Whole-genome sequencing in the prediction of antimicrobial resistance

KEYWORDS:

• Next generation sequencing
• bioinformatics
• metagenome
• multidrug-resistant bacteria
• single-cell transcriptome

The discovery of antibiotics is one of the major achievements in biomedical research. Regrettably, the abuse of antibiotics has caused the emergence of multidrug-resistant pathogens. Since then, infections and mortality caused by multidrug-resistant pathogens have increased globally [1]. To date, antimicrobial resistance is a global threat in public health and clinical medicine. It is estimated that the deaths caused by antibiotic-resistant bacteria in the European Union and United States of America amount to 131 cases per day, a rate that will certainly pose a serious socioeconomic burden [1].

Antimicrobial resistance can be caused by genetic mutations that gives rise to, for example, (i) the production of a protein or enzyme that inactivates, modifies or degrades the antimicrobial agent; (ii) the presence of an alternative enzyme that is not inhibited by the antimicrobial agent; (iii) mutation-mediated target modification; (iv) posttranscriptional and/or posttranslational modification of the target which may or may not be caused by mutation [2]; (v) reduced uptake of the antimicrobial agent [3]; (vi) drug‐specific and multidrug efflux pumps [3]; and (vii) excessive production of the target.

It is also possible that bacteria acquire antimicrobial resistance via horizontally-acquired resistance, for example plasmids, transposons and integron systems mediating antimicrobial resistance [4]; It is also possible that antimicrobial resistance may be caused by hitherto unknown mechanisms [5].

Mutation is a common cause of antimicrobial resistance, or change of susceptibility to a specific drug but multiple mutations are often observed for strong resistance towards a drug candidate [6]. This issue is further complicated by the presence of a large number of genes which influence drug resistance and this may include many genes that are not directly involved in the antimicrobial resistance phenotype [7]. This complex link between antimicrobial resistance and genetic events remains enigmatic [6].

The accurate and rapid determination of antimicrobial resistance is crucial not only for the treatment of infections but also for minimizing the risk of antibiotic abuse. Antimicrobial resistance detection methods can be broadly divided into nucleic acid-based and phenotype-based. It is generally accepted that nucleic acid-based techniques are more accurate. These techniques allow the detection of mutations down to the single gene level. Consequently, nucleic acid-based techniques became widely used in diagnostic and research laboratories worldwide.

More recently, however, advances in next generation sequencing (NGS) created a breakthrough in the study of antimicrobial resistance. Depending on the type of information needed, NGS can include DNA and RNA (in the form of cDNA) as sequencing material. DNA NGS can reveal, for example, the presence of antimicrobial genes while RNA NGS (or RNA-sequencing) detects global gene expression, including the expression of antimicrobial resistance genes. Coupled with appropriate bioinformatics pipelines such as the ResFinder tool for the identification of acquired antimicrobial resistance genes [8], NGS offers the unprecedented advantage of providing genetic information at the whole genome level, thus making it ideal for uncovering all possible genetic determinants of antimicrobial resistance in a single microbial genome. Recent studies involving large NGS studies of pathogen genomes have identified at least 70 genes that might be involved in the drug resistance of Mycobacterium tuberculosis [9], which are likely to be missed by most routine molecular techniques.

Of note, high throughout NGS instruments such as the HiSeq series from Illumina Inc. (CA, USA) are capable of sequencing hundreds of whole bacterial genomes simultaneously. This has accelerated the generation of sequence data and enhanced the study of phylogenomes of large numbers of clinical isolates to enable the discovery of novel antimicrobial resistance genes [8,10–12]. Much work has been done on bacterial pathogens such as Escherichia coli, Klebsiella pneumoniae, Salmonella spp. and Neisseria meningitidis [13]. Furthermore, NGS has been used to study the rate of emergence of antibiotic resistance in M. tuberculosis [12] as well as to detect mega plasmids and multiple plasmids using bioinformatics tools such as the PlasmidFinder [14–17]. Another application of NGS is metagenome analysis, either de novo or with the use of 16S rRNA gene sequencing [11,18]. Samples such as blood, sputum and stool, and even waste water [18] can be subjected to metagenomic analysis that allows the study of all members in a bacterial community. This ‘landscape genome’ will allow the detection of pathogens which are typical in particular diseases or environments. To illustrate, in the author’s laboratory, samples from bronchoscopy were subjected to shot gun metagenome analysis leading to the simultaneous identification of pathogens and the generation of population genomics data that are valuable for the prediction of antimicrobial resistance.

Advancements in NGS instrumentation has enabled NGS to be performed using single molecules that generate long read lengths to yield complete genomes instead of draft genomes [19,20]. A complete genome will ease the downstream bioinformatics analysis for the accurate prediction of antimicrobial resistance genes, but, at present, the cost of obtaining a complete genome is still exorbitant [21–23]. As the single-cell NGS circumvents the need for pure microbial cultures, it can be applied to the study of unculturable microorganisms such as human pathogens surviving in the environment but are unable to grow on artificial culture media. In the clinical setting, this technique should benefit the examination of biopsy specimens as the study of the genome at the single-cell level will be independent of the biopsy sample size.

Improvement on sample preparation has shortened the turn-around time from culture-on-the plate to NGS data generation from weeks to within 4–5 days as compared to the previous NGS technology [24]. This speed is crucial because the prediction and diagnosis of antimicrobial resistance is time sensitive. The prediction of antimicrobial resistance involves genome data quality control checks, data trimming, genome assembly, annotation and antimicrobial resistance genes prediction. At present, all these procedures can be done via bioinformatic analysis within a very short time, depending on, among others, the computer processing power, quality of database and also the internet speed. Useful web tools such as CARD are available, that can provide automated prediction of antibiotic resistance genes [25].

Recently, it was reported that, in the case of Staphylococcus aureus and M. tuberculosis, the whole process of generating a clinician-friendly report on antimicrobial resistance from raw sequence reads takes merely 3 min on a laptop [26]. With decreasing cost and turn-around time, improved sample preparation work flow, and a small footprint format, the next generation sequencing platform, coupled with clinician-friendly bioinformatics tools, NGS is rapidly becoming the technique of choice for the investigation and prediction of antimicrobial resistance at an unprecedented scale [27,28]. More importantly, as compared to other molecular techniques, NGS provides information that goes beyond merely detecting resistance mechanisms; it also allows phylogenomic studies for genotyping which is important for clinical epidemiology and public health, tracking onward transmission of pathogens, and plasmidome analysis.

To conclude, in the near future, NGS will prove its worth as the platform technology of choice for the detection and characterization of antimicrobial resistance.

Declaration of interest

This work received financial funding from the University of Malaya- High Impact Research Grants, (UM-MOHE HIR Grant UM.C/625/1/HIR/MOHE/CHAN/14/1, no. H-50001-A000027; UM-MOHE HIR Grant UM.C/625/1/HIR/MOHE/CHAN/01, no. A000001-50001) which are awarded to K.G Chan. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. The author thanks Senior Professor Dr Yun Fong Ngeow from Universiti Tunku Abdul Rahman (Faculty of Medicine and Health Sciences) for proofreading the manuscript.