Abstract 609: Contribution of the tumor and stroma compartments for TNBC molecular classification using spatial transcriptomics analysis

Abstract

Abstract Introduction Triple negative breast cancer (TNBC) is a heterogeneous disease associated at least with five distinct molecular subtypes namely basal like (BL), immunomodulatory (IM), luminal AR (LAR), mesenchymal (M) and mesenchymal-stem like (MSL). Spatial transcriptomics (ST) interrogates gene expression in spatially defined spots. Here, we aimed to assess the contribution of tumor and stroma compartments in defining each TNBC molecular subtype by comparing bulk tumor RNAseq, pseudo-bulk (PB) derived from ST, tumor and stroma ST spots. Methods Spatially resolved gene expression profiles were obtained from 94 TNBC samples using ST, spots with artifacts or less than 500 reads being excluded from further analysis. Forty-one TNBC samples were manually annotated by a breast pathologist into 3 different classes: tumor, stroma and the rest. Manual histological annotation was used to build a classifier to assign each ST spot to tumor or stroma compartment. Bulk RNAseq data was obtained from consecutive sections for all tumors using RiboZero kit. PB profiles were obtained by summing the reads from all the spots of a given tumor, as well as from all spots classified as tumor and stroma compartment. A linear booster classifier was used to classify the spots based on gene expression. Each bulk and PB was assigned to one of 5 TNBC molecular subtypes. Results PB data derived from ST analysis identified 18 BL, 26 IM, 10 LAR, 27 M and 13 MSL among the 94 TNBC tumor samples. An 87% concordance in the subtyping was observed between the bulk and PB data, with half of the discordant cases being between BL and IM subtypes. All LAR and BL tumors, as well as most M (96%) were consistently classified using tumor spots only whereas all MSL (100%) and the majority of LAR (91%) and IM (92%) samples were concordant using stroma spots only. In half of the IM samples, the signal derived from both the tumor and the stroma compartments while in the other half derived from the stroma. Misclassification mainly occurred between BL and IM subtypes, as well as between M and MSL. LAR subtype appeared to be the most stable one. Conclusion The use of spatial transcriptomics allowed to show the relative contribution of the tumor and stroma compartments for TNBC molecular classification. The LAR, BL and M subtypes were driven by signals from the tumor compartment while MSL was driven by the stroma compartment. The IM signal was derived by both the tumor and stroma compartments. Citation Format: David Venet, Xiaoxiao Wang, Floriane Dupont, Ghizlane Rouas, Linnea Stenbeck, Annelie Mollbrink, Denis Larsimont, Joakim Lundeberg, Françoise Rothé, Christos Sotiriou. Contribution of the tumor and stroma compartments for TNBC molecular classification using spatial transcriptomics analysis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 609.