Genetic dissection of tissue composition in genetically diverse mouse populations
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Single-cell RNAseq data can be harnessed to infer cell-type compositions from spatial or bulk transcriptomes, providing greater biological context underlying systemic tissue or organ dynamics. Here, we used a transcriptome-based systems genetics approach to assess biological factors that influence the cellular composition of tissues, including sex, age and genetic variation in two multiparental mouse resource populations derived from the same 8 parental strains.
Methods
We leveraged publicly available bulk RNAseq data from kidney, heart and bone from 188 Diversity Outbred (DO) mice and male/female pairs from 58 inbred strains of the Collaborative Cross (CC). DO mice were aged 6, 12 or 18 months. We used single-cell RNAseq data from healthy C57BL/6 mice as reference for cell-type decomposition of bulk transcriptomes. We tested for differences in cell type composition based on age and sex. We then performed genetic analysis on the cell-type composition phenotypes, estimating the heritability of cell composition levels, mapping cell composition quantitative trait loci (ccQTL), and assessing candidate gene intermediates of ccQTL based on a genetic mediation approach and support in the literature.
Results
The heritability of cell-type compositions differed between the inbred CC and outbred DO. Contrasting this, we observed strong and consistent sex differences between the CC and DO. For example, we observed consistency in renal proximal straight tubule (higher in males), T cell abundance in kidney (higher in females), ventricular cardiomyocyte (higher in males) and endocardial cells (higher in females). DO mice exhibited increased levels B and T cells with age, representing an age-related increase in overall renal immune cell content. In heart, DO mice showed decreased ventricular cardiomyocyte composition and increasing endocardial cell content with age. In bone from DO mice, vascular endothelial cells decreased with age and bone marrow stromal cells increased with age. Genome-wide association yielded several cell composition QTLs (ccQTLs) in the three tissues, likely related to known biological phenomena. These include a ccQTL for cardiac fibroblasts in the CC that contains the Frg2 gene family and fibroblast growth factor inhibitor Spry2 , and a bone vascular endothelial cell ccQTL in the DO that encompasses 10 genes encoding members of the S100a protein family. In addition, our analysis revealed several ccQTLs for different renal tubular segments, immune cell content, and approximately two thirds of cardiac cell types.
Conclusion
We establish a role for transcriptome-inferred cellular states or composition for QTL mapping, which adds further utility to existing resources. Furthermore, differences in heritability and QTL between the CC and DO highlight how variation in allele frequencies and inbred vs outbred genetic backgrounds allow for population-specific genetic effects to be discovered. Several of the renal tubular segment and cardiac cell ccQTLs have narrow confidence regions with the potential to harbor novel biological insight. Overall, we propose that our ccQTL mapping approach can be applied systematically across readily available genetic resource transcriptomes to enable new biological discoveries.
