The application of second generation sequencing technology to the characterization of complex microbial communities has profoundly affected our appreciation of microbial diversity. The explosive growth of microbial sequence data has also necessitated advances in bioinformatic methods for profiling microbial communities. Data aggregation strategies should allow the relation of metagenomic sequence data to our understanding of microbial taxonomy, while also facilitating the discovery of novel taxa. For eukaryotes, a method has been established that links DNA sequences to the identification of organisms: DNA Barcoding. A similar approach has been developed for prokaryotes using target genic regions as markers for species identification and to profile communities. A key difference in these efforts is that within DNA barcoding there is a formalized framework for the evaluation of barcoding targets, whereas for prokaryotes the 16S rRNA gene target has become the de facto barcode without formal evaluation. Using the framework developed for evaluating DNA barcodes in eukaryotes, a study was undertaken to formally evaluate 16S rRNA and cpn60 as DNA barcodes for Bacteria. Both 16S rRNA and cpn60 were found to meet the criteria for DNA barcodes, with cpn60 a preferred barcode based on its superior resolution of closely related taxa. The high resolution of cpn60 enabled a method of sequence data aggregation through sequence assembly: microbial profiling using metagenomic assembly (mPUMA). The scoring of metagenomic assemblies in terms of sensitivity and specificity of the operational taxonomic units formed was used to evaluate and optimize the assembly of cpn60 barcodes. Using optimized parameters, mPUMA was demonstrated to faithfully reconstruct a synthetic community in terms of richness and abundance. To facilitate the use of mPUMA, a software package was developed and released under an open source license. The utility of mPUMA was further examined through the characterization of the epiphytic seed microbiomes of Triticum and Brassica species. A microbiome shared across both crop genera including fungi and bacteria was detected: a particularly important observation as it implies that seeds may serve as a vector for microbes that could include both pathogenic and beneficial organisms. The relative abundances of taxa identified by mPUMA were confirmed by qPCR for multiple cases of both fungal and bacterial taxa. By culturing isolates of both bacteria and fungi from the seed surfaces it was demonstrated that mPUMA faithfully assembled consensus sequences for OTUs that were 100% identical to isolated fungi and bacteria. Patterns observed in the relative abundances of the shared microbiome OTUs were used to generate the hypothesis that an Pantoea-like bacterium and an Alternaria-like fungus had an antagonistic relationship, since sequences corresponding to these organisms showed reciprocal abundance patterns on Triticum and Brassica seeds. Studies of the interactions of cultured isolates revealed fungistatic interactions that could account for their reciprocal abundances. These interactions could be directly relevant to plant health, given that Alternaria-like fungi are linked to grain spoilage in wheat, and diseases in canola. Taken together, results of this thesis demonstrate the superiority of the cpn60 universal target as a barcode for Bacteria, forming the basis for an assembly-based strategy for microbial profiling of bacterial and eukaryotic microbial communities that can lead to the discovery of novel taxa and microbial interactions.

Complex microbiomes are part of the food we eat: they are naturally present on the raw material, they merge along the food system, or they can be intentionally inoculated. Whether their presence is desired, such as in case of fermentation or probiotic supplementation, or undesired, in case of pathogenic or spoilage microbes, depends on who they are and what they are doing and therefore several studies investigated the microbiota of specific foods. However, the diversity of food microbiomes remains largely unexplored and similar studies present inconsistencies in methods and results. The study of the food microbiome is relevant also in light of the human microbiome and its multifaceted connection to hosts’ health status. Diet is one of the main factors influencing the human microbiome and many studies investigated how nutrition impacts the endogenous microbial communities both in the gut and in the oral cavity. Nevertheless, they largely overlooked the possibility of direct contribution of food-origin microorganisms. The primary aim of my PhD was the comprehensive characterization of foodborne microbial communities with the ultimate goal of estimating their impact on the human microbiome. This research intended to be humble contribution to the global effort in understanding the microbial sources building these composite ecosystems inhabiting the human body. In order to explore the food microbiome diversity, I selected and collected 583 publicly available food (shotgun) metagenomes and integrated them with 1950 newly sequenced food metagenomes. Through an assembly-based pipeline, I reconstructed >10,000 metagenome-assembled genomes (MAGs) that resulted in 290 previously undescribed taxa and, hence, firstly observed in this work. I characterized the composition of microbial communities in food, proving strong specificity across food categories and types through statistical analysis and machine learning approaches. The uniformly and coherently processed curated metadata, taxonomic profiles and reconstructed genomes are publicly available in a resource called curatedFoodMetagenomicData (cFMD). To investigate the presence of food-associated bacteria among human oral and gut microbiomes, I analyzed 20,000 human metagenomes available in curatedMetagenomicData (cMD) through the same expanded pipeline used for food samples. The overlap between food and human microbiomes showed high variations according to host characteristics and the food prevalent species accounted on average for 3% of relative abundance in adult microbiomes. I recognized 43 bacterial species prevalent in both environments that were investigated at the strain level, showing close genomic similarities of strains found both in food and humans.To our knowledge this was the first attempt to investigate the global food microbiome and to estimate its involvement in human microbiome at a large-scale. Our results showedan expansion of known and yet-to-be-isolated species associated with food microbiomes, their characterization to uncover microbial diversity and provide insights on links with the human microbiome, and the release of a publicly-available resource as cFMD that will support the use of metagenomics in food microbiology and food safety, certificationand quality control applications.