Abstract Introduction Sepsis, a dysregulated immune response to infection, accounts for nearly 11 million deaths worldwide each year and remains a major cause of hospital mortality. The absence of precision diagnostics limits efforts to characterize immune dynamics and predict clinical outcomes. We evaluated whether liquid biopsies collected at ICU admission could provide insight into sepsis biology and enable precision diagnostics. Methods We applied cell-free RNA (cfRNA) liquid biopsies as a noninvasive method to highlight signaling networks, identify tissue-specific injury, and predict immune exhaustion. We collected serial samples throughout hospitalization from a cohort of 100 patients admitted with likely sepsis to a medical ICU. Comprehensive clinical data, including inpatient laboratory values and key outcome measures such as survival and ventilator-free days, were curated. Circulating cfRNA was isolated and purified from these samples and prepared for sequencing. We generated data from 254 samples at a minimum sequencing depth of 30x. We then designed a novel deep-learning computational method to enable deconvolution to better analyze and interpret these data. Results In our cohort, a total of 84 patients were adjudicated to have sepsis by the Sepsis-3 criteria, and 25 patients died within 90 days. Differential analysis was performed comparing circulating cfRNA in these septic patients to that of healthy controls, stratified by mortality along with specified subgroups including shock on admission, ventilator dependence, and preexisting immunosuppression. We identified multiple differentially expressed transcripts between septic patients and healthy controls, with significant alterations in signaling cascades associated with mitochondrial signaling, including alterations in oxidative phosphorylation and fatty acid metabolism. Using defined ligand-receptor interaction databases, we identified key pathways associated with immune and endothelial dysfunction in sepsis. Specifically, our analysis identified numerous alterations in CXCL8 (IL-8)-associated signaling cascades (engaging CXCR1, CXCR2, SDC1-3, and ACKR). Our analyses also highlighted key tissue factor pathway inhibitor (TFPI)-associated signaling via interactions with tissue factor and low-density lipoprotein receptor-related protein 1 (LRP1). These analyses are corroborated by our new deep-learning augmented approach to deconvolution, where we demonstrate the identification of rare tissue- and cell-specific signals that are associated with increased mortality in our cohort. Conclusion Liquid biopsy approaches such as cfRNA profiling can provide insight into mechanisms underlying sepsis and may uncover underlying drivers of endothelial dysfunction. These findings support ongoing efforts to develop liquid biopsy-based diagnostics and personalized therapeutic strategies for patients at high risk for sepsis and related mortality. This abstract is funded by: NIH
Viswanathan et al. (Fri,) studied this question.