Quarterly ReportOct 1 – Dec 31, 2025
Sepsis, Q4 2025 edition
We read 554 papers and selected 10.
Summary
Sepsis research during 2025-Q4 shifted decisively from syndrome-level treatment toward dynamic, biologically stratified care. Large human studies established that infection source, age, longitudinal molecular state, and treatment-response signatures are major determinants of heterogeneity. The quarter also produced important methodological advances, including goal-directed multi-omics subgrouping, rapid time-resolved transcriptomics, explainable AI, and integrated single-cell immune mapping. Clinically, capillary-refill-time-guided resuscitation and scalable maternal infection interventions provided the strongest randomized evidence, although the latter was not selected among the top ten because its primary contribution was implementation rather than a major scientific advance. Mechanistic studies identified actionable endothelial, immunocoagulation, PANoptosis, and immunometabolic pathways, but most targeted therapies remain preclinical. Overall, the quarter strengthened the case for predictive enrichment and repeated biological assessment in future sepsis trials.
Selected Articles
1. Single-cell multi-omic landscape reveals anatomical-specific immune features in adult and pediatric sepsis.
This human study integrated single-cell transcriptomics, T-cell and B-cell receptor sequencing, CITE-seq, bulk RNA sequencing, and plasma proteomics in adults and children with sepsis. It showed that infection source imprints distinct immune programs while preserving shared core features across ages.
Impact: It provides a comprehensive human reference map that reframes sepsis heterogeneity around infection source and age rather than a uniform systemic syndrome.
Clinical Implications: Source- and age-specific immune signatures may improve biomarker selection and trial enrichment, but prospective clinical utility and assay standardization are still required.
Key Findings
- Integrated multi-omics identified infection-site-specific immune programs.
- Adult and pediatric sepsis shared core features but also showed distinct age-related states.
- An NR4A2-associated signature emerged as a candidate marker of immune heterogeneity.
2. Longitudinal multi-omic signatures of ARDS and sepsis inflammatory phenotypes identify pathways associated with mortality.
Longitudinal metabolomic and transcriptomic profiling identified four persistent mortality-associated molecular signatures involving innate immunity, glycolysis, fatty-acid oxidation, interferon suppression, redox biology, and mitochondrial dysfunction. The signatures were externally validated in an independent critically ill sepsis cohort.
Impact: It advances the field from static inflammatory phenotyping to persistent, externally validated molecular pathways that can guide longitudinal endotype research.
Clinical Implications: Repeated molecular profiling could identify patients with active mitochondrial, metabolic, interferon, or redox dysfunction, although simplified assays and prospective treatment studies are needed.
Key Findings
- Four mortality-associated molecular signatures were identified.
- The signatures persisted from day 0 to day 2 and converged on mitochondrial dysfunction.
- Findings were externally validated in an independent critically ill sepsis cohort.
3. A rapid time-resolved host gene expression signature predicts responses to antibiotic treatment in neonatal bacterial sepsis.
A time-resolved host gene-expression signature reversed within 24 hours of vancomycin initiation and tracked clinical improvement. Reproducibility across pediatric and adult cohorts suggests that host biology can provide an early measure of antibiotic response.
Impact: It introduces a dynamic biological approach to monitoring antibiotic response, addressing nonresponse detection and antimicrobial stewardship.
Clinical Implications: After rapid-platform validation, the signature could support early reassessment, de-escalation, and duration decisions, but neonatal-specific performance and turnaround time must be established.
Key Findings
- The signature reversed within 24 hours of antibiotic initiation.
- Signature changes correlated with clinical improvement.
- The signal was reproducible across pediatric and adult sepsis cohorts.
4. DLL4+ neutrophils promote Notch1-mediated endothelial PANoptosis to exacerbate acute lung injury in sepsis.
The study identified an eCIRP-induced DLL4-positive neutrophil population that activated Notch1 on pulmonary endothelial cells, causing ZBP1-mediated PANoptosis and barrier failure. A Notch1-DLL4 inhibitor reduced lung injury and improved survival in sepsis models.
Impact: It defines a novel neutrophil-endothelial pathway and links a specific molecular interaction to a candidate endothelial-protective intervention.
Clinical Implications: Notch1-DLL4 blockade may become an adjunctive therapy for sepsis-associated acute lung injury, but human validation, pharmacokinetics, and infection-related safety testing are required.
Key Findings
- eCIRP induced DLL4-positive neutrophils.
- DLL4-Notch1 signaling triggered endothelial PANoptosis and vascular leakage.
- A Notch1-derived inhibitor reduced injury and improved survival in vivo.
5. A Critical Role for MALAT1 in Gram-negative Bacteria-induced Coagulation via Regulation of Caspase-11 signaling.
Admission IFNβ predicted septic DIC within 48 hours, while HMGB1 did not. Mechanistic experiments showed that IFNβ induced MALAT1, suppressed GPX4, and promoted caspase-11-dependent phosphatidylserine exposure and immunocoagulation.
Impact: It bridges a clinically measurable early biomarker with a detailed immunocoagulation mechanism and a potential therapeutic axis.
Clinical Implications: IFNβ may help identify patients at risk for early DIC, but external validation and assessment of bleeding and infection-control consequences are essential.
Key Findings
- Admission IFNβ correlated with DIC onset 48 hours later.
- IFNβ induced MALAT1 and reduced GPX4 activity.
- MALAT1 deletion reduced phosphatidylserine exposure and bacteria-induced coagulation.
6. Deriving consensus sepsis clusters via goal-directed subgroup identification in multi-omics study.
This framework integrated longitudinal transcriptomic, proteomic, metabolomic, and phenomic data from 1,327 patients across 43 hospitals. Unlike conventional clustering, it optimized subgroups for differential treatment response and externally validated survival differences related to fluid strategies and ulinastatin.
Impact: It introduces a methodological solution to treatment-effect heterogeneity by deriving subgroups according to differential benefit rather than descriptive phenotype alone.
Clinical Implications: The approach could support predictive-enrichment trials, but prospective assay-standardized validation is required before therapy is assigned using these clusters.
Key Findings
- The framework integrated multi-omic data from 1,327 patients across 43 hospitals.
- Derived subgroups predicted differential outcomes with restrictive versus liberal fluids.
- The method identified differential benefit signals for ulinastatin and was externally validated.
7. Unexpected Protective Role of Thrombosis in Lung Injury via Endothelial Alox15.
Using genetic, lipidomic, and sepsis-model approaches, this study found that mild pulmonary thrombosis could preserve endothelial integrity through ALOX15-dependent protective lipids, whereas severe thrombosis or thrombocytopenia worsened injury. The findings challenge indiscriminate anticoagulation.
Impact: It overturns the assumption that all pulmonary thrombosis is harmful and provides a mechanistic explanation for heterogeneous anticoagulation effects.
Clinical Implications: The findings support caution against blanket anticoagulation and motivate phenotype-specific studies of ALOX15-regulated lipid mediators.
Key Findings
- Mild pulmonary thrombosis reduced endothelial apoptosis and lung injury in sepsis models.
- Endothelial ALOX15 manipulation altered injury severity.
- Severe thrombosis or thrombocytopenia worsened lung injury.
8. Personalized Hemodynamic Resuscitation Targeting Capillary Refill Time in Early Septic Shock: The ANDROMEDA-SHOCK-2 Randomized Clinical Trial.
This multinational randomized trial compared capillary-refill-time-targeted resuscitation with usual care in 1,467 patients across 86 centers. The personalized strategy improved a hierarchical 28-day patient-centered outcome, mainly through shorter vital-organ support duration rather than a clear mortality reduction.
Impact: It provides high-level evidence that a simple bedside perfusion measure can guide individualized early resuscitation.
Clinical Implications: Structured CRT-guided care may reduce vasopressor and organ-support exposure, but implementation requires training, patient selection, and outcome surveillance because mortality benefit was not definitive.
Key Findings
- The trial enrolled 1,467 patients from 86 centers in 19 countries.
- The CRT-guided strategy produced a win ratio of 1.16.
- The benefit was driven mainly by shorter vital-organ support duration.
9. Development of a gene panel for immune status assessment in sepsis.
A four-gene panel consisting of TBX21, GNLY, PRF1, and IL2RB was developed and externally validated for immune-status assessment. The panel identified an endotype in which hydrocortisone or thymosin was associated with higher mortality, providing a potential safeguard against indiscriminate immunotherapy.
Impact: It translates complex immune biology into a compact and potentially deployable biomarker panel for treatment stratification.
Clinical Implications: The panel may enrich biomarker-stratified trials and identify patients at risk from immunomodulation, but treatment-interaction findings remain hypothesis-generating.
Key Findings
- The four-gene panel achieved AUROCs of 0.891-0.909.
- The panel separated clinically meaningful immune endotypes.
- One endotype showed higher mortality with hydrocortisone or thymosin treatment.
10. Explainable AI unravels sepsis heterogeneity via coagulation-inflammation profiles for prognosis and stratification.
SepsisFormer and the simpler SMART tool were developed in 12,408 patients using coagulation-inflammatory laboratory data. The models defined risk tiers and subphenotypes and generated testable hypotheses about differential anticoagulant benefit.
Impact: It combines predictive performance, explainability, and routine-data accessibility, making precision stratification more clinically deployable.
Clinical Implications: SMART may support triage and monitoring, while treatment-benefit signals require prospective randomized confirmation before guiding anticoagulant use.
Key Findings
- SepsisFormer achieved an AUC of 0.9301 in a multicenter cohort.
- SMART used seven laboratory variables and age to define four risk tiers.
- Observational analyses suggested differential anticoagulant benefit across subphenotypes.