Infectious diseases have a substantial global health impact. Clinicians need rapid and accurate diagnoses of infections to direct patient treatment and improve antibiotic stewardship, but current methodologies face severe limitations in this regard. In the first funding cycle of our MPI grant ?GM111066 - MS diagnostic bacterial identification library,? we produced a novel diagnostic platform in which microbial membrane glycolipids analyzed by mass spectrometry represent chemical ?fingerprints? that were then used to differentiate Gram- negative and ?positive and fungal isolates after mono- or poly-microbial growth in standard laboratory medias or complex biological (urine, blood bottles, and would effluent). In the second funding cycle, we aim to improve the diagnostic as discussed below. At the start this project, it had not been previously shown that bacterial or fungal membrane lipids could provide a unique chemical signature or barcode that could be used for reliable pathogen identification. The fact that these lipids (Gram-: LPS/lipid A, Gram+: Lipoteichoic acid/cardiolipin, Fungi: glycerophospholipids, sphingolipids, and sterols) are present in high abundance (~106 copies per cell) makes them easily extractable with a single rapid LPS-based protocol (less than 60 minutes from sample to MS identification). Importantly, for clinical use, we successfully used our platform to solve these four major unmet needs from the protein-based phenotyping approach: 1) removed the need for growth prior to MS analysis, 2) identification of bacterial and fungal isolates with a single extraction protocol, 3) identification directly from complex biological fluids, including urine, BAL fluid, wound effluent, and blood bottles, and 4) antimicrobial resistant strains could be distinguished from the related susceptible strain. Finally, based on our thirteen peer-reviewed publications from the first funding period and extensive preliminary data, we believe we have proven our highly innovative original hypothesis and even advanced it past the original aims by using a design of experiment (DOE) process to allow identification in under an hour direct from specimen. In the second funding cycle, we propose to further innovate by i) using DOE to improve limit of detection (LOD) from 106 to 103 which is the threshold for urinary tract infections; ii) extend the assay to direct analysis of urine and stool samples without culture; iii) develop machine learning approaches to improve identification of individual bacteria from polymicrobial infections; iv) expand detection of antimicrobial resistance beyond colistin; v) develop a method for identification and structure analysis of lipids isolated from 100-1000 cells; and vi) vastly expand our ability to identify pathogenic fungi, which are a growing healthcare issue, and Gram- positive organisms.
Failure to identify bacterial and fungal infections rapidly places an enormous financial burden on the U.S. health system totaling billions annually with complications from minor events extending hospital stays to death from sepsis. We have developed a lipid-based MS method for bacterial and fungal identification that can be carried out in under an hour direct from specimen. Here, we will further develop this method for use in the study of complex biological fluids, including stool, urine, and blood bottles without culture. Specifically, we will improve the limit of detection in order to distinguish microbial load at or below 103 CFU/ml ! !
