Abstract
The tumor microenvironment (TME) shapes response to immune checkpoint blockade (ICB). Several pan-cancer single-cell RNA-sequencing (scRNA-seq) analyses have reported how TME heterogeneity profoundly differs between cancer types. These studies mainly focused on one cell type (e.g., T-cells) and combined different technologies and bioinformatics pipelines with data being collected from both published and newly generated datasets. A comprehensive pan-cancer scRNA-seq map of the TME involving all cell types is therefore still lacking.
We obtained scRNA-seq (10x Genomics) on 234 fresh tissue samples from 161 treatment-naïve patients. Samples were collected from 9 cancer types and subjected to a uniform in-house optimized protocol of tissue dissociation, sequencing and bioinformatics analysis. Abundancies of cell types and subtypes were correlated with each other and a tumor-reactive T-cell signature.
From 683,184 high-quality single cells, we identified 9 cell types and 71 subtypes of T-cells, B-cells, dendritic cells (DCs), monocytes/macrophages and endothelial-cells (ECs), shared between cancer types. PDCD1 (PD1) was expressed by differentiated T-cells subtypes (e.g., CD8+ exhausted and CD4+ T-helper-1 subclusters), while CD274 and PDCD1LG (PDL-1/PDL-2) were mainly expressed by regulatory B-cells, immune-regulatory DCs, CXCL10+ and CCL2+ monocyte-derived macrophages as well as inflammatory ECs. Pairwise analyses showed positive correlations between PD1-expressing T-cell subclusters, CD4+ T-regulatory cells, plasma B-cells (plasmablasts, IgA and IgG plasma cells), immune-regulatory DCs, CXCL10+ and CCL2+ macrophages and lymphatic ECs. On the other hand, negative correlations were observed with naïve T- and B-cells, conventional DC2 (cDC2), monocytes, CX3CR1+ macrophages, as well as arterial and capillary ECs. When ranking individual tumors based on a tumor-reactive T-cell signature, we found reactive tumors to correlate with positively interacting subtypes, while they anti-correlated with subclusters negatively interacting with PD1-expressing differentiated T-cells. Upon deconvolution of bulk RNA-seq data using gene signatures derived from each subcluster, these correlations were replicated in TCGA datasets across cancers. These signature scores correlated with tumor mutation burden (TMB) and other immunological features, while they were also predictive of response in clinical trials involving ICB.
We provide insights into the TME complexity at unprecedented level, identifying numerous subclusters enriched in immune-reactive (hot) or -suppressive (cold) tumors across cancer types. We validate these observations in TCGA by deconvolution of bulk RNA-seq data, and show that already pre-treatment several TME subclusters predict response to ICB.
Citation Format: Francesca Lodi, Sam Vanmassenhove, Elena Donders, Pierre Van Mol, Amelie Franken, Sarah Cappuyns, Ayse Bassez, Siel Olbrecht, Liselore Loverix, Michel Bila, Hanne Vos, Joanna Pozniak, Kevin Punie, Diether Lambrechts. A pan-cancer single-cell tumor micro-environment atlas predictive of immunotherapy response. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5785.