Genomic and Ecogenomic Characterization of Proteus mirabilis Bacteriophages

Abstract

Proteus mirabilis often complicates the care of catheterized patients through the formation of crystalline biofilms which block urine flow. Bacteriophage therapy has been highlighted as a promising approach to control this problem, but relatively few phages infecting P. mirabilis have been characterized. Here we characterize five phages capable of infecting P. mirabilis, including those shown to reduce biofilm formation, and provide insights regarding the wider ecological and evolutionary relationships of these phages. Transmission electron microscopy (TEM) imaging of phages vB_PmiP_RS1pmA, vB_PmiP_RS1pmB, vB_PmiP_RS3pmA, and vB_PmiP_RS8pmA showed that all share morphologies characteristic of the Podoviridae family. The genome sequences of vB_PmiP_RS1pmA, vB_PmiP_RS1pmB, and vB_PmiP_RS3pmA showed these are species of the same phage differing only by point mutations, and are closely related to vB_PmiP_RS8pmA. Podophages characterized in this study were also found to share similarity in genome architecture and composition to other previously described P. mirabilis podophages (PM16 and PM75). In contrast, vB_PimP_RS51pmB showed morphology characteristic of the Myoviridae family, with no notable similarity to other phage genomes examined. Ecogenomic profiling of all phages revealed no association with human urinary tract viromes, but sequences similar to vB_PimP_RS51pmB were found within human gut, and human oral microbiomes. Investigation of wider host-phage evolutionary relationships through tetranucleotide profiling of phage genomes and bacterial chromosomes, indicated vB_PimP_RS51pmB has a relatively recent association with Morganella morganii and other non-Proteus members of the Morganellaceae family. Subsequent host range assays confirmed vB_PimP_RS51pmB can infect M. morganii.

Keywords: bacteriophage; biofilms; catheters; ecogenomics; phage therapy.

FIGURE 1

Capsid and plaque morphology of Proteus mirabilis phages. Images show transmission electron microscopy (TEM) projections of negatively stained phage (left), and examples of plaques formed by each phage on lawns of host bacteria (right). Phage capsid structures were congruent with membership of the Podoviridae family (RS1pmA and RS8pmB), and the Myoviridae family (RS51pmB). RS1pmA is used to represent the previously described group of phages RS1pmA, RS1pmB and RS3pmA, which all have analogous capsid morphology (Nzakizwanayo et al., 2016). Scale bars show 50 nm on TEM images and 1 cm on images of plaques. TEM images shown are representative of at least 10 fields of view containing one or more virions.

FIGURE 2

Functional content and characteristics of P. mirabilis phage genomes. Maps show the physical gene architecture of P. mirabilis phages RS1pmA, RS51pmB, and RS8pmA. Block arrows represent positions of open reading frames (ORFs) with colors indicating affiliation to broad functional groups relevant to phage replication, as denoted by the associated key: Cell Lysis, predicted to encode for activities involved in lysis of host cells; Replication and Regulation, predicted to encode for activities involved in replication of phage genomes and regulation of this process; Structure and Packaging, predicted to encode components of the phage capsid or support packaging of new phage genomes during replication; Unknown Phage related, unknown function but homologous sequences identified in other phage genomes; Unknown, generated no valid hits in BlastP or Conserved Domain searches. Putative functions of each ORF were predicted using BlastP and conserved domains (CD) searches against the nr database (BlastP: minimum 20% identity, 1e^{– 5} or lower; CD: 1e^{– 2} or lower).

FIGURE 3

Representation of P. mirabilis phage-encoded ORFs in other phage genomes. The representation of P. mirabilis ORF homologs in other phage genomes was explored using tBLASTn searches of 715 complete phage genomes. The proportion of ORFs affiliated to other phage genomes is based on top hit by bit score (min 35% identity, over ≥25 amino acids, 1e^{– 5} or lower). Hits were categorized by genus of host bacterial species for phage genomes generating each hit. Legends associated with charts describe bacterial genera represented. The “Gamma Proteobacteria” group included Hamiltonia sp., Listonella sp., and Pseudoalteromonas sp., which were each represented by less than two hits across all phage genomes. RS51pmB also only generated a single hit to phage infecting Mycobacterial sp. ORFs generating no valid hits in tBlastn searches are designated as “Unknown.” ^*RS1pmA is used to represent the previously described group of phages RS1pmA, RS1pmB, and RS3pmA (Nzakizwanayo et al., 2016), which were found to differ only by point mutations (see Supplementary Figure S3).

FIGURE 4

Evaluation of broader evolutionary relationships and host affiliations. The potential for broader evolutionary relationships between P. mirabilis phages and other bacterial hosts were explored by comparing P. mirabilis phages with bacterial genomes belonging to genera containing the most common Gram-negative pathogens of the catheterized urinary tract (Proteus, Providencia, Morganella, Pseudomonas, Escherichia, Klebsiella, Enterobacteria, and Serratia) (Stickler, 2014). (A) Unsupervised ordination using nMDS was initially used to visualize relationships between all sequences based on tetranucleotide usage profiles (Ogilvie et al., 2012, 2013). Bacterial genomes were classified by family. For nMDS, points show position of individual genomes in the final ordination, with connecting lines indicating relationship to the group centroid. Filled ellipses show standard deviation of group dispersion relative to the group centroid. Ordination were based on an Euclidean distance matrix of tetranucleotide usage profiles from each sequence. nMDS analysis was conducted with 1000 random starts using the R package Vegan. (B) ANOSIM between groups of sequences represented in the nMDS ordination. Charts show the ANOSIM R Statistic for each group of bacterial genomes vs. the P. mirabilis phage group, where an increasing strength of separation between groups is indicated as the R Statistic approaches 1. ^∗∗Denotes statistical significance of R Statistics between groups, P ≤ 0.001 (calculated as part of the ANOSIM in the R Package Vegan). (C) Cladogram based on tetranucleotide usage profiles showing relationship between P. mirabilis phage sequences and genomes from Morganellaceae and Yersiniaceae groups represented in nMDS ordinations. The cladograms were constructed from a Pearson dissimilarity matrix of tetranucleotide profiles from all sequences used in nMDS analyses as majority consensus trees from 500 bootstrap replicates (as calculated by the R package BioDist). The cladogram presented shows a sub-region of the larger tree populated by P. mirabilis phages and the most closely related bacterial genomes by this analysis. Chart Inset: shows differences between P. mirabilis phage sequences, based on scatter plots of phage RS1pmA and RS51pmB tetranucleotide correlation scores with other phage sequences. (D) To further test relationships in phage RS51pmB indicated by tetranucleotide analyses, ORFs from this phage were searched against all sequences from genus level taxonomic groups represented in cladograms (Proteus, Providencia, Morganella, and Serratia) using BlastP, and hits used to assign a taxonomic affiliation. Only hits with a maximum e-value of 1e^{– 3} were considered valid. Charts show the proportion of assignable ORFs affiliated to each genus, and the proportion of assigned and unassigned ORFs. (E) To test if the relationships with non-Proteus species indicated in tetranucleotide and Blast analyses were relevant to the host range of this phage, the ability of RS51pmB to replicate in species other than P. mirabilis was tested (Supplementary Table S1). The image presented demonstrates the ability of phage RS51pmB to use M. morganii as an alternate host species (Supplementary Table S1). Host range assays were performed in triplicate at phage titres ranging from ∼10 to 10⁷ PFU, with the image shown representative of M. morganii lawns exposed to 10⁷ PFU of RS51pmB.

FIGURE 5

Ecogenomic profiling of P. mirabilis phage encoded functions. The representation of functions encoded by P. mirabilis phages in a range of microbial habitats was investigated by calculating the relative abundance of similar sequences in 809 metagenomic data sets. Valid hits from tBlastn searches with P. mirabilis phage sequences (≥35% identity, ≥50% query coverage, ≤1e^{– 5}) were used to calculate the average relative abundance of ORF homologs in each dataset (expressed as Hits/Mb). (A) The heatmap shows average relative abundance of ORFs from each phage in metagenomes or viromes from various habitats. Rows represent datasets grouped by habitat, and columns represent individual phage. The intensity of shading in cells indicate cumulative relative abundance of ORFs from a given phage in metagenomes from the corresponding habitat (according to the scale shown). The associated histogram shows average relative abundance of phage ORFs across all datasets. (B) Heatmaps show cumulative relative abundance of individual ORFs from each phage (Rows) in metagenomes from various habitats (Columns), and shading of cells indicates relative abundance according to the associated scale. Associated histograms show the proportion of ORFs in each phage with homologs detected in various habitats represented by metagenomes surveyed. ^∗∗∗∗P < 0.0001 (Kruskal-Wallis test with Dunn’s correction), error bars show standard error of the mean. RS1pmA is used to represent the previously described group of phage RS1pmA, RS1pmB, and RS3pmA (Nzakizwanayo et al., 2016). ENV, whole community metagenomes with non-host associated environmental origin (n = 16); HGUT, HORAL, and HBODY, whole community metagenomes from the human gut microbiome, the human oral cavity, or various human body sites (primarily external) (n = 285, 295, and 181, respectively). HGV, viral metagenomes from the human gut (n = 12). UTV, viral metagenomes from the human urinary tract (n = 20). Details of datasets used can be found in Supplementary Table S2.