Abstract Background Gram-negative bacterial bloodstream infections (GNB-BSI) are common and frequently lethal. Many patients experience recurrent GNB-BSI for unclear reasons. This study explores how antibiotic tolerance may lead to infection relapse.Figure 1:Flow diagram of relapsed versus recurrent infection. Approximately 63% of recurrent GNB-BSI are due to relapse (almost genetically identical isolate) and 37% due to reinfection (genetically distinct isolate).Figure 2:Relapsed GNB-BSI isolate exhibits increased antibiotic tolerance in murine bacteremia model. BALB/c mice infected with paired initial or relapsed GNB-BSI isolate. 50 mg/kg ertapenem added at 3 hours post infection (hpi). Graph shows recovery of viable bacteria from liver tissue. Methods We used a prospective cohort of patients with GNB-BSI at Duke Hospital to identify patients with >1 episode of GNB-BSI due to the same bacterial species. We used whole genome sequencing (WGS) to distinguish reinfection (recurrent infection with different isolate) from relapse (recurrent infection with near-identical isolate). Time-kill curves with meropenem were used to determine the development of antibiotic tolerance between the initial and relapsed Escherichia coli isolates. The biological relevance of antibiotic tolerance was tested using a murine bacteremia model. Results We determined that 63% (30/48) of recurrent GNB-BSI episodes were due to relapse and 37% (18/48) were due to reinfection. We screened 10 relapsed E. coli pairs (initial and relapsed isolates) using meropenem to identify increases in antibiotic tolerance. One isolate pair showed a 100-fold increase in multidrug antibiotic tolerance. We determined the decreased antibiotic killing was due to a loss-of-function mutation in the ptsI gene encoding Enzyme I of the phosphoenolpyruvate phosphotransferase system. To test if the in-vitro tolerance phenotype translated to decreased antibiotic efficacy in-vivo, we developed a murine model of E. coli bacteremia. In our murine bacteremia model, the ptsI mutant was equally virulent as the wild-type, but exhibited 10-fold less killing during antibiotic treatment. Conclusion Our work provides a unique insight into the molecular changes occurring during GNB-BSI and how the pathogen adapts to the host through acquisition of antibiotic tolerance. We address the controversy regarding the clinical relevance of antibiotic tolerance in-vivo by providing compelling data that not only do these mutations arise during bloodstream infection in humans, but the presence of antibiotic tolerance in-vitro likely leads to decreased antibiotic efficacy in-vivo. Further work is required to determine if early detection of antibiotic tolerance could lead to alterations in medical management and ultimately improve patient outcomes. Disclosures Vance G. Fowler, MD, MHS, Amphliphi Biosciences, Integrated Biotherapeutics; C3J, Armata, Valanbio; Akagera, Aridis, Roche, Astra Zeneca: Advisor/Consultant|Genentech, Regeneron, Deep Blue, Basilea, Janssen;: Grant/Research Support|Infectious Diseases Society of America: Honoraria|MedImmune, Allergan, Pfizer, Advanced Liquid Logics, Theravance, Novartis, Merck; Medical Biosurfaces; Locus; Affinergy; Contrafect; Karius;: Grant/Research Support|Novartis, Debiopharm, Genentech, Achaogen, Affinium, Medicines Co., MedImmune, Bayer, Basilea, Affinergy, Janssen, Contrafect, Regeneron, Destiny,: Advisor/Consultant|Sepsis diagnostic: Patent pending|UpToDate: Royalties|Valanbio and ArcBio: Stock Options Joshua T. Thaden, MD, PhD, Resonantia Diagnostics, Inc: Advisor/Consultant