Using Genetic Distance from Archived Samples for the Prediction of Antibiotic Resistance in Escherichia coli

Rapid inference of antibiotic susceptibility phenotype of uropathogens using metagenomic sequencing with neighbor typing

Timely diagnostic tools are needed to improve antibiotic treatment. Pairing metagenomic sequencing with genomic neighbor typing algorithms may support rapid clinically actionable results. We created resistance-associated sequence elements (RASE) databases for Escherichia coli and Klebsiella spp. and used them to predict antibiotic susceptibility in directly sequenced (Oxford Nanopore) urine specimens from critically ill patients. RASE analysis was performed on pathogen-specific reads from metagenomic sequencing. We evaluated the ability to predict (i) multi-locus sequence type (MLST) and (ii) susceptibility profiles. We used neighbor typing to predict MLST and susceptibility phenotype of E. coli (64/80) and Klebsiella spp. (16/80) from urine samples. When optimized by lineage score, MLST predictions were concordant for 73% of samples. Similarly, a RASE-susceptible prediction for a given isolate was associated with a specificity and a positive likelihood ratio (LR+) for susceptibility of 0.65 (95% CI, 0.54-0.76) and 2.26 (95% CI, 1.75-2.92), respectively, with an increase in the probability of susceptibility of 10%. A RASE-non-susceptible prediction was associated with a sensitivity and a negative likelihood ratio (LR-) for susceptibility of 0.79 (95% CI, 0.74-0.84) and 0.32 (95% CI, 0.24-0.43) respectively, with a decrease in the probability of susceptibility of 20%. Numerous antibiotic classes could reasonably be reconsidered empiric therapy by shifting empiric probabilities of susceptibility across relevant treatment thresholds. Moreover, these predictions can be available within 6 h. Metagenomic sequencing of urine specimens with neighbor typing provides rapid and informative predictions of lineage and antibiotic susceptibility with the potential to impact clinical decision-making.

IMPORTANCE

Urinary tract infections (UTIs) are a common diagnosis in hospitals and are often treated empirically with broad-spectrum antibiotics. These broad-spectrum agents can select for resistance in these bacteria and co-colonizing organisms. The use of narrow-spectrum agents is desirable as an antibiotic stewardship measure; however, it is counterbalanced by the need for adequate therapy. Identification of causative organisms and their antibiotic susceptibility can help direct treatment; however, conventional testing requires days to produce actionable results. Methods to quickly and accurately predict susceptibility phenotypes for pathogens causing UTI could thus improve both patient outcomes and antibiotic stewardship. Here, expanding on previous work showing accurate prediction for certain Gram-positive pathogens, we demonstrate how the use of RASE from metagenomic sequencing can provide informative and rapid phenotype prediction results for common Gram-negative pathogens in UTI, highlighting the future potential of this method to be used in clinical settings to guide empiric antibiotic selection.

Next-generation rapid phenotypic antimicrobial susceptibility testing

Slow progress towards implementation of conventional clinical bacteriology in low resource settings and strong interest in greater speed for antimicrobial susceptibility testing (AST) more generally has focused attention on next-generation rapid AST technologies. In this Review, we systematically synthesize publications and submissions to regulatory agencies describing technologies that provide phenotypic AST faster than conventional methods. We characterize over ninety technologies in terms of underlying technical innovations, technology readiness level, extent of clinical validation, and time-to-results. This work provides a guide for technology developers and clinical microbiologists to understand the rapid phenotypic AST technology landscape, current development pipeline, and AST-specific validation milestones.

A nested cohort 5-year Canadian surveillance of Gram-negative antimicrobial resistance for optimized antimicrobial therapy

We analyzed 5 years (2016-2020) of nested Canadian data from the Study for Monitoring Antimicrobial Resistance Trends (SMART) to identify pathogen predominance and antimicrobial resistance (AMR) patterns of adult Gram-negative infections in Canadian health care and to complement other public surveillance programs and studies in Canada. A total of 6853 isolates were analyzed from medical (44%), surgical (18%), intensive care (22%) and emergency units (15%) and from respiratory tract (36%), intra-abdominal (25%), urinary tract (24%) and bloodstream (15%) infections. Overall, E. coli (36%), P. aeruginosa (18%) and K. pneumoniae (12%) were the most frequent isolates and P. aeruginosa was the most common respiratory pathogen. 18% of Enterobacterales species were ESBL positive. Collective susceptibility profiles showed that P. aeruginosa isolates were highly susceptible (> 95%) to ceftolozane/tazobactam and colistin, though markedly less susceptible (58-74%) to other antimicrobials tested. Multi-drug resistance (MDR) was present in 10% of P. aeruginosa isolates and was more frequent in those from respiratory infections and from ICU than non-ICU locations. Of P. aeruginosa isolates that were resistant to combinations of ceftazidime, piperacillin/tazobactam and meropenem, 73-96% were susceptible to ceftolozane/tazobactam over the period of the study. These national data can now be combined with clinical prediction rules and genomic data to enable expert antimicrobial stewardship applications and guide treatment policies to optimize adult patient care.

Performance Characteristics of Next-Generation Sequencing for the Detection of Antimicrobial Resistance Determinants in Escherichia coli Genomes and Metagenomes

Short-read sequencing can provide detection of multiple genomic determinants of antimicrobial resistance from single bacterial genomes and metagenomic samples. Despite its increasing application in human, animal, and environmental microbiology, including human clinical trials, the performance of short-read Illumina sequencing for antimicrobial resistance gene (ARG) detection, including resistance-conferring single nucleotide polymorphisms (SNPs), has not been systematically characterized. Using paired-end 2 × 150 bp (base pair) Illumina sequencing and an assembly-based method for ARG prediction, we determined sensitivity, positive predictive value (PPV), and sequencing depths required for ARG detection in an Escherichia coli isolate of sequence type (ST) 38 spiked into a synthetic microbial community at varying abundances. Approximately 300,000 reads or 15× genome coverage was sufficient to detect ARGs in E. coli ST38, with comparable sensitivity and PPV to ~100× genome coverage. Using metagenome assembly of mixed microbial communities, ARG detection at E. coli relative abundances of 1% would require assembly of approximately 30 million reads to achieve 15× target coverage. The minimum sequencing depths were validated using public data sets of 948 E. coli genomes and 10 metagenomic rectal swab samples. A read-based approach using k-mer alignment (KMA) for ARG prediction did not substantially improve minimum sequencing depths for ARG detection compared to assembly of the E. coli ST38 genome or the combined metagenomic samples. Analysis of sequencing depths from recent studies assessing ARG content in metagenomic samples demonstrated that sequencing depths had a median estimated detection frequency of 84% (interquartile range: 30%-92%) for a relative abundance of 1%. IMPORTANCE Systematically determining Illumina sequencing performance characteristics for detection of ARGs in metagenomic samples is essential to inform study design and appraisal of human, animal, and environmental metagenomic antimicrobial resistance studies. In this study, we quantified the performance characteristics of ARG detection in E. coli genomes and metagenomes and established a benchmark of ~15× coverage for ARG detection for E. coli in metagenomes. We demonstrate that for low relative abundances, sequencing depths of ~30 million reads or more may be required for adequate sensitivity for many applications.

Systems-Based Approach for Optimization of Assembly-Free Bacterial MLST Mapping

Epidemiological surveillance of bacterial pathogens requires real-time data analysis with a fast turnaround, while aiming at generating two main outcomes: (1) species-level identification and (2) variant mapping at different levels of genotypic resolution for population-based tracking and surveillance, in addition to predicting traits such as antimicrobial resistance (AMR). Multi-locus sequence typing (MLST) aids this process by identifying sequence types (ST) based on seven ubiquitous genome-scattered loci. In this paper, we selected one assembly-dependent and one assembly-free method for ST mapping and applied them with the default settings and ST schemes they are distributed with, and systematically assessed their accuracy and scalability across a wide array of phylogenetically divergent Public Health-relevant bacterial pathogens with available MLST databases. Our data show that the optimal k-mer length for stringMLST is species-specific and that genome-intrinsic and -extrinsic features can affect the performance and accuracy of the program. Although suitable parameters could be identified for most organisms, there were instances where this program may not be directly deployable in its current format. Next, we integrated stringMLST into our freely available and scalable hierarchical-based population genomics platform, ProkEvo, and further demonstrated how the implementation facilitates automated, reproducible bacterial population analysis.

Heuristic and Hierarchical-Based Population Mining of Salmonella enterica Lineage I Pan-Genomes as a Platform to Enhance Food Safety

The recent incorporation of bacterial whole-genome sequencing (WGS) into Public Health laboratories has enhanced foodborne outbreak detection and source attribution. As a result, large volumes of publicly available datasets can be used to study the biology of foodborne pathogen populations at an unprecedented scale. To demonstrate the application of a heuristic and agnostic hierarchical population structure guided pan-genome enrichment analysis (PANGEA), we used populations of S. enterica lineage I to achieve two main objectives: (i) show how hierarchical population inquiry at different scales of resolution can enhance ecological and epidemiological inquiries; and (ii) identify population-specific inferable traits that could provide selective advantages in food production environments. Publicly available WGS data were obtained from NCBI database for three serovars of Salmonella enterica subsp. enterica lineage I (S. Typhimurium, S. Newport, and S. Infantis). Using the hierarchical genotypic classifications (Serovar, BAPS1, ST, cgMLST), datasets from each of the three serovars showed varying degrees of clonal structuring. When the accessory genome (PANGEA) was mapped onto these hierarchical structures, accessory loci could be linked with specific genotypes. A large heavy-metal resistance mobile element was found in the Monophasic ST34 lineage of S. Typhimurium, and laboratory testing showed that Monophasic isolates have on average a higher degree of copper resistance than the Biphasic ones. In S. Newport, an extra sugE gene copy was found among most isolates of the ST45 lineage, and laboratory testing of multiple isolates confirmed that isolates of S. Newport ST45 were on average less sensitive to the disinfectant cetylpyridimium chloride than non-ST45 isolates. Lastly, data-mining of the accessory genomic content of S. Infantis revealed two cryptic Ecotypes with distinct accessory genomic content and distinct ecological patterns. Poultry appears to be the major reservoir for Ecotype 1, and temporal analysis further suggested a recent ecological succession, with Ecotype 2 apparently being displaced by Ecotype 1. Altogether, the use of a heuristic hierarchical-based population structure analysis that includes bacterial pan-genomes (core and accessory genomes) can (1) improve genomic resolution for mapping populations and accessing epidemiological patterns; and (2) define lineage-specific informative loci that may be associated with survival in the food chain.