Genomics as a tool for natural product structure elucidation

Natural product discovery is in the midst of a transition from a largely serendipity-based effort to an informatics-driven one. For most of the 20th century, natural product discovery relied on genome-blind bioassay-guided isolation. This was initially exceptionally productive, yielding the golden age of antibiotics. The fact that a majority of all medicines—especially antibiotics—are in some way derived from or inspired by natural products is a testament to the importance of understanding and harnessing the chemical strategies for biological interaction that have evolved over millions of years. Unfortunately, the overwhelmingly frequent rediscovery rate of known compounds among screened natural extracts meant that what was initially a life-saving torrent of new drugs eventually dried up into a costly trickle. Unfortunately, this has coincided with the rise of drug-resistance superbugs as our initial stockpiles of antibiotics have become overdeployed. Fortunately, we are now poised to enact an antibiotic renaissance powered by the ease and affordability of large-scale genomic analysis. The ability to genome-gaze has not only revealed hundreds of thousands of yet-untapped secondary metabolites in sequenced organisms but also can facilitate strain prioritization, novelty determination (dereplication), structure elucidation, three principal bottlenecks in the discovery process, as reviewed in Chapter 1. We report here progress in the use of genomics to facilitate the discovery and contextualization of new chemical matter. In Chapter 2, we report the discovery, isolation, and structural elucidation of streptomonomicin (STM), an antibiotic lasso peptide from Streptomonospora alba, and report the genome for its producing organism. STM-resistant clones of Bacillus anthracis harbor mutations to walR, the gene encoding a response regulator for the only known widely-distributed and essential two-component signal transduction system in Firmicutes. Our results demonstrate that understudied microbes remain fruitful reservoirs for the rapid discovery of novel, bioactive natural product and also highlight the usefulness of genomics in combination with NMR and HR-MS/MS for determining the structure of ribosomal natural products. In Chapter 3, we use HR-MS/MS, reactivity-based screening, NMR, and bioinformatic analysis to identify Streptomyces varsoviensis as a novel producer of JBIR-100, a fumarate-containing hygrolide. Using a combination of NMR and bioinformatic analysis, we elucidated the stereochemistry of the natural product. We investigated the antimicrobial activity of JBIR-100, with preliminary insight into mode of action indicating that it perturbs the membrane of Bacillus subtilis. S. varsoviensis is known to produce compounds from multiple hygrolide sub-families, namely hygrobafilomycins (JBIR-100 and hygrobafilomycin) and bafilomycins (bafilomycin C1 and D). In light of this, we identified the biosynthetic gene cluster for JBIR-100, which, to our knowledge, represents the first reported for a hygrobafilomycin. Finally, we performed a bioinformatic analysis of the hygrolide family using our RODEO algorithm from Chapter 4, describing clusters from known and predicted producers. Our results indicate that potential remains for the Actinobacteria to yield novel hygrolide congeners and provides a survey of the hygrolide landscape. In Chapter 4, we report RODEO (Rapid ORF Description and Evaluation Online), an algorithm which combines hidden Markov model-based analysis, heuristic scoring, and machine learning to identify biosynthetic gene clusters and predict RiPP precursor peptides. We initially focused on lasso peptides, which display intriguing physiochemical properties and bioactivities, but their hypervariability renders them challenging prospects for automated mining. Our approach yielded the most comprehensive mapping of lasso peptide space, revealing >1,300 compounds. We characterized the structures and bioactivities of six lasso peptides, prioritized based on predicted structural novelty, including an unprecedented handcuff-like topology and another with a citrulline modification exceptionally rare among bacteria. These combined insights significantly expand the knowledge of lasso peptides, and more broadly, provide a framework for future high-throughput genome mining. In addition to lasso peptides, RODEO provides the ability to analyze local genomic regions using custom profile hidden Markov models (pHMMs) and is suitable for RiPP, polyketide (PKS), nonribosomal peptide (NRPS), and other natural product biosynthetic gene cluster types; as part of an effort to make it available as a community resource we have created a web portal with its code and tutorials