Differential Survival and Therapy Benefit of Patients with Breast Cancer Are Characterized by Distinct Epithelial and Immune Cell Microenvironments

Abstract

Purpose: Extensive work in preclinical models has shown that microenvironmental cells influence many aspects of cancer cell behavior, including metastatic potential and their sensitivity to therapeutics. In the human setting, this behavior is mainly correlated with the presence of immune cells. Here, in addition to T cells, B cells, macrophages, and mast cells, we identified the relevance of nonimmune cell types for breast cancer survival and therapy benefit, including fibroblasts, myoepithelial cells, muscle cells, endothelial cells, and seven distinct epithelial cell types.

Experimental design: Using single-cell sequencing data, we generated reference profiles for all these cell types. We used these reference profiles in deconvolution algorithms to optimally detangle the cellular composition of more than 3,500 primary breast tumors of patients that were enrolled in the SCAN-B and MATADOR clinical trials, and for which bulk mRNA sequencing data were available.

Results: This large data set enables us to identify and subsequently validate the cellular composition of microenvironments that distinguish differential survival and treatment benefit for different treatment regimens in patients with primary breast cancer. In addition to immune cells, we have identified that survival and therapy benefit are characterized by various contributions of distinct epithelial cell types.

Conclusions: From our study, we conclude that differential survival and therapy benefit of patients with breast cancer are characterized by distinct microenvironments that include specific populations of immune and epithelial cells.

Figure 1.

Heterogeneity of breast tumors analyzed through single-cell mRNA sequencing. A, UMAP dimensional reduction of all cells using the Seurat analysis suite. Each color and each number represent an individual cluster. B, The same UMAP plot as in A highlighting the patient of origin for each single cell. C, Expression of cell type–specific marker genes across all cell clusters. Size of the dot indicates the percentage of cells within a cluster that express a marker gene; the color indicates the average level of expression in a cluster. The marker genes are examples of significantly enriched genes that can be found in each individual cell cluster.

Figure 2.

Deconvolution of the MATADOR bulk mRNA sequencing data. A, Results of the deconvolution of the bulk mRNA sequencing data from the MATADOR trial using TCD. The top bars indicate the pathologic subtype, PAM50 classification, total immune content, and 5-year recurrence status of each patient. The middle bars show the contribution of all the ECs. The bottom set of bars shows the contribution from the microenvironmental cell types as defined by TCD. B, Overall survival probabilities for triple-negative patients with low (<median) or high (>median) immune infiltrate as determined by TCD.

Figure 3.

Deconvolution of the SCAN-B bulk mRNA sequencing data. A, Results of the deconvolution of the bulk mRNA sequencing data from the SCAN-B trial using TCD. The top bars indicate the pathologic subtype, PAM50 classification, total immune content, and 5-year recurrence status of each patient. The middle bars show the contribution of all the ECs. The bottom set of bars shows the contribution from the microenvironmental cell types as defined by TCD. B, OS probabilities for triple-negative patients with low (<median) or high (>median) immune infiltrate as determined by TCD.

Figure 4.

Considering the complete cellular heterogeneity has prognostic value. A, Survival probabilities for the three different patient groups in the SCAN-B training set as ranked by CTS. Determination of the three patient groups is based on quartiles: 25% of patients with the lowest risk, 25% of patients with the highest risk, and 50% of the patients in the intermediate-risk group. B, Survival probabilities for the three different patient groups in the SCAN-B validation set. C, Survival probabilities for the three different patient groups in the MATADOR cohort. D and E, EC-II, V, VI and myoepithelial-, B-, T cells, and macrophages are the main contributors for determining CTS. The contribution of these cell types in the three different patient groups for the SCAN-B validation set (D) and MATADOR set (E).

Figure 5.

Considering the complete cellular heterogeneity for chemotherapy selection. A, Cox proportional hazard model for patients treated with TAC (orange) or ddAC (blue) based on their DTS^CT as determined by the Cox model with feature selection (see Methods). x-axis depicts the DTS^CT, y-axis depicts the logarithm of the HR, gray areas indicate the 95% confidence interval of the HR. P value depicted in the panel is for the entire Cox proportional hazard model. P value for the interaction between DTS^CT and the chemotherapy treatment is <0.001 and <0.001 for uncorrected and models corrected for clinicopathologic factors, respectively. B, Two subgroups were assigned: 50% of the patients with the lowest DTS^CT and 50% of the patients with the highest DTS^CT. Shown is the forest plot for all patients that had an event in the MATADOR trial. The dotted line indicates the separation between the high-risk (high DTS^CT) and low-risk (high DTS^CT) patients. C, Survival probabilities for the high-risk and low-risk patient groups treated with either TAC or ddAC. D, Results of the deconvolution of the bulk mRNA sequencing data from the MATADOR trial using TCD. The bars show the contribution from the cell types as defined by TCD. E, EC-V, myoepithelial cells, and T cells are the main contributors for determining DTS^CT. The plot illustrates the contribution of these cell types in the two patient groups.