Rapid antibiotic susceptibility testing based on bacterial motion patterns with long short-term memory neural networks.

Antibiotic resistance is an increasing public health threat. To combat it, a fast method to determine the antibiotic susceptibility of infecting pathogens is required. Here we present an optical imaging-based method to track the motion of single bacterial cells and generate a model to classify active and inactive cells based on the motion patterns of the individual cells. The model includes an image-processing algorithm to segment individual bacterial cells and track the motion of the cells over time, and a deep learning algorithm (Long Short-Term Memory network) to learn and determine if a bacterial cell is active or inactive. By applying the model to human urine specimens spiked with an Escherichia coli lab strain, we show that the method can accurately perform antibiotic susceptibility testing as fast as 30 minutes for five commonly used antibiotics.

Phenotypic Antimicrobial Susceptibility Testing with Deep Learning Video Microscopy.

Timely determination of antimicrobial susceptibility for a bacterial infection enables precision prescription, shortens treatment time, and helps minimize the spread of antibiotic resistant infections. Current antimicrobial susceptibility testing (AST) methods often take several days and thus impede these clinical and health benefits. Here, we present an AST method by imaging freely moving bacterial cells in urine in real time and analyzing the videos with a deep learning algorithm. The deep learning algorithm determines if an antibiotic inhibits a bacterial cell by learning multiple phenotypic features of the cell without the need for defining and quantifying each feature. We apply the method to urinary tract infection, a common infection that affects millions of people, to determine the minimum inhibitory concentration of pathogens from human urine specimens spiked with lab strain E. coli (ATCC 43888) and an E. coli strain isolated from a clinical urine sample for different antibiotics within 30 min and validate the results with the gold standard broth macrodilution method. The deep learning video microscopy-based AST holds great potential to contribute to the solution of increasing drug-resistant infections.

Real-time detection of antibiotic activity by measuring nanometer-scale bacterial deformation.

Diagnosing antibiotic-resistant bacteria currently requires sensitive detection of phenotypic changes associated with antibiotic action on bacteria. Here, we present an optical imaging-based approach to quantify bacterial membrane deformation as a phenotypic feature in real-time with a nanometer scale (∼9 nm) detection limit. Using this approach, we found two types of antibiotic-induced membrane deformations in different bacterial strains: polymyxin B induced relatively uniform spatial deformation of Escherichia coli O157:H7 cells leading to change in cellular volume and ampicillin-induced localized spatial deformation leading to the formation of bulges or protrusions on uropathogenic E. coli CFT073 cells. We anticipate that the approach will contribute to understanding of antibiotic phenotypic effects on bacteria with a potential for applications in rapid antibiotic susceptibility testing.

Rapid Antibiotic Susceptibility Testing of Uropathogenic E. coli by Tracking Submicron Scale Motion of Single Bacterial Cells.

To combat antibiotic resistance, a rapid antibiotic susceptibility testing (AST) technology that can identify resistant infections at disease onset is required. Current clinical AST technologies take 1-3 days, which is often too slow for accurate treatment. Here we demonstrate a rapid AST method by tracking sub-µm scale bacterial motion with an optical imaging and tracking technique. We apply the method to clinically relevant bacterial pathogens, Escherichia coli O157: H7 and uropathogenic E. coli (UPEC) loosely tethered to a glass surface. By analyzing dose-dependent sub-µm motion changes in a population of bacterial cells, we obtain the minimum bactericidal concentration within 2 h using human urine samples spiked with UPEC. We validate the AST method using the standard culture-based AST methods. In addition to population studies, the method allows single cell analysis, which can identify subpopulations of resistance strains within a sample.

Current and emerging techniques for antibiotic susceptibility tests.

Infectious diseases caused by bacterial pathogens are a worldwide burden. Serious bacterial infection-related complications, such as sepsis, affect over a million people every year with mortality rates ranging from 30% to 50%. Crucial clinical microbiology laboratory responsibilities associated with patient management and treatment include isolating and identifying the causative bacterium and performing antibiotic susceptibility tests (ASTs), which are labor-intensive, complex, imprecise, and slow (taking days, depending on the growth rate of the pathogen). Considering the life-threatening condition of a septic patient and the increasing prevalence of antibiotic-resistant bacteria in hospitals, rapid and automated diagnostic tools are needed. This review summarizes the existing commercial AST methods and discusses some of the promising emerging AST tools that will empower humans to win the evolutionary war between microbial genes and human wits.

Antimicrobial Susceptibility Test with Plasmonic Imaging and Tracking of Single Bacterial Motions on Nanometer Scale.

Antimicrobial susceptibility tests (ASTs) are important for confirming susceptibility to empirical antibiotics and detecting resistance in bacterial isolates. Currently, most ASTs performed in clinical microbiology laboratories are based on bacterial culturing, which take days to complete for slowly growing microorganisms. A faster AST will reduce morbidity and mortality rates and help healthcare providers administer narrow spectrum antibiotics at the earliest possible treatment stage. We report the development of a nonculture-based AST using a plasmonic imaging and tracking (PIT) technology. We track the motion of individual bacterial cells tethered to a surface with nanometer (nm) precision and correlate the phenotypic motion with bacterial metabolism and antibiotic action. We show that antibiotic action significantly slows down bacterial motion, which can be quantified for development of a rapid phenotypic-based AST.

Note: An automated image analysis method for high-throughput classification of surface-bound bacterial cell motions.

We present a Single-Cell Motion Characterization System (SiCMoCS) to automatically extract bacterial cell morphological features from microscope images and use those features to automatically classify cell motion for rod shaped motile bacterial cells. In some imaging based studies, bacteria cells need to be attached to the surface for time-lapse observation of cellular processes such as cell membrane-protein interactions and membrane elasticity. These studies often generate large volumes of images. Extracting accurate bacterial cell morphology features from these images is critical for quantitative assessment. Using SiCMoCS, we demonstrated simultaneous and automated motion tracking and classification of hundreds of individual cells in an image sequence of several hundred frames. This is a significant improvement from traditional manual and semi-automated approaches to segmenting bacterial cells based on empirical thresholds, and a first attempt to automatically classify bacterial motion types for motile rod shaped bacterial cells, which enables rapid and quantitative analysis of various types of bacterial motion.

Plasmonic imaging of protein interactions with single bacterial cells.

Quantifying the interactions of bacteria with external ligands is fundamental to the understanding of pathogenesis, antibiotic resistance, immune evasion, and mechanism of antimicrobial action. Due to inherent cell-to-cell heterogeneity in a microbial population, each bacterium interacts differently with its environment. This large variability is washed out in bulk assays, and there is a need of techniques that can quantify interactions of bacteria with ligands at the single bacterium level. In this work, we present a label-free and real-time plasmonic imaging technique to measure the binding kinetics of ligand interactions with single bacteria, and perform statistical analysis of the heterogeneity. Using the technique, we have studied interactions of antibodies with single Escherichia coli O157:H7 cells and demonstrated a capability of determining the binding kinetic constants of single live bacteria with ligands, and quantify heterogeneity in a microbial population.

High density diffusion-free nanowell arrays.

Proteomics aspires to elucidate the functions of all proteins. Protein microarrays provide an important step by enabling high-throughput studies of displayed proteins. However, many functional assays of proteins include untethered intermediates or products, which could frustrate the use of planar arrays at very high densities because of diffusion to neighboring features. The nucleic acid programmable protein array (NAPPA) is a robust in situ synthesis method for producing functional proteins just-in-time, which includes steps with diffusible intermediates. We determined that diffusion of expressed proteins led to cross-binding at neighboring spots at very high densities with reduced interspot spacing. To address this limitation, we have developed an innovative platform using photolithographically etched discrete silicon nanowells and used NAPPA as a test case. This arrested protein diffusion and cross-binding. We present confined high density protein expression and display, as well as functional protein-protein interactions, in 8000 nanowell arrays. This is the highest density of individual proteins in nanovessels demonstrated on a single slide. We further present proof of principle results on ultrahigh density protein arrays capable of up to 24000 nanowells on a single slide.