Diversity and distribution of nuclease bacteriocins in bacterial genomes revealed using Hidden Markov Models

Abstract

Bacteria exploit an arsenal of antimicrobial peptides and proteins to compete with each other. Three main competition systems have been described: type six secretion systems (T6SS); contact dependent inhibition (CDI); and bacteriocins. Unlike T6SS and CDI systems, bacteriocins do not require contact between bacteria but are diffusible toxins released into the environment. Identified almost a century ago, our understanding of bacteriocin distribution and prevalence in bacterial populations remains poor. In the case of protein bacteriocins, this is because of high levels of sequence diversity and difficulties in distinguishing their killing domains from those of other competition systems. Here, we develop a robust bioinformatics pipeline exploiting Hidden Markov Models for the identification of nuclease bacteriocins (NBs) in bacteria of which, to-date, only a handful are known. NBs are large (>60 kDa) toxins that target nucleic acids (DNA, tRNA or rRNA) in the cytoplasm of susceptible bacteria, usually closely related to the producing organism. We identified >3000 NB genes located on plasmids or on the chromosome from 53 bacterial species distributed across different ecological niches, including human, animals, plants, and the environment. A newly identified NB predicted to be specific for Pseudomonas aeruginosa (pyocin Sn) was produced and shown to kill P. aeruginosa thereby validating our pipeline. Intriguingly, while the genes encoding the machinery needed for NB translocation across the cell envelope are widespread in Gram-negative bacteria, NBs are found exclusively in γ-proteobacteria. Similarity network analysis demonstrated that NBs fall into eight groups each with a distinct arrangement of protein domains involved in import. The only structural feature conserved across all groups was a sequence motif critical for cell-killing that is generally not found in bacteriocins targeting the periplasm, implying a specific role in translocating the nuclease to the cytoplasm. Finally, we demonstrate a significant association between nuclease colicins, NBs specific for Escherichia coli, and virulence factors, suggesting NBs play a role in infection processes, most likely by enabling pathogens to outcompete commensal bacteria.

Fig 1. Identifying genetically linked conserved cytotoxic and immunity motifs is a powerful and accurate way to identify NB operons.

a, Gene/protein organisation of a typical nuclease bacteriocin from E. coli. b, Left-hand panel, key interactions of conserved catalytic residues of the HNH motif of DNase bacteriocins. The two histidine residues of the HNH motif are involved in coordinating a divalent metal ion and the asparagine constrains the metal binding loop. The phosphate anion denotes the position of the scissile phosphate in substrate DNA (PDB code, 1V14 [34]). Right-hand panel, the helical immunity protein (green) showing the conserved aromatic residues of the α-helix III, which forms a critical part of the binding site for the DNase domain [15]. c, Conserved residues used to form the HMM profile of each protein are highlighted in the sequence alignments.

Fig 2. Distribution of NBs is restricted to the γ-proteobacteria.

Taxonomic tree representing all γ-proteobacteria species in the pubMLST that have over 15 genomes, constructed using NCBI taxonomy commontree. The presence of different cytotoxic domains is indicated in the color key associated with each species. NBs are found throughout Enterobacteriaceae and Pseudomonadaceae.

Fig 3. NBs cluster into 8 groups of differing domain arrangements.

Top panel, sequences of NBs from Enterobacteriaceae and Pseudomonas aeruginosa had cytotoxic domains removed before clustering using CLANS [51]. Bottom panel, PFAM profiles, coiled-coil regions, disordered regions and the conserved DPY motif in the most common domain organisations observed for each species. Sequences within clades tend to share a similar predicted domain structure. Block arrows indicate proteolytic processing sites that have been identified for a select few NBs [52, 53]. Box inset contains pyocin S9 sequences which were too distantly removed to show at the correct distance. Connections show High Scoring Segment Pairs (HSP) of <5x10^-20. The figure also highlights previously identified NBs as reference points. In addition to identifying many more NBs within the groups containing such characterized NBs we also identify three NB groups (III, IV and VI) that have completely novel domain organisations not previously described in the literature.

Fig 4. A conserved translocation motif was identified in all NBs.

a, Protein structure-based sequence alignment using PROMALS 3D indicates the conserved β-sheet secondary structure of a conserved domain identified in nuclease bacteriocins. Alignment features bacteriocins from E. coli (Colicin B, E9 and Cloacin DF13); Klebsiella pneumoniae (klebicin B), Pseudomonas aeruginosa (Pyocin AP41) and the pyocin_s domain from Erwinia carotovora The DPY motif was identified using MEME and is shown by a LOGO plot at the C-terminal end of the T-domain. b, Crystal structure of colicin E9, with its constituent domains identified, in complex with its immunity protein Im9 (PDB code, 5EW5). c, The conserved segment of the T-domain (blue) is annotated as the pyocin_S domain in the PFAM database (PFAM 06958), which is usually part of a larger T-domain, annotated as PFAM 03515 (green). Inset, Alignment of resides at the core of PFAM 06958 showing a conserved hydrogen bond network formed between the residues of the DPY motif; Asp270 and Tyr285 (colicin E9 numbering) and Arg185 at the beginning of PFAM 06958. d, Cytotoxic plate killing assay of DPY motif mutations. 100-fold serial dilutions of colicin E9 and DPY motif alanine mutants were spotted onto a lawn of sensitive E. coli showing that only the Tyr285Ala mutant abolishes colicin activity.

Fig 5. Pangenome analysis of colicinogenic bacteria shows evidence of an association between NBs and virulence factors.

Association of pathogenicity and colicinogenicity genes based on a Cochran-Mantel-haenszel test. Left-hand panel, RAxML tree of a core genome alignment. Population structure was calculated using BAPS and tree nodes are coloured by cluster as predicted by BAPS. Right-hand panel, presence and absence of genes associated with colicinogenicity. Red genes show nuclease colicin of different cytotoxic domains. dnaC is a core gene and included as a control.