Abstract Rationale Sepsis is a common and heterogeneous syndrome of organ damage and immune dysregulation due to infection, and a major cause of morbidity and mortality worldwide. In recent years, substantial effort has been devoted to subphenotyping sepsis to understand the biological basis of its heterogeneity. This work has led to the discovery of hyperinflammatory and hypoinflammatory subphenotypes, which are associated with distinct prognoses and respond differently to common sepsis treatments. However, the biologic factors determining these subphenotypes remain incompletely understood. The gut microbiome plays a major role in regulating host immunity and is variably altered by antibiotic exposure, yet its contribution to sepsis subphenotypes remains unknown. We hypothesized that prior antibiotic treatments may influence these subphenotypes through gut microbiome effects. Methods We studied a previously described cohort of 7,569 adults hospitalized with sepsis at the University of Michigan (2014-2018), of whom 1,398 (18%) were subsequently readmitted with sepsis. To isolate the effect of prior antibiotic exposure from confounding by indication, we leveraged a natural experiment: a 2015-2016 national shortage of piperacillin-tazobactam, which shifted empiric prescribing between piperacillin-tazobactam (anti-anaerobic) and cefepime (anaerobe-sparing) at the first (index) admission. Each patient’s readmission was classified as hyperinflammatory or hypoinflammatory using the Sinha et al. clinical classifier derived from gradient-boosted machine learning models. Then, using mouse models, we pretreated mice with vehicle, cefepime, piperacillin-tazobactam, or cefepime + metronidazole and then injected them with 5 mg/kg lipopolysaccharide and measured serum cytokines 24 hours post-injection. Results Among patients readmitted with sepsis, 712 (51%) had received piperacillin-tazobactam and 686 (49%) received cefepime during their first admission. Baseline comorbidity, age, and illness severity were similar across groups. Prior piperacillin-tazobactam exposure was associated with a 15.7% lower probability of the hyperinflammatory phenotype on readmission (β = -0.16; P 0.001). Mice pretreated with anti-anaerobic antibiotics (piperacillin-tazobactam or metronidazole) had significantly lower levels of the cytokines TNFR1, IL-6, and CXCL1 compared to animals treated with vehicle or anaerobe-sparing antibiotic (cefepime). Conclusions Prior antibiotic exposure shapes the host inflammatory response in subsequent sepsis. In humans and mice, depletion of gut anaerobic bacteria produced a markedly blunted inflammatory phenotype in subsequent sepsis. These findings identify past antibiotic exposure as a key driver of physiologic heterogeneity in sepsis via microbiome-mediated effects. This abstract is funded by: National Institutes of Health, ZOLL Foundation
Bongers et al. (Fri,) studied this question.