Plasma proteomics reveals shared biologic signatures of acute respiratory failure across etiology and severity in COVID-19 and non-COVID-19 cohorts

Acute respiratory failure (ARF) is common, exhibits variable outcomes, and arises from diverse causes that incompletely reflect underlying biology. Although the pandemic accelerated molecular investigation of ARF in COVID-19, whether these biologic signals generalize to ARF in hospital populations of varying etiology and severity remains unknown. We analyzed aptamer-based plasma proteomics collected within 48 hours of admission across two multihospital studies of varying etiology and severity, including individuals aged ≥14 years with acute viral or respiratory illness, stratified by COVID-19 and non-COVID-19 status (N = 1,096 total across four cohorts). The primary outcome was critical ARF, defined as advanced respiratory support or death by day 7. We identified proteins reproducibly associated with critical ARF in all cohorts, then contextualized protein associations with pathway analysis. Additionally, we developed proteomics-based models for two clinical purposes: predicting critical ARF and identifying heterogeneity of treatment response. We identify 9 proteins associated with critical ARF across cohorts (FDR < 0.05). Seven proteins related to innate immunity, oxidative stress, and barrier disruption are higher in critical ARF, while two proteins linked to immune signaling are lower. Pathway analysis identifies dysregulated inflammatory and fibrotic signaling pathways. Protein-based models improve discrimination of critical ARF beyond clinical variables in non-COVID-19 cohorts. A causal forests model finds elevated IL1RL1 (soluble ST2) is associated with greater benefit from imatinib. A proteomic signature linked to ARF-related outcomes is reproducible across four cohorts with a breadth of etiologies and severity. The proteins may inform design of novel treatments and enriched clinical trials. Acute respiratory failure (ARF) is a life-threatening condition triggered by many different illnesses, and the biological reasons why some patients experience poor outcomes remain incompletely understood. To better understand this biology, we analyzed relationships between thousands of blood proteins and ARF outcomes in two multi-hospital studies of over 1,000 patients with COVID-19 and non-COVID-19 illness. We identified consistent protein signatures of ARF across diverse causes and severity, reflecting shared processes of inflammation, stress, and tissue injury. To develop clinical uses for blood proteins we demonstrated: 1) proteins improve outcome prediction beyond standard clinical information, and 2) elevations in one immune response protein identify patients who benefit from imatinib, a drug previously tested for COVID-19 ARF with mixed results. Together, the findings yield new opportunities for more precise ARF treatment. Sathe et al define plasma proteomic signatures of acute respiratory failure outcomes which are reproducible across COVID-19 and non-COVID-19 cohorts of varying severity. They find signatures reflecting inflammation, oxidative stress, and barrier disruption are detectable at hospital admission and offer biomarker candidates for developing and testing targeted treatment.