Towards the discovery and evolutionary analysis of Nonribosomal Peptide Synthetases (NRPSs) within protist lineages

This work aimed to study the taxonomic distribution of Nonribosomal Peptide Synthetases (NRPSs) present within protists and to investigate their evolutionary history and origin. The research was conducted on available genomic data from major protist lineages, representing eukaryotic organisms with very different cellular organizations, morphologies, and lifestyles. To support this effort, two new software packages were developed. First, AMAW (Automated MAKER annotation wrapper), a dedicated pipeline for the gene annotation of emerging eukaryotic organisms. This tool was used to infer protein datasets from protist DNA genomes for which gene structures and protein sequences were not yet annotated. Second, Palantir, a large Perl library and associated scripts, was designed to make possible the analysis of thousands of NRPS predictions from the genome datasets, refine the NRPS domain annotations, and store their multidimensional information (i.e., NRPS domain, module, gene and cluster scales) into SQL relational databases.

The presence of NRPSs was surveyed in protist genomic data and highlighted a broad taxonomic coverage in terms of lineages, revealing a broad taxonomic distribution across the eukaryotic Tree of Life. However, the presence of these enzymes was observed to be scarce at lower taxonomic levels (i.e., genus and species levels). In an attempt to better understand the nature and properties of eukaryotic NRPSs, qualitative and statistical comparisons were performed with those of Actinobacteria, Cyanobacteria, Firmicutes, Proteobacteria, and Fungi, considered as reference lineages. Identified protist NRPSs showed characteristics mainly similar to those of bacterial NRPSs with respect to domain composition and architecture.

The evolutionary history and origin of protist NRPSs were explored with the use of comparative genomics and phylogenetic methods. On the one hand, two metrics were designed to score the pairwise comparison of entire NRPS gene clusters and build similarity networks: (i) a global score attempting to reflect how two NRPS gene clusters are close through the shared monomer composition of their nonribosomal peptides, and (ii) a local score focusing on the sequence similarity between NRPS domains. The global score considers three properties from NRPSs, rewarding shared domain similarity sequence and synteny, while penalizing the presence of domains unique to one NRPS. A Machine Learning model was trained to predict this global score based on a labeled dataset combining NRPS data of these three genomic features with the Tanimoto coefficient of the monomer composition of their nonribosomal peptides (as proxy scores of their global resemblance). On the other hand, the local score allows the incomplete and repetitive alignment of the compared sets of domains. The global similarity network showed heteroclite NRPS associations from a taxonomic perspective. This result was interpreted evidence for the network grouping NRPSs of similar domain architecture encoding likely related potential products, rather than being informative on the evolution and origin of these megasynthetases. In contrast, the local similarity network connected genetically close NRPSs, highlighting recent horizontal gene transfer (HGT) events and/or possible genomic contamination.

Regarding phylogenetic methods, these were used to infer the evolutionary relationships between NRPSs at the domain and module scales rather than gene clusters. Phylogenetic trees of homologous domains and modules showed that bacterial lineages tended to cluster in subtrees presenting high numbers of inclusions from other bacterial lineages. In comparison, Fungi were more taxonomically homogeneous, with very limited inclusions from bacterial lineages. Phylogenetic trees revealed that protist NRPSs were not closely related between them and instead mainly associated to various bacterial NRPSs. A notable exception concerned a fungal HGT into Aphanomyces (Stramenopiles molds), demonstrating that NRPSs can be transferred between eukaryotes. These results discard both a universal (i.e., in LUCA) origin of NRPSs and their introduction in early eukaryotes through a single HGT from bacteria. Instead, they support multiple (maybe recent) HGTs that affected modern protist lineages, from distinct bacterial donors. Based on this work, we formulated a hypothesis with regard to a more complete evolutionary scenario for the evolution of protist NRPSs. Despite the insights this scenario offers into the origin of NRPSs, some uncertainties remain. The low amount of genomes and their uneven distribution across protist lineages precludes a complete picture of the number of origins and occurrences over the course of evolution within each lineage, as well as the extent of their vertical inheritance. Furthermore, no precise identification of the donors of NRPSs to protists could have been identified. It is possible these donors were not represented in our dataset of reference lineages, which may require to be extended to other bacterial lineages for a definitive answer. Finally, tracking the number and order of HGTs that occurred over time is a very challenging task, given the complex evolutionary mechanisms of these large modular enzymes.