This proof-of-concept cohort study (IMMERSE) enrolled 69 critically ill adults with sepsis and 30 elective cardiac surgery patients, sampling longitudinal peripheral blood transcriptomics (pan-leukocyte RNA-seq), mass cytometry (CyTOF), flow cytometry, and plasma cytokines at up to four time points from ICU admission through discharge or death, to map immune response trajectories rather than single-timepoint subtypes.
Pseudotime analysis of 718 differentially expressed genes identified three Sepsis Temporal Immune States (STImS1–3): STImS1 (n=44 samples) was innate-immune-dominant with highest IL-6, mortality-associated gene signatures, and lowest immune resilience; STImS3 (n=60) was adaptive-immune-dominant with improving resilience; all three STImS appeared at every clinical time point, demonstrating that immune state and clinical time are partially decoupled. STImS explained 32.8% of transcriptomic variance vs. 1.04% for age and 0.19% for sex. Hospital mortality was 15% (10/69), with non-survivors most enriched in STImS1 (39%) and least in STImS3 (20%).
- Transcriptome derivation used only 198 samples from a single center; external validation relied on mapping to existing public datasets rather than a prospective replication cohort. - No STImS classifier model was developed, limiting immediate clinical application. - Sampling began at clinical sepsis recognition, so pre-sepsis immune responses during the infection-to-sepsis transition were not captured.
Clinical time point (e.g., ICU admission = 'early sepsis') does not reliably reflect a patient's actual immune state — some admitted patients are already in late immune states, and some 'recovered' patients remain in early states. Before enrolling patients in immunomodulatory trials or initiating targeted therapies (e.g., IL-6 blockade, JAK inhibitors, corticosteroids), aligning treatment to the patient's immune trajectory state rather than clinical time may improve outcomes.
Explore related topics
