Tumor Microenvironment-Mediated Immune Profiles Characterized by Distinct Survival Outcome and Immunotherapeutic Efficacy in Breast Cancer

Abstract

Background: Numerous reports have highlighted that the tumor microenvironment (TME) is closely linked to survival outcome and therapeutic efficacy. However, a comprehensive investigation of the TME feature in breast cancer (BC) has not been performed. Methods: Here, we performed consensus clustering analysis based on TME cell expression profiles to construct TME pattern clusters and TME-related gene signature in BC. GSVA combined with CIBERSORT and ssGSEA algorithms were applied to evaluate the differences in biological pathway and immune cell infiltration level, respectively. The PCA method was employed to construct TME-score to quantify the TME-mediated pattern level in individual BC patients. Results: We determined two distinct TME gene clusters among 3,738 BC samples, which exhibited distinct survival outcome and enriched biological processes. The TME features demonstrated that these two clusters corresponded to the established immune profiles: hot and cold tumor phenotypes, respectively. Based on TME-related signature genes, we constructed the TME-score and stratified BC patients into low and high TME-score groups. Patients with high TME-score exhibited favorable outcome and increased infiltration of immune cells. Further investigation revealed that high TME-score was also related with high expression of immunosuppressive molecules, decreased tumor mutation burden (TMB), and high rate of mutation in significantly mutated genes (SMGs) (e.g., PIK3CA and CDH1). Conclusion: Assessing the TME-mediated pattern level of individual BC patients will assist us in better understanding the responses of BC patients to immunotherapies and directing more effective immunotherapeutic approaches.

Keywords: TCGA; breast cancer; immune profiles; immunotherapy; tumor microenvironment.

FIGURE 1

Tumor microenvironment-mediated pattern clusters in breast cancer. (A) The interaction between TME-infiltrating immune cells in BC. The lines connecting immune cells represented their interaction with each other. The size of each circle represented the prognostic effect of each immune cell and scaled by p-value. Protective factors for patients’ survival were indicated by a green dot in the circle center and risk factors indicated by the purple dot in the circle center. (B) Kaplan–Meier curves of OS and PFS for BC patients in the meta-cohort and TCGA-BRCA dataset with distinct TME clusters. The numbers of patients in TME cluster A and B were 1,682 and 545, respectively. (C,D) Unsupervised clustering of immune cells in the TCGA-BRCA and METABRIC datasets. Clinicopathological information, including age, gender, ER, PR, HER2, molecular subtype, stage, grade and TME cluster were used as patient annotations. Red represented the high expression of immune cells, and blue represented the low expression of immune cells. (E, F) The fraction of tumor-infiltrating lymphocytes in distinct TME clusters using the ssGSEA and CIBERSORT algorithms. Within each group, the scattered dots represented TME cell expression values. The thick line represented the median value. The bottom and top of the boxes were the 25th and 75th percentiles (interquartile range). The statistical difference of two TME clusters was compared through the Kruskal–Wallis H test. *p < 0.05; **p < 0.01; ***p < 0.001. (G) PCA of TME-infiltrating immune cells to distinguish TME cluster A from B.

FIGURE 2

Functional enrichment analysis of differentially expressed genes between distinct TME clusters. (A, B) GO and KEGG analysis of DEGs. (C) PPI of DEGs. The high confidence (0.700) was considered as the minimum required interaction score. (D) Identification of hub genes. Genes with degree of interaction more than 40 were recognized as hub genes. The X-axis label means the degree of interaction of each gene.

FIGURE 3

The landscape of genetic alteration of differentially expressed genes in breast cancer. (A) 320 of the 986 BC samples experienced genetic alterations, mostly including missense mutation, nonsense mutation, and splice site mutation. The number on the right indicated the mutation frequency of each DEG. Each column represented the individual BC sample. (B) The CNV mutation frequency of DEGs was prevalent. The column represented the alteration frequency. The amplification frequency, pink dot; the deletion frequency, green dot. Because of the size of the gene list (205 genes), we presented the CNV results in 3 plots: one for CNV deletion and the other two for amplified CNV. (C) PCA and t-SNE analysis based on the expression of DEGs to distinguish BC samples (red dots) from normal ones (green dots). (D) The differences in mRNA expression level of DEGs between tumor and normal samples was compared by R “limma” package. The asterisks represented the statistical p-value (*p < 0.05; **p < 0.01; ***p < 0.001). For graphical reasons, we presented the gene lists (upregulated and downregulated genes in BC samples) in 2 plots and ordered them based on the log FC.

FIGURE 4

Construction of TME-related gene signature. (A, B) Unsupervised clustering of TME-related signature genes in the TCGA and METABRIC datasets. Clinicopathological information including age, gender, ER, PR, HER2, stage, grade, and molecular subtype is shown in annotations above. Red represented the high expression, while blue represented the low expression. (C) PCA based on the TME-related gene signature could satisfactorily distinguish between TME gene cluster I and II. (D) The Kaplan–Meier curves of OS and PFS for BC samples in the meta-cohort and TCGA-BRCA dataset, respectively.

FIGURE 5

Biological pathways and tumor microenvironment characteristics in distinct gene clusters. (A) Heatmap showed the GSVA score of representative hallmark pathways curated from MSigDB between distinct TME gene clusters. (B) The differences in mRNA expression of transcripts associated with immune activation between distinct gene clusters. (C, D) The fraction of tumor-infiltrating immune cells between distinct gene clusters using the CIBERSORT and ssGSEA algorithms. Within each group, the scattered dots represented TME cell expression values. The thick line represented the median value. The bottom and top of the boxes were the 25th and 75th percentiles (interquartile range), respectively. The statistical difference of the three gene clusters was compared through the Kruskal–Wallis H test. *p < 0.05; **p < 0.01; ***p < 0.001. (E) Comparison of the expression level of immunosuppressive molecules between distinct TME gene clusters.

FIGURE 6

Construction of the TME-score and exploration of its clinical significance. (A) Alluvial diagram of the TME cluster, TME gene cluster, and TME-score. (B) Relative distribution of TME-score between distinct TME clusters and TME gene clusters. (C) Correlations between TME-score and tumor-infiltrating immune cells using Spearman analysis. The asterisks represented the statistical p-value (*p < 0.05). (D) Representative results of KEGG pathways between high and low TME-score groups via GSEA. (E) Kaplan–Meier curves for patients with high and low TME-score groups in the meta-cohort and METABRIC, TCGA-BRCA, GSE20685, GSE58812, GSE42568, GSE88770, and GSE20711 datasets. The median TME-score obtained from the METABRIC dataset (discovery cohort) was utilized to separate BC samples into high- and low-score groups. Independent validations were then performed using TCGA-BRCA and GSE datasets (external validation cohort).

FIGURE 7

Characteristics of the TME-score in tumor somatic mutation. (A) Correlation analysis between TME-score and TMB. (B) Relative distribution of TMB in high versus low TME-score groups. (C) Kaplan–Meier curve for high and low TMB patient groups (p = 0.001). (D) Kaplan–Meier curve for subgroup patients stratified by both TME-score and TMB (p < 0.001). (E, F) Mutational landscape of SMGs in the TCGA-BRCA dataset stratified by high (left panel) versus low TME-score (right panel) groups. Individual patients were represented in each column. The upper bar plot showed TMB, and the right bar plot showed the mutation frequency of each gene in separate TME-score groups.

FIGURE 8

Correlation between the TME-score and immunotherapeutic benefits and chemotherapeutic efficacy. (A) The relative distribution of immunosuppressive molecules was compared between TME-score high versus low groups in the meta-cohort. (B) Relative distribution of immunotherapeutic efficacy in high TME-score versus low TME-score groups. (C) Relative distribution of IC50 for chemotherapeutic drugs commonly used in clinical practice between TME-score high versus low groups.