BACKGROUND Sepsis encompasses considerable biological and clinical heterogeneity. Previously, 2 phenotypes (“hyperinflammatory” and “hypoinflammatory”) have been consistently identified within sepsis via latent class analysis. These phenotypes differ in their biological features, clinical outcomes, and therapeutic responses to interventions. Prior studies of sepsis heterogeneity have focused primarily on the host response. Here, we investigate the potential influence of the causative pathogen on sepsis heterogeneity and pathobiology.METHODS We performed a retrospective observational analysis of 8,280 critically ill patients with sepsis to identify associations between pathogen characteristics and the hyperinflammatory and hypoinflammatory patient phenotypes. We also performed controlled murine and swine modeling of sepsis and lung injury and a secondary analysis of 449 patients enrolled in the EUPHRATES randomized controlled trial.RESULTS Pathogen characteristics (pathogen identity, burden, virulence, and anatomic site of infection) were strongly and independently associated with the previously reported phenotypes. In a cohort of critically ill patients with sepsis, infection with gram-negative pathogens, primarily Enterobacterales spp. (e.g., Escherichia coli, Klebsiella pneumoniae), was strongly associated with the hyperinflammatory phenotype. The hyperinflammatory phenotype was also independently associated with increased pathogen burden, virulence, and initial anatomic site of infection. In controlled murine and swine modeling, both the identity and burden of the pathogen provoked key biological features of the hyperinflammatory phenotype. Among patients with sepsis, the prognostic value of lactate clearance varied substantially by phenotype. In a secondary analysis of a randomized trial of polymyxin B hemoadsorption (which removes circulating endotoxin), hypoinflammatory patients experienced worse survival.CONCLUSIONS Our results demonstrate the central importance of pathogen features in the clinical and biological heterogeneity of sepsis. Future studies of sepsis pathobiology and heterogeneity should expand their scope beyond the host response, as understanding pathogen-host interactions will be crucial in the development of precision therapeutic strategies to improve patient outcomes.TRIAL REGISTRATION EUPHRATES trial NCT01046669.FUNDING 5P30AG024824, IK2CX002766, R01HL144599, K24HL159247, R01HL158626, R01HL173531, R35GM142992, R35GM145330, R35GM136312, K23HL166880, R35HL140026.
Rishi Chanderraj, Brian Bartek, Kathleen A. Stringer, Mohamad H. Tiba, Michael W. Sjoding, Ying He, Mark Nuppnau, Kale S. Bongers, Mark D. Adame, Sunny S. Lou, V. Eric Kerschberger, Matthew M. Churpek, Carolyn S. Calfee, Sandhya Tripathi, Debra M. Foster, John A. Kellum, Robert P. Dickson, Pratik Sinha
Usage data is cumulative from March 2026 through June 2026.
| Usage | JCI | PMC |
|---|---|---|
| Text version | 2,645 | 0 |
| 614 | 0 | |
| Figure | 729 | 0 |
| Table | 244 | 0 |
| Supplemental data | 288 | 0 |
| Citation downloads | 124 | 0 |
| Totals | 4,644 | 0 |
| Total Views | 4,644 | |
Usage information is collected from two different sources: this site (JCI) and Pubmed Central (PMC). JCI information (compiled daily) shows human readership based on methods we employ to screen out robotic usage. PMC information (aggregated monthly) is also similarly screened of robotic usage.
Various methods are used to distinguish robotic usage. For example, Google automatically scans articles to add to its search index and identifies itself as robotic; other services might not clearly identify themselves as robotic, or they are new or unknown as robotic. Because this activity can be misinterpreted as human readership, data may be re-processed periodically to reflect an improved understanding of robotic activity. Because of these factors, readers should consider usage information illustrative but subject to change.