Abstract B209: Transcriptomic analysis of patient-derived xenografts reveals heterogeneity in human and mouse stroma/immune compartments

Abstract

Abstract Introduction: Cancers are increasingly recognized as diverse collections of not only genetic diseases but also immune diseases. The heterogeneous tumor microenvironment (TME), including immune components, plays critical roles in tumor growth/progression and response to pharmaceutics, particularly immuno-oncology therapeutics. However, investigating stroma-specific components is rather challenging due to the difficulty in separating stroma from tumor cells, either physically via microdissection or in silico via bioinformatics. Experimental animal models are excellent system to investigate TME1. For example, patient-derived xenograft (PDX), where human and mouse content can readily be separated in silico2, provides a model to study tumor microenvironment, including tumor-stroma interactions. Methods: Whole transcriptome sequencing was performed on subcutaneous PDXs in athymic mice3 and all reads were aligned to human and mouse genomes to discriminate human and mouse content, respectively. Low expressed/less variable genes were removed across the two components. MCP-counter algorithm4 was used to estimate infiltration by various human immune and other stromal cell populations from human gene expression data. The gene expression profiles of 35 cell types from Immunological Genome Consortium (ImmGen)5 were used to create lineage-specific signatures and compute mouse stroma composition by CIBERSORT6. Weighted gene coexpression network analysis (WGCNA)7 was applied on human and mouse gene expression data, respectively, to identify distinct gene modules associated with different biologic functions. Human-mouse gene correlation analysis was performed to construct cross-species expression networks. Results: We have established and profiled ~1800 subcutaneous PDXs, with ~1500 of them whole transcriptome-sequenced (http://crownbio.com/eortc17/stromalposter)8. The average mouse-to-human sequencing read ratio is around 11%, consistent with the previous report2. By deconvolution analysis of human and mouse gene expression data, we identified all types of adaptive and innate immune cells in mouse stroma of these immune-deficient mice. Most surprisingly, a variety of human immune components were also observed in human expression contents. The corresponding fractions vary across cancer subtypes and individual models. The heterogeneity has been also validated by flow cytometry. Distinct coexpressed gene modules from human (50x) and mouse (28x) revealed that human clusters were more correlated to cancer type than mouse clusters, particularly strong in pancreatic, colorectal, and lung cancers. Knowledge-based functional analysis of these gene modules revealed the association of diverse biologic processes. Moreover, cross-species correlation network identified a novel relationship between human and mouse genes over all PDXs and in specific cancers. Conclusions: Both human and mouse stroma/immune composition exists and varies across cancers in subcutaneous PDXs, with each cancer type and each individual model within subtypes, suggesting PDX as a useful experimental model to study TME. Citation Format: Jia Xue, Wubin Qian, Sheng Guo, Jie Cai, Xuesong Ouyang, Henry Qixiang Li. Transcriptomic analysis of patient-derived xenografts reveals heterogeneity in human and mouse stroma/immune compartments [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2017 Oct 26-30; Philadelphia, PA. Philadelphia (PA): AACR; Mol Cancer Ther 2018;17(1 Suppl):Abstract nr B209.