The study design and sequencing cohort
Figure 1 illustrates our study design. In this study, we integrated multi-omics data of TNBC from FUSCC and single-cell sequencing data from Ayse Bassez et al. for subsequent analysis (19, 21). There were 2217 TNBC patients with clinical prognosis data treated at our center, used for analyzing the clinical prognostic features of low Ki-67 TNBC. RNA-seq data from 360 cases were used for analyzing differentially expressed genes, metabolic pathways and the immune microenvironment. Additionally, 279 patients with available whole-exome sequencing (WES) data were used for elucidating somatic mutation characteristics. The single-cell sequencing (scRNA-seq) data included 13 cases, comprising 8172 tumor cells and 3177 myeloid cells, which were used for cell annotation, cell chat and evolutionary trajectory analysis.
Clinical and Prognosis Features of low Ki-67 TNBC
Our data indicated that, compared to Luminal and HER-2 overexpression breast cancer, TNBC had a higher proportion of patients with high expression of Ki-67 tumors (p < 0.001) (Fig. 2A). Only 20.5% of TNBC cases exhibited tumors with low expression of Ki-67. Low Ki-67 TNBC is characterized by an older age at diagnosis, lower tumor grade, and a lower prevalence of the invasive ductal carcinoma (IDC) histological type (p < 0.001). However, there is no difference in tumor grade between the low Ki-67 and high Ki-67 groups (Fig. 2B). Figure 2C showed the distinct treatment strategies and survival outcomes between low Ki-67 and high Ki-67 TNBC. Consistently, the survival analyses indicated that breast cancers with low Ki-67 expression displayed improved recurrence-free survival (RFS) and overall survival (OS), regardless of the molecular subtypes (all p < 0.001) (Figure. S1,S2).
Previous studies have classified TNBC into different subtypes based on gene expression characteristics or metabolic pathway features (19, 22). We further investigated the distribution of molecular subtypes in TNBC with different Ki-67 index. The results revealed that low Ki-67 TNBC was mainly composed of the metabolic-pathway-based subtype 1 (MPS1) subtype (61.4%), characterized by upregulated lipid metabolism; In contrast, high Ki-67 TNBC was predominantly composed of the glycolytic subtype MPS2 (44.7%) and the mixed subtype MPS3 (44.7%) (Fig. 2E). We determined the Fudan transcriptome-based TNBC subtypes for each sample and investigated their overlap with the Ki-67 groups. Low Ki-67 group primarily contained the luminal androgen receptor (LAR) subtype (50.9% in Ki-67 Low vs. 15.4% in Ki-67 High), while the high Ki-67 group had a relatively larger proportion of the basal-like immune-suppressed (BLIS) subtype (50.9% in Ki-67 High vs. 15.4% in Ki-67 Low) (Fig. 2E).
Mutation Profile of low Ki-67 TNBC
Genomic alterations contribute to the high heterogeneity of malignant tumors, and exploring these mutations shows potential in providing valuable biomarkers for precise cancer therapy. Figure 3A illustrated the landscape of the top 15 somatic mutations in TNBC, and these findings demonstrated a significant difference between low Ki-67 TNBC and high Ki-67 TNBC. Low Ki-67 TNBC had a higher frequency of AKT1 mutations (low Ki-67 vs. high Ki-67: 11% vs 2%, p < 0.05), while high Ki-67 TNBC exhibited a higher proportion of TP53 mutations (high Ki-67 vs. low Ki-67: 79% vs 47%, p < 0.001) (Fig. 3A).
To gain a more comprehensive understanding of the mutation profile, we conducted an analysis of mutational profiles based on pathways. Low Ki-67 TNBC had more somatic mutations in PI3K-AKT-mTOR signaling pathway (low Ki-67 vs. high Ki-67: 50% vs 31%, p < 0.05) and RTK-RAS signaling pathway (low Ki-67 vs. high Ki-67: 29% vs 26%, p < 0.05); while high Ki-67 TNBC exhibited a higher proportion of TP53 signaling pathway (high Ki-67 vs. low Ki-67: 80% vs 47%, p < 0.001) (Fig. 3B). Furthermore, we investigated the significant co-occurring and mutually exclusive alterations within the different pathways. We observed a significant mutually exclusive pattern unique to the high Ki-67 group, involving the TP53 and TGF-β signaling pathways (p < 0.05) (Fig. 3C). Additionally, significant co-occurring alterations were exclusively found in the low Ki-67 group: TP53 and NOTCH signaling pathways (p < 0.05); PI3K-AKT-mTOR and RTK-RAS signaling pathways (p < 0.05) (Fig. 3C). Additionally, we explored the genomic stability and observed that the low Ki-67 TNBC group exhibited lower homologous recombination deficiency (HRD) scores (p < 0.01) and lower loss of heterozygosity (LOH) scores (p < 0.05) (Fig. 3D).
Dysregulated metabolic pathways and targetable metabolic genes in the low Ki-67 group
In this study, we performed differential expression analysis between the low and high Ki-67 groups. The volcano plot displayed 1425 upregulated genes represented in red and 2266 downregulated genes shown in green for the low Ki-67 group, with a threshold of |log2FC| > 2.5 and adjusted p-value < 0.05 (Fig. 4A). These results indicated significant differences in gene expression between the two groups and suggested potential candidate genes associated with the Ki-67 index in TNBC. Subsequently, we utilized the hallmark gene sets from the Molecular Signatures Database (MsigDB) to analyze the pathway alterations (23). In the low Ki-67 group, we observed a significant upregulation of various metabolic pathways, including Fatty Acid Metabolism, Cholesterol Homeostasis and Adipogenesis (Fig. 4B). Additionally, in the high Ki-67 group, there was an upregulation of cell cycle signaling pathways containing Mitotic Spindle, G2M Checkpoint, etc. (Fig. 4B).
In order to uncover the metabolic heterogeneity of low Ki-67 TNBC, we used gene set variation analysis (GSVA) to estimate the enrichment scores of metabolic pathways (24). Consistent with the GSVA results, the low Ki-67 group exhibited an upregulation of lipid metabolism, while the high Ki-67 group showed an upregulation of nucleotide metabolism (Fig. 4C, S3A). The RNA-seq results revealed that metabolic genes in the Fatty Acid Biosynthesis and Biosynthesis of Unsaturated Fatty Acids pathways were upregulated in the low Ki-67 group (Fig. S3B, S3C). We further validated the upregulation of metabolic pathways in the low Ki-67 group using qRT-PCR experiments. Three TNBC cell lines were selected for analysis, and it was observed that the E0771 cell line had low Ki-67 expression, while the AT3 and 4T1 cell lines exhibited relatively higher Ki-67 expression. Additionally, the cell lines with low Ki-67 expression showed increased expression of metabolic genes in the Fatty Acid Biosynthesis and Biosynthesis of Unsaturated Fatty Acids pathways (Fig. 4D, 4E).
Tumor microenvironment and heterogeneity of myeloid cells
To explore the features of the tumor immune microenvironment (TIME) in the low Ki-67 group, we utilized the CIBERSORT method to calculate the abundance of 22 identified immune cells (25). The results revealed significant differences in the abundance of three immune cells between the low Ki-67 and high Ki-67 groups, including Macrophage 2 (M2) (p < 0.01), memory B cells (p < 0.001), and activated Dendritic cells (p < 0.05) (Fig. 5A, S4A). ScRNA-seq data were utilized to comprehensively assess the expression profile of myeloid cells. After quality filtering, a total of 3177 myeloid cells were analyzed. Using t-distributed stochastic neighbor embedding (t-SNE), we identified four clusters, including Macrophage1, Macrophage2, Monocyte, and Dendritic cells (Fig. 5B-5D). Figures F and G depicted the clustering markers, facilitating the linkage of clusters to specific cell types, including M1-like macrophages (CD163, CD68, MRC-), M2-like macrophages (CD163, CD68, MRC+), Monocyte (FCN1), and Dendritic cells (BIRC3) (Fig. S4C-S4F).
M1-like macrophages (anti-tumor) and M2-like macrophages (pro-tumor) are two major macrophage subtypes (26). Our results revealed an increased prevalence of M2-like macrophages in the low Ki-67 group, indicating the presence of an immunosuppressive microenvironment (Fig. 5E). Pathway analysis of macrophages revealed that in the high Ki-67 group, M1-like macrophages exhibited upregulation of immune-related signaling pathways, including the Complement pathway and Interferon Alpha/Gamma Response. Conversely, in the low Ki-67 group, M2-like macrophages showed upregulation of pro-tumorigenic signaling pathways, such as Wnt Beta-Catenin Signaling pathway (Fig. 5H). Gene analysis further revealed that M1-like macrophages in the high Ki-67 group exhibited upregulation of immune activation-related genes such as C1QA, ISG15, C1QC, C1QB, and others (Fig. 5I).
In summary, our results suggested that the low Ki-67 group was predominantly infiltrated by M2-like macrophages, promoting tumor progression, while the high Ki-67 group was mainly infiltrated by M1-like macrophages, exhibiting an immune-activating anti-tumor effect.
Tumor cell heterogeneity and interactions between tumor cells and myeloid cells
To investigate the heterogeneity of tumor cells in TNBC, we identified four clusters of tumor cells in TNBC through t-SNE analysis, and each cluster exhibited distinct transcriptomic signatures (Fig. 6A, 6B, S5A). Among them, cluster 1 was the major cell population in the low Ki-67 group, while clusters 2 and 3 were specifically enriched in the high Ki-67 group (Fig. 6C). Gene ontology (GO) analyses revealed that cluster 0 was mainly enriched in Extracellular matrix structural constituent; cluster 1 showed enrichment in Oxidative phosphorylation and Aerobic respiration; cluster 2 was primarily enriched in Extracellular matrix structural constituent; and cluster 3 exhibited enrichment in Pyridine nucleotide metabolic process and Nicotinamide nucleotide metabolic processes (Fig. 6D-6G). These results suggested differential intrinsic biological characteristics in tumor cells between low and high Ki-67 groups, with upregulated lipid metabolism in low Ki-67 cells and increased proliferation and nucleotide metabolism in high Ki-67 cells.
We conducted pseudo-time analysis to investigate the cell lineage trajectory and found that the low Ki-67 tumor cells were in the early stages of cell differentiation (Fig. 6H-6J). Cell Chat analysis indicated that tumor cells and macrophages mostly interacted through the MIF (CD74 + CD4) receptor-ligand interaction (Fig. 6K, S5B, S5C). Moreover, cluster 1, as the sender, exhibited the most significant impact on M2 macrophages through the MIF signaling pathway (Fig. 6L).