Culture dependent and independent analysis of bacterial communities associated with commercial salad leaf vegetables

Abstract

Background: Plants harbor a diverse bacterial community, both as epiphytes on the plant surface and as endophytes within plant tissue. While some plant-associated bacteria act as plant pathogens or promote plant growth, others may be human pathogens. The aim of the current study was to determine the bacterial community composition of organic and conventionally grown leafy salad vegetables at the point of consumption using both culture-dependent and culture-independent methods.

Results: Total culturable bacteria on salad vegetables ranged from 8.0 × 10(3) to 5.5 × 10(8) CFU g(-1). The number of culturable endophytic bacteria from surface sterilized plants was significantly lower, ranging from 2.2 × 10(3) to 5.8 × 10(5) CFU g(-1). Cultured isolates belonged to six major bacterial phyla, and included representatives of Pseudomonas, Pantoea, Chryseobacterium, and Flavobacterium. Eleven different phyla and subphyla were identified by culture-independent pyrosequencing, with Gammaproteobacteria, Betaproteobacteria, and Bacteroidetes being the most dominant lineages. Other bacterial lineages identified (e.g. Firmicutes, Alphaproteobacteria, Acidobacteria, and Actinobacteria) typically represented less than 1% of sequences obtained. At the genus level, sequences classified as Pseudomonas were identified in all samples and this was often the most prevalent genus. Ralstonia sequences made up a greater portion of the community in surface sterilized than non-surface sterilized samples, indicating that it was largely endophytic, while Acinetobacter sequences appeared to be primarily associated with the leaf surface. Analysis of molecular variance indicated there were no significant differences in bacterial community composition between organic versus conventionally grown, or surface-sterilized versus non-sterilized leaf vegetables. While culture-independent pyrosequencing identified significantly more bacterial taxa, the dominant taxa from pyrosequence data were also detected by traditional culture-dependent methods.

Conclusions: The use of pyrosequencing allowed for the identification of low abundance bacteria in leaf salad vegetables not detected by culture-dependent methods. The presence of a range of bacterial populations as endophytes presents an interesting phenomenon as these microorganisms cannot be removed by washing and are thus ingested during salad consumption.

Figure 1

Viable counts of culturable bacteria obtained from leafy salad vegetables. Samples were plated on TSA (A) and R2A (B) media and are baby spinach, romaine lettuce, red leaf lettuce, iceberg lettuce, and green leaf lettuce of conventionally (C) and organically (O) grown varieties. Subsamples of each type were also subjected to surface sterilization (s) prior to processing. Counts represent means (+/− SE) of three analytical replicate plates per sample.

Figure 2

Relative abundance of bacterial phyla associated with leafy salad vegetables as determined from pyrosequencing. Samples are organically (Org) and conventionally grown baby spinach (Spi), romaine lettuce (Rom), red leaf lettuce (Red), iceberg lettuce (Ice), and green leaf lettuce (Gre) and include intact and surface sterilized (S) subsamples. Percentages represent the portion of 16S rRNA gene 454 reads (mean 2,515 per sample) that were classified to that phylum (or subphylum in the case of Proteobacteria).

Figure 3

Similarities of bacterial communities associated with leafy salad vegetables as derived from pyrosequencing. Samples are organically (Org) and conventionally grown baby spinach (Spi), romaine lettuce (Rom), red leaf lettuce (Red), iceberg lettuce (Ice), and green leaf lettuce (Gre) and include intact and surface sterilized (S) subsamples. Community similarity is determined from Jaccard similarity scores followed by nonmetric multidimensional scaling (A) or UPGMA dendrogram construction (B). Analyses are run on subsamples of 1507 sequences from each sample, and show the mean outcome of 1000 individual subsampling runs.