Infectious Diseases and Whole-Genome Sequencing: The “One Day in Denmark” project

Infectious diseases are responsible for approximately 15% of all deaths worldwide. From the more than one thousand organisms able to cause human disease identified so far, 38% correspond to bacteria. Although bacterial species are intrinsically resistant to particular antimicrobial agents, they are able to acquire and transmit new mechanisms of resistance to drugs that have, in the past, been effective for treatment. This acquisition and dissemination of new antimicrobial resistance (AMR) mechanisms is currently one the of most concerning public health threats, not only due to the depletion of effective antimicrobial agents but also due to the more severe outcomes of infections caused by resistant organisms, when compared with their susceptible counterparts.

In clinical microbiology laboratories different diagnostics methods are used to characterize these organisms. Species identification, antimicrobial susceptibility testing and typing are frequently performed, either through phenotypic methods or molecular methods. These methods vary in cost, turnover time, accuracy and discriminatory power.

Whole-genome sequencing (WGS) technologies are a promising diagnostics tool due to the large amount of data produced by one single laboratory protocol, which can be analysed to completely characterize bacterial isolates, instead of the multiple routine protocols currently in place. The increase of sequencing accuracy, decrease in cost and improvement of bioinformatics tools and databases make the application of these technologies in clinical settings a possibility, but there are still significant gaps in research impeding their implementation. The most important shortcoming of current literature is that the majority of benchmarking and comparative studies only focus on pre-selected bacterial taxa, thus not truly reflecting the variety of organisms processed in clinical settings.

The main aim of this PhD project and thesis was to resolve that important gap in scientific knowledge through the comparison of routine diagnostics results with WGS-based analyses in clinical isolates collected from all Danish Clinical Microbiology Laboratories (DCM), without any pre-selection of species.

Manuscript I describes the point-prevalence analysis of all bacteria processed in all the DCM during one day and evaluates the ability of WGS to perform species identification, using two bioinformatics tools (KmerFinder and rMLST). The WGS datasets were compared with results from routine analysis provided by the DCM, obtained through matrix-assisted laser desorption/ionization – time-of-flight mass spectrometry (MALDI-TOF MS). We have found that the most prevalent bacterial species analysed in the DCM were Escherichia coli and Staphylococcus aureus, recovered from urine samples and skin or soft tissues samples, respectively. WGS-based species identification and routine MALDI-TOF MS results were completely in agreement for 95.7% of all cases. Results of the two bioinformatics tools and those obtained through MALDI-TOF MS were furthermore concordant at genus level for 99.7% of all cases. We also performed a comparison of point-prevalence and the research focus attributed to those species in scientific literature, concluding (with several limitations, as stated in the discussion) that for certain species there is a discrepancy between their prevalence and the research focus they receive.

In Manuscript II we provide a description of phenotypic antimicrobial susceptibility profiles for a random selection of 500 isolates, as determined by the standard method of broth microdilution. Those were then compared with WGS-based prediction of AMR, determined with ResFinder. AMR was relatively low in Denmark, and more prevalent in Gram-negative bacteria (10.9%) than in Gram-positive bacteria (6.1%). In silico antibiograms were in agreement with standard phenotypic results in 91.7% of all cases, with most discordances being observed for macrolides and tetracycline in streptococci, ciprofloxacin and β-lactams in combination with β-lactamase inhibitors in Enterobacterales, and most antimicrobials in Pseudomonas aeruginosa.

Manuscript III presents in silico antibiograms of all E. coli isolates collected in the same context as Manuscript I, as well as the respective typing results and clustering analysis performed with MLST, cgMLST and CSIPhylogeny bioinformatics tools. We detected AMR genes or chromosomal point mutations in 56.2% of the isolates, with the most prevalent being β-lactams resistance determinants (40.1%). The percentages of detected AMR determinants were similar to the percentages of phenotypically resistant isolates usually observed in the country, with the exception of ciprofloxacin. MLSTs were assigned to 99% of the isolates and these were distributed through 182 sequence types. cgMLST analysis assigned alleles to a high percentage of the loci present in the E. coli cgMLST scheme for 99.6% of the isolates. With a conservative distance of 15 alleles as the threshold to classify isolates as being epidemiologically related, we were able to identify 23 clusters, containing from two to six isolates each. Results of within-cluster single-nucleotide polymorphisms (SNP)-based analysis corroborated the clustering achieved through the cgMLST approach, except for two clusters which revealed 111 and 461 SNPs.

This thesis has expanded the scientific knowledge regarding the applicability of WGS technologies as clinical microbiology diagnostics methods. We have shown the adequate performance of these technologies as a bacterial species identification method, and furthermore clearly identified the current shortcomings of WGS for antimicrobial susceptibility testing, directing future research efforts. We demonstrated the potential of WGS-based typing and clustering analysis for clinical isolates, with advantages for surveillance systems and outbreak detection. Together, all these findings support the possibility of implementing WGS in clinical settings. Further targeted research and harmonisation of protocols will allow these technologies to be confidently used in public health.