Predicting Cell Type-Specific Extracellular Vesicle Biology Using an Organism-Wide Single Cell Transcriptomic Atlas - Insights from the Tabula Muris

Abstract

Background: Extracellular vesicles (EVs) like exosomes are functional nanoparticles traffcked between cells and found in every biofluid. An incomplete understanding of which cells, from which tissues, are traffcking EVs in vivo has limited our ability to use EVs as biomarkers and therapeutics. However, recent discoveries have linked EV secretion to expression of genes and proteins responsible for EV biogenesis and found as cargo, which suggests that emerging “cell atlas” datasets could be used to begin understanding EV biology at the level of the organism and possibly in rare cell populations. To explore this possibility, here we analyzed 67 genes that are directly implicated in EV biogenesis and secretion, or carried as cargo, in ~44,000 cells obtained from 117 cell populations of the Tabula Muris. Methods: We obtained normalized count (FPKM) data from 44,779 FACS-sorted cells described in the Tabula Muris. These data were freely obtained from the Tabula Muris web portal operated by the Chan-Zuckerberg Biohub. We selected 67 EV genes based on their direct role in the regulation of endosome biogenesis, protein sorting, secretion or cargo. These genes include the endosome sorting complexes required for transport (ESCRT), multiple Rab GTPases, tetraspanins and syndecans. To identify the most abundant EV genes, we plotted mean expression (FPKM) across all cells against the fraction of cells with gene expression > 50 FPKM. To measure gene expression variance, we calculated %CV for each gene within each cell population and across all cell populations. For gene co-expression analysis, we performed linear regressions comparing selected EV genes (Cd9, Cd63, Cd81, Sdc4 and Sdcbp) to all other genes, within each cell population. Results: We found that the most abundant EV cargo proteins (tetraspanins (Cd9, Cd63 and Cd81) and syndecans (Sdc4 and Sdcbp)) are also the most abundant EV genes expressed across all cell populations. However, the expression of these genes varied greatly among cell populations. For example, Cd81 is most abundant in brain glial populations and multiple endothelial cell populations. By contrast, Cd9 expression is highest in "barrier" cells that support innate immunity while Cd63, a classic marker of exosomes, is most abundantly expressed by stem cells. Expression variance analysis was used to compare the transcriptional behavior of these abundant EV genes (Cd9, Cd63, Cd81, Sdc4 and Sdcbp). We found that Cd63 is the most dynamic EV gene while others are more constitutively expressed. This finding may help explain why Cd63 is typically found either within the cytoplasm or outside of the cell on EVs, whereas Cd9 and Cd81 are abundant both on the cell surface and on EVs. Finally, we used EV gene co-expression analysis to define cell population-specific transcriptional networks. We discovered that EV gene co-expression networks are heavily dependent on the EV cargo gene examined with different cell populations driving clustering. Conclusions: Our analysis has provided at least four independent observations which advance our understanding of EV biology. These are that: 1) abundant EV cargo genes are differentially expressed among cell populations, 2) skeletal muscle stem cells display the highest expression of both Cd63 and Sdc4, 3) Cd81 may be a primary EV cargo traffcked in the brain whiand 4) that rare cell populations can secrete EVs and thereby influence tissue biology. This analysis is the first, to our knowledge, to predict tissue- and cell type-specific EV biology at the level of the organism. As such, we expect this resource to be the first of many valuable tools for predicting the endogenous impact of specific cell populations on EV function in health and disease. National Institutes on Aging (R01AG078859). This is the full abstract presented at the American Physiology Summit 2024 meeting and is only available in HTML format. There are no additional versions or additional content available for this abstract. Physiology was not involved in the peer review process.