Dissecting disease-suppressive rhizosphere microbiomes using metagenomics

Plants and microbes have coexisted for hundreds of millions of years and have developed deeply intertwined mutually beneficial relations. Among the many benefits of a selected microbial community, pathogen suppression is a particularly desirable trait for both the plant and agronomy industry. Suppressive soils have been described for many years but technology to develop a deep understanding of this phenomenon was only recently introduced. In this thesis, I applied different metagenomic sequencing approaches to study the biological basis of suppressive soils with particular interest toward the biosynthetic potential of the suppressive rhizosphere community. In the introduction, I describe ecology-inspired solutions for biosynthetic gene cluster (BGC) mining and the existing tools and sequencing technologies that can be used to this end. We take a structured approach to the dissection of the suppressive-associated rhizosphere communities. In the first part of the work, we perform the first large-scale soil survey aiming at establishing a soil collection to identify unique characteristics of suppressive soils. Through a combination of phenotyping and marker gene sequencing, we identify four soils with strong Fusarium culmorum-suppressive characteristics. We compare taxonomy composition and volatile profile of both suppressive and non-suppressive soils to identify features that distinguish suppressive soils. The suppressive soils found in this collection do not share physiochemical or categorical characteristics. In addition, diversity and community structure do not separate suppressive and non-suppressive soils. However, network-based analysis shows a group of acidobacteria which co-occur in a suppressive-soil-associated hub. Then, to better understand the secondary metabolite diversity of the suppressive samples we characterize of nonribosomal peptide synthetase diversity in suppressive and conducive soils with the use of functional amplicon sequencing of NRPS adenylation domains. To this end, we developed dom2BGC, a pipeline for the annotation of domains associated with BGCs. We also perform cooccurrence-based clustering of the sequenced domains to restore, through guilt by association, the physical clustering of the different domains annotated to the same (BGC). We identified multiple NRPSes with potentially antifungal activity that occur exclusively in suppressive soils. Furthermore, we sequenced one suppressive sample with 10X metagenomic sequencing, which was used to confirm the presence of dom2BGC reconstructed clusters.After extensive study of suppressive rhizosphere communities, we zoom in and perform a dilution to extinction experiment with a microbial extract from a suppressive soil which progressively loses its phenotype in accordance with the dilution. We evaluate the effect of dilution on the microbial composition and functional profile of the community. Genetic characteristics and taxonomic groups that correlate with the dilution and phenotype loss are investigated for links that can shed light on the key players of suppressive soils. We found multiple metagenome-assembled genomes rich in BGCs and chitin-degrading ability that closely correlate with the loss in pathogen suppression. This work then proceeds to describe the characteristics of a suppressive endosphere. We compare Rhizoctonia solani suppressive and conducive endosphere communities of sugar beet. Shotgun metagenome sequencing showed significant differences in taxonomic abundance and genes associated with Chitinophaga and Flavobacterium bacteria. Additionally, we compose a synthetic community from endosphere isolates that provides disease suppression against Rhizoctoria solani infection. Finally, disease suppression of the synthetic community is lost upon site-directed mutagenesis of a candidate suppressive NRPS in a flavobacterial isolate, providing a model to explain the phenotype.Finally, we detail the development of an assembly tool that aims to improve the assembly of complex BGCs. BGCs often contain repetitive domains that are hard to assemble, but are still very informative as they strongly influence the predicted natural product. Such repetitive domains are sometimes inadvertedly collapsed during the assembly graph formation, which inevitably leads to an erroneous or incomplete cluster. These problems are exacerbated in complex metagenomics assemblies. BiosyntheticSPAdes is designed to identify and isolate BGC-harboring neighborhoods in the assembly graph by finding multiple adjacent BGC-associated domains. Once identified, the BGC subgraph is extracted and collapsed domains are restored based on local coverage. Depending on the subgraph, multiple paths can be traversed to produce a BGC sequence. When multiple putative BGCs are produced, a ranking pipeline shows which candidate BGC is structurally similar to previously assembled BGCs based on sequence similarity and domain order. To conclude, in the discussion I offer some considerations on the effects of sequencing technologies on microbiology and microbial ecology, to then propose experimental and computational strategies that are best fit to identify microbial natural products from complex ecosystems