High resolution analysis of the tumour microenvironment of high grade serous ovarian cancer (HGSOC) using single cell transcriptomics and quantitative histopathological examination

Abstract

Ovarian cancer (OC) is a highly lethal gynaecologic malignancy with high grade serous ovarian carcinoma (HGSOC) accounting for more than 70% of all OC deaths. Although the prognosis for HGSOC patients has shown some improvement over the last 30 years with an overall five-year survival increasing from 31% to 46%, the disease remains highly recurrent with progressively shorter survival after each relapse. The remitting and relapsing course of this disease arises in part due to the cellular and molecular diversity within the tumour microenvironment (TME) accounting for the highly heterogenous nature of the disease. Bulk gene expression profiling has identified novel subgroups of HGSOC but only interrogates the average signal of cells within a tumour. Single cell RNA-sequencing (scRNA-seq) enables the quantification of gene expression from individual cells, allowing assessment of potential rare subpopulations including chemo-resistant tumour cells.

A major part of this study comprised a thorough methodological optimisation of tumour dissociation using fresh and cryopreserved samples in order to obtain high quality single cell suspensions (>70% viable cells) essential for further scRNA-seq. To investigate the heterogenous landscape of HGSOC, this study utilised droplet-based scRNA-seq to profile ~134,000 cells from multi-site tumour specimens with the representation of a high-resolution analysis of the HGSOC TME and further demonstration of the cellular subclonal phenotypes and their functional importance. Rare cell types were identified such as T cells with unique metallothionein expression profile in treatment resistant samples, and ovarian cancer stem cells (CSCs) which may be targeted in future personalised therapies. Bulk TCGA HGSOC transcriptomic subtypes were found to be driven by the different cellular composition within the HGSOC TME with cancer epithelial cells accounting for the ‘Differentiated’ subtype, immune cells for the ‘Immunoreactive’ subtype, fibroblasts for the ‘Mesenchymal’ subtype and endothelial cells for the ‘Proliferative’ subtype.

Quantitative digital histopathological analysis performed here using clinically and molecularly well-annotated Edinburgh cohorts demonstrated the value of using tumour stromal proportion (TSP) and necrosis as a histological feature for helping to stratify and prognosticate the disease. Patients with stroma-rich and lack of necrosis tumours at pre-chemotherapy stage were found to have poorer prognosis. Combined with bulk transcriptomic profiling of these tumours, ‘stroma-rich’ signature was associated with pro-tumourigenic biological processes including epithelial-mesenchymal-transition (EMT); while ‘necrosis’ signature was linked with anti-tumourigenic processes such as upregulation of T cell proliferation.

Finally, integration of digital histopathological stratification using TSP with molecular markers from bulk ‘stroma-rich’ signature and single cell cancer-associated fibroblast (CAF) markers was performed to deconvolute the clinical significance of the stromal compartment using a multi-model characterisation. Patients with TSP of 0.5 or greater had increased expression of ‘stroma-rich’ and pro-tumourigenic CAF markers. A prognostic multi-gene signature (comprised of 26 genes) was constructed from the intersection of bulk and single cell stromal biomarkers which may aid risk stratification for HGSOC patients and facilitate future personalised therapies.

In summary, these data expanded our knowledge of the heterogenous TME of HGSOC with extensive cross-talk occurring between the diverse cellular and molecular components which should be better interrogated using multi-omics studies.