Uncovering the unexplored diversity of thioamidated ribosomal peptides in Actinobacteria using the RiPPER genome mining tool

Abstract

The rational discovery of new specialized metabolites by genome mining represents a very promising strategy in the quest for new bioactive molecules. Ribosomally synthesized and post-translationally modified peptides (RiPPs) are a major class of natural product that derive from genetically encoded precursor peptides. However, RiPP gene clusters are particularly refractory to reliable bioinformatic predictions due to the absence of a common biosynthetic feature across all pathways. Here, we describe RiPPER, a new tool for the family-independent identification of RiPP precursor peptides and apply this methodology to search for novel thioamidated RiPPs in Actinobacteria. Until now, thioamidation was believed to be a rare post-translational modification, which is catalyzed by a pair of proteins (YcaO and TfuA) in Archaea. In Actinobacteria, the thioviridamide-like molecules are a family of cytotoxic RiPPs that feature multiple thioamides, and it has been proposed that a YcaO-TfuA pair of proteins also catalyzes their formation. Potential biosynthetic gene clusters encoding YcaO and TfuA protein pairs are common in Actinobacteria but the chemical diversity generated by these pathways is almost completely unexplored. A RiPPER analysis reveals a highly diverse landscape of precursor peptides encoded in previously undescribed gene clusters that are predicted to make thioamidated RiPPs. To illustrate this strategy, we describe the first rational discovery of a new family of thioamidated natural products, the thiovarsolins from Streptomyces varsoviensis.

NeuRiPP: Neural network identification of RiPP precursor peptides

“…Email: emzodls@gmail.com products with a variety of different bioactivities. In contrast to peptide natural products from assembly-line non-ribosomal peptide synthetase (NRPS) pathways, RiPPs are derived from a ribosomally encoded precursor peptide (PP) that is extensively modified by RiPP tailoring enzymes (RTEs) (4,5). Beginning from a ribosomally encoded peptide makes RiPPs an attractive target for bioengineering as RTEs can be highly selective for recognition sequences in the PP but promiscuously process other regions of the sequence (6).…”

“…Identification of putative PP sequences is more challenging as they are frequently missed in genome annotation due to their short size (6). However, their proper identification is an important aspect of in silico RiPP BGC analysis as knowledge of the PP sequence can aid in structure elucidation and provide information of the molecular interactions between the RTEs and the PP (5). To this end, several methods have been developed to identify putative PPs in regions in proximity to RTEs, this typically involves a two-step process where sequences to be screened are first identified either through the use of gene-finding software (5,7), or from identifying open reading frames (ORFs) of specified length in the proximity of RTEs (6,8).…”

“…However, their proper identification is an important aspect of in silico RiPP BGC analysis as knowledge of the PP sequence can aid in structure elucidation and provide information of the molecular interactions between the RTEs and the PP (5). To this end, several methods have been developed to identify putative PPs in regions in proximity to RTEs, this typically involves a two-step process where sequences to be screened are first identified either through the use of gene-finding software (5,7), or from identifying open reading frames (ORFs) of specified length in the proximity of RTEs (6,8). The likelihood of these sequences to be PPs is then evaluated by different methods such as looking at similarity to known PPs by BLAST (7), hidden Markov models (HMMs) (5,9), or machine learning approaches such as Support Vector Machine (SVM) classifiers that are trained to identify likely PPs for different classes based on characteristics of PPs in the specified class (6,10).…”

“…To this end, several methods have been developed to identify putative PPs in regions in proximity to RTEs, this typically involves a two-step process where sequences to be screened are first identified either through the use of gene-finding software (5,7), or from identifying open reading frames (ORFs) of specified length in the proximity of RTEs (6,8). The likelihood of these sequences to be PPs is then evaluated by different methods such as looking at similarity to known PPs by BLAST (7), hidden Markov models (HMMs) (5,9), or machine learning approaches such as Support Vector Machine (SVM) classifiers that are trained to identify likely PPs for different classes based on characteristics of PPs in the specified class (6,10). While successful in identifying PP sequences and even identifying sequences that are different from known PPs of a specified RiPP class, these approaches have limitations which include only recognizing similar enough sequences to known PPs and being class-specific or context-dependent on the genes surrounding the putative PP.…”

“…A recent study presented a pipeline for identifying new RiPP clusters that included a modified version of the gene finding software prodigal (11), prodigal-short. Prodigal-short was used to find putative PPs in proximity to RTEs, and peptide similarity network analysis of the identified PPs was used to identify new RiPP classes (5). This demonstrated the potential of using gene-finding software as a starting point for identifying novel RiPPs; however, the number of likely coding sequences from this approach was still large and the researchers used proximity to known RTEs, restricting searches by phylogeny, and looking at only large similarity networks to reduce the number of putative BGCs to a size where manual curation was tractable.…”