Plasma Metabolomic Signatures in Patients with Multidrug-Resistant Bacterial Sepsis

Abstract Background and Objective: Multidrug-resistant (MDR) bacterial infections are a major cause of sepsis-related death. This study aimed to characterize the distinct plasma metabolomic signatures associated with MDR gram-positive and gram-negative bacterial infections to facilitate early sepsis classification and diagnosis and to provide insights for optimizing antimicrobial therapy. Methods: Two cohorts of septic patients were recruited, with 215 subjects (98 with MDR and 117 with susceptible strain infections) in the discovery cohort, and 198 patients (95 MDR and 103 susceptible bacteria) in the validation cohort. Plasma metabolomic profiling was performed using liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). Multiple machine learning was used to identify the distinct metabolomic signatures associated with MDR infections and to develop predictive models for early MDR bacterial identification. Results: The plasma metabolomic profiles of patients with MDR G- and G+ bacteria were both separated from those of non-MDR bacterial infection in the discovery and validation cohorts, respectively. Moreover, Variable importance in projection (VIP) analysis further confirmed MDR-specific metabolic alterations. Random forest identified the top 15 most discriminative metabolites between MDR and non-MDR septic patients. In MDR G- infections, key perturbed pathways included sphingomyelin metabolism, ceramide metabolism, and purine metabolism, whereas in MDR G+ infections, sphingomyelin metabolism and arginine biosynthesis were the most significantly altered pathways. The predictive models demonstrated strong diagnostic performance, with AUROC values of 0.885 and 0.878 in the G- discovery and validation cohorts, respectively, and 0.763 and 0.715 in the G+ cohorts. Conclusion: This study identifies distinct plasma metabolomic signatures in patients with MDR bacterial sepsis, highlighting key metabolic alterations that differentiate MDR from non-MDR infections. These findings not only provide a basis for the early identification of MDR bacterial infections but also suggest potential host metabolic targets for adjunctive therapeutic strategies beyond conventional antibiotic treatment.