Composition of three major breast cancer subtypes at cellular resolution.
To unambiguously investigate the distinct cellular composition among ER+, HER2 + and TNBC tumors, we reanalyzed the public human scRNA-Seq data (13 ER+, 6 HER2 + and 8 TNBC samples)[20]. A total of 150,293 cells were detected with an average of 5,566 cells per patient. After normalization, data from individual tumors were integrated using Harmony alignment[27] and cell clusters were further unbiasedly identified by Louvain graph-based clustering. According to the expression of canonical cell type gene markers, 150,293 cells were annotated as 8 major cell types, including epithelial cells (EPCAM and CDH1), mesenchymal cells (PDGFRB, TAGLN), fibroblasts (DCN, POSTN), endothelial cells (VWF, FABP4), macrophages (ITGAM, CSF1R), T cells (CD3D, CD3E), B cells (MS4A1, CD79A) and Plasma cells (JCHAIN, IGKC) (Fig. 1A-C). The known pan-epithelial cell marker, EPCAM, was expressed in five distinct breast epithelial subpopulations, luminal progenitor cells (3.42% of all cells; KIT, SOX10), mature luminal cells (45.60%; ESR1, FOXA1), basal (myoepithelial) cells (1.79%; KRT5, MYLK), ISG+ epithelial cells (2.53%; IFIT3, ISG15) and cycling epithelial cells (10.51%; MKI67, TOP2A) (Fig. 1B and C). Among 5 epithelial subpopulations, basal cells had lower EPCAM expression than other epithelial cells (Fig. 1B), as reported in the previous study[28]. The initiation of metastasis relies on epithelial to mesenchymal transition (EMT)[29]. In contrast to ER + and HER2 + tumors, the EMT transcription factor, SNAI2 (SLUG), was abundantly expressed in the epithelial cells of TNBC samples (Fig. 1C). In addition, TNBC tumors have lower proportion of mature luminal cells while higher cycling epithelial cell proportion than ER + or HER2 + tumors (Fig. 1D and S1), indicating that epithelial cells within TNBC tumors have higher stemness properties, which endows TNBC cells with higher invasive and metastatic potential.
TREM2 + macrophages play as a key regulator in TNBC microenvironment.
Due to the highest mutational burden in TNBC among breast cancer subtypes, TNBC has an increased propensity to generate neoantigens for immune activation[4], resulting in a greater number of tumor-infiltrating lymphocytes[30]. Besides T cells, we also observed elevated macrophages in TNBC compared to ER + and HER2 + tumors (Fig. 1D). TAMs have been reported to play an important role in breast cancer progression[31]. Based on the cell map in ER+, HER2 + and TNBC tumors, we revealed two dominant macrophage clusters that highly expressed MHC class II (MHC-II) genes (HLA-DRA) and TREM2 respectively, and a smaller subpopulation with high expression levels of ACP5, a histochemical marker of osteoblasts (Fig. 1A and B), suggesting the high heterogeneity of the macrophage lineage in breast cancer. Intriguingly, we noticed that the proportion of TREM2+ macrophages in the TNBC samples was higher than that in HER2 + and ER + tumors, but without statistical significance, which might be due to a large degree of variation of the tumor immune landscape in different patient specimens (Fig. 1D and S1).
In order to further investigate the intratumoral interaction differences among distinct breast cancer subtypes, we utilized CellChat to analyze the ligand-receptor interactions between 15 cell types[25]. Remarkably, T cells showed the highest interaction strength with other cells in both ER + and HER2 + tumors, while in TNBC, TREM2+ macrophages received stronger incoming signals than T cells (Fig. 2A). Due to the prominent role of TREM2+ macrophages in the cellular interaction network in TNBC, we then examined the communication probability (information flow) of the incoming signaling to TREM2+ macrophages across ER+, HER2 + and TNBC subtypes, and found 9 pathways were specifically active in TNBC tumors, including PECAM1, GAS and ADGRE5 (Fig. 2B and C). Of interest, PECAM1 (CD31), an endothelial cell marker, is essential for adhesion and accumulation of platelets and has been demonstrated to be capable of inducing EMT in cancer[32, 33]. All these results suggest the key role of TREM2+ macrophages in the TME of TNBC and the potential correlation between the macrophage subset and TNBC metastasis.
Characterization of TREM2 + macrophages in TNBC.
To establish whether comparable changes in TREM2+ macrophages occur in TNBC, we performed differential gene analysis using the MAST algorithm[34]. This analysis identified the significantly elevated gene expression of CLL18, CCL13 and basal cell markers of KRT14 and KRT15 in the TREM2+ macrophages of TNBC subtype (Fig. 3A and S2), wherein basal cytokeratin KRT14 is a hallmark feature of TNBC[35], and has been reported to promote TNBC peritoneal metastasis[36]. In addition, the previous study revealed that chemokine CCL18 secreted by TAMs could promote the invasion and metastasis of breast cancer[37]. In order to further determine the pathway level differences driving TREM2+ macrophage heterogeneity between TNBC and other breast cancer subtypes, we performed enrichment analysis using the differentially expressed genes, and revealed that the TREM2+ macrophages in TNBC were characterized by the up-regulation of cornification, keratinization, epidermal cell differentiation and response to cytokine (Fig. 3B). All these results support that TREM2+ macrophages might contribute to the progression and metastasis of TNBC.
Prognostic abilities of TREM2 + macrophage gene signature in human TNBCs.
To investigate the influence of TREM2 + macrophages in TNBC, we estimated the relative abundance of cell clusters from single cell data in the TCGA BRCA cohort using CIBERSORTx[26]. As shown in Fig. 4A, there was a significantly higher proportion of TREM2 + macrophages in the TCGA TNBC patients than non-TNBC patients. We further performed differential gene analysis in the comparison of tumor vs. normal samples and TNBC vs. Non-TNBC samples. Enrichment analysis was performed to determine the enrichment of TREM2+ macrophage gene signature derived from single cell data (Supplemental Table 1) in the whole-genome gene list ranked by differential gene expression score. The results showed that TREM2+ macrophage gene signature was only enriched in the ranked gene list based on expression differences between TNBC and Non-TNBC samples (Fig. 4B), but not between tumor and normal samples (Fig. 4C). Collectively, these observations demonstrate that TREM2+ macrophages are more abundant in TNBC patients.
As our data indicate that TREM2+ macrophage subset is an essential effector population in the microenvironment of TNBC, we explored the clinical value of TREM2+ macrophages in TNBC patients. 115 TNBC patients in the TCGA BRCA cohort were stratified according to the expression of TREM2+ macrophage gene signature derived from single-cell data. Interestingly, high TREM2+ macrophage score in TNBC patients correlated with worse 2-year survival outcomes (Fig. 5A), compared with low TREM2+ macrophage score (p = 0.015, Hazard Ratio [HR] = 9.221, 95% Confidence Interval [CI] of HR: 8.125–10.317). Nevertheless, TREM2+ macrophage score wasn’t the significant factor of the 5-year survival outcomes (Fig. 5B; p = 0.528, HR = 1.366, 95% CI of HR: 0.872–1.860). In addition, there were no statistically significant differences in neither 2-year nor 5-year survival outcomes of all TCGA BRCA patients (n = 1,092) between high and low TREM2+ macrophage score groups (Fig. 5C and 5D). These observations suggest that TREM2+ macrophages are only associated with the short-term disease outcomes in the TNBC patients, which might be due to the fact that several factors including treatment could influence the long-term survival outcomes of TNBC patients.