Preliminary work on the integration of selective enrichment and Nanopore sequencing for simultaneous Salmonella detection and subtyping

To detect Salmonella in food, animal feed, and environmental samples, methods based on the enrichment of Salmonella spp. in selective media have been standardized internationally. While these methods are highly reliable and are the basis of public health surveillance, detailed Salmonella monitoring requires laborious downstream procedures to identify target serotypes. Further, classical serotyping procedures are limited in their specificity and do not provide ancillary genomic information for monitoring. Fortunately, direct long-read sequencing and bioinformatics analysis of Salmonella enrichment cultures can overcome these issues. Still, deciding on a sampling, sequencing, and analysis approach is affected by several biological and technical factors that are poorly understood. For instance, we know little about how the dynamics of different Salmonella subtypes in the enrichment cultures change depending on interactions with co-enriched microbes, which potentially involve both inter- and intra-species competition. It is also unclear how the specificity of existing bioinformatics methods depends on the number of reads sequenced (i.e., coverage). Here, we first conduct Salmonella-spike experiments that assess the impact of microbial interactions. In doing so, we provide insight into the effects of different enrichment media and steps. We provide, for example, preliminary data showing how non-selective preenrichment steps potentially amplify the effects of inter-species competition. Subsequently, we show how Salmonella spp. are maximally enriched in selective media before 24 hours. More surprisingly, we find that intraspecies competition between closely related subtypes, which is influenced by the sample matrix, can lead to a 40-fold difference in relative enrichment. Then, we assess existing bioinformatics methods for subtyping and argue for methods grounded in population genomics that classify single reads. Towards the development of such methods, we use a recent model to show that Salmonella strains from the same serotypes often share a large fraction of the genome that is the result of a clonal descent. Lastly, we provide some proof-of-concept for leveraging this observation by building a database of genomes clustered by their clonal fraction and using it to attempt to classify individual sequencing reads. The work here ultimately highlights the importance of competition dynamics during selective Salmonella enrichment. It also makes recommendations for choosing samples and developing an integrated approach for monitoring Salmonella serotypes with the use of standardized selective enrichment cultures and long-read sequencing.

• VetCore Infrastructure