Abstract 3762: Single-cell analysis of cancer-associated fibroblast heterogeneity in non-small cell lung cancer: Mapping molecular phenotypes in tumors

Abstract

Abstract This work aims to characterise the heterogeneity and spatial relationships of the cancer-associated fibroblast (CAF) population in non-small cell lung cancer. Fresh human lung tissue was dissociated for sixty minutes to extract the maximum possible proportion of fibroblasts. Single-cell RNA sequencing was performed using a droplet-barcoded platform (Drop-seq). Quality control was performed on the raw sequencing data and resulting gene expression matrix. Bioinformatic analysis was performed using multiple packages in R. Spatial relationships between cell types were assessed using a multi-immunohistochemical (IHC) staining technique. We developed a workflow for efficient processing of raw Drop-seq data including quality control, normalisation and visualisation. Low-quality events were identified by integrating previously-described and novel quality-control metrics into a machine learning (random forest) model, and demonstrated that this approach improves clustering quality. Applying this method to samples from twelve non-small cell lung cancer (NSCLC) patients, we identified 5 distinct fibroblast subtypes; 3 predominantly derived from normal tissue and 2 largely from tumor samples. Of the normal subtypes, one showed gene expression consistent with the previously-described "inflammatory" fibroblast phenotype. Trajectory analysis identified a branched differentiation process from normal to CAF phenotypes, suggesting that these cells share a common initial activation before differentiation to either a "matrix remodelling" or "hypoxic" subtype. The prevalence and impact of these sub-populations appears to differ between NSCLC subtypes. The "matrix remodelling" subtype is present in both adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC), but confers a negative prognostic effect in LUAD only; the "hypoxic" phenotype appears relatively LUSC-specific. Multiplexed IHC using identified cluster markers demonstrated that these subtypes have different spatial distributions and relationships to other cell types. CAF remain a poorly-characterised population, despite their abundance in most solid cancers. No single molecular marker identifies all CAF, and there has been a scarcity of evidence regarding the existence of distinct subtypes and whether such subgroups have different functions. Our analysis has revealed five distinct CAF subtypes in NSCLC. In addition to divergent differentiation pathways, these subtypes have differential gene set enrichment, indicative of functional differences. In keeping with this, the phenotypes show distinct prognostic impact across NSCLC subtypes. Characterisation of CAF subgroups associated with aggressive tumor progression may facilitate identification of novel stromal targeting strategies. Citation Format: Sara Waise, Christopher J. Hanley, Rachel Parker, Christian H. Ottensmeier, Matthew Rose-Zerilli, Gareth J. Thomas. Single-cell analysis of cancer-associated fibroblast heterogeneity in non-small cell lung cancer: Mapping molecular phenotypes in tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 3762.