Salmonella serotype determination utilizing high-throughput genome sequencing data

Abstract

Serotyping forms the basis of national and international surveillance networks for Salmonella, one of the most prevalent foodborne pathogens worldwide (1-3). Public health microbiology is currently being transformed by whole-genome sequencing (WGS), which opens the door to serotype determination using WGS data. SeqSero (www.denglab.info/SeqSero) is a novel Web-based tool for determining Salmonella serotypes using high-throughput genome sequencing data. SeqSero is based on curated databases of Salmonella serotype determinants (rfb gene cluster, fliC and fljB alleles) and is predicted to determine serotype rapidly and accurately for nearly the full spectrum of Salmonella serotypes (more than 2,300 serotypes), from both raw sequencing reads and genome assemblies. The performance of SeqSero was evaluated by testing (i) raw reads from genomes of 308 Salmonella isolates of known serotype; (ii) raw reads from genomes of 3,306 Salmonella isolates sequenced and made publicly available by GenomeTrakr, a U.S. national monitoring network operated by the Food and Drug Administration; and (iii) 354 other publicly available draft or complete Salmonella genomes. We also demonstrated Salmonella serotype determination from raw sequencing reads of fecal metagenomes from mice orally infected with this pathogen. SeqSero can help to maintain the well-established utility of Salmonella serotyping when integrated into a platform of WGS-based pathogen subtyping and characterization.

FIG 1

An example workflow of fliC H antigen prediction. A detailed description can be found in Materials and Methods. fliC_eh>fliC_ir>fliC_z35^a, predefined antigen clusters are summarized in Table S5 in the supplemental material.

FIG 2

Major components and workflows of SeqSero. Two workflows are represented, including serotype determination from (i) genome assembly and (ii) raw sequencing reads.

FIG 3

Predicted incorrect H antigen identification using reads mapping with 95% confidence limits. (A) Prediction for fliC identification. (B) Prediction for fljB identification. Logistic regression was used to estimate the probability of making an incorrect identification given the size of the mapped reads difference scaled by total number of reads sequenced from a genome. The GenomeTrakr data set selected for SeqSero validation was used for this analysis. Observed correct and incorrect antigens calls were based on the first round of reads mapping.

FIG 4

Phylogenetic relationship among detected Salmonella enterica serotype Typhimurium strains from fecal metagenomes of mice. A maximum likelihood tree shows the phylogenetic distance among the Salmonella strains serotyped from stool metagenomes of mice before and after oral infection. Raw reads from each metagenome were mapped to the genome (GenBank accession number CP001363) of the infection strain (str. 14028s), high-quality single nucleotide polymorphisms (SNPs) were identified, and a core genome SNP maximum likelihood tree was built using methods similar to those previously described in reference .