Co-enriching microflora associated with culture based methods to detect Salmonella from tomato phyllosphere

Abstract

The ability to detect a specific organism from a complex environment is vitally important to many fields of public health, including food safety. For example, tomatoes have been implicated numerous times as vehicles of foodborne outbreaks due to strains of Salmonella but few studies have ever recovered Salmonella from a tomato phyllosphere environment. Precision of culturing techniques that target agents associated with outbreaks depend on numerous factors. One important factor to better understand is which species co-enrich during enrichment procedures and how microbial dynamics may impede or enhance detection of target pathogens. We used a shotgun sequence approach to describe taxa associated with samples pre-enrichment and throughout the enrichment steps of the Bacteriological Analytical Manual's (BAM) protocol for detection of Salmonella from environmental tomato samples. Recent work has shown that during efforts to enrich Salmonella (Proteobacteria) from tomato field samples, Firmicute genera are also co-enriched and at least one co-enriching Firmicute genus (Paenibacillus sp.) can inhibit and even kills strains of Salmonella. Here we provide a baseline description of microflora that co-culture during detection efforts and the utility of a bioinformatic approach to detect specific taxa from metagenomic sequence data. We observed that uncultured samples clustered together with distinct taxonomic profiles relative to the three cultured treatments (Universal Pre-enrichment broth (UPB), Tetrathionate (TT), and Rappaport-Vassiliadis (RV)). There was little consistency among samples exposed to the same culturing medias, suggesting significant microbial differences in starting matrices or stochasticity associated with enrichment processes. Interestingly, Paenibacillus sp. (Salmonella inhibitor) was significantly enriched from uncultured to cultured (UPB) samples. Also of interest was the sequence based identification of a number of sequences as Salmonella despite indication by all media, that samples were culture negative for Salmonella. Our results substantiate the nascent utility of metagenomic methods to improve both biological and bioinformatic pathogen detection methods.

Figure 1. Rarefaction plots illustrating the increase in taxa identified as a function of different sequencing depths for a) flashed and b) Meta-Velvetg reads.

Figure 2. Principal Coordinates Analysis (PCoA) depicting the taxonomic (a and b) and functional (c and d) differences among the replicates and treatments.

Figure 3. Taxonomic description of Phyla in uncultured and cultured samples using Flash and Meta Velvetg for assembly and the lowest common ancestor taxonomic rank of best hits (maximum e-value cutoff of 1.0⁻⁵, minimum percent identity of 95%, and minimum alignment length 99 bp).

Figure 4. Taxonomic classification of the nine most prevalent Proteobacteria genera in uncultured and cultured samples based on FLASHed and Meta Velvetg assembly and the lowest common ancestor taxonomic rank among the best hits (maximum e-value cutoff of 1.0⁻⁵, minimum percent identity of 95%, and minimum alignment length 99 bp).

Figure 5. Taxonomic classification of Firmicute genera associated with metagenomic and shotgun sequenced enrichments using FLASHed and Meta Velvetg assembly and the lowest common ancestor taxonomic rank among the best hits (maximum e-value cutoff of 1.0⁻⁵, minimum percent identity of 95%, and minimum alignment length 99 bp).

Figure 6. Boxplots of the relative abundance of a) Salmonella, b) Paenibacillus c) Proteobacteria, and d) Firmicutes among the different treatments using the FLASHed and Meta Velvetg assemblies.

Boxes show the interquartile range, bars illustrate the median, and the whiskers extend out to 1.5 times the interquartile range.

Figure 7. Results of the IMG pipeline assigning reads to either only Salmonella (Salmonella Only, orange), both Salmonella and the other database but with greater confidence to the former (Salmonella ↑ + IMG, white), both databases with equal confidence (both, black), or the other database only (IMG Only, grey) for a) flashed and b) Meta-Velvetg reads.

Figure 8. Plots of the average number of base-pairs (in millions) observed and estimates of quantity necessary to achieve approximately 1X coverage across all genomes present in cultured and uncultured samples.