Basal-like breast cancer with low TGFβ and high TNFα pathway activity is rich in activated memory CD4 T cells and has a good prognosis

Abstract

Basal-like breast cancer (BLBC) is a type of high-grade invasive breast cancer with high risk of recurrence, metastases, and poor survival. Immune activation in BLBC is a key factor that influences both cancer progression and therapeutic response, although its molecular mechanisms are not well clarified. In this study, we examined five cancer immunity-related pathways (IFNα, IFNγ, STAT3, TGFβ and TNFα) in four large independent breast cancer cohorts (n = 6,381) and their associations with the prognosis of breast cancer subtypes. Activities of the 5 pathways were calculated based on corresponding pathway signatures and associations between pathways and clinical outcomes were examined by survival analysis. Among the five PAM50-based subtypes, BLBC had the highest IFNα, IFNγ, TNFα pathway activities, and the lowest TGFβ activity. The IFNα, IFNγ, TNFα pathway activities were negatively correlated with BLBC recurrence. In contrast, positive association and no association with BLBC recurrence were observed for TGFβ and STAT3 pathways, respectively. TNFα/TGFβ pathway combination improved the prediction of recurrence and chemotherapy response of BLBCs. Immune cell subset analysis in BLBC showed that M0, M1 and M2 macrophage levels were associated with either TNFα or TGFβ pathways, whereas the level of activated memory CD4 T cells were associated with both pathways. Moreover, this T cell subset was most abundant in BLBCs with low TGFβ and high TNFα pathway activities. These results suggested that cooperation of TNFα and TGFβ signaling may be involved in the regulation of memory T cells and anti-cancer immunity in BLBCs. Our data also demonstrate that TNFα/TGFβ pathway combination may represent a better biomarker for BLBC prognosis and clinical management.

Keywords: Basal-like breast cancer; chemotherapy; immunity-related pathways; prognosis.

Figure 1

Comparison of IFNα, IFNγ, TNFα, TGFβ and STAT3 pathway activities among PAM50-based intrinsic subtypes of breast cancers. A) Patient cohort 1. B) Cohort 2. C) Cohort 3. D) Cohort 4. The four BC cohorts were merged from 42 Affymetrix microarray datasets as described in Materials and methods. Box-Whisker plots were used to show pathway activity, and the five statistics (5th, 25th, 50th, 75th and 95th percentile) were represented by the lower whisker, the lower box part, the solid line, the upper box part and the upper whisker, respectively. Notably, three datasets (GSE16446, GSE25055 and GSE25065) were present in both Cohort 1 and Cohort 4. To avoid repeated counting of sample data here, these three datasets were removed from cohort 1 when calculating pathway activity. Lum A: luminal A; Lum B: luminal B; Normal: Normal-like; Her2: HER2-enriched; Basal: basal-like.

Figure 2

Cox regression analysis of the associations of the IFNα, IFNγ, TNFα and TGFβ pathways with recurrence risk in different subtypes of breast cancers. A) IFNα pathway. B) IFNγ pathway. C) TNFα pathway. D) TGFβ pathway. The three BC cohorts annotated with patient's survival information were analyzed here. The pathway activities were used as continuous variables. The recurrence risk with the increase of the pathway activity was indicated by HR (presented per one-SD increment) as shown in forest plot. The overall effect of HR was calculated using a random-effects model, and the significance of the overall effects across multiple cohorts was estimated by Z test. HRs are shown in forest plots, in which the squares and horizontal lines represent the HR and 95% CI for the individual variables, while the diamonds represent the HR and 95% CI for the overall estimate. IFNa: IFNα; IFNg: IFNγ; TNFa: TNFα; TGFb: TGFβ.

Figure 3

Kaplan-Meier analysis of the associations of the IFNα, IFNγ, TNFα and TGFβ pathway activities with recurrence risk of BLBC. The three BC cohorts annotated with patient's survival information were analyzed, including cohort 2 (A, D, G, J), cohort 3 (B, E, H, K) and cohort 4 (C, F, I, L). The BLBCs were stratified into three subgroups based on tertile splits of predicted activities of the IFNα (A-C), IFNγ (D-F), TNFα (G-I) or TGFβ (J-L) pathway (high: > 2/3 percentile; low ≤ 1/3 percentile; intermediate: ≤ 2/3 percentile and >1/3 percentile) for each cohort, and Kaplan-Meier analysis was performed to compare the probabilities of DFS among each of the three subgroups. IFNa: IFNα; IFNg: IFNγ; TNFa: TNFα; TGFb: TGFβ.

Figure 4

Pathway combinations improve the prediction of BLBC prognosis. A). Comparison of prediction efficiency of various pathway combinations in five PAM50 subtypes BC. Combination of the TGFβ pathway with the IFNα, IFNγ or TNFα pathways was tested here as indicated. The overall HR for recurrence risk shown in forest plot was based on Cox regression analysis of three breast cancer cohorts as described in Methods. Scores for pathway combinations, as described in Methods, were used as continuous variable in Cox regression analysis and the HR [95% CI] is presented per one-SD increment. B) Forest plot showing the prediction efficiency of pathway combinations for BLBC in three individual cohorts. C) Forest plot showing the prediction efficiency of pathway combinations for HER2-enriched BC in three individual cohorts.

Figure 5

Kaplan-Meier analysis of the synergistic effect of TNFα and TGFβ pathway in prognosis prediction of BLBCs. Three BC cohorts with survival information were analyzed, including cohort 2 (A-C), cohort 3 (D-F) and cohort 4 (G-I). The patients were stratified into two groups based on median splits of predicted pathway activities or pathway combination scores in each cohort. A, D, G) Stratification based on TNFα pathway activity. B, E, H) Stratification based on TGFβ pathway activity. C, F, I) Stratification based on the combination scores of TNFα and TGFβ pathway. TNFa: TNFα; TGFb: TGFβ.

Figure 6

Associations of TNFα and TGFβ pathway status with activities of six BLBC-related pathways and infiltrations of four immune cell subsets in BLBC. BLBC from four BC patient cohorts were tested. A) Cohort 1. B) Cohort 2. C) Cohort 3. D) Cohort 4. Four BLBC subgroups, including high-TNFα/low-TGFβ, low-TNFα/high-TGFβ, high-TNFα/high-TGFβ and low-TNFα/low-TGFβ as indicated, were stratified from each cohort based on median values of the TNFα and TGFβ pathway activities, the heatmap was used to depict relative pathway activities and immune cell subset levels across the four BLBC subgroups. Each row represents one pathway or an immune cell subset as indicated. Each column represents one BLBC sample.