A Novel Three-Gene Score as a Predictive Biomarker for Pathologically Complete Response after Neoadjuvant Chemotherapy in Triple-Negative Breast Cancer

Abstract

Although triple-negative breast cancer (TNBC) typically responds better to neoadjuvant chemotherapy (NAC) compared to the other subtypes, a pathological complete response (pCR) is achieved in less than half of the cases. We established a novel three-gene score using genes based on the E2F target gene set that identified pCR after NAC, which showed robust performance in both training and validation cohorts (total of n = 3862 breast cancer patients). We found that the three-gene score was elevated in TNBC compared to the other subtypes. A high score was associated with Nottingham histological grade 3 in TNBC. Across multiple cohorts, high-score TNBC enriched not only E2F targets but also G2M checkpoint and mitotic spindle, which are all cell proliferation-related gene sets. High-score TNBC was associated with homologous recombination deficiency, high mutation load, and high infiltration of Th1, Th2, and gamma-delta T cells. However, the score did not correlate with drug sensitivity for paclitaxel, 5-fluorouracil, cyclophosphamide, and doxorubicin in TNBC human cell lines. High-score TNBC was significantly associated with a high rate of pCR not only in the training cohort but also in the validation cohorts. High-score TNBC was significantly associated with better survival in patients who received chemotherapy but not in patients who did not receive chemotherapy. The three-gene score is associated with a high mutation rate, immune cell infiltration, and predicts response to NAC in TNBC.

Keywords: neoadjuvant chemotherapy; predictive biomarker; prognosis; three gene; triple-negative breast cancer; tumor immune microenvironment.

Figure 1

Establishment and association of the 3-gene score with response to neoadjuvant chemotherapy (NAC) for triple-negative breast cancer (TNBC). (A) Volcano plots illustrating the differentially expressed mRNAs between pathological complete response (pCR) (n = 57) and non pCR groups (n = 113) of TNBC in the GSE25066 cohort. X-axes; log₂ (fold change), Y-axes; −log₁₀ p-value from limma analysis. mRNA with adjusted p-value < 0.05 are marked in blue, and top three genes of p-value are marked in red. (B) Receiver operating characteristic (ROC) curve of the 3-gene score and E2F targets score with the area under the curve (AUC) in the GSE25066 cohort. (C) ROC curve of the 3-gene score with AUC in the GSE20194 and HESS cohorts.

Figure 2

Association of the 3-gene score level with clinical characteristics in the GSE25066 and METABRIC cohorts. Boxplots of the 3-gene score level by (A) subtype in the whole cohort, and (B) Nottingham pathological grade (grade 1 and 2 vs. grade 3) and American Joint Committee on Cancer pathological stages in TNBC. Correlation plots between the 3-gene score and MKI67 gene expression. The Kruskal–Wallis test or Mann–Whitney U test was used accordingly.

Figure 3

Gene set enrichment analysis (GSEA) of high 3-gene score triple-negative breast cancer (TNBC) in the GSE25066, METABRIC, and TCGA cohorts. Enrichment plots of hallmark E2F targets, G2M checkpoints, and mitotic spindle gene sets in the GSE25066, METABRIC, and TCGA cohorts superimposed with a normalized enrichment score (NES) and false discovery rate (FDR) are shown. NES and FDR were determined with the classical GSEA method, where FDR < 0.25 is considered significant.

Figure 4

Association of the 3-gene score with mutation load and tumor-infiltrating immune cells. (A) Boxplots of the level of the mutation-related score; homologous recombination deficiency (HRD), silent and non-silent mutation load, fraction altered, single-nucleotide variant (SNV), and indel neoantigens by low and high 3-gene score triple-negative breast cancer in the TCGA cohorts. Boxplots of the fraction of (B) anti-cancer immune cells; CD8⁺ T cells, CD4⁺ T cells, T helper type 1 (Th1) cells, M1 macrophages, γδT cells and (C) pro-cancer immune cells; Regulatory T cells (Tregs), T helper 2 (Th2) cells, M2 macrophages, by low and high 3-gene scores TNBC in the GSE25066 and METABRIC cohorts. (D) Boxplots of the 3-gene score by tumor, immune, and stromal cells in single-cell sequence data (GSE75688 cohort). Mann–Whitney U or Kruskal–Wallis test were used to calculate p values.

Figure 5

Correlation of the 3-gene score with treatment response in cell lines. Correlation plots between the 3-gene score level and area under the curve (AUC) of several drug sensitivity, paclitaxel, 5-fluorouracil, cyclophosphamide, and doxorubicin, for TNBC cell lines. Spearman rank correlation was used for the analysis.

Figure 6

Association of the 3-gene score with drug response for TNBC cell lines and patients. Bar plots of the comparison of the pCR rate after NAC between the 3-gene score low (blue) and high (orange) groups in the GSE25066 (n = 170), GSE20194 (n = 68), and HESS (n = 27) cohorts. Fisher’s exact test was used for the analysis. Group sizes are shown underneath the bar.

Figure 7

Association between the 3-gene score and the survival of patients with triple-negative breast cancer (TNBC) with or without chemotherapy. (A) Kaplan–Meier plots of comparison between low (blue line) and high (red line) 3-gene score groups for disease-free survival (DFS) in the GSE25066 cohort. (B) Kaplan–Meier plots of the comparison between low and high 3-gene score groups for overall survival (OS), DFS, and disease-specific survival (DSS) in the treatment group and non-treatment group in the METABRIC cohort. The top one-third was defined as the high-score group within the cohort. The log rank test was used to calculate the p values.