Article Text
PDF
Abstract
Introduction/Background High grade serous ovarian cancer (HGSOC) is generally detected in an advanced stage with poor long-term survival. This therapeutic resistance is attributed to its extensive inter- and intratumoural heterogeneity. Several non-tumoural cells within the tumour specimen — such as immune cells, fibroblasts and endothelial cells — are increasingly recognized as important contributors to tumoural development as they shape the microenvironment and the tumour's ability to metastasize.
Methodology In this project, we employed ´3 single-cell RNA-sequencing (scRNA-seq) on randomly dissociated cells from tumour biopsies of seven HGSOC patients before first-line treatment.
Results Based on gene expression and principal component analysis, we clustered 18,403 cells into 11 tumoural and 35 non-tumoural cell subtypes (figure 1). Analysis of differential gene expression enabled us to correlate known and new biological functions to most subtypes. Furthermore, we selected specific markers genes (MG) for each subtype by differential gene expression analysis within and across cell subtypes. As a downstream analysis, we evaluated these subtype-specific MG in publically available datasets containing bulk expression and curated clinical annotation of 1467 HGSOC patients. We performed a meta-analysis using multivariate Cox proportional hazards regression, adjusting for age, residual disease and FIGO stage. This allowed us to discover a prognostic effect for 7 cell subtypes across cohorts, of which 3 were associated with improved overall survival (OS) (1 type of dendritic cell, b-cell and t-cell) and 4 were associated with worsened OS (2 types of fibroblasts, 1 type of tumour cell and endothelial cell).
Conclusion ScRNA-seq enabled us to disentangle the heterogeneous microenvironment of HGSOC and detect several relevant tumoural and non-tumoural cells that influence OS. These subtypes can act as novel biomarkers for treatment stratification. Quantification of these novel cell types by specific MG in bulk RNA data can make clinical implementation feasible and pave the way to a more personalized therapy.
Disclosure Nothing to disclose
t-distributed stochastic Neighbour Embedding of the 18,403 profiled single cells with each colour coding for a cell subtype based on differential expression analysis (n=46)