Graph-based pathway analysis of T cell populations in hepatocellular carcinoma reveals novel metabolic regulators of tumor-infiltration lymphocyte activity

Abstract With advances in single cell RNA sequencing (scRNAseq), accurate detection of perturbed pathways between conditions or cell types of interest becomes a critical analytical step. The ability to model alterations in a set of genes corresponding of a biological function is particularly useful when comparing cells between healthy and tumor tissues. Yet, few tools exist to detect changes in the multivariate distribution of genes corresponding to a given pathway. We developed a novel, graph-based statistical framework based on optimal matching for testing differential distribution of biological pathways in scRNAseq. We applied our method to data generated from >4,000 T cells isolated from six individuals with hepatocellular carcinoma (HCC). The T cell populations were purified from three tissue locations: peripheral blood, tumor-infiltrating immune cells (TIICs), and normal tissue adjacent to the tumor. We examined the distribution of gene sets that belong to a particular metabolic pathway across T cell subtypes to address the following questions: (1) Which pathways have a similar distribution across T cell subtypes in a given tissue? (2) Are there pathways that have a stable distribution across T cell subtypes in a normal/healthy tissue, but a perturbed distribution in HCC? (3) For pathways that have a disparate distribution across the T cell subtypes, which subtypes show the most distinct distribution? Of the 86 metabolic pathways compared, 41, 63 and 76 metabolic pathways were indistinguishable across T cell subtypes in the tumor, peripheral blood and adjacent normal tissues, respectively. We observed that most metabolic pathways do not show evidence for dissimilar distribution across cell types, suggesting that T cell subtypes might be more similar than previously appreciated in terms of how they regulate their basic metabolic machinery. Further, for each pair of tissue locations, we computed the overlap in perturbed pathways, and found that the concordance was substantially higher for blood and adjacent normal, than what either of these tissues had with the tumor tissue. Interestingly, five metabolic pathways were differentially distributed across the T cell subtypes in each tissue examined in HCC: glycolysis, purine metabolism, glycosphingolipid biosynthesis, pyruvate metabolism, and glycerophospholipid metabolism. Our model also found that CD4+ regulatory T cells were the strongest contributors, driving the differential distribution of these pathways between T cell subtypes in HCC. Altogether, our approach allows for a systems level characterization of pathway activity across multiple cell types with a variety of applications in single cell pathway analysis in oncology. Our work here also highlights unexpected regulatory mechanisms of regulatory T cells that might play in role in the immunobiology of HCC. Citation Format: Divyansh Agarwal, Tina Bharani, Somabha Mukherjee. Graph-based pathway analysis of T cell populations in hepatocellular carcinoma reveals novel metabolic regulators of tumor-infiltration lymphocyte activity [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 2042.