Ó³»­´«Ã½

BACKGROUND: Elevated circulating endothelial cells (CECs), released from monolayers after insult, have been implicated in worse outcomes in ARDS and COVID-19, however there is no consensus proteomic phenotype that define CECs. We queried whether a transcriptomic approach would alternatively support the presence of endothelial cells in circulation and correlate with worsening respiratory failure.METHODS: To test whether elevated endothelial cell signatures (ECS) in circulation plays a role in worse respiratory outcomes, we used unsupervised bulk-transcriptome deconvolution to quantify ECS% in two cohorts. Our pilot analysis included pediatric patients requiring invasive mechanical ventilation (CAF-PINT, NCT01892969). Our validation cohort included adult hospitalized patients with COVID-19 (IMPACC, NCT04378777), testing the association of ECS% to outcomes in patients at risk of acute respiratory failure/ARDS. Primary outcome was 28-day mortality.RESULTS: In CAF-PINT, day 0 ECS% was higher in non-survivors compared to survivors of respiratory failure (2.8%, IQR 2.4-3.4% versus 2.6%, IQR 2.2-3.0% n = 244, p < 0.05, Wilcoxon rank-sum). In IMPACC, baseline ECS% (< 72 h of hospitalization) was higher in COVID-19 non-survivors versus survivors (2.9%, IQR 2.6-3.4%, versus 2.7%, IQR 2.3-3.1%, n = 932, p < 0.001, Wilcoxon rank-sum). Each 1% increase in baseline ECS% was significantly associated with mortality (adjusted OR 1.36, CI 1.03-1.79) by multivariable logistic regression. Increased baseline ECS% was associated with worse respiratory trajectories (2.5%, IQR 2.2-2.8% for trajectory with no oxygen requirements, 2.9%, IQR 2.6-3.4% for the trajectory with fatal outcome by day 28, n = 932, p < 0.001, one-way ANOVA).CONCLUSION: Quantifying ECS by deconvolution supports a transcriptomics-driven approach towards the non-invasive evaluation of endothelial damage in respiratory outcomes. This is a first step towards elucidating mechanistic components linking endothelial damage to ARDS utilizing non-invasive, circulating transcriptomic data by leveraging a novel deconvolution approach.