 19 mutated genes were identified for research after selected simultaneously from the top 31 mutated genes of two databases, including TP53, PIK3CA, KDM6A, RB1 and so on (Figure 2C). Then the TMB calculation of 19 included genes was caught out to analyze the difference between wild type groups and mutant type groups. Statistically, TMBs of mutant type groups were higher than wild type groups for almost every set, except for FGFR3 and CREBBP (Figure 2D), verifying that genomic mutation associates positively with TMB in BC.
Overall survival analysis and prognostic factors identification
In order to figure out the influence of mutated genes on prognosis, we performed survival analyses of the 19 genes mentioned above. 19 K-M curves of overall survival were drafted to compare survival rates and times showing as Figure 3A-S. We found that, among all of the 19 genes, wild group presents significant survival advantage to mutation group only in RB1 (log rank p=0.039), which meant the high mutation frequency on RB1 had a negative effect on prognosis of BC patients.
For further investigation of whether RB1 could be identified as an independent survival factor, random-size effects models combining with clinical characteristics were separately performed to comprehensively analyze the HRs of multiple potential prognostic factors. As is shown in Figure 4A, age (p=0.001, HR=1.703), TNM stage (p<0.001, HR=2.266), TMB (p<0.001, HR=0.917) and mutation of RB1 (p=0.043, HR=1.488) are demonstrated as statistically significant survival factors, using univariate Cox regression analysis. Multivariate Cox regression analysis was conducted based on the consequence of single factor analysis displaying on another forest plot (Figure 4B). It was implied that age (p=0.001, HR=1.704), TNM stage (p<0.001, HR=2.285), TMB (p<0.001, HR=0.911) and mutation of RB1 (p=0.004, HR=1.776) could further be identified as independent prognostic factors affecting overall survival, combining with the results of univariate Cox regression analysis (Table 1). More evidently, according to the regression analyses, either TMB or mutation of RB1 associates negatively with prognosis. Moreover, this discovery is prognosticly consistent with the conclusion mentioned in Figure 2D that genomic mutation associates positively with TMB in BC.
Biological roles of RB1 in KEGG pathways
To further predict the potential biological roles of RB1 it plays in BC, the GSEA analysis was employed for KEGG pathways research. In accordance to the GSEA score, KEGG pathways like DNA replication, mismatch repair, cell cycle, oocyte meiosis and regulation of autophagy (Figure 5A), which mostly relate to cell proliferation and cell differentiation, enrich on the mutation region of RB1 band. This finding signifies that mutation types of RB1 might be more inclined to take part in cell proliferation and development of BC.
Patterns of immune cell infiltration in BC tumor
The population proportions of 22 types of immune cells in 398 samples were visualized as a bar plot in BC tumors, and these types include B cells, T cells, plasma cells, monocytes, macrophages, natural killer cells, eosinophils, neutrophils, etc. As is presented in the bar plot, 398 samples from TCGA database display the respective distributions of 22 immune cell infiltration(Figure 6A). Generally, the proportions of every type of immune cell are neither even nor stable in different BC samples.
As for the correlation patterns, relevance between each of the 22 immune cells is multifarious, or even loose in general. While it is worth noting that CD4 memory activated T cell shows significantly positive correlations with CD8 T cell, NK resting cell, Macrophage 1 (M1) and follicular helper T cell. Whereas, both of CD8 T cell and CD4 memory activated T cell associate respectively negatively with CD4 memory resting T cell and Macrophage 0 (M0) (Figure 6B). This finding indicates that immune cells cooperating together are inclined to infiltrate in tumors. In contrast, those with conversed function tend to exist mutually opposite to each other in the tumor microenvironment of BC.
In terms of RB1, fractions of 22 immune cell types were established in the vioplot. Wilcoxon rank-sum test reveals that the amount of regulatory T cells (Tregs) infiltrating in wild type group was significantly higher than in mutant type group (Wilcoxon rank-sum test, p = 0.031, Figure 6C). One probable inference is that the occurrence of genomic mutation on RB1 decreases the antigens Tregs used to be able to recognize, which suppresses its specific immune response to the BC tumor while strengthen the tumor immune escaping.