The pros and cons of tracking the microbial contamination in infant food processing chain through amplicon sequencing and metagenomics

The targeted sequencing of the 16S rRNA gene’s amplicon has been extensively exploited in the last decades to characterise composition and structure of microbiotas in a vast number of habitats, ranging from the human body to the most extreme ecological niches on Earth. Now, more attention is being paid to metagenomics, since it adds pivotal information on microbiome functionality, although it requires higher analytical costs and more advanced computational approaches. The transition from observational studies to practical application of these two omics has begun in medical research, where specific profiles of patient’s microbiomes are nowadays linked to their health conditions and the probability to develop diseases. In food science possible applications need to be assessed in diverse food processing plants, in order to be fully understood and thus effectively applied by the industries in microbial control. To address this, we have profiled the microbiota and the microbiome of an infant food processing plant, for more than a year, through DNA-based amplicon sequencing and metagenomics. A total of twelve productions of infant cereal-based formula were followed and one hundred and fifty samples were collected from raw materials, environment, intermediate and final products. In parallel, the presence of microbial populations of interest has been verified by qPCR assays. Both amplicon sequencing and metagenomics highlighted high prevalence of Bacillus spp. in raw materials and environment, which was confirmed by targeted qPCR detection. The resident microbiome changed over time, with a rapid turnover of major taxa and a fluctuation of communities’ biodiversity mainly linked to the incoming of new raw materials, in the presence of a high level of airborne microbial contamination. Greater resolution of taxonomic assignments was achieved by metagenomics compared to the amplicon sequencing approach, but only in samples with high DNA quality and quantity. As to be expected, the hindering effect of non-bacterial DNA was less evident in amplicon sequencing data than with metagenomic data. Alternative alignment strategies were therefore tested to maximise the taxonomic and function assignments of metagenomic data and to detect potential signatures of targeted microbes. The outcomes underline the need to develop specific approaches for the analysis of the microbiomes in food processing plants.