A104-24 Persistent Blood DNA Methylation Signatures Distinguish Sepsis-associated ARDS Across Acute and Convalescent Time Points

Rationale

Sepsis-associated acute respiratory distress syndrome (ARDS) is associated with high mortality and long-term morbidity, yet early and reliable identification remains challenging. Blood-based biomarkers have shown promise, but their clinical utility is limited by inter-individual molecular heterogeneity and strong dependence on timing of sample collection. Because DNA methylation changes can persist beyond acute illness, we hypothesized that differentially methylated regions (DMRs) distinguishing sepsis-associated ARDS from sepsis alone would be detectable across time points, enabling development of a time-robust epigenetic biomarker prediction model.

Methods

Whole blood was obtained from 22 sepsis patients and age-, sex-, and race-matched controls from the Mass General Brigham (MGB) Biobank. Sepsis-associated ARDS, defined by the Global definition, was adjudicated by board-certified pulmonary and critical care physicians. Samples were stratified by timing and ARDS status: early (day 0-2) sepsis (n = 4) and sepsis-ARDS (n = 2); mid (day 3-10) sepsis (n = 7); late (day 22-62) sepsis (n = 4); and late sepsis-ARDS (day 17-68, n = 5). DNA methylation profiling was performed using the Illumina EPIC v2.0 BeadChip platform. DMRs were identified relative to controls at each time point. DMR-associated genes specific to sepsis-ARDS across all time points were subjected to Gene Ontology (GO) and KEGG pathway enrichment analyses. Normalized beta values were used for principal components analysis (PCA), followed by logistic regression to classify sepsis-ARDS vs. sepsis alone, and model performance was assessed.

Results

A large number of DMRs were identified, particularly at early time points: early sepsis (7,798 DMRs; 2,054 genes); early sepsis-ARDS (31,333 DMRs; 6,863 genes); mid sepsis (3,856 DMRs, 935 genes); late sepsis (1,997 DMRs; 551 genes); and late sepsis-ARDS (327 DMRs; 90 genes). Seventy-two DMRs were specific to sepsis-ARDS across all time points. Pathway enrichment analysis identified terms, “positive regulation of monocyte differentiation” and “propanoate metabolism.” A classification model based on these DMRs achieved an AUROC (95% CI) of 0.952 (0.872-0.952), with sensitivity 0.714, specificity 0.933, positive predictive value 0.833, and negative predictive value 0.875.

Conclusions

Sepsis and sepsis-associated ARDS are associated with extensive, time-dependent epigenetic alterations in peripheral blood. By identifying DMRs that persist across acute and convalescent phases, we developed a robust epigenetic classifier that distinguishes sepsis-ARDS from sepsis alone across a wide range of sampling times. These findings support the potential utility of persistent DNA methylation signatures for post-sepsis ARDS phenotyping and risk stratification. Despite modest sample size, consistency of ARDS-specific DMRs across time points supports biological robustness. Validation in larger, independent cohorts is warranted.

This abstract is funded by: 1R21HL175093-01A1; Massachusetts General Hospital Executive Committee on Research: ISF Award