Poster #24 - Avantika R. Diwadkar

Genetic variants regulating immune cell-cell interactions are key drivers for autoimmune disease risk

Avantika R. Diwadkar (1,2), Lida Wang (2), Havell Markus (1,3), Natashia J. Benjamin (1,2), Jeniece M. Regan (4), Paige E. Bond (4), Bibo Jiang (2), Laura Carrel (1,4), Dajiang J. Liu (1,2,3) 1. Bioinformatics and Genomics PhD Program; Pennsylvania State University College of Medicine; Hershey, Pennsylvania, 17033; USA. 2. Department of Public Health Sciences; Pennsylvania State University College of Medicine; Hershey, Pennsylvania, 17033; USA. 3. Institute for Personalized Medicine; Pennsylvania State University College of Medicine; Hershey, Pennsylvania, 17033; USA. 4. Department of Molecular and Precision Medicine; Pennsylvania State University College of Medicine; Hershey, Pennsylvania, 17033; USA

GWAS has identified hundreds of variants associated with autoimmune diseases, most of which are non-coding and may function by modulating gene expression. Linking GWAS loci to the target genes has been challenging. Despite growing sample sizes and granularity of single-cell RNASeq data, many GWAS loci still fail to colocalize with bulk or single-cell (sc)-eQTLs. It motivates us to re-examine ways to integrate GWAS with sc-RNASeq data. Importantly, immune cells do not function in isolation but interact with each other via antigen presentation, cytokine signaling, etc. When cells interact, the receptor gene expression will vary with the ligand. We hypothesize that genetic variants (which we define as cci-QTLs) may interact with ligand genes to influence receptor gene expression. We develop a method, GUCCI (Genetic determinants Underlying Cell-Cell Interaction), to identify cci-QTLs. GUCCI views the standard sc-eQTLs for receptor genes as weighted averages of cci-QTLs when the ligand gene expression is high and low. To improve power, we further propose to jointly model bulk-eQTLs with cci-QTL and sc-eQTLs, as bulk-eQTLs are weighted averages of sc-eQTLs from constituent cell types. We applied GUCCI to the OneK1K dataset across 14 immune cell types, using publicly available ligand-receptor pairs and the eQTLGen bulk-eQTL whole blood dataset. GUCCI identified 69k cci-QTLs across 154 cell type pairs and 168 ligand-receptor pairs, with only 11% identified previously as sc-eQTLs. 84% of cci-QTLs co-localize with at least one GWAS locus of autoimmune diseases. We successfully fine-mapped causal SNPs and cell type pairs for 86 ligand-receptor interactions, with 95% credible sets containing fewer than 5 variants per cell type pair. Intriguingly, cci-QTLs were enriched in regions of the lymph node spatial transcriptomics where they are known to interact e.g., cci-QTLs for B naïve cell interactions are enriched in B cell zones. Additionally, 32% of ligand-receptor pairs show significantly higher interactions in spots with fine-mapped cell types. The cci-QTLs are also enriched in ENCODE cCREs and have comparable Enformer functional scores as sc-eQTLs. Amongst our findings, interleukin-6 (IL-6) and receptor (IL6R) interact between CD4 naïve T cells and uniquely explain an ankylosing spondylitis locus, which is missed by sc-eQTLs. Our results demonstrate the validity of our method and its potential to bring unique mechanistic insight into risk genes. This is a first-of-its-kind analysis that can advance biological insights for autoimmune diseases.