Breast cancer is a heterogeneous disease at multiple levels, ranging from subtype differences between patients (inter-patient heterogeneity) to the diverse composition of malignant cells, heterogeneity of hormone receptor (HR) expression and cellular makeup within single breast cancer samples (intra-tumour heterogeneity). Despite an increase in the number of effective therapies available, many patients will experience an incomplete treatment response and subsequent relapse. These adverse treatment outcomes may be attributed to the often overlooked but critical factor of cellular heterogeneity.
To gain deeper insights into breast cancer heterogeneity, we applied single-cell technologies to a cohort of 250 primary, untreated breast cancers. We optimized tissue cryopreservation and dissociation methods, improved capture designs to account for sample biology, and incorporated PBMC spike-ins. These advancements eliminated the need for fresh sample processing, enhanced tissue dissociation, and improved handling of small tissues, such as biopsies. Together, resulting in a process that reduces batch effects, is cost-efficient, and yields better sample recovery and cell quality. For accurate and reliable data analysis, we developed a scalable computational workflow incorporating benchmarked methods for SNP-demultiplexing, high-resolution cell annotation, and cellular integration. Building on these methods, we extended our approach to study the cellular heterogeneity of breast cancers. Our method, scSubtyper, examines phenotypic differences among malignant cells within tumours by comparing single-cell features to distinct molecular subtypes, assigning each cell to one of these subtypes. Preliminary results revealed that over 90% of samples exhibit a mix of malignant cells from different subtypes, and 50% of samples contain cells representing all four molecular subtypes. This highlights the substantial heterogeneity among malignant cells within a tumour. In addition, our ecotyping approach groups samples based on patterns of cell type frequencies and their co-occurrences. Preliminary results identified five ecotypes across all subtypes and four within the ER-positive subtype. Interestingly, these ecotypes lack significant associations with clinical subtypes but are characterized by distinct abundances of immune and stromal cells. This suggests that ecotypes are not merely surrogates for clinical or molecular subtypes but may play a role in driving differential treatment responses. Lastly, further analysis of luminal tumours revealed an association between ER heterogeneity and survival, as well as ER and PR co-expression and recurrence events.
Together, our high-throughput tissue processing and computational approaches to studying intra-tumour heterogeneity are now being applied to our large, extensively annotated, clinical cohort. Supported by the preliminary results, we hypothesize that this study will play a vital role in optimizing breast cancer patient stratification to improve treatment management and outcome.