Transforming clinical microbiology with bacterial genome sequencing

Abstract

Whole-genome sequencing of bacteria has recently emerged as a cost-effective and convenient approach for addressing many microbiological questions. Here, we review the current status of clinical microbiology and how it has already begun to be transformed by using next-generation sequencing. We focus on three essential tasks: identifying the species of an isolate, testing its properties, such as resistance to antibiotics and virulence, and monitoring the emergence and spread of bacterial pathogens. We predict that the application of next-generation sequencing will soon be sufficiently fast, accurate and cheap to be used in routine clinical microbiology practice, where it could replace many complex current techniques with a single, more efficient workflow.

Figure 1. Principles of current processing of bacterial pathogens

Schematic representation of the current workflow for processing samples for bacterial pathogens is presented, with high complexity and a typical timescale of a few weeks to a few months. The schematic is an approximation that highlights the principle steps in the workflow; it is not intended to be a comprehensive or precise description. Samples that are likely to be normally sterile are often cultured on rich medium that will support the grown of any culturable organism. Those from samples contaminated with colonising flora present a challenge for growing the infecting pathogen. Many types of culture media (referred to as selective media) are used to favour the growth of the suspected pathogen; this approach is particularly important for culturing pathogens from faeces. Boxes A to H arbitrarily represents the many different media for culture. The medium H represents a medium designed for growing mycobacteria that have specific growth requirements. Once an organism is growing, the morphological appearance and density of growth are properties that need specialist knowledge for deciding whether it is likely to be pathogenic. The likely pathogens are then processed through a complex pathway that has many contingencies to determine species and antimicrobial susceptibility. Broadly, there are two approaches. One approach uses MALDI-TOF for species identification prior to setting up susceptibility testing. The other uses Gram staining followed by biochemical testing to determine species; susceptibility testing is often set up simultaneously with doing biochemical tests. Categorisation of pathogens into groups of species is needed to choose the appropriate susceptibility testing panel. Lastly, depending on the species and perceived likelihood of an outbreak, a small subset of isolates may be chosen for further investigation using a wide range of typing tests often only provided by reference laboratories. The dashed lines and question marks are positioned arbitrarily to indicate that the further investigation is varied and happens only in a small number of cases.

Figure 2. Hypothetical workflow based on whole genome sequencing

Schematic representation of the workflow anticipated after adoption of whole genome sequencing, with an expected timescale that could fit within a single day. The culture steps would be the same as currently used in a routine microbiology laboratory. Some types of sample might be directly sequenced (see ‘future directions’, not shown here). Once a sample or likely pathogen is ready for sequencing, DNA will be extracted. This procedure is becoming simpler, as the input required for successful sequencing is reducing; it is now possible to use as little as 5 ng and to purify this in <30 minutes. For current bench-top machines it can take as little as 2 hours to prepare the DNA for sequencing, and new platforms (Box 1) could enable sequencing without preparation. Therefore, bacterial genome sequencing in hours and possibly even minutes is a realistic prospect. After sequencing, the main processes for yielding information will be computational. The development of software and databases is a major challenge to overcome before a pathogen sequencing can be deployed in clinical microbiology. Automated sequence assembly algorithms will be necessary for processing the raw sequence data (Box 1). This assembled sequence would then be analysed by modular software to determine species, relationship to other isolates of the same species, antimicrobial resistance profile and virulence gene content. Results of this analysis will be reported through hospital information systems. All the results will also be used for outbreak detection and infectious diseases surveillance. These developments will require new large database and other informatics technology and will take time to develop. In particular, it will need ‘intelligent systems’ which will incorporate elements of machine learning to enable automatic updating of key knowledge bases for species identification, antimicrobial resistance determination and virulence detection. Formal evaluation of such a solution will also need robust testing to ensure it performs at least as well as current methods.