Daily Sepsis Research Analysis
A Cell study reveals that extracellular vesicles can transplant gasdermin D pores to bystander cells, propagating pyroptosis and amplifying inflammation. Preclinical research identifies kynurenic acid as a tryptophan-derived metabolite that shifts macrophages toward an anti-inflammatory phenotype via PPARγ/NF-κB to mitigate septic colonic injury. A real-world machine learning model modestly improves empirical beta-lactam activity in Enterobacterales bloodstream infections, supporting antimicrobi
Summary
A Cell study reveals that extracellular vesicles can transplant gasdermin D pores to bystander cells, propagating pyroptosis and amplifying inflammation. Preclinical research identifies kynurenic acid as a tryptophan-derived metabolite that shifts macrophages toward an anti-inflammatory phenotype via PPARγ/NF-κB to mitigate septic colonic injury. A real-world machine learning model modestly improves empirical beta-lactam activity in Enterobacterales bloodstream infections, supporting antimicrobial stewardship.
Research Themes
- Pyroptosis propagation via extracellular vesicles
- Immunometabolism (kynurenine pathway) and macrophage polarization
- Machine learning for antimicrobial stewardship in bloodstream infections
Selected Articles
1. Transplantation of gasdermin pores by extracellular vesicles propagates pyroptosis to bystander cells.
Pyroptotic cells release extracellular vesicles carrying GSDMD pores that insert into bystander cell membranes and induce lytic death, propagating inflammation. Using DNA-PAINT and immuno-EM, the study structurally visualizes GSDMD pores on EVs and demonstrates intercellular propagation of pyroptosis in vitro and in vivo.
Impact: This work uncovers a previously unrecognized mechanism of inflammatory cell death spread via EV-mediated transplantation of gasdermin pores. It reframes how pyroptosis escalates tissue damage and identifies potential points for therapeutic intervention.
Clinical Implications: Targeting EV release, cargo loading, or GSDMD pore assembly could mitigate bystander tissue injury in infections and sepsis. The findings also motivate biomarker development based on EV-associated GSDMD.
Key Findings
- Extracellular vesicles from pyroptotic cells carry GSDMD pore structures visualized by DNA-PAINT and immunoelectron microscopy.
- Pyroptotic EVs transplant GSDMD pores onto bystander cell membranes, inducing lytic death and amplifying inflammation.
- Intercellular propagation of pyroptosis was demonstrated both in vitro and in vivo, revealing a domino-like spread mechanism.
Methodological Strengths
- Use of super-resolution DNA-PAINT and immunoelectron microscopy to structurally verify GSDMD pores on EVs
- Functional validation of pyroptosis propagation across cells in both in vitro and in vivo systems
Limitations
- Preclinical models may not capture the full complexity of human sepsis and tissue microenvironments
- Therapeutic blockade of EV-mediated pore transfer was not evaluated for efficacy
Future Directions: Test pharmacologic or genetic inhibition of EV biogenesis/cargo loading or GSDMD pore assembly in infection models, and explore EV-associated GSDMD as a circulating biomarker.
Pyroptosis mediated by gasdermins (GSDMs) plays crucial roles in infection and inflammation. Pyroptosis triggers the release of inflammatory molecules, including damage-associated molecular patterns (DAMPs). However, the consequences of pyroptosis-especially beyond interleukin (IL)-1 cytokines and DAMPs-that govern inflammation are poorly defined. Here, we show intercellular propagation of pyroptosis from dying cells to bystander cells in vitro and in vivo. We identified extracellular vesicles (EVs) released by pyroptotic cells as the propagator of lytic death to naive cells, promoting inflammation. DNA-PAINT super-resolution and immunoelectron microscopy revealed GSDMD pore structures on EVs released by pyroptotic cells. Importantly, pyroptotic EVs transplant GSDMD pores on the plasma membrane of bystander cells and kill them. Overall, we demonstrate that cell-to-cell vesicular transplantation of GSDMD pores disseminates pyroptosis, revealing a domino-like effect governing disease-associated bystander cell death.
2. The tryptophan metabolite kynurenic acid ameliorates septic colonic injury through activation of the PPARγ signaling pathway.
Multiomics identified kynurenic acid as enriched in M2 macrophage supernatant and decreased in sepsis serum, inversely correlating with inflammatory mediators. KYNA administration mitigated septic colonic injury in mice by promoting M2 polarization through PPARγ activation and NF-κB inhibition.
Impact: It links a specific tryptophan metabolite to macrophage polarization and demonstrates organ-protective effects in sepsis models, suggesting a tractable immunometabolic therapy.
Clinical Implications: KYNA levels may serve as a biomarker for inflammatory status, and PPARγ-targeted strategies or KYNA analogs could be explored to restore macrophage balance and reduce gastrointestinal injury in sepsis.
Key Findings
- Non-targeted metabolomics revealed KYNA is significantly enriched in M2 macrophage supernatants.
- Serum KYNA levels are reduced in sepsis in both humans and mice and inversely correlate with inflammatory cytokines.
- Exogenous KYNA alleviates septic colonic injury in mice via PPARγ activation and NF-κB inhibition, promoting M1-to-M2 polarization.
Methodological Strengths
- Integrated multiomics with human and mouse serum data and in vivo functional validation
- Mechanistic dissection of the PPARγ/NF-κB axis linking metabolite signaling to macrophage polarization
Limitations
- Findings are primarily preclinical; no human interventional data on KYNA supplementation
- Organ-specific focus on colon; systemic sepsis outcomes, dosing, and safety profiles remain undefined
Future Directions: Validate KYNA as a biomarker in clinical sepsis cohorts, test PPARγ-targeted or KYNA-based therapies in larger animal models, and define pharmacokinetics and safety.
Sepsis is the leading cause of death among critically ill patients in clinical practice, making it urgent to reduce its incidence and mortality rates. In sepsis, macrophage dysfunction often worsens and complicates the condition. M1 and M2 macrophages, two distinct types, contribute to pro-inflammatory and anti-inflammatory effects, respectively. An imbalance between them is a major cause of sepsis. The aim of this study was to explore the potential of a differential metabolite between M1 and M2 macrophages in mitigating septic colonic injury via multiomics in combination with clinical data and animal experiments. Using nontargeted metabolomics analysis, we found that Kynurenic acid (KYNA), a metabolite of tryptophan metabolism, was significantly upregulated in the supernatant of M2 macrophages. Furthermore, we discovered that the level of KYNA was significantly decreased in sepsis in both human and mouse serum and was negatively correlated with inflammatory factor levels. In vivo experiments demonstrated that KYNA can effectively alleviate septic colon injury and reduce inflammatory factor levels in mice, indicating that KYNA plays a very important protective role in sepsis. Mechanistically, KYNA promotes the transition of M1 macrophages to M2 macrophages by inhibiting the NF-κB signaling pathway and alleviates septic colonic injury through the PPARγ/NF-κB axis. This article reveals that KYNA, a differentially abundant metabolite between M1 and M2 macrophages, can become a new strategy for alleviating septic colon injury.
3. Machine learning and clinician predictions of antibiotic resistance in Enterobacterales bloodstream infections.
Using 4,709 Enterobacterales bloodstream infection episodes, XGBoost models achieved AUCs of 0.68–0.74 for resistance prediction, improving to 0.72–0.83 with species identification. Compared with clinician prescribing, model-guided choices could raise active beta-lactam coverage from 70% to 79% while reducing undertreatment.
Impact: This real-world analysis shows that even modest ML performance can improve empirical antibiotic coverage and stewardship compared with current practice.
Clinical Implications: Embedded decision support could guide initial beta-lactam selection before culture results, reducing undertreatment without excessive broad-spectrum use. Prospective validation is needed prior to deployment.
Key Findings
- Trained and evaluated on 4,709 Enterobacterales bloodstream infection episodes with resistance rates 7–67%.
- Held-out AUCs were 0.680–0.737 without species data and improved to 0.723–0.827 when species identification was included.
- Simulation suggested active beta-lactam coverage could increase from 70% (clinicians) to 79% (model-guided), reducing undertreatment from 30% to 21%.
Methodological Strengths
- Large, contemporary cohort with temporal external validation on held-out 2022–2023 data
- Direct comparison to clinician prescribing decisions, quantifying potential stewardship impact
Limitations
- Model performance was modest (AUCs ~0.68–0.83) and derived from a single region, limiting generalizability
- Retrospective simulation without prospective clinical outcome validation
Future Directions: Prospective, multicenter trials to assess clinical outcomes and safety of ML-guided empiric therapy; EHR integration, model calibration, and fairness assessment.
BACKGROUND: Patients with Gram-negative bloodstream infections are at risk of serious adverse outcomes without active treatment, but identifying who has antimicrobial resistance (AMR) to target empirical treatment is challenging. METHODS: We used XGBoost machine learning models to predict antimicrobial resistance to seven antibiotics in patients with Enterobacterales bloodstream infection. Models were trained using hospital and community data from Oxfordshire, UK, for patients with positive blood cultures between 01-January-2017 and 31-December-2021. Model performance was evaluated by comparing predictions to final microbiology results in test datasets from 01-January-2022 to 31-December-2023 and to clinicians' prescribing. FINDINGS: 4709 infection episodes were used for model training and evaluation; antibiotic resistance rates ranged from 7-67%. In held-out test data, resistance prediction performance was similar for the seven antibiotics (AUCs 0.680 [95%CI 0.641-0.720] to 0.737 [0.674-0.797]). Performance improved for most antibiotics when species identifications (available ∼24 h later) were included as model inputs (AUCs 0.723 [0.652-0.791] to 0.827 [0.797-0.857]). In patients treated with a beta-lactam, clinician prescribing led to 70% receiving an active beta-lactam: 44% were over-treated (broader spectrum treatment than needed), 26% optimally-treated (narrowest spectrum active agent), and 30% under-treated (inactive beta-lactam). Model predictions without species data could have led to 79% of patients receiving an active beta-lactam: 45% over-treated, 34% optimally-treated, and 21% under-treated. CONCLUSIONS: Predicting AMR in bloodstream infections is challenging for both clinicians and models. Despite modest performance, machine learning models could still increase the proportion of patients receiving active empirical treatment by up to 9% over current clinical practice in an environment prioritising antimicrobial stewardship.