Metabolic preference assay for rapid diagnosis of bloodstream infections

Bloodstream infections (BSIs) are a leading cause of mortality in North America and Europe, resulting in over 80,000 deaths annually in North America alone. The delay between symptom onset and appropriate antimicrobial treatment is a critical factor determining patient survival. A single day's delay can increase mortality by up to 5%, while delays in septic shock increase mortality by over 70% per hour. Current diagnostic techniques, including microbial identification (ID) and antimicrobial susceptibility testing (AST), typically require 2–5 days, hindering timely treatment. This delay also contributes to the misuse of broad-spectrum antibiotics and the selection of antimicrobial-resistant strains. The current clinical workflow involves several time-consuming steps: incubation to achieve detectable microbial densities, subculturing to obtain single colonies, identification using MALDI-TOF MS, and AST using automated systems. While some time savings have been achieved through methods like direct MALDI-TOF-MS, these do not address the time-consuming AST workflow. DNA-based technologies offer some speed but lack the capacity for complete antimicrobial susceptibility profiles. The study explores the use of metabolomics, a sensitive and high-throughput approach, to rapidly identify pathogens and assess their susceptibility to antimicrobials, offering a potential solution to this critical challenge. The high abundance of metabolites compared to proteins and compatibility with existing clinical mass spectrometry platforms make metabolomics a promising candidate for minimizing rate-limiting steps in the existing clinical workflow. Previous metabolomics approaches have focused on identifying biomarkers in patient blood, which is challenging due to the intrinsic variability of human metabolism. This study proposes a different approach focusing on the inherent metabolic differences of microbial cultures.

Existing literature highlights the critical need for rapid diagnostic tools to combat the high mortality associated with bloodstream infections (BSIs). The lengthy turnaround times of current diagnostic methods, which typically involve 2–5 days for both microbial identification and antimicrobial susceptibility testing, directly impact patient outcomes and contribute to antibiotic resistance. Studies have consistently shown a strong correlation between delayed treatment and increased mortality rates in sepsis and septic shock. The current diagnostic toolkit suffers from inherent variability and multiple processing steps. Various efforts to accelerate diagnostics have been explored, including streamlining existing workflows, employing direct MALDI-TOF-MS, and utilizing DNA-based technologies. However, these approaches have limitations, either in their speed or in their ability to provide comprehensive antimicrobial susceptibility profiles. Molecular-based assays can sometimes identify pathogens faster, but often require culture-based isolation which slows them down and typically requires further testing to determine antibiotic susceptibility. This necessitates a new diagnostic strategy capable of integrating pathogen identification and antimicrobial susceptibility testing into a single, faster workflow. The literature suggests that metabolomics may provide such a solution, given its high throughput and sensitivity. Previous studies have explored the use of metabolomics in identifying infections, but these primarily focused on analyzing host biomarkers in blood, which can be hampered by inherent metabolic variability. This study innovates by investigating the metabolic profiles of microbial cultures themselves.

The study introduces the metabolic preference assay (MPA) which uses patterns of consumed and excreted metabolites in ex vivo microbial cultures for pathogen identification and antimicrobial susceptibility testing. Initially, metabolic boundary fluxes (rates of nutrient and waste product flow) of seven common bloodstream pathogens were measured using LC-MS/MS. Three independent experiments were performed, each with three clinical isolates of each species, incubated in triplicate and analyzed at 0 h and 4 h. Untargeted analysis using MAVEN software identified 53 putative markers, which were further processed to identify 210 biomarkers using in-house software and validated by MS/MS fragmentation and comparison to commercial standards. These biomarkers successfully differentiated the seven target species in the discovery dataset. The diagnostic robustness of the assay was validated using an independent cohort of 696 clinical isolates, which confirmed that the 210 markers identified in the discovery phase retained their species-specific patterns. The quantitative reliability was also assessed by analyzing mixtures of chemical standards alongside clinical samples, resulting in RMSE values that did not significantly impact assay performance. In addition to the MPA, a metabolic inhibition assay (MIA) was developed to assess antimicrobial susceptibility. MIA monitored changes in microbial culture supernatants following a 4-hour incubation with and without antibiotics. Results demonstrated that drug-sensitive and resistant strains produced significantly different metabolic profiles. This assay was validated against traditional growth-based methods, showing 96% consistency across different antimicrobial mechanisms of action. For high-throughput analysis, a rapid HILIC method was optimized for the MIA. A larger cohort (n=246) was used to further validate MIA, achieving 95.2% agreement with VITEK 2 results in antibiotic sensitivity predictions. Head-to-head comparisons between the MPA/MIA workflow and VITEK 2 showed a 2-3 fold decrease in total testing time. Chromatography was performed using a Thermo Fisher Scientific Vanquish UPLC platform with HILIC, and mass spectrometry was performed on a Thermo Scientific Q Exactive™ HF. Data analysis involved statistical methods such as ANOVA, Tukey-Kramer post-hoc tests, and ROC curve analysis.

The study identified seven metabolites sufficient for robustly differentiating seven common BSI pathogens: arachidonic acid (C. albicans), urocanate, succinate, xanthine, mevalonate, N',N'-diacetylspermine (+FA-H), and lactate. These metabolites, assessed via MPA, accurately distinguished between species in both discovery (three independent experiments, three isolates of each of seven species) and validation (696 clinical isolates) cohorts. The MPA assay showed high quantitative reliability, with minimal error rates that did not affect performance. The metabolic inhibition assay (MIA) accurately determined antimicrobial susceptibility profiles, agreeing with traditional growth-based assays in 96% of cases. The MIA was validated using a cohort of 246 isolates, showing 95.2% agreement with VITEK 2 results. Head-to-head comparisons between the MPA/MIA workflow and VITEK 2 demonstrated a significant reduction in testing time – an average of 24.3 h for S. aureus and 22.4 h for E. coli. This represents a 2.2- and 2.3-fold decrease in total testing time, respectively, with an even greater reduction (4.7- to 5-fold) for identification and AST alone. The study found no significant correlation between patient demographics (age and sex) and the identified biomarkers.

The MPA/MIA workflow offers a significant advance in BSI diagnostics. The ability to rapidly identify pathogens and determine their susceptibility profiles within a timeframe significantly shorter than existing methods has the potential to drastically improve patient outcomes. Timely administration of appropriate antibiotics is crucial in reducing mortality rates associated with BSIs, particularly septic shock. The MPA's ability to distinguish between species, even closely related ones, highlights its accuracy and precision. The high concordance between MIA and VITEK 2 validates the reliability of the metabolic inhibition approach for assessing antimicrobial susceptibility. The reduced testing time, as demonstrated in the head-to-head comparison, translates to considerable time savings in clinical settings, allowing for faster treatment decisions and improved patient management. The inherent flexibility of the MPA/MIA workflow, its integration of ID and AST, and its minimal sample handling requirements make it a strong contender for clinical adoption. The use of readily available mass spectrometry platforms further facilitates its integration into existing clinical laboratories.

This study successfully demonstrates a rapid and accurate metabolomics-based workflow (MPA/MIA) for diagnosing bloodstream infections. The MPA accurately identifies common BSI pathogens, while the MIA precisely determines antimicrobial susceptibility. This combined approach significantly reduces diagnostic turnaround time compared to conventional methods, potentially leading to improved patient outcomes and optimized antibiotic stewardship. Future research could focus on expanding the panel of targeted pathogens, refining the assay for even higher throughput, and validating the workflow in diverse clinical settings. Exploring the use of different mass spectrometry platforms and optimization for various microbial species are other directions for continued development.

While the study demonstrated excellent performance with common BSI pathogens, further validation is needed with a broader range of pathogens and clinical isolates. The study primarily focused on seven species, limiting generalizability to less frequent BSI causative agents. Although the quantitative reliability was assessed, potential bias associated with the specific clinical isolates used in the study should be considered. The head-to-head comparison was performed with a limited number of samples; larger-scale studies are warranted for more robust validation.