Metagenomic identification of disease-causing Salmonella enterica serovars and antimicrobial resistance genes from paediatric faecal samples

TL;DR

This study explores using metagenomics to detect Salmonella and antibiotic resistance genes in children's fecal samples, showing promise but highlighting challenges in accurate identification.

Contribution

The study evaluates the effectiveness of metagenomic methods for identifying Salmonella serovars and AMR genes in pediatric cases compared to controls.

Findings

Salmonellosis cases had significantly higher Enterobacteriaceae abundance than controls.

Centrifuge was the most sensitive method for identifying Salmonella enterica sequences in case metagenomes.

No significant difference in AMR gene counts or classes was found between cases and controls.

Abstract

Background. Nontyphoidal Salmonella (NTS) is a common cause of enterocolitis and a major cause of death in children in low- and middle-income countries (LMICs). High antimicrobial resistance (AMR) prevalence in LMICs reduces treatment options for individuals at risk of severe infections.

Methods. We investigated the use of metagenomics to identify NTS and associated AMR genes in 28 faecal metagenomes from children with culture-confirmed salmonellosis in Vietnam, using accompanying NTS genomes from isolated serovars (one per metagenome). Read-based and assembly-based methods were utilised for NTS and AMR detection. Case metagenomes were compared to healthy control metagenomes (n=21) with respect to the microbiome composition, NTS relative abundances, number of unique AMR genes and antimicrobial classes to which the genes confer resistance, including classes used in Salmonella treatment.

Results. Salmonellosis cases displayed significantly higher relative abundances of Enterobacteriaceae than controls. Bracken and Centrifuge analysis facilitated the identification of Salmonella enterica sequences in case metagenomes at varying relative abundances (0.00259–27.7 % of total reads), which were significantly higher than controls. MetaPhlAn4 did not detect S. enterica in any control metagenomes, though 12 case metagenomes were also negative. The isolated serovars were identified in 78.6% of the associated case metagenomes with Centrifuge, suggesting this method is the most sensitive; however, the isolated genome serovar was the most abundant in only six case metagenomes, and serovar sequences were also identified in control metagenomes. Alignment to a Salmonella reference database, followed by local assembly and realignment, predicted the isolated serovar as the most likely serovar present in 35.7% of metagenomes, whereas Salmonella in silico typing resource classification of the local assembly was concordant with the isolate genome in 28.6% of cases. Metagenome-assembled genomes produced using two tools following de novo assembly identified the isolated serovar in 17.8–21.4% of cases. The percentage of NTS AMR genes identified in each case metagenome ranged between 0.00 and 100%. There was no significant difference in the number of unique AMR genes or antimicrobial classes between cases and controls, indicating comparable resistomes between cohorts.

Conclusions. This study highlights the potential of metagenomics for NTS identification in faecal samples, although overlap in S. enterica relative abundance between cohorts calls for further work to identify a diagnostic cutoff. Reliable characterisation of the organism to the serovar and AMR genotype level is affected by the complexity of the microbiome, sequencing and analysis approaches. Increased sequencing depth, for example through improved host DNA depletion, may facilitate enhanced characterisation. Detection of multiple serovars within individual samples with the Centrifuge suggests inaccurate classification or the presence of multiple serovars, making characterisation difficult.

Figures