Spatial transcriptomics reveal pitfalls and opportunities for the detection of rare high-plasticity breast cancer subtypes

Abstract

Breast cancer is one of the most prominent types of cancers, in which therapeutic resistance is still a major clinical hurdle. Specific subtypes like Claudin-low (CL) and metaplastic breast cancers (MpBC) have been associated with high non-genetic plasticity, which can facilitate resistance. The overlaps and differences between these orthogonal subtypes, respectively identified by molecular and histopathological analyses, are however still insufficiently characterised. Adequate methods to identify high-plasticity tumours to better anticipate resistance are furthermore still lacking. Here we analysed 11 triple negative breast tumours, including 3 CL and 4 MpBC samples, via high-resolution spatial transcriptomics. We combined pathological annotations and deconvolution approaches to precisely identify tumour spots, on which we performed signature enrichment, differential expression and copy-number analyses. We used the TCGA and CCLE public databases for external validation of expression markers. By levying spatial transcriptomics to focus analyses only to tumour cells in MpBC samples, and therefore bypassing the negative impact of stromal contamination, we could identify specific markers that are not expressed in other subtypes nor stromal cells. Three markers (BMPER, POPDC3 and SH3RF3) could furthermore be validated in external expression databases encompassing bulk tumour material and stroma-free cell lines. We find that existing bulk expression signatures of high-plasticity breast cancers are relevant in mesenchymal transdifferentiated compartments but can be hindered by stromal cell prevalence in tumour samples, negatively impacting their clinical applicability. Spatial transcriptomics analyses can however help identify more specific expression markers, and could thus enhance diagnosis and clinical care of rare high-plasticity breast cancers.