The tumour microenvironment is a complex cellular ecosystem where heterotypic interactions can play central roles in disease progression and response to therapy. For instance, there is emerging evidence that subsets of stromal cells can contribute to immune evasion within tumours. However, our knowledge of the tumour microenvironment, its composition and organisation remains unclear, limiting the development of effective targeted therapies.
Here we resolved the cellular architecture of primary human breast cancer tissues using single cell multi-omics (scRNA-Seq, CITE-Seq, snRNA-Seq & scATAC-Seq) and spatial transcriptomics (Visium). Our analysis revealed heterogeneity for stromal phenotypes existing within every tumour, which has confounded previous ‘bulk’ studies. Using pseudo-temporal analysis, we provide evidence that differentiation can drive transitions between the major stromal cell states. In particular, we describe functionally distinct states of cancer-associated fibroblasts (CAFs) which shared features of mesenchymal-stem cells and were enriched for immune-regulatory pathways, cell surface checkpoint molecules and prognostic associations with immune evasion. Spatial mapping revealed that CAF niches also co-localised with T-cell subsets in tumour tissues, where signalling pathways related to immune evasion were further enriched. We next developed a strategy to prospectively isolate and culture human CAFs ex vivo. Using CAF-T-cell co-culture systems, we show that CAFs have a strong immunosuppressive effect on T-cell phenotypes. Furthermore, integrated analysis identified important transcriptional regulators of CAFs, identifying candidate gene regulatory networks mediating CAF phenotypes, for future experimental studies.
Our multi-omics analysis provides important advances in generating a comprehensive stromal taxonomy of carcinomas. We envisage such insights to guide the development of therapies targeting stromal-directed immune evasion in cancer.