UC Berkeley

A wide range of specialized (or “secondary”) metabolites are produced by bacteria in naturalecosystems with varying functions, including antibiotics, siderophores, signalling molecules, and antifungals. These specialized metabolites are produced using operonic sets of genes that work in concert, known as biosynthetic gene clusters. In this work, genome-resolved metagenomic approaches were applied to understand the distribution and ecology of biosynthetic genes and the bacteria that possess them. Bacterial genomes were assembled and binned from deeply sequenced metagenomes from a California grassland meadow soil and a permanently wet vernal pool soil, and clusters of biosynthetic genes were identified in each genome. Genomes were reconstructed for novel species that belong to rarely cultivated but ubiquitous soil phyla, including the Acidobacteria, Verrucomicrobia, and the candidate phylum Rokubacteria. Bacteria from the grassland meadow soil were shown to have unexpectedly large numbers of biosynthetic gene clusters. In particular, two novel lineages of Acidobacteria were identified that possessed an unusual genomic capacity for specialized metabolite biosynthesis - up to 15% of their genomes were predicted to be dedicated to the production of nonribosomal peptides and polyketides. Sampling a second study site of soils from a vernal pool, Another three species were obtained from one of these uncultivated lineages — the candidate genus Angelobacter ​ ​ — which also possessed a large genomic repertoire of diverse biosynthetic genes. By mining public soil metagenomes, additional high quality draft genomes from this candidate genus were also analyzed, confirming that species in this genus are widespread across soil environments. It was therefore established that Angelobacter spp. ​with a substantial capacity for specialized metabolite biosynthesis are widespread in soils with a range of moisture contents and vegetation types. Transcriptional activity of nonribosomal peptide synthetase and polyketide synthase genes of abundant organisms from the grassland soil was tracked over time using 120 metatranscriptomic samples from soil microcosms. For several bacterial species within the samples, unsupervised clustering of genes by co-expression across samples identified modules of biosynthetic genes that were tightly co-expressed with genes involved in transcriptional regulation, environmental sensing, and secretion. For some vernal pool samples where ​ Angelobacter ​ were the most abundant microbial community members, metatranscriptomics demonstrated clear transcriptional activity ​ in situ ​ . Transcription of many ​ Angelobacter ​ biosynthetic genes was detected, extending findings from the grassland soil microcosms. Genetic variation in soil bacteria and their biosynthetic genes was investigated using population genomics methods that leverage genetic variation within sequencing reads that map to genomes from metagenomes. Metagenomic methods to track genetic variation within populations in a spatial context were applied to study the most abundant bacterial species across the grassland meadow study site. Genetic variation specifically within biosynthetic genes was elevated, indicating that there can be substantial allelic diversity in the biosynthetic genes of an abundant species in a local soil ecosystem. For about half of the bacterial populations studied, strong genetic population structure associated with spatial scale was observed. Genomes and gene variants were more genetically similar if they were from the same meadow plot. Simultaneously, while genetic gradients were observed across the meadow, within sample genetic diversity was also found to be high. Genomic signatures of recombination and gene-specific selection were also identified, indicating that ongoing selection and recombination may shape genetic divergence of populations on local spatial scales in soils. While biosynthetic gene clusters can be outlined and annotated with confidence in microbial genomes, prediction of the function of the metabolites produced for novel gene clusters is often an unsolved problem. Colocalized transporter genes associated with biosynthetic gene clusters may help predict metabolite function, due to their intimate association with the metabolite(s) they are transporting. This hypothesis was tested and benchmarked on a dataset of characterized biosynthetic gene clusters. In particular, a strong specificity of transporter genes for siderophore export and re-uptake was quantified as a signal of siderophore production. Using this specific genomic signal, putative siderophore BGCs were annotated across bacterial genomes recovered from soil, as well as from better characterized microbes from the adult and premature infant microbiomes. Surprisingly few genomes from soil bacteria contained transporter genes associated with siderophore biosynthesis. While 23% of microbial genomes from premature infant microbiomes possess at least one siderophore-like biosynthetic gene cluster, only 3% of those from adult gut microbiomes do. In sum, this thesis presented a metagenomic perspective on specialized metabolisms, contributed to discovery of novel species, examined evolutionary processes, and improved genomic functional predictions. The strength of this approach lies in its ability to investigate microbes in ​ in situ community contexts and detect ecological trends among the uncultivated microbial majority.