Genomic, proteomic and bioinformatic analysis of two temperate phages in Roseobacter clade bacteria isolated from the deep-sea water

Background Marine phages are spectacularly diverse in nature. Dozens of roseophages infecting members of Roseobacter clade bacteria were isolated and characterized, exhibiting a very high degree of genetic diversity. In the present study, the induction of two temperate bacteriophages, namely, vB_ThpS-P1 and vB_PeaS-P1, was performed in Roseobacter clade bacteria isolated from the deep-sea water, Thiobacimonas profunda JLT2016 and Pelagibaca abyssi JLT2014, respectively. Two novel phages in morphological, genomic and proteomic features were presented, and their phylogeny and evolutionary relationships were explored by bioinformatic analysis. Results Electron microscopy showed that the morphology of the two phages were similar to that of siphoviruses. Genome sequencing indicated that the two phages were similar in size, organization, and content, thereby suggesting that these shared a common ancestor. Despite the presence of Mu-like phage head genes, the phages are more closely related to Rhodobacter phage RC1 than Mu phages in terms of gene content and sequence similarity. Based on comparative genomic and phylogenetic analysis, we propose a Mu-like head phage group to allow for the inclusion of Mu-like phages and two newly phages. The sequences of the Mu-like head phage group were widespread, occurring in each investigated metagenomes. Furthermore, the horizontal exchange of genetic material within the Mu-like head phage group might have involved a gene that was associated with phage phenotypic characteristics. Conclusions This study is the first report on the complete genome sequences of temperate phages that infect deep-sea roseobacters, belonging to the Mu-like head phage group. The Mu-like head phage group might represent a small but ubiquitous fraction of marine viral diversity. Electronic supplementary material The online version of this article (doi:10.1186/s12864-017-3886-0) contains supplementary material, which is available to authorized users.


Background
Marine phages are one of the most abundant biological components of marine environments and are believed to significantly contribute to the microbial loop and biogeochemical cycles of the ocean [1][2][3]. Although the emergence of cultivation-independent tools such as metagenomics have expanded our understanding of viral community composition and their genetic diversity [4], research studies in the field of marine virology have focused on identification of phages infecting ecologically important environmental bacteria [5]. The importance of phage isolation is exemplified by studies on phages infecting the ubiquitous marine bacteria such as cyanophages of Cyanobacteria [6], SAR11 clade viruses [7], and roseophages of Roseobacter clade bacteria (RCB) [8].
Recent viral ecological studies have indicated that phages provide an important, yet previously ignored contribution to deep-sea ecosystems functioning and environmental adaptation to its hosts [3,[23][24][25]. Compared to viruses in coastal and estuarine environments, lysogeny seems to be more prevalent in the deep biosphere, as indicated by the presence of high amounts of temperate phages [25]. Although attempts to isolate phages from deep-sea bacteria have been successful in several cases [26][27][28], these remain largely unexplored because only a few hosts have been cultivated. To date, our understanding of deep-sea roseobacter phages is limited. Pelagibaca abyssi JLT2014 [29] and Thiobacimonas profunda JLT2016 [30] are two RCB members that have been isolated from the deep-seawater (water depth: 2000 m and 2571 m) of the Southeastern Pacific Ocean. A recent study suggests that two deep-sea roseobacter bacteria have mixotrophic capacities that these may be potentially utilized in chemolithotrophic carbon dioxide fixation [31].
The present study characterized phages of deep-sea roseobacters by DNA sequencing and proteomics analysis, resulting in the identification of two mitomycin Cinduced temperate phages that contain Mu-like elements and transposases, hereby designated as Thiobacimonas phage vB_ThpS-P1 and Pelagibaca phage vB_PeaS-P1. Mu-like bacteriophages are phylogenetically related to Mu phages and have been isolated primarily from Gammaproteobacteria such as Escherichia phage D108 and Haemophilus phage SuMu belonging to the Myoviridae family [32,33], and Pseudomonas phages D3112 and B3 affiliated to the Siphoviridae family [34,35]. The Mu-like phage RcapMu with siphoviruslike morphology was induced using high temperature from Rhodobacter capsulatus SB1003, which belongs to the Rhodobacteraceae family [36] and is the first reported transposing bacteriophage that infects Alphaproteobacteria. Mu-like prophages are generally not inducible by mitomycin C [32]. However, lysogenic phage vB_CibM-P1 with Mu-like elements was induced by mitomycin C from Citromicrobium sp. JLT354 within marine Alphaproteobacteria, and it showed a myovirus-like morphology [37]. Two novel phages in roseobacters, vB_ThpS-P1 and vB_PeaS-P1, contain structural modules and proteomes similar to those of the Mu and Mu-like phages. However, unlike Mu and Mu-like phages, these are incapable of carrying variable amounts of host DNA during both lytic and lysogenic development. The present study compared phages containing Mu-like elements to typical Mu and Mu-like phages, herein designated as the "Mu-like head phage group", to resolve discrepancies between function and phylogeny of transposable phages.

Phage induction
P. abyssi JLT2014 and T. profunda JLT2016 were cultured in rich organic medium (1 g yeast extract, 1 g Bacto-peptone, and 1 g sodium acetate per liter of artificial seawater with vitamins and trace elements) at 28°C at a constant rotation of 160 rpm. The induction process and sampling were performed as earlier described [21,37]. Briefly, bacterial suspensions were cultured in two 500-mL conical flasks until these reached a stable growth phase; mitomycin C (final concentration: 0.5 μg/mL) was added to one, whereas the other served as the control. After mitomycin C treatment for 30 min, the cells in both the control and treatment tubes were centrifuged, washed, and resuspended in 500 mL of fresh rich organic medium. Samples (2 mL) for viral and bacterial counting were immediately fixed with glutaraldehyde (final concentration: 1%) for 15-20 min in the dark and then stored in an −80°C refrigerator for flow cytometry analysis. Virus counting was conducted using an Epics Altra II flow cytometer (Beckman-Coulter, USA), and bacterial counts were determined by using a BD Accuri C6 flow cytometer. Samples were diluted in 0.2-μm filtered TE buffer (Tris-EDTA, pH 8), and stained with the DNA dye SYBR Green I (Molecular Probes, Inc., USA). The bacterial and viral particles were identified and counted as described elsewhere [38,39]. All reagents used in the experiments were obtained from Sigma-Aldrich (USA) unless otherwise specified.

Phage purification
Phage particles in lysates were harvested and purified as previously described [21,37]. Phage lysates were treated with RNase A (final concentration: 2 μg/mL) and DNase I (final concentration: 2 μg/mL) by incubating for 1 h and then centrifuging at 10,947×g for 10 min in a Thermo Scientific Sorvall ST-16R. Supernatants were filtered through a 0.45-μm pore size filter (type HA, Millipore, USA) to remove host cells and cellular debris. Phage particles in the filtrate were treated with polyethylene glycol 8000 (final concentration: 100 g/L) overnight at 4°C and precipitated by centrifugation at 10,947×g for 60 min. The pellets were resuspended in 6 mL of SM buffer (10 mM NaCl, 50 mM Tris, 10 mM MgSO 4 , and 0.1% gelatin) and then incubated overnight at 4°C. The phage suspension was mixed with CsCl (final concentration: 0.6 g/mL) and centrifuged in an Optima™ L-100XP (200,000×g for 24 h at 4°C). Visible bands were extracted and then dialyzed (molecular weight: 530 kDa) twice in SM buffer overnight at 4°C.

Transmission electron microscopy for phage morphology
One drop of purified phage suspension was adsorbed to a Formvar/carbon-coated 200-mesh copper grid for 10 min and negatively stained with 2% (wt/L) phosphotungstic acid in the dark for 30 s. After 30 min of drying, the grid was examined using a JEM-2100 transmission electron microscope (JEOL, Japan) or a Tecnai G2 Spirit transmission electron microscope at 120 KeV (Thermo Fisher Scientific, USA). Images were captured using a GATAN INC CCD image transmission system.

DNA preparation and genome sequencing
Phage DNA was extracted as described elsewhere [21], dissolved in TE buffer (10 mM Tris, 1 mM EDTA), and stored at 4°C. DNA library preparation was performed according to the NEBNext® Ultra™ DNA Library Prep Kit for Illumina (NEB, USA). Approximately 10 ng of the DNA sequencing library was used to generate a cluster in cBot using a TruSeq PE Cluster Kit (Illumina, USA) and then sequenced in an Illumina HiSeq™ 2500 system for 2 × 125 bp data. The raw data were filtered using a FASTX-Toolkit to remove the adapters, N bases, and low-quality reads (http://hannonlab.cshl.edu/fastx_toolkit/ ). Clean reads were mapped to the bacterial complete genome sequences [31] using Bowtie 2 [40], and then two prophage regions with an average coverage of 8000× and 7500× were compared to the whole genome, which had a depth of about 10× and 15×, respectively.
The sequences of phages vB_ThpS-P1 and vB_PeaS-P1 have been deposited in the GenBank database under Accession Number KT381864 and KT381865, respectively.

Proteomics analysis
The purified phages were treated with a lysis buffer (1 mM EDTA, 250 mM Tris-HCI (pH 6.8), 4% 2mercaptoethanol, 4% SDS, 50% glycerol, and 0.02% bromphenol blue) at 100°C for 10 min, and stored at −20°C for further protein analysis. Protein in-solution digestion was performed according to the FASP procedure [41]. Proteomics analysis was performed on a Q Exactive mass spectrometer that was coupled to an Easy nLC (Thermo Fisher Scientific, USA). The instrument was run with the peptide recognition mode enabled. MS/MS spectra were searched using a MASCOT engine (Matrix Science, London, UK; version 2.2) against phage genomes. For protein identification, the following options were used. Peptide mass tolerance = 20 ppm, MS/MS tolerance = 0.1 Da, enzyme = trypsin, missed cleavage = 2, fixed modification: carbamidomethyl (C), and variable modification: oxidation (M).

Bioinformatics analysis
Clean high-depth mapped datasets were assembled using Velvet (v1.2.03) [42]. The final assembled phage genome was automatically annotated, then manually corrected through the RAST server using SEED annotation tools [43]. Prophage-like sequences at the gene cluster level in NCBI GenBank bacterial genomes (01/2015) were detected by using a MultiGeneBlast (v1.1.14) architecture search with vB_ThpS-P1, vB_PeaS-P1, and vB_CibM-P1 genomes as queries [44]. The prophage identification tool PHAge Search Tool (PHAST) was used to determine the region containing prophage like elements in bacterial genomes [45]. The clustering of sequences into homologous families was performed using SiLiX (v1.2.8) using a minimum identity threshold of 80% and default values for the remaining parameters [46]. The gene content of phages and phage-like elements were visualized by a hierarchical clustering method using Gene-E tool (https://software.broadinstitute.org/ GENE-E/download.html). Phylogenetic trees were based on maximum-likelihood and neighbor-joining methods and constructed using MEGA 6.0with a JTT model, with gamma set to 4 [47]. Bootstrap resampling was performed for 1000 replications. BLAST-based average nucleotide identity was determined using JSpecies (v1.2.1) [48]. An InterPro database search of gpT and gp23 genes sequences in the Tara Oceans metagenomic datasets was performed with the protein domain IPR018774 and IPR010762 as queries (https://www.ebi.ac.uk/metagenomics/projects/ ERP001736) [49,50]. To remove bias to the average genome size with gene sampling of genes from a given metagenomic community and the effects of gene size on hit retrieval, the abundance of gpT gene relative to the number of single-copy genes (recA) hits for each site was calculated as previously described [51], which was as follows: Number of single-copy genes (recA) = Number of size-normalized gpT gene hits/Average number of recA gene hits. All sequences in NCBI viral genomes and Mu-like head phages were Blast against the Pacific Ocean Virome, respectively (E-value 0.00001) [52].

Phage Induction and Morphology
T. profunda JLT2016 growth was apparently inhibited after the addition of mitomycin C at the exponential growth phase, whereas that of virus-like particles (VLPs) rapidly increased to 4.9 × 10 10 particles/mL after 15 h ( Fig. 1A and B). A dramatic increase in the number of inducible VLPs (4.4 × 10 10 particles/mL) in P. abyssi JLT2014 was observed within 10 h of mitomycin C treatment ( Fig. 1C and D). The two transposable prophages were not induced at a high temperature (42°C), whereas the transposable coliphage Mu is usually induced by high temperature rather than mitomycin C [32]. When these induced VLPs were used to re-infect two strains, lytic interactions between the phages and strains were not observed. A lysogenic bacterium is resistant to reinfection by the same or related phages because an "immunity" is conferred by the presence of the prophage [53]. The induced bacteriophage vB_ThpS-P1 exhibited a siphovirus-like morphology with a long flexible and non-contractile tail ( Fig. 2A). The average particle had a head size of approximately 63 ± 3 nm and tail length of approximately 205 ± 4 nm. The inducible phage vB_PeaS-P1 was morphologically identical to that of vB_ThpS-P1 phages, with only a slight difference in length and width (head size: 64 ± 2 nm; and tail length: 211 ± 3 nm; Fig. 2B). The tail features of the two deepsea roseobacters phages were similar to those of the well-characterized Mu-like phage RcapMu [36].

Genomic Features
vB_ThpS-P1 contained double-stranded (ds) DNA of 39,591 bp in size and a GC content of 66.7%, which was similar to that of the complete genome of T. profunda JLT2016 (67.1%) [31]. The phage vB_PeaS-P1 genome consisted of 38,686 bp of dsDNA and a GC content of 63.8%, which was similar to that of its host DNA (66%) [31]. No tRNA sequences were detected in their genomes.
A total of 52 and 51 open reading frames (ORFs) were identified in vB_ThpS-P1 and vB_PeaS-P1, respectively. A total of 24 ORFs and 19 ORFs were functionally  (Table 1). ORFs in the genomes of the two phages were mostly oriented in a single direction, whereas two phage repressor genes (ORF5 in vB_ThpS-P1 and ORF1 in vB_PeaS-P1) were transcribed in the reverse direction (Fig. 3). They showed almost the same functional genes, except for phage vB_ThpS-P1 lacked the gene encoding for DNA-binding protein HU-beta. These also harbored Mu-like phage gp36 and gp29 genes, which might play a role in virulence [33]. The phage genomes both possessed ParB homologs and two putative transposase genes, which might function in phage integration into the host genome [54]. Each phage genome contained a repressor (Cro/CI family) gene, which prevents transcription and translation of lysis and other late genes [27]. These genes, which act as a lysogeny module, are common features of temperate phages. Based on the putative functions of ORFs, the genome of vB_ThpS-P1 and vB_PeaS-P1 shared a similar backbone that could be roughly divided into four functional modules (Fig. 3): the right half largely encodes functions required for structure and lysis (putative phage gene expression late region: tail, head, and lysis modules), whereas the left half mainly encodes proteins that modulate phage gene expression or host response and lysogeny (putative phage gene expression in the early or middle region). The genomes of the two phages were distinct from that of other known roseophages [11][12][13][14][15][16][17][18][19][20].
The genomes of vB_ThpS-P1 and vB_PeaS-P1 had Mu-like phage homologs, which include the head morphogenesis and transposases genes. Although the modular organization and gene content of its structural head module were similar to that of Mu-like phages, some differences between its genomes and that of the Mu phage were identified. The two phages shared more sequence homologs with a Rhodobacter phage RC1 (23 homologs between vB_ThpS-P1 and RC1, 22 homologs between vB_PeaS-P1 and RC1) than with the Mu or RcapMu phages, suggesting that the two phages were closely related to RC1 at the genomic level (Fig. 4). Rhodobacter phage RC1 was induced from Rhodobactersp. E32, which was isolated from the deep-sea sediment (water depth: 5086 m) and belonged to the Siphoviridae family (GenBank Accession NumberNC _020839.1). The two phages also lacked the hostnuclease inhibitor protein, Gam, and the Mor transcription activator, which are high-frequency proteins that exist in Mu and Mu-like phages [55], whereas the two novel phages have the ParB protein, which was not detected in the Mu phage. In addition, the viruslike particle did not harbor any random host-derived sequences (>1 kb) at its genomic DNA termini, which is a unique feature of Mu phage-related phages [32,34].

Proteomic features
A detailed proteomic characterization of virion particles by using high-resolution LC-mass spectrometry identified 15 and 18 proteins in vB_ThpS-P1 and vB_PeaS-P1, respectively ( Table 1). The same functional proteins were identified in the two phages, including one tail protein, four Mu-like head structural proteins, one Mu-like phage gp26 protein, one lysis-related protein, and two hypothetical proteins. Based on mass spectrometry spectral count, the most abundant structural protein detected in both phages was the major head protein, Mu-like phage gpT protein. Other proteins that predominated in both phages were encoded in the genomic tail and morphogenesis modules. Mu-like phage gp36 protein and gp29 protein were identified in both. Of the hypothetical proteins in the two phages, an unknown protein with a DUF3164 domain was detected. A transposase B protein was detected in vB_ThpS-P1. A transposase B protein normally promotes efficient transposition and is directly involved in the choice of DNA target sites and immunity to self-integration [56,57].

Phylogenetic analysis
Phylogenetic analysis based on the conserved amino acid sequences of the Mu-like phage gpT protein supported the finding that among all the known phages but distinct from previously characterized Mu and Mu-like, vB_ThpS-P1 and vB_PeaS-P1 were most closely related to RC1 (Fig. 5). This tree resolved one subgrouping of the siphovirus phages, including the known RC1 and Pseudomonas phages and one subgrouping of the myovirus phages, including Mu-like phage SuMu [33] and Mu phage [32]. However, this tree placed the Mu-like phage gpT protein of myovirus-like phage vB_CibM-P1 [37] in the clade that included RC1, vB_ThpS-P1, and vB_PeaS-P1. In addition, their similar phylogenetic relationships were recovered with the reconstruction of other five other head structural proteins (Additional file 1: Fig. S1), transposase A or B, and transcriptional regulator proteins, respectively (Additional file 1: Fig. S2). To reconcile possible discrepancies between phylogeny and morphologyof transposable phages, a new family Saltoviridae of the order Caudovirales was recently proposed, which included subfamilies Myosaltovirinae and Siphosaltovirinae [58]. Phylogenetic analysis combined with morphological assessment indicated that the two phages could be taxonomically classified into the Siphosaltovirinae. The myovirus genomes have a notably lower GC content than the siphoviruses, with the exception of vB_CibM-P1. Furthermore, these have similar genome sizes and number of ORFs (Fig. 5).

Distribution of the Mu-like phage elements in bacterial genomes
Mu-like phage elements occur in more than 130 bacterial genera (Additional file 2: Table S1). The predicted Mu-like phage regions are not only found in species of Escherichia, Vibiro, Haemophilus, Burkholderia, Neisseria, Pseudomonas, and Rhodobacter, in which Mu-like prophages were described [32][33][34][35][36][59][60][61][62], but also in unexplored species such as strains of RCB (Additional file 2: Table S1). Most of the identified phage elements contain the Mu-like phage head genes, which suggest that head structural genes in Mu-like elements are relatively stable at the genomic level (Additional file 2: Table S1). These have at least one transposase and one transcriptional regulator (Additional file 3: Table S2). Prophage-like elements frequently harbor the DUF3164 and Mu-like phage gp26 proteins (Table 1). Based on the gene content of phage and phage-like elements, the Mu and RcapMu phages were classified into a separate branch from that of the clade containing phages vB_ThpS-P1, vB_PeaS-P1, RC1, and vB_CibM-P1 (Fig. 6). We proposed a "Mu-like head phage group" allowing the inclusion of phages that contain Mu-like head structural genes. The phages vB_ThpS-P1, vB_PeaS-P1, RC1, and vB_CibM-P1 were classified as members of the Mu-like head phage group, which reflects its evolutionary relationship with known Mu-like phages that, in turn, could be included in the group. For instance, Mu-like head phages infecting P. aeruginosa contain Mulike phages (such as D3112, DMS3, and MP22) and the other transposable phages (Fig. 5). Their genomes are similar and share nearly identical Mu-like head proteins (Additional file 1: Figure S1).
Geographic distribution of members of the mu-like head phage group Hosts of phage vB_CibM-P1 were obtained from a seawater depth of 75 m [37], whereas hosts of vB_ThpS-P1 and vB_PeaS-P1 were isolated from two distinct stations at depths of 2000 m and 2571 m in the bathypelagic water, respectively [29,30]. The average relative abundances of gpT genes of Mu-like phages in the Tara ocean metagenomic datasets for surface and mesopelagic water samples were 1.3% and 0.4%, respectively, with an average ratio of 1% between the gpT genes of Mu-like phages and gp23 genes of T4-like bacteriophages (Additional file 4: Table S3), implying that the hosts of Mu-like head prophages rarely occur in the surface water down to the mesopelagic zone. The highest relative abundance of gpT genes (10.0%) was observed in one station of the South Atlantic at a depth of 800 m (Additional file 4: Table S3). The sequences of Mu-like head phage group were represented in 0.4% of the deep-sea viral metagenomesin the Pacific Ocean. A Mu-like head phage Rhizobium phage RR1-B was found to significantly contribute to temperate phages abundance in the deep subseafloor sediments [23]. These analyses revealed that this Evolutionary implications for the mu-like head phage group The vB_ThpS-P1 and vB_PeaS-P1 genomes exhibited an average nucleotide identity (ANIb) of 65.1% (69% genome involved in alignment), which was derived from random fragment BLAST, whereas the ANIb value of vB_ThpS-P1 and RC1, vB_PeaS-P1 and RC1 was 62.9% and 63.1%, respectively (both >60% genome alignment). Their low level of nucleotide identity and the observation that only five pairs of homologous sequences showing >50% amino acid identity (Fig. 3) suggest that a significant number of mutations might have occurred in the phage genome, similar to that observed in other tail bacteriophages [63]. In addition, the genomes of the novel phages in the two roseobacters exhibited a mosaic relationship with the other phages (Additional file 5: Table S4), which was similar to other dsDNA tailed bacteriophages [63,64]. Several Siphoviridae-like tail genes in vB_ThpS-P1, vB_PeaS-P1 and RC1 have homologous relationships with each other (Fig. 6). Majority of head ORFs in vB_CibM-P1 possess a close phylogenetic relationship with those in RC1 ( Fig. 5 and Additional file 1: Figure  S1). Furthermore, they were closely clustered with RC1 based on gene content (Fig. 6). However, vB_CibM-P1 has a mosaic genomic structure that includes a Myoviridaelike tail and a Siphoviridae-like head. One tail protein of vB_CibM-P1 showed a low amino acid identity with the corresponding Rhizobium phage protein [37]. Based on these observations, we hypothesized that an ancestor of vB_CibM-P1 might belong to siphovirus, and their tail genes were subsequently replaced with those of other phages through the horizontal exchange of genetic material.
On the other hand, domestication of prophages by bacteria can result in the loss of substantial amounts of genes in the host genome [65]. Most of the identified prophage regions in bacterial genomes (Additional file 2: Table S1) are more likely to be prophage remnants because its genomic length and number of ORFs were respectively shorter and lower than those of the characterized Mu-like head phages (Fig. 5). A Mu-like phage remnant in the RCB strain Oceanicola sp. S124 harbored 27 ORFs that were homologous to vB_ThpS-P1, which was indicative of a close evolutionary relationship (Fig. 6). The prophage remnants were identified in other RCB genera, including Phaeobacter, Roseovarius, Fig. 6 Hierarchical clustering of vB_ThpS-P1, vB_PeaS-P1, and prophage-like elements among bacteria in terms of gene content, which was based on SiliX homology clustering. Green represents the presence of a gene. Two prophages in this study were labeled with blue dots and other reference phages are labeled as red dots. ORFs are color-coded according to predicted function: thistle, tail; yellow, head; dark violet, lysis; light pink, restriction-modification evasion/regulation of gene expression/replicative transposition/integration; white, hypothetical proteins. For the complete hierarchical clustering map, please refer to Additional file 5: Table S4 Hoeflea, Silicbacter, Leisingera, and Ruegeria (Additional file 2: Table S1). Several lysogenic genes or head structural genes were identified in Citreicella sp. SE45 and Roseobacter denitrificans OCh114 (Additional file 2: Table S1). Thus, genetic mutation, gene acquisition and gene loss might contribute to the diversity of this phage group.