Abstract
Understanding the contribution of different diarrhoeagenic Escherichia coli pathotypes to disease burden is critical to mapping risk and informing vaccine development. Targeting select virulence genes by PCR is the diagnostic approach of choice in high-burden, least-resourced African settings. We compared the performance of a commonly-used multiplex protocol to whole genome sequencing (WGS). PCR was applied to 3,815 E. coli isolates from 120 children with diarrhoea and 357 healthy controls. Three or more isolates per specimen were also Illumina-sequenced. Following quality assurance, ARIBA and Virulencefinder database were used to identify virulence targets. Root cause analysis of deviant PCR results was performed by examining target sensitivity using BLAST, Sanger sequencing false-positive amplicons, and identifying lineages prone to false-positivity using in-silico multilocus sequence typing and a Single Nucleotide Polymorphism phylogeny constructed using IQTree. The sensitivity and positive predictive value of PCR compared to WGS ranged from 0-77.8% while specificity ranged from 74.5-94.7% for different pathotypes. WGS identified more enteroaggregative E. coli (EAEC), fewer enterotoxigenic E. coli (ETEC) and none of the Shiga toxin-producing E. coli detected by PCR, painting a considerably different epidemiological picture. Use of the CVD432 target resulted in EAEC under-detection, and enteropathogenic E. coli eae primers mismatched more recently described intimin alleles common in our setting. False positive ETEC were over-represented among West Africa-predominant ST8746 complex strains. PCR precision varies with pathogen genome so primers optimized for use in one part of the world may have noticeably lower sensitivity and specificity in settings where different pathogen lineages predominate.