Breast cancer is a clinically and genetically heterogeneous disease. Whilst single-cell studies have advanced our understanding of the underlying biological diversity of this disease, a major limitation in previous studies is that they fail to capture the broad spectrum of breast cancer subtypes and lack the statistical power to resolve subtype-specific differences. To address this gap, we generated a comprehensive breast cancer single-cell atlas designed to robustly characterise transcriptomic and genomic heterogeneity across all major clinical subtypes.
Focusing on treatment-naïve tissues to capture disease biology prior to therapeutic intervention, we profiled 200 patient samples to generate a comprehensive multiomics atlas of scRNA-seq, whole transcriptome sequencing (WTS), whole genome sequencing (WGS), and clinicopathological data, such as age, histological grade and type, ER/PR/HER2 measurements, stromal tumour-infiltrating lymphocyte (sTIL) and PD-L1 status.
To quantify inter- and intra-tumoral heterogeneity, we developed a framework of breast cancer archetyping that positions malignant epithelial cells along axes capturing their intrinsic properties. Archetyping successfully resolves biologically meaningful gradients and systematically quantifies transcriptomic heterogeneity. Our results show that luminal cancers exhibit a large degree of intra- and inter-tumoural gene expression diversity. We define the key gene programs that contribute to this variability and identify immune programs among luminal cancers. Using sTIL status as a measure of immunogenicity, we identified an underappreciated subset of luminal tumours exhibiting high immune infiltration—challenging the prevailing view that luminal cancers are uniformly “immune cold.” Using WGS we identified distinct mutagenic processes driving these immune hot and cold disease phenotypes, independent of tumour mutational burden or homologous recombination definiency.
Our single-cell multiomics breast cancer atlas spanning all major clinical subtypes provides a high-resolution framework for dissecting the multilayered heterogeneity of breast cancer. By maximizing patient representation and integrating rich clinical and genomic metadata, this resource enables robust identification of subtype-specific biology and reveals striking diversity within luminal breast cancers, including a substantial subset with immune-hot characteristics. Ongoing work will leverage this atlas and our deeply curated metadata, including survival outcomes, to refine breast cancer stratification and identify clinically actionable tumour ecosystems.