Plasma multi-omics reveals pathogen-associated mechanisms and diagnostic signatures in Escherichia coli and Klebsiella pneumoniae bloodstream infections.

OBJECTIVE

Escherichia coli and Klebsiella pneumoniae are major Gram-negative pathogens causing bloodstream infection (BSI) and sepsis, but their plasma host-response patterns remain incompletely characterized. This study aimed to explore shared and pathogen-associated proteomic and metabolomic features in E. coli BSI and K. pneumoniae BSI.

METHODS

We performed plasma proteomic and metabolomic profiling in 100 participants, including patients with E. coli BSI (n = 26), K. pneumoniae BSI (n = 15), blood culture-negative sepsis (NegBC, n = 29), and healthy controls (HC, n = 30). Differential molecular analyses were performed with additional adjustment for baseline clinical covariates where applicable. Downstream GO, KEGG, PPI, and protein-metabolite network analyses were based on molecules that remained significant after covariate adjustment. A leakage-free internal validation framework was used to construct an exploratory plasma proteome-based model for distinguishing E. coli BSI from non-E. coli sepsis.

RESULTS

Both E. coli BSI and K. pneumoniae BSI showed prominent coagulation-immune alterations involving complement/coagulation-related pathways. E. coli BSI showed a more evident platelet-centered immunometabolic pattern, particularly relative to NegBC, whereas K. pneumoniae BSI showed coagulation-immune activation coupled with lipid and broader metabolic remodeling. In the direct comparison between E. coli BSI and K. pneumoniae BSI, no proteins remained significantly different after covariate adjustment, whereas metabolomic differences persisted, involving glycine, serine and threonine metabolism, terpenoid backbone biosynthesis, apoptosis, and folate biosynthesis. An exploratory E. coli BSI model based on FCGR3A, COPS3, NT5C3A, and ADO showed preliminary discrimination between E. coli BSI and non-E. coli sepsis in leakage-free internal validation (AUC = 0.806, 95% CI: 0.706-0.906).

CONCLUSION

Plasma multi-omics suggested shared thromboinflammatory features and partially distinct platelet-associated and metabolic patterns in E. coli BSI and K. pneumoniae BSI. However, findings involving direct pathogen comparison and the K. pneumoniae BSI cohort should be interpreted cautiously because of baseline clinical heterogeneity and limited sample size. The candidate diagnostic model remains exploratory and requires external validation.