Longitudinal multiomic signatures of ARDS and sepsis inflammatory phenotypes identify pathways associated with mortality
Narges Alipanah-Lechner, Lucile Neyton, Pratik Sinha, Carolyn Leroux, Kim Bardillon, Sidney A. Carrillo, Suzanna Chak, Olivia Chao, Taarini Hariharan, Carolyn Hendrickson, Kirsten Kangelaris, Charles R. Langelier, Deanna Lee, Chelsea Lin, Kathleen Liu, Liam Magee, Angelika Ringor, Aartik Sarma, Emma Schmiege, Natasha Spottiswoode, Kathryn Sullivan, Melanie F. Weingart, Andrew Willmore, Hanjing Zhuo, Angela J. Rogers, Kathleen A. Stringer, Michael A. Matthay, Carolyn S. Calfee
Narges Alipanah-Lechner, Lucile Neyton, Pratik Sinha, Carolyn Leroux, Kim Bardillon, Sidney A. Carrillo, Suzanna Chak, Olivia Chao, Taarini Hariharan, Carolyn Hendrickson, Kirsten Kangelaris, Charles R. Langelier, Deanna Lee, Chelsea Lin, Kathleen Liu, Liam Magee, Angelika Ringor, Aartik Sarma, Emma Schmiege, Natasha Spottiswoode, Kathryn Sullivan, Melanie F. Weingart, Andrew Willmore, Hanjing Zhuo, Angela J. Rogers, Kathleen A. Stringer, Michael A. Matthay, Carolyn S. Calfee
View: Text | PDF
Clinical Research and Public Health Clinical Research Inflammation Pulmonology

Longitudinal multiomic signatures of ARDS and sepsis inflammatory phenotypes identify pathways associated with mortality

Abstract

BACKGROUND Critically ill patients with acute respiratory distress syndrome (ARDS) and sepsis exhibit distinct inflammatory phenotypes with divergent clinical outcomes, but the underlying molecular mechanisms remain poorly understood. These phenotypes, derived from clinical data and protein biomarkers, were associated with metabolic differences in a pilot study.METHODS We performed integrative multiomics analysis of blood samples from 160 patients with ARDS in the ROSE trial, randomly selecting 80 patients from each latent class analysis–defined inflammatory phenotype (hyperinflammatory and hypoinflammatory) with phenotype probability greater than 0.9. Untargeted plasma metabolomics and whole-blood transcriptomics at day 0 and day 2 were analyzed using multimodal factor analysis (MEFISTO). The primary outcome was 90-day mortality, with validation in an independent critically ill sepsis cohort (EARLI).RESULTS Multiomics integration revealed 4 molecular signatures associated with mortality: (a) enhanced innate immune activation coupled with increased glycolysis (associated with hyperinflammatory phenotype), (b) hepatic dysfunction and immune dysfunction paired with impaired fatty acid β-oxidation (associated with hyperinflammatory phenotype), (c) interferon program suppression coupled with altered mitochondrial respiration (associated with hyperinflammatory phenotype), and (d) redox impairment and cell proliferation pathways (not associated with inflammatory phenotype). These signatures persisted through day 2 of trial enrollment. Within-phenotype analysis revealed distinct mortality-associated pathways in each group. All molecular signatures were validated in the independent EARLI cohort.CONCLUSION Inflammatory phenotypes of ARDS reflect distinct underlying biological processes with both phenotype-specific and phenotype-independent pathways influencing patient outcomes, all characterized by mitochondrial dysfunction. These findings suggest potential therapeutic targets for precise treatment strategies in critical illness.FUNDING NIH National Heart, Lung, and Blood Institute and National Institute of General Medical Sciences.

Authors

Narges Alipanah-Lechner, Lucile Neyton, Pratik Sinha, Carolyn Leroux, Kim Bardillon, Sidney A. Carrillo, Suzanna Chak, Olivia Chao, Taarini Hariharan, Carolyn Hendrickson, Kirsten Kangelaris, Charles R. Langelier, Deanna Lee, Chelsea Lin, Kathleen Liu, Liam Magee, Angelika Ringor, Aartik Sarma, Emma Schmiege, Natasha Spottiswoode, Kathryn Sullivan, Melanie F. Weingart, Andrew Willmore, Hanjing Zhuo, Angela J. Rogers, Kathleen A. Stringer, Michael A. Matthay, Carolyn S. Calfee

×

Figure 8

ROSE MEFISTO factor projections in the EARLI observational cohort.

Options: View larger image (or click on image) Download as PowerPoint
ROSE MEFISTO factor projections in the EARLI observational cohort. (A) S...
(A) Scatterplot of paired projected factor values per patient sample, colored by LCA phenotype designation, in patients with both transcriptomic and metabolomic data available (N = 61). (B) Association of projected MEFISTO factor values with clinical variables as determined via Spearman’s correlation for continuous predictors and linear regression for categorical predictors. (C) Projected factor values per patient sample comparing hyperinflammatory and hypoinflammatory phenotype. P value determined by Wilcoxon’s rank-sum. N = 189 patients with transcriptomic data (factors 1 and 3), and N = 183 patients with metabolomic data (factor 2). (D) Projected factor values per patient sample comparing hospital survivors and non-survivors in those with physician-adjudicated sepsis and transcriptomic data (N = 196 for transcriptomic factors 1, 3, and 5; N = 195 for metabolomic factor 2). P value determined by Wilcoxon’s rank-sum. (E) Projected phenotype-specific MEFISTO factors associated with mortality onto the extreme (P > 0.9) hypoinflammatory (left, N = 101) and extreme hyperinflammatory (right, N = 61) patients in EARLI. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. APACHE, Acute Physiology and Chronic Health Evaluation III score; GFR, glomerular filtration rate.