Inflammation Is Associated with Worse Outcome in the Whole Cohort but with Better Outcome in Triple-Negative Subtype of Breast Cancer Patients

Abstract

Inflammation has been linked with cancer, but whether it is part of the problem or part of the solution remains to be a matter of debate in breast cancer. Our group and others have demonstrated that inflammation aggravates cancer progression; however, some claim that inflammation may support immune cell infiltration and suppress cancer. We defined the gene set variation analysis of the Molecular Signatures Database Hallmark inflammatory response gene set as the inflammatory pathway score and analyzed 3632 tumors in total from 4 breast cancer cohorts (METABRIC, TCGA, GSE25066, and GSE21094). In the whole breast cancer cohort, high-score tumors were associated with aggressive clinical characteristics, such as worse disease specific survival, higher Nottingham histological grade, and younger age. Inflammatory score was significantly higher in triple-negative (TNBC) as well as basal and normal subtypes compared with the other subtypes, which suggest that the detrimental effect of high level of inflammation may be because it includes a more aggressive subtype. On the contrary, high score within TNBC was significantly associated with better survival. TNBC with high score enriched not only IFN-α, IFN-γ response, IL-2/STAT5 signaling, Allograft rejection, Complement, p53 pathway, Reactive Oxygen, and Apoptosis but also TNF-α signaling, IL6-JAK-STAT signaling, TGF-β signaling, Coagulation, Angiogenesis, EMT, KRAS signaling, and PI3K-AKT-MTOR signaling gene sets. High score was associated with mainly favorable anticancerous immune cell infiltration as well as Leukocyte fraction, TIL regional fraction, Lymphocyte infiltration, IFN-γ response, TGF-β response, and cytolytic activity scores. Although the inflammatory pathway score was not associated with neoadjuvant treatment response, it associated with expressions of immune checkpoint molecules. In conclusion, inflammation was associated with worse outcome in the whole breast cancer cohort, but with better outcome in TNBC, which was associated with favorable anticancerous immune response and immune cell infiltrations.

Figure 1

Association between inflammatory pathway score and the clinical factors in breast cancer. (a) Disease-specific (DSS), disease-free (DFS), and overall survival (OS) of inflammatory score high (red line) and low (blue line) in whole METABRIC breast cancer cohort. Log-rank test was used to compare the groups with Kaplan-Meier survival curves. (b) Boxplots of the inflammatory scores by AJCC cancer stage, Nottingham pathological grade, and age at diagnosis (<40 yo, 40-70 yo, and <70 yo) in both METABRIC and TCGA cohorts. (c) Gene expression levels of SphK1, SphK2, and S1P receptor 1 by high vs. low inflammatory score. (d) Boxplots of the inflammatory scores by subtype and PAM50 classifications. Tukey-type boxplots show median and interquartile level values, and the ANOVA test is used to calculate p values.

Figure 2

Association between the inflammatory pathway score and tumor aggressiveness by the subtypes. (a) Boxplots of the inflammatory scores by Nottingham pathological grade in triple-negative breast cancer (TNBC), estrogen receptor-positive/human epidermal growth factor receptor 2-negative (ER+/HER2-), and HER2-positive subtypes of the METABRIC and TCGA breast cancer cohorts. Tukey-type boxplots show median and interquartile level values, and the ANOVA test was used to calculate p values. (b) Tumors with low (blue) and high (red) inflammatory scores in both cohorts of patients with each subgroup are compared for survival DSS, DFS, and OS of inflammatory score high and low inflammatory groups in each breast cancer subtype of METABRIC cohort. Log-rank test used to calculate p values to compare two groups with Kaplan-Meier survival curves.

Figure 3

Gene Set Enrichment Assay (GSEA) of high inflammatory pathway score in triple-negative breast cancer. Enrichment plots are shown for (a) favorable immune-related and (b) unfavorable Hallmark gene sets for which highly enriched in the high inflammatory pathway score compared to low-score group in both the TCGA and METABRIC cohorts, along with normalized enrichment score (NES) and false discovery rate (FDR). The statistical significance of GSEA was determined using FDR of 0.25.

Figure 4

Composition of tumor infiltrating immune cells and immune function scores by high or low inflammatory scores in METABRIC and TCGA cohorts. (a) anticancerous immune cells including CD8 T cells, CD4 memory T cells, M1 macrophages, and dendritic cells (DC); (b) procancerous immune cells including regulatory T cells, Type 2 helper T cells (Th2), and M2 macrophages; and (c) B cells and plasma cells in the TCGA and METABRIC cohorts. Comparison of high vs. low inflammatory scores in (d) cytolytic activity score (CYT). (e) Leukocyte fraction, tumor infiltrating lymphocyte (TIL) regional fraction, lymphocyte infiltration, Interferon-gamma (IFN-γ) response, T cell receptor (TCR) Shannon, TCR Richness, B cell receptor (BCR) Shannon, BCR Richness, and transforming growth factor-beta (TGF-β) response. Inflammation score in each cohort were divided into low and high groups by the median value. Tukey-type boxplots show median and interquartile level values, and the ANOVA test is used to calculate p values.

Figure 5

Relationships between the inflammatory score and neoadjuvant chemotherapy or expression of immune checkpoint molecules. (a) Percentage of achievement of pathological complete response (pCR) between low (blue bar) and high (red bar) inflammatory score in whole, TNBC and ER+/HER2- in the GSE25066 (n = 467) and GSE20194 (n = 248) breast cancer cohorts that underwent neoadjuvant chemotherapy. The number of patients who achieved pCR is shown below the plots. Fisher's exact test is used to compare pCR rates between two groups. (b) Comparison of low (blue) and high (red) inflammatory score groups in gene expression of immune checkpoint molecules (log₂ transcripts per million) in METABRIC and TCGA cohorts. Inflammatory score in each cohort was divided into low and high groups by the median values. Tukey-type boxplots show median and interquartile level values, and the ANOVA test is used to calculate p values. BTLA: B- and T-lymphocyte attenuator; CTLA4: cytotoxic T-lymphocyte-associated protein 4; IDO1/2: indoleamine dioxygenase 1/2; LAG3: lymphocyte activation gene 3; PD-1: programmed death-1; PD-L1: programmed death ligand 1; TIGIT: tyrosine-based inhibitory motif domain.