15P Portraying HER2-positive Breast Cancer Heterogeneity with Spatial Transcriptomics

HER2-positive breast cancer is a heterogeneous disease, presenting tumor and microenvironment features which can impact prognosis and treatment response. Here, we aimed at better understanding the heterogeneity of this disease by performing spatial transcriptomics (ST) on HER2-positive breast cancer samples. Spatial transcriptomics (Visium) was performed on 33 frozen HER2-positive breast cancer surgical samples, including 6 residual disease samples. H&E images of the ST slides were annotated for morphological structures at the single-cell/structure level (QuPath software). Clusters identified on integrated data (harmony R package) were characterized calculating gene expression signatures, including HER2DX gene modules, at the spot level, and by morphological annotation. Gene signatures were computed on pseudo-bulk RNA data as well. A total of 25 integrated clusters were identified (range 15-21 in each sample). As each spot/cluster represents a mixture of different cell types, using gene expression and morphological data we defined a total of 9 tumor-enriched clusters (of which 5 sample-specific), as well as 12 clusters mainly enriched for stroma, 3 for adipose tissue, and 1 for tumor-infiltrating lymphocytes. All samples presented more than 1 tumor cluster, in various proportions. Interestingly, when comparing tumor clusters, levels of HER2DX signatures depicting HER2 amplicon, luminal phenotype, proliferation, B cell infiltration, as well as signatures related to stroma activation, signaling pathways and metabolism differed, demonstrating heterogeneity in tumor-enriched areas. Of note, within the same sample, tumor clusters with high/low levels of the HER2DX modules and other signatures could co-exist, and samples presenting signature scores above/below the cohort median at the pseudo-bulk level (also influenced by the stroma composition) showed the co-presence of tumor clusters with high/low signature levels. Our findings highlight the heterogeneity of HER2-positive breast cancer. Spatial transcriptomics may help in refining gene expression signatures computed on bulk RNA, and these results open to further analyses aimed at better understanding the tumor microenvironment in this disease.