Machine learning approaches offer the potential to systematically identify transcriptional regulatory interactions from a compendium of microarray expression profiles. However, experimental validation of the performance of these methods at the genome scale has remained elusive. Here we assess the global performance of four existing classes of inference algorithms using 445 Escherichia coli Affymetrix arrays and 3,216 known E. coli regulatory interactions from RegulonDB. We also developed and applied the context likelihood of relatedness (CLR) algorithm, a novel extension of the relevance networks class of algorithms. CLR demonstrates an average precision gain of 36% relative to the next-best performing algorithm. At a 60% true positive rate, CLR identifies 1,079 regulatory interactions, of which 338 were in the previously known network and 741 were novel predictions. We tested the predicted interactions for three transcription factors with chromatin immunoprecipitation, confirming 21 novel interactions and verifying our RegulonDB-based performance estimates. CLR also identified a regulatory link providing central metabolic control of iron transport, which we confirmed with real-time quantitative PCR. The compendium of expression data compiled in this study, coupled with RegulonDB, provides a valuable model system for further improvement of network inference algorithms using experimental data.

Antibiotic resistance has increased markedly in gram-negative bacteria over the last two decades, and in many cases has been associated with increased mortality and healthcare costs. The adoption of genotyping and next generation whole genome sequencing of large sets of clinical bacterial isolates has greatly expanded our understanding of how antibiotic resistance develops and transmits among bacteria and between patients. Diverse mechanisms of resistance, including antibiotic degradation, antibiotic target modification, and modulation of permeability through the bacterial membrane have been demonstrated. These fundamental insights into the mechanisms of gram-negative antibiotic resistance have influenced the development of novel antibiotics and treatment practices in highly resistant infections. Here, we review the mechanisms and global epidemiology of antibiotic resistance in some of the most clinically important resistance phenotypes, including carbapenem resistant Enterobacteriaceae, extensively drug resistant (XDR) Pseudomonas aeruginosa, and XDR Acinetobacter baumannii. Understanding the resistance mechanisms and epidemiology of these pathogens is critical for the development of novel antibacterials and for individual treatment decisions, which often involve alternatives to β-lactam antibiotics.

Objectives: The role of follow-up blood cultures (FUBCs) in the management of Gram-negative bacteraemia (GNB) is poorly understood. We aimed to determine the utility of FUBCs in identifying patients with increased mortality risk. Methods: An observational study with a prospectively enrolled cohort of adult inpatients with GNB was conducted at Duke University Health System from 2002 to 2015. FUBCs were defined as blood cultures performed from 24 hours to 7 days from initial positive blood culture. Results: Among 1702 patients with GNB, 1164 (68%) had FUBCs performed. When performed, FUBCs were positive in 20% (228/1113) of cases. FUBC acquisition was associated with lower all-cause in-hospital mortality (108/538, 20%, vs. 176/1164, 15%; p 0.01) and attributable in-hospital mortality (78/538, 15%, vs. 98/1164, 8%; p < 0.0001). Propensity scoreeweighted Cox proportional hazards models revealed that obtaining FUBCs was associated with reductions in all-cause (hazard ratio (HR) 0.629; 95% confidence interval (CI), 0.511e0.772; p < 0.0001) and attributable mortality (HR 0.628; 95% CI, 0.480e0.820; p 0.0007). Positive FUBCs were associated with increased all-cause mortality (49/228, 21%, vs. 110/885, 11%; p 0.0005) and attributable mortality (27/228, 12%, vs. 61/885, 7%; p 0.01) relative to negative FUBCs. Propensity scoreeweighted Cox proportional hazards models revealed that positive FUBCs were associated with increased all-cause (HR 2.099; 95% CI, 1.567e2.811; p < 0.0001) and attributable mortality (HR 1.800; 95% CI, 1.245e2.603; p 0.002). In a calibration analysis, a scoring system accurately identified patients at high risk of positive FUBCs. Conclusions: Rates of positive FUBCs were high and identified patients at increased risk for mortality. Clinical variables can identify patients at high risk for positive FUBCs. FUBCs should be considered in the management of GNB.

OBJECTIVE To describe the epidemiology of extended-spectrum β-lactamase (ESBL)-producing Escherichia coli (ESBL-EC) and Klebsiella pneumoniae (ESBL-KP) infections DESIGN Retrospective cohort SETTING Inpatient care at community hospitals PATIENTS All patients with ESBL-EC or ESBL-KP infections METHODS ESBL-EC and ESBL-KP infections from 26 community hospitals were prospectively entered into a centralized database from January 2009 to December 2014. RESULTS A total of 925 infections caused by ESBL-EC (10.5 infections per 100,000 patient days) and 463 infections caused by ESBL-KP (5.3 infections per 100,000 patient days) were identified during 8,791,243 patient days of surveillance. The incidence of ESBL-EC infections increased from 5.28 to 10.5 patients per 100,000 patient days during the study period (P =.006). The number of community hospitals with ESBL-EC infections increased from 17 (65%) in 2009 to 20 (77%) in 2014. The median ESBL-EC infection rates among individual hospitals with ≥1 ESBL-EC infection increased from 11.1 infections/100,000 patient days (range, 2.2–33.9 days) in 2009 to 22.1 infections per 100,000 patient days (range, 0.66–134 days) in 2014 (P =.05). The incidence of ESBL-KP infections remained constant over the study period (P = .14). Community-associated and healthcare-associated ESBL-EC infections trended upward (P =.006 and P = .02, respectively), while hospital-onset infections remained stable (P = .07). ESBL-EC infections were more common in females (54% vs 44%, P < .001) and Caucasians (50% vs 40%, P < .0001), and were more likely to be isolated from the urinary tract (61% vs 52%, P < .0001) than ESBL-KP infections. CONCLUSIONS The incidence of ESBL-EC infection has increased in community hospitals throughout the southeastern United States, while the incidence of ESBL-KP infection has remained stable. Community- and healthcare-associated ESBL-EC infections are driving the upward trend.

The impact of bacterial species on outcome in bloodstream infections (BSI) is incompletely understood. We evaluated the impact of bacterial species on BSI mortality, with adjustment for patient, bacterial, and treatment factors. From 2002 to 2015, all adult inpatients with monomicrobial BSI caused by Staphylococcus aureus or Gram-negative bacteria at Duke University Medical Center were prospectively enrolled. Kaplan-Meier curves and multivariable Cox regression with propensity score models were used to examine species-specific bacterial BSI mortality. Of the 2,659 enrolled patients, 999 (38%) were infected with S. aureus, and 1,660 (62%) were infected with Gram-negative bacteria. Among patients with Gram-negative BSI, Enterobacteriaceae (81% [1,343/1,660]) were most commonly isolated, followed by non-lactose-fermenting Gram-negative bacteria (16% [262/1,660]). Of the 999 S. aureus BSI isolates, 507 (51%) were methicillin resistant. Of the 1,660 Gram-negative BSI isolates, 500 (30%) were multidrug resistant. The unadjusted time-to-mortality among patients with Gram-negative BSI was shorter than that of patients with S. aureus BSI (P ϭ 0.003), due to increased mortality in patients with non-lactose-fermenting Gramnegative BSI generally (P Ͻ 0.0001) and Pseudomonas aeruginosa BSI (n ϭ 158) in particular (P Ͻ 0.0001). After adjustment for patient demographics, medical comorbidities, bacterial antibiotic resistance, timing of appropriate antibiotic therapy, and source control in patients with line-associated BSI, P. aeruginosa BSI remained significantly associated with increased mortality (hazard ratio ϭ 1.435; 95% confidence interval ϭ 1.043 to 1.933; P ϭ 0.02). P. aeruginosa BSI was associated with increased mortality relative to S. aureus or other Gram-negative BSI. This effect persisted after adjustment for patient, bacterial, and treatment factors. KEYWORDS Gram negative, Pseudomonas aeruginosa, Staphylococcus aureus, bloodstream infections Bacterial bloodstream infections (BSI) are a leading cause of morbidity and mortality in contemporary medical practice (1-3). The determinants of poor outcome in patients with BSI are among the most intensely investigated areas of infectious diseases. Although the intrinsic characteristics of bacterial species are thought to influence clinical outcomes, there is surprisingly little clinical data to support this impression. Several studies have shown that patients with Pseudomonas aeruginosa BSI experienced increased crude mortality relative to BSI caused by other bacterial species (4-9). However, these studies were unable to discriminate how much of this clinical outcome, if any, was attributable to the intrinsic virulence of P. aeruginosa rather than patient or