Whole-Genome Sequencing Data for Serotyping Escherichia coli—It's Time for a Change!

Abstract

The accessibility of whole-genome sequencing (WGS) presents the opportunity for national reference laboratories to provide a state-of-the-art public health surveillance service. The replacement of traditional serology-based typing of Escherichia coli by WGS is supported by user-friendly, freely available data analysis Web tools. An article in this issue of the Journal of Clinical Microbiology (K. G. Joensen, A. M. M. Tetzschner, A. Iguchi, F. M. Aarestrup, and F. Scheutz, J Clin Microbiol, 53:2410–2426, 2015, http://dx.doi.org/10.1128/JCM.00008-15) describes SerotypeFinder, an essential guide to serotyping E. coli in the 21st century.

TEXT

In public health microbiology, whole-genome sequencing (WGS) is a valuable resource used to combat emerging infectious disease and investigate outbreaks caused by a wide range of pathogens, including Escherichia coli (1, 2). The unprecedented level of discrimination and the evolutionary context that WGS provides has had a major impact on our understanding of the underlying epidemiology and pathogenicity of bacteria that are a threat to public health (3–5). However, the prospect of implementing WGS for public health surveillance and routine microbial typing can be daunting. Radical changes to conventional thinking, organizational infrastructure, and employee skill sets are required. Additional challenges include maintaining a consistent, high-quality service and producing results that are compatible with historical data and data generated by other stakeholders in the field. With respect to these additional challenges, Katrine Joensen and colleagues' comprehensive and detailed article in this edition of the Journal of Clinical Microbiology reveals why replacing phenotypic serotyping of E. coli with in silico serotyping using WGS is an easy switch (6).

In the traditional phenotypic serotyping scheme for E. coli, antisera to a combination of immunogenic structures, including lipopolysaccharide (LPS) (O) and flagellar (H) antigens, are raised in rabbits. The scheme, first developed in the 1940s (7), has stood the test of time because it is applicable across the E. coli species, offers an appropriate level of discrimination for outbreak detection and investigation, and provides a user-friendly designation for taxonomic differentiation and pathogenic groups. However, establishing, maintaining, and developing the scheme (comprising more than 188 O and 53 H antisera) is expensive and laborious and requires specialist resources and expertise. Consequently, phenotypic serotyping is provided by a limited number of reference laboratories worldwide.

Colleagues at the Center for Genomic Epidemiology (CGE) at the Danish Technical University (DTU), Lyngby, Denmark (http://www.genomicepidemiology.org), have developed a Web tool, SerotypeFinder, for WGS serotype prediction of E. coli based on the O- and H-antigen-processing genes. It is a freely available web tool that requires little or no expertise in bioinformatics to operate. The database was constructed using complete O and H antigen genes from the NCBI nucleotide collection, a comprehensive set of O antigen genes reported by Iguchi et al. (8), and WGS of E. coli reference strains from the WHO Collaborating Centre for E. coli at the Statens Serum Institute, Copenhagen, Denmark. All database genes are compared against the genome of each test strain using the BLAST algorithm, and the output is the predicted O and H serotypes, based on the best-matching genes.

SerotypeFinder comprises all valid O types, O1 to O187 (except O14 and O57), and tolerates a wide range of O antigen gene variants, with the wzx and wzy gene variants clustering according to their O type. Certain pairs of O-antigen-encoding regions have identical or highly similar (>99.8%) wzx and/or wzy genes and are, therefore, potentially difficult to distinguish using SerotypeFinder, but in practice, the validation data demonstrated that the majority of these difficult identifications could be resolved and were assigned correctly despite the high similarity. Furthermore, the majority of these pairs represent known phenotypic serological cross-reactions and are difficult to differentiate using traditional methods.

Phenotypic H typing can be even more labor-intensive and time-consuming than O typing, as strains of E. coli must first be induced to express flagella in vitro. This initial step often requires multiple passages through a Craigie tube, which can take several days or even weeks. SerotypeFinder covers all 53 known H antigen genes, including over 100 variants. Sequence similarity of >99% was observed only between H4 and H17 variants, and this relationship was described previously (9).

There are many advantages to adopting the in silico WGS serotyping approach for typing E. coli. Once implemented, WGS can be faster and more cost-effective than traditional methods. Studies show that increasing numbers of strains of E. coli are reported as “O group unidentifiable” due to antisera failing quality control procedures, unresolvable cross-reactions, or novel serogroups (10, 11). In silico WGS serotyping avoids the need for the resource-intensive antiserum production process and the inherent quality control issues. Using WGS, problematic phenotypic cross-reactions appear, for the most part, to be resolved, and identifying and establishing novel O groups is much less demanding and operationally complex than producing and verifying new rabbit antisera for the phenotypic scheme. Strains of E. coli that are phenotypically untypeable due to lack of expression of O antigens (designated “rough”) or H antigens (designated “nonmotile”) are fully typeable by WGS. WGS data offers valuable insights into the degree of variation in the O- and H-antigen-encoding genes within each O and H type and the significance of cross-reactions between O groups. Furthermore, this approach provides the opportunity to examine the effect of mobile genetic elements, such as prophage, plasmids, and genomic islands, on O-antigen expression and evolution.

Without doubt, the unprecedented level of discrimination provided by WGS will inspire the community to develop and improve typing nomenclature. However, in this early phase of WGS implementation, the option of using the traditional serotype designation will minimize disruption for reference laboratory service users and facilitate data exchange with colleagues in the field. The in silico serotyping of E. coli using WGS data approach will help keep all our stakeholders “on board” at this time of radical change. The detailed validation and evaluation of SerotypeFinder in the article by Joensen et al. highlight both the positive features and the likely pitfalls of the approach and deliver an essential guide to serotyping E. coli in the 21st century.

The views expressed in this Commentary do not necessarily reflect the views of the journal or of ASM.