Rapid culture-free analysis of bloodstream pathogens for transforming sepsis care

PROJECT SUMMARY Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to an infection. Majority of sepsis is bacterial, and is typically accompanied by bacteremia, i.e., the presence of bacteria in blood at 1-104 CFU/mL. Each year, sepsis kills over 10 million people globally, and is the most expensive condition treated in US hospitals costing over $60 billion annually. It is a medical emergency requiring early recognition and immediate treatment, but the lack of rapid and actionable tests leads to delays in recognition and treatment of sepsis, non-optimal treatment, and overuse of broad-spectrum antibiotics resulting in emergence of drug- resistant pathogens. We propose an integrated “DIMS-Raman” platform to rapidly isolate and analyze bacterial cells from blood samples by machine learning-enabled Raman spectroscopy within approximately 1-2 h to provide actionable information to guide treatment in a timely manner. The proposed system combines a novel yet simple fluidic platform called Density-shift Immunocapture Separation (DIMS) that enables direct separation of microbial cells from whole blood, with Raman spectroscopy – an analytical technique that generates molecular fingerprints of targets from the inelastic scattering of light in a sample – combined with machine learning-based spectral analysis and classification. We envision that target cells isolated by DIMS are concentrated into micro/nanoliter volumes for direct microscopic observation and Raman spectroscopy at the single cell level, enabling culture-free enumeration, and machine learning-enabled identification of bacterial strains and assessment of antibiotic resistance. Our preliminary study using DIMS was able to isolate, concentrate by 10,000x, and image salmonella (~30 CFU/mL) at the single cell level from blood within 3 h. Bacterial cells have unique Raman signatures, which, using machine learning based spectral analysis have enabled differentiation of the 30 most common sepsis related pathogens (including methicillin resistant and susceptible S. aureus) with a 99% classification accuracy using just 10 cells. Moreover, our work demonstrated applications in heterogeneous mixtures of bacteria species and blood cells in liquid droplets with 92% accuracy. In addition, we further demonstrated antibiotic co-incubation free susceptibility testing across the four major anti-tuberculosis drugs with 98% accuracy. The proposed work aims to develop an integrated DIMS-Raman platform for sepsis by demonstrating integrated isolation, concentration, detection, and Raman analysis at the single cell level of the most common sepsis-causing bacteria with a sensitivity of 1 CFU/mL from whole blood within 1-2 h. If successful, the proposed work could eventually lead to a platform capable of returning pathogen type, concentration, and antibiotic susceptibility of clinically-relevant pathogens from blood without relying on culture, thus transforming sepsis detection and informing appropriate antibiotic treatment in a timely manner, saving human lives and reducing the cost of care.