Genome mining of uncharted ribosomally synthesized and post-translationally modified peptides

Abstract

Ribosomally Synthesized and Post-translationally Modified Peptides (RiPPs) are among the largest natural product families. RiPPs exhibit remarkable structural diversity, allowing them to be promising candidates for drug discovery. With over forty known RiPP families and new family members being reported yearly, RiPPs represent a largely unexplored dark matter in microbial secondary metabolites. Although a wealth of excellent bioinformatic tools have been developed to annotate RiPP biosynthetic gene clusters (BGCs), predicting outside the box is an intrinsic limitation of these methods. We here provide a new genome mining strategy for RiPP BGC discovery and demonstrate our method by systematically analyzing two enzyme families from bacterial genomes. Chapter 1 introduces the biosynthesis of RiPPs and current genome mining strategies for RiPP BGC discovery. Chapter 2 reports a biosynthetic logic-based bioinformatic workflow SPECO (small peptide and enzyme co-occurrence) for RiPP BGC discovery. We applied this method to all the available bacterial genomes and comprehensively analyzed radical S-adenosyl-L-methionine (SAM)-dependent enzyme (rSAM)-catalyzed RiPPs. The bioinformatic data demonstrate that rSAM enzyme has potential to catalyze diverse ribosomal peptide cyclization and over 50% of them are not characterized. Our data provided a valuable source for exploring new rSAM enzymology in peptide modification. Chapter 3 introduces the characterization of three rSAM enzyme families that were selected from our comprehensive bioinformatic analysis. To achieve this, we cloned and heterologously expressed three BGCs, including sca, vgu and bla in E. coli, which enabled us to obtain modified peptides for further analysis. Structure elucidation and in vitro enzymatic reconstitution proved that the rSAM enzyme from the sca cluster catalyzes C(sp2)-C(sp3) bond formation between histidine and alanine via radical chemistry. Notably, this enzyme can also construct tyrosine-, tryptophan- and phenylalanine-alanine crosslink, implying a broad substrate tolerance. The rSAM enzyme from the vgu family catalyzes two crosslinks between tyrosine-arginine and histidine-arginine. The bla cluster was found to biosynthesize a new sactipeptide, featuring two C-S bonds between two cysteine-glycine pairs. Our characterization of these novel rSAM enzymes and their corresponding cyclized peptides has significantly expanded our understanding of ribosomal peptide macrocyclization. In Chapter 4, we further investigated bacterial cytochrome P450-catalyzed RiPPs, which is sporadically reported due to the difficulty of the current genome mining strategy in annotating this type of BGC. We incorporated AlphaFold2 into the SPECO workflow to predict small peptide-P450 complex, based on which we can infer the binding pose of small peptide and potential modification sites of the precursor. This combined method was applied to analyze all the genomes of actinobacteria, yielding 1962 putative P450-catalyzed BGCs. We characterized four representative BGCs and found that these P450s exhibit diverse catalytic capabilities, catalyzing unseen aromatic crosslinks by forming C-C, C-N, and C-O bonds. Notably, these P450 enzymes can cyclize non-native precursor peptides, thus highlighting their immense potential in biocatalysis. This work presents an alternative biosynthetic pathway for the construction of structurally complex macrocycles. Our genome mining approach, along with the discovery of these novel P450 enzymes and identified cyclic peptides contribute to the development of genome mining tools, enzyme discovery and engineering.