Survey of Thirteen Novel Pseudomonas putida Bacteriophages

Abstract

Bacteriophages have been widely investigated as a promising treatment of food, medical equipment, and humans colonized by antibiotic-resistant bacteria. Phages pose particular interest in combating those bacteria which form biofilms, such as the medically important human pathogen Pseudomonas aeruginosa and several plant pathogens, including P. syringae. In an undergraduate lab course, P. putida was used as the host to isolate novel anti-pseudomonal bacteriophages. Environmental samples of soil and water were collected, and purified phage isolates were obtained. After Illumina sequencing, genomes of these phages were assembled de novo and annotated. Assembled genomes were compared with known genomes in the literature and GenBank to identify taxonomic relations and to refine their functional annotations. The thirteen phages described are sipho-, myo-, and podoviruses in several families of Caudoviricetes, spanning several novel genera, with genomes ranging from 40,000 to 96,000 bp. One phage (DDSR119) is unique and is the first reported P. putida siphovirus. The remaining 12 can be clustered into four distinct groups. Six are highly related to each other and to previously described Autotranscriptaviridae phages: Waldo5, PlaquesPlease, and Laces98 all belong to the Waldovirus genus, whereas Stalingrad, Bosely, and Stamos belong to the Troedvirus genus. Zuri was previously classified as the founding member of a new genus Zurivirus within the family Schitoviridae. Ebordelon and Holyagarpour each represent different species within Zurivirus, whereas Meara is a more distantly related member of the Schitoviridae. Dolphis and Jeremy are similar enough to form a genus but have only a few distant relatives among sequenced phages and are notable for being temperate. We identified the lysis cassettes in all 13 phages, compared tail spike structures, and found auxiliary metabolic genes in several. Studies like these, which isolate and characterize infectious virions, enable the identification of novel proteins and molecular systems and also provide the raw materials for further study, evaluation, and manipulation of phage proteins and their hosts.

1. Introduction

Bacteria of the genus Pseudomonas are ubiquitous in soil, water, and the surfaces of living organisms and inanimate objects [1,2,3,4]. The medically important pseudomonad, P. aeruginosa, causes severe, often fatal infections, primarily in hospital environments [3,5,6,7]. P. syringae poses a serious threat in agriculture, including over 60 pathovars that infect and destroy economically important crops [8]. Both P. aeruginosa and P. syringae are notorious for their ability to form biofilms, which can serve to increase infectivity and allow bacterial cells to resist damage from antibiotics [9,10].

The host used in this discovery-based study, P. putida, is a fast-growing bacterium found mostly in temperate soil and water habitats. It colonizes plant roots and provides benefits to the root microbiome [11]. For many years, P. putida has been used for bioremediation of environmental contaminants [12,13,14,15,16,17]. P. putida possesses a broad spectrum of metabolic pathways and can tolerate a wide range of stressors, making it an organism of interest for practical applications in biomanufacturing and biotechnology [18]. Proposals have included manipulating P. putida populations in the rhizosphere to benefit agriculture [19] and bioremediation strategies such as using P. putida proteins to degrade aromatic pollutants captured via nanoparticles [20,21]. Some of the molecular and genetic tools that could enable and support these types of applications include using bacteriophage-derived depolymerase enzymes to degrade P. putida biofilms [22] and using bacteriophage vectors for delivery of large genetic cassettes into P. putida genomes [17,23,24].

Bacteriophages (“phages”) are among the most genetically diverse organisms on the planet, serving as a reservoir of unique DNA sequences [25]. They are being investigated as possible antibacterial agents to help address the rising problem of antibiotic-resistant bacterial pathogens [26,27,28,29] and have been administered directly to humans in clinical trials and as experimental treatment for individual patients [26,30,31]. Some phage products are marketed for decontamination of food products or management of plant diseases, including plant diseases caused by pseudomonads [23,27,28,32]. In addition, individual phage components such as endolysins are being developed for use as antibacterial drugs [33,34].

Relatively few bacteriophages have been isolated that infect P. putida, especially when compared to human pathogens such as P. aeruginosa [35,36]. The first reported phage to be isolated on P. putida, gh-1, was published in 1966 [37] and the genome sequence reported decades later [38]. A review of the literature prior to 2024 in PubMed and DNA sequences in GenBank yielded only nine publications describing phages of P. putida that also had sequences available through GenBank, and four additional papers describing approximately 30 phages were not accessible. The diversity within this small group of phages is remarkable; finding two or more phages similar enough to constitute a genetically related group was unusual. Only myovirus and podovirus morphologies were seen, all are double-stranded DNA phages, and no temperate phages were reported. As of 14 April 2025, GenBank contained a total of 91 submitted dsDNA genomes from P. putida bacteriophages, of which 72 belong to the broadly T7-like Autographivirales (Supplemental Table S1), mostly from a single study that reported a collection of phages infecting P. putida KT2440 [35]. Although the number of phages substantially increased, the same trends hold true. Further, according to previous reports, P. putida phages infection is mostly limited to P. putida host strains, but there have been reports of infection in P. syringae, and P. fluorescens [39].

Collecting new viruses expands the repertoire of unique P. putida phages, not only for genomic information but also for potential biotechnology applications such as engineering broader host ranges [40], manipulating P. putida populations in the rhizosphere [19], degrading biofilms through phage-encoded depolymerase enzymes [22], and serving as vectors for delivery of large genetic cassettes into P. putida genomes [23]. Importantly, at the time of writing, over 259 prophages have been identified in P. putida isolates, but few have been isolated as free phages [36]. It is unknown how many of these annotated prophages can be induced to yield whole virions. Indeed, it is not known whether these even represent complete phage genomes. Thus, many relevant aspects of their biology cannot be studied [41].

The goal of this study was to isolate and analyze phages using a locally isolated strain of P. putida. Here we describe the discovery and characterization by undergraduates in a research course of thirteen novel P. putida phages, including two temperate phages, the first shown to infect P. putida. We provide in-depth analysis of these phage genomes including comparative genomics between our phages and other published phages, lytic and lysogenic cassettes, spike proteins that may help with degradation of biofilms common for pseudomonads, and auxiliary metabolic genes that would be useful for many environmental remediation applications.

2. Materials and Methods

2.1. Phage Isolation, Culturing, and Imaging

Phages were discovered from random environmental samples (see

Table 1. P. putida phages isolated in this study, ordered by year of isolation and sample type. All environmental samples were collected in Harrisonburg, VA (38′ N, 78′ W) except for PlaquesPlease, which was collected in Nokesville, VA (38′ N, 77′ W).

Phage Name	GenBank Accession	Year Isolated	Source
Ebordelon	PV565042.1	2017	Soil
Holyagarpour	PV565043.1	2017	Soil
Jeremy	PV565041.1	2017	Soil
Laces98	PV477980.1	2017	Soil
Meara	PV565044.1	2017	Soil
Stamos	PV565045.1	2017	Soil
Waldo5	MT711889.1	2017	Soil
Bosely	PV329693.1	2018	Soil
Zuri	MK863032.2	2018	Soil
Dolphis	MT711888.1	2018	Soil
Stalingrad	MT711887.2	2018	Stream water
PlaquesPlease	MT711890.1	2018	Streambed clay
DDSR119	MT663720.1	2019	Tree root soil

for locations and type) through enrichment of extracts with P. putida strain ISAT203. P. putida ISAT203 was sequenced and confirmed to be a novel strain using PacBio sequencing (Supplemental Methods). The closest relative in GenBank is P. putida ISoF (CP072013.1, ref. [42]). Bacterial suspensions for initial phage propagation consisted of approximately ½ cup of soil/water mixed with 50 mL Luria–Bertani (LB) broth (10 g/L peptone, 5 g/L yeast extract, 10 g/L NaCl). After overnight shaking incubation at 37 °C, enriched cultures were sterilized with 0.22 µm filters and 5 μL of filtrate was spotted on a lawn of the host bacteria in 0.7% LB agar. Phage populations were purified by standard SEA-PHAGES methods (https://seaphagesphagediscoveryguide.helpdocsonline.com/home, accessed on 3 February 2021) [43], using SM buffer (100 mM NaCl, 8 mM MgSO₄·7H₂O, 50 mM pH 7.5 Tris-HCl), serially diluting, infecting into culture aliquots and plating with LB top agar. After purification, a single isolated plaque was picked into SM buffer and the sample diluted to a concentration predicted to produce near confluent lysis after infection of the appropriate host. Phage lysates were harvested by flooding confluent plates with SM buffer, held at 4 °C for 4 to 12 h stationary, then collected and filtered. Suspensions of phage were stored at 4 °C in SM buffer.

Phages from lysates were negatively stained with 1% uranyl acetate on formvar-coated copper grids and photographed by a FEI Morgagni 268 transmission electron microscope (TEM) (Hillsboro, OR, USA), Copenhagen, Denmark, at Mary Washington University, Fredericksburg, VA, USA.

2.2. DNA Isolation

DNA was purified from 0.2 micron-filtered lysates of ≥1 × 10⁹ plaque-forming units (PFU)/mL that had been treated with 1 unit/μL of DNAse (Ambion cat. #AM2224, Austin, TX, USA) for 1 h at room temperature to degrade residual bacterial DNA. DNA for PlaquesPlease, Stalingrad, DDSR119, and Dolphis was prepared using a protocol developed at JMU. Briefly, 100 μL of DNA prep solution (10 mM EDTA, 2.5% Ficoll^® 400, 3.3 mM pH 8.0 Tris-HCl, 0.08% SDS) was added to 100 μL of each lysate, and the suspension was mixed and held at room temperature for 10 min. Next, 600 μL of cold 100% isopropanol was added, and the solution was kept on ice for 5 min. The mixture was centrifuged at 4 °C at 13,000 RPM for 30 min, the supernatant was removed, and the pellet was allowed to dry. DNA was resuspended in 100 μL of diH₂O. DNA for Waldo5 and Zuri was prepared with a DNA purification kit (Promega, Madison WI, USA) [44] using a modified procedure as described in the SEA-PHAGES manual (https://seaphagesphagediscoveryguide.helpdocsonline.com/home, accessed on 3 February 2021) [43]. DNA for Laces98, Meara, Stamos, Jeremy, Ebordelon, Bosely, and Holyagarpour was isolated using the Norgen Biotek Phage DNA Isolation kit (cat. #46800, Thoroid, ON, Canada) following the manufacturer’s protocol.

2.3. DNA Sequencing and Assembly

DNA libraries were prepared using Illumina DNA prep fragmentation kit (cat. #20060059, San Diego, CA, USA) with Illumina Nextera DNA Unique Dual index adapters (set D, cat. #20091660, San Diego, CA, USA). DNA libraries were pooled and sequenced using Illumina MiSeq or HiSeq 2000 technology through core facilities located at Admera Health Genome Center (South Plainfield, NJ, USA), the North Carolina State Genomic Sciences Laboratory (Raleigh, NC, USA), or Sandia National Laboratories (Livermore, CA, USA).

Raw reads were processed to remove low quality reads with BBDuk (https://sourceforge.net/projects/bbmap/, accessed on 3 February 2021) using the following parameters: ktrim  =  r, k  =  21, mink  =  11, hdist  =  1. Paired-end reads (150 or 300 bp depending on platform and kit) were assembled into whole genome consensus sequences using Newbler v2.9 [45] or SPAdes v3.12 [46] at ~50X coverage for each assembly. Each genome assembled into a linear sequence, most with defined ends. Genomes were manually examined for coverage and integrity of sequence using Consed v.14 [47], consistent with SEA-PHAGES standards for genome completeness (https://seaphagesphagediscoveryguide.helpdocsonline.com/home, accessed on 3 February 2021) [43].

Short inverted repeat ends were determined by direct inspection of genome sequences. Terminal repeats on phages with defined ends were determined by observing pileups in read alignments, using the PhageTerm program [48], or by comparison to related phages in GenBank.

2.4. Genome Annotation and Other Analyses

Laces98, Stamos, Meara, Holyagarpour, Jeremy, Ebordelon, and Bosely were annotated for ORFs and functions using DRAM-v (version 0.1.2) [49] at KBASE (version 1.2.0) [50]. Additional functional annotation was performed using web based BLASTp (version 2.17.0) [51]. All thirteen phage genomes were run through DRAM-v at KBase to determine AMGs.

Genomes for Zuri, PlaquesPlease, Waldo5, Dolphis, Stalingrad, and DDSR119 were annotated using DNA Master 5 (https://phagesdb.org/DNAMaster/, accessed on 10 January 2019) [52], and PECAAN (https://seaphages.org/media/docs/PECAAN_User_Guide_Dec7_2016.pdf, accessed on 3 February 2021). ORFs were identified using Glimmer [53], GeneMark.hmm [54], and features built into DNA Master 5 and PECAAN, and were refined through homology to ORFs in previously discovered phages using the current version of BLASTn (version 2.17.0). The function of each gene product was classified using homology to known proteins in the current versions of BLASTp, HHpred (version 3.3.0) [55], and Protein Data Bank [56]. Prediction of transmembrane domains (TMDs) with TMpred (version 1.0) [57] and DeepTMHMM (version 1.0.24) [58] was used to refine the calling of protein functions. Using the default parameters, TMHMM (version 2.0) [59] and lipoP (version 1.0) [60] were used to evaluate all the phage coding sequences for potential holin, endolysin, and spanin candidates based on sequence homology, presence of TMDs, and lipoprotein membrane anchor signals. Aragorn was used to identify tRNA genes [61]. Phage genome maps were prepared using web based Proksee and colored according to gene type [62].

2.5. Comparative Genomics

Taxonomic information is based on Master Species List #40v.1, released by the International Committee on Taxonomy of Viruses (ICTV) on 3 March 2025 (https://ictv.global/taxonomy, accessed on 5 April 2025). Initially, each phage’s closest relatives were identified using nucleotide-level BLASTn comparisons in GenBank, with comparisons limited to phages with contractile tails (taxid:2731619), long non-contractile tails (taxid:2731619), and short tails (taxid:2731619). Subsequently, VIRIDIC [63] and ViPTree [64] were run with the default parameters to explore possible evolutionary relationships among our phages and previously described phages. VIRIDIC generates percent intergenomic similarity values for all possible pairwise genome comparisons based on genome identity and coverage. ViPTree builds phylogenetic trees based on genome-wide protein-level similarity generated with tBLASTx [51]. More detailed comparisons and alignments of related genomes were performed using Phamerator [65]. In addition to displaying nucleotide sequence similarities, Phamerator displays protein phamily (group) assignments calculated using PhaMMseqs (version 1.0.4) and can derive shared proteins [66].

2.6. Lifestyle Prediction

Bacphlip was used to predict temperate versus virulent phage lifestyles [67]. Bacphlip is a random forest classifier that was trained primarily on tailed phages and uses a broad array of protein signatures associated with temperate phages to predict lifestyle.

2.7. Protein Structure Predictions

Prediction of secondary structure was performed with the SABLE program (http://sable.cchmc.org/, accessed on 10 March 2025). Prediction of tertiary structure of all available spike proteins in the new collection and several from Pseudomonas, Escherichia coli, and Salmonella was performed using AlphaFold3 (https://alphafold.com/, accessed on 20 January 2025) [68,69], which uses a trained machine-learning architecture to model intra- and inter-protein interactions. Visualizations of these structures were prepared in MolStar (https://molstar.org/, accessed on 6 April 2025) [70], which includes an option to overlay multiple proteins. For pairwise comparisons, we highlighted the amino acids on each protein that correspond to the desired region (C- or N- terminal regions in this study) and superimposed. This rotates one of the two figures to be situated where the highlighted regions are as close as possible without altering the individual structures.

2.8. Phages Added from GenBank

Eight phages related by genome or proteome similarity to some of our thirteen newly isolated phages were added to the analyses described herein, for the purpose of situating ours within new or existing classifications and observing conservation of proteins, lifestyles, and morphology (

Table 2. Previously isolated phages used in our analyses.

Phage Name	GenBank Accession	Bacterial Host	Publication PMID
phi15	NC_015208.1	P. putida	21526174 [71]
Arace01	PP179312.1	P. putida	38833289 [72]
gh-1	NC_004665.1	P. putida	12842620 [38]
Henninger	NC_047922.1	P. sp.	Unpublished
PHB04	NC_047861.1	Bordetella bronchiseptica	30229303 [73]
ST4	OR261032.1	Aeromonas hydrophila	unpublished
Torfinnsbu	PQ464596.1	P. poae	39742975 [74]
Touem01	PP179325.1	P. syringae	38833289 [72]

). When identifying these relatives, we excluded sequences that were only deposited as MAGs (metagenome-assembled sequences) and typically excluded genomes with regions of nucleotide identity covering less than 10% of our query genomes.

2.9. Sequencing and Analysis of P. putida Host

P. putida ISAT203 was grown overnight at 30 °C in LB broth. Genomic DNA was isolated following manufacturers’ protocols using the Pacbio Nanobind CBB kit (PN-301-102-900, Menlo Park, CA, USA). PacBio HiFi DNA library was prepared following the manufacturer’s protocol. In brief, DNA was sheared using COVARIS g-tubes and cleaned up using the SMRTbell cleanup beads (PacBio 102-158-300). The HMW library was prepared using SMRTbell Express Template prep kit 3.2 with SMRTbell Adapters (PacBio) for barcoding. Size selection was performed with AMPure PB beads, and the final library was pooled with other barcoded samples. The HMW DNA library size was measured with an Agilent Technologies (Santa Clara, CA, USA) 4200 Tape Station to be 11.1 kbp. The concentration of the HMW DNA library was 82.6 ng/µL determined by the Qubit Flex fluorometer (Thermo Fisher, Waltham, MA, USA). DNA libraries were run on a single 8 M SMRT cell on a PacBio Sequel IIe.

There were 74 HMW resultant reads from the long-read sequencing effort ranging in size from 1506 bp to 16,217 bp. Whole genome assemblies were attempted with Canu v2.2 (https://github.com/marbl/canu, accessed on 12 April 2025) (genome size = 6.2 m) but failed due to low coverage. To determine the host strain taxonomy, we used BLASTn against the core_nt database for each read larger than 5000 bp [51]. We further analyzed through bowtie alignment of raw reads to the published P. putida ISoF genome (Genbank Accession: GCA_023101365.1) and found a 97.3% alignment rate.

3. Results

3.1. Genome Features and Relatedness to Other Phages

In this study, we report the discovery and analysis of thirteen novel phages that infect P. putida strain ISAT203, a new bacterium isolated from soil in the same geographical region as the phages. “Novel” denotes that there are no identical phage genomes present in the GenBank non-redundant nucleic acid database. Major genomic features are summarized in

Table 3. Genomic features of thirteen novel Pseudomonas putida phages.

Phages *	GenBank Accession	Genome Length (bp)	GC Content	Protein- Coding Genes (tRNA Genes)	Proteins with Putative Functions	Packaged Genome Structure **	Lifestyle
Autotranscriptaviridae
Waldovirus
PlaquesPlease	MT711890.1	41,456	58.07%	51 (0)	30	218 bp DTR	Virulent
Waldo5	MT711889.1	41,195	57.73%	53 (0)	29	219 bp DTR	Virulent
Laces98	PV477980.1	41,870	58.15%	50 (0)	32	219 bp DTR	Virulent
Troedvirus
Bosely	PV329693.1	40,504	57.82%	50 (0)	32	253 bp DTR	Virulent
Stamos	PV565045.1	40,667	58.38%	48 (0)	32	253 bp DTR	Virulent
Stalingrad	MT711887.2	40,723	57.85%	49 (0)	32	253 bp DTR	Virulent
Unclassified Caudoviricetes
Zurivirus
Zuri	NC_049456.2	75,873	53.53%	104 (3)	37	457 bp DTR	Virulent
Ebordelon	PV565042.1	74,043	53.70%	93 (3)	41	466 bp DTR	Virulent
Holyagarpour	PV565043.1	75,462	53.55%	95 (3)	43	467 bp DTR	Virulent
Novel genus related to Zurivirus
Meara	PV565044.1	74,705	53.68%	97 (3)	34	Circularly permuted, TR	Virulent
Novel genus, no close relative
Dolphis	MT711888.1	93,507	63.71%	125 (0)	56	Circularly permuted, TR	Temperate
Jeremy	PV565041.1	95,487	63.53%	126 (0)	54	Circularly permuted, TR	Temperate
Novel genus, no close relative
DDSR119	MT663720.1	52,905	60.39%	79 (2)	34	Circularly permuted, TR	Virulent

* Phages are grouped based on existing or proposed membership in ICTV-recognized taxa. ** DTR = direct terminal repeats; TR = terminally redundant.

and genome maps are provided in Supplemental Figures S1 and S2. While we did not conduct extensive host-range testing of our phages, our preliminary data showed that none of these phages were able to plaque on P. putida strains KT2440, S12, or eight other strains belonging to a different clade of P. putida than ISAT203.

Figure 1A is a dotplot comparison of the thirteen phages that we isolated and characterized in this study. We also used VIRIDIC to calculate the percent intergenomic similarity of all of our phages plus their closest relatives in GenBank. Figure 1B,C correspond to the two related groups of phages boxed in blue in the upper left of the dotplot (the Waldovirus and Troedvirus genera). Waldo5 and PlaquesPlease each represent different ICTV-recognized species within the genus Waldovirus (ICTV MSL#40v.1). Based on the commonly accepted thresholds of 70% and 95% intergenomic similarity for genus and species inclusion, respectively [75], Laces98 belongs to the same species as PlaquesPlease (Figure 1B). Similarly, Stalingrad (Figure 1C) was recently found to represent new species within the Troedvirus genus. The only other ICTV-recognized species within Troedvirus is represented by phage phi15. One of the two closely related phages in our collection, Bosely, belongs to the same species as Stalingrad, whereas Stamos represents a new species within Troedvirus that is distinct from both Stalingrad and phi15. Collectively, all of these phages belong to the Studiervirinae subfamily in the Autotranscriptaviridae family of lytic podoviruses that encode their own single-subunit RNA polymerases.

The four phages boxed in pink in the center of the dotplot are related to Zuri. Zuri is the founding member of the genus Zurivirus (ICTV MSL#36), within the Schitoviridae family of lytic N4-like podoviruses [76]. We propose that Ebordelon and Holyagarpour each represent new species within Zurivirus (Figure 1D). Based on BLASTn comparisons, Zuri is also the closest relative for our novel phage Meara. Only one other complete phage genome in GenBank, Arace01 (Table 2), shares any significant nucleotide-level identity with Zuri across >10% of the genome. Based on a whole-proteome tree, Meara and Arace01 cluster within the Schitoviridae family (Supplemental Figure S3A), and we therefore propose that they each represent a novel genus within that family.

We propose that Dolphis and Jeremy represent two new species within a new genus (Figure 1E, boxed in green in Figure 1A), based on the 95% and 70% intergenomic similarity thresholds described above. No higher-level classification seems possible within the Caudoviricetes due to a dearth of close relatives. In a tBLASTx search of Caudoviricetes genomes, we only identified three phages with any notable protein-level similarity to Dolphis or Jeremy: Bordetella phage PHB04, P. syringae phage Touem01, and Aeromonas phage ST4. Using ViPTree proteome trees, these five phages cluster together, separately from other phages. ViPTree can be useful to define higher-order taxa such as Families and Subfamilies but in this case, the available phages are too few and too distantly related to clearly support such intermediate taxa (Supplemental Figure S3B).

DDSR119 is a singleton, with no close relatives in GenBank. A recent study [72] reported finding a Pseudomonas poae siphovirus called Torfinnsbu (PQ464596.1) that shares 20% intergenomic similarity with DDSR119, mostly involving structural genes.

3.2. Phage Morphology

We obtained electron micrographs of several of our phages. Representative images are shown in Figure 2. Virion dimensions are approximately 78 nm capsid and 170 nm tail for DDSR119 (Figure 2A) and 94 nm capsid and 145 nm tail for the Dolphis-like phages (Figure 2B). For the Waldovirus (Figure 2C), Troedvirus (Figure 2D) and Zuri-like (Figure 2E) podoviruses, capsid diameters are approximately 52 nm, 52 nm, and 69 nm, respectively, correlating with the genome lengths ranging from 40,504 bp to 75,873 bp. The virion morphologies shown in these TEMs are all consistent with expectations based on the taxonomic comparisons described above (e.g., Autrotranscriptoviridae phages have the expected podovirus morphology). DDSR119 appears to be a siphovirus, like Torfinnsbu, and thus would be the first reported P. putida siphovirus.

3.3. Protein Features of our Phages and Selected Relatives Using Phamerator

Having identified the closest relatives of our phages, we conducted more detailed genome-wide comparisons and alignments using Phamerator [65]. The phages are then sorted into clusters based on those phamily assignments. In Figure 3, Figure 4, Figure 5 and Figure 6, DNA sequence similarity was determined using pairwise BLASTn with E-value visualized according to the color spectrum from violet (E-value of 0) to red (E-value of 10⁻⁴). Protein phamilies are shown in color-coded rectangles. Selected proteins with predicted functions are shown above or below the boxed and colored predicted proteins. All other proteins with predicated functions assigned can be seen in the associated GenBank files. Typical in phage genomes, fewer than 40% of genes can be assigned a function, even though they can be organized into Phamilies. Those who cannot be assigned a function with confidence are named “hypothetical protein.” White boxes, “orphams,” represent predicted proteins not shared with any other in the specific database used to generate the Phamerator output. The more unique the phage is, the fewer functions can be defined; note phage DDSR117 in Figure 6. Consistent coloration of proteins across these genomes indicates conserved protein functions, even when nucleic acid sequences may not be conserved; many examples of these are seen in Figure 3, Figure 4, Figure 5 and Figure 6 and are most obvious in comparisons of distantly related phages (Figure 5 and Figure 6). In the image, the ruler represents the DNA and is divided into segments of 100 bp (small ticks) and 1000 bp (numbered ticks), allowing visual estimates of gene sizes.

The results of genome clustering matched the taxonomic groupings that we inferred from genomic data and are grouped this way throughout Figure 3, Figure 4, Figure 5 and Figure 6. In Figure 3, we compared our three Waldovirus phages, our three Troedvirus phages along with phi15, and two phages from other genera in the Autotranscriptaviridae family that were chosen somewhat randomly to help visualize the higher-level relatedness within the family. Overall, 25 phams were shared by all nine phages in Figure 3, encompassing morphogenetic proteins and essential phage enzymes; see examples identified by function name. The Waldovirus phages alone share 43 phams and the Troedvirus phages share 36 phams. Despite each belonging to different genera, phages gh-1 and Henninger are more similar to the three Waldovirus phages than any of them are to the Troedvirus phages, such that these five phages share 29 phams.

Our three Zurivirus phages were highly conserved at the DNA level, whereas Meara, a member of our new collection within Schitoviridae, and Arace01 from GenBank, were highly dissimilar over >50% of the genome (Figure 4). However, many of the proteins (68 phams) were still shared, most notably structural genes, lysis cassette proteins, and genes for nucleotide metabolism. Only one, Holyagarpour, contains an identifiable spike protein gene. All 5 phages in this group each have 2 tape measure genes encoding proteins involved in the length of tails [77]. The larger one (Pham57, 401 AA) is at location ~gp76. The smaller one (Pham43, 109 AA) is at location ~gp82. There is no similarity with these tape measure proteins and those of Waldoviruses or Troedviruses. One possible explanation is that the tape measure proteins are performing some other function than annotated. In E. coli phage HK97, the protein annotated as a tape measure was shown to be involved in the genome injection process [78]. In Mycobacterium phages, tail tape measure proteins may act as signaling molecules that promote growth of dormant bacterial hosts [79].

As described in Section 3.1, Dolphis and Jeremy do not have close relatives within our collection or in GenBank, but we did find three distantly related phages with protein-level similarity: Bordetella phage PHB04, Aeromonas phage ST4, and P. syringae phage Touem01 (Phage ST4 was not included in the comparative genomics studies). The three Pseudomonas phages, Dolphis, Jeremy, and Touem01, share 46 phams. These include some structural genes, as would be expected among related myoviruses, and also include several DNA-interacting proteins distributed across the lengths of all three genomes (see examples in Figure 5A). In contrast, Bordetella phage PHB04, which is more distantly related, shares only 26 phams with Dolphis and Jeremy, and these are limited to a much smaller segment of the genome (Figure 5B). The Jeremy genome contains a 715 bp tract of randomly ordered purine bases at coordinates 70,489–71,204, a highly unusual phenomenon. We confirmed the presence of this sequence by read walking to rule out an assembly artifact. A protein predicted within this region would contain glycine, glutamic acid, lysine, and arginine, similarly in random order. A similar phenomenon has been reported in Bacillus podoviruses, where a 150 bp run of A’s was observed in Bacillus phage Stitch (KX349901.1) [79] and a 100 bp run of purines in Bacillus phage Porch (unpublished data). The significance of these is unknown and nothing else resembling such DNA or protein sequences is found in the GenBank NR database. However, there is one report of Mycobacterium phage Bxz1 having a run of 150 GC pairs (not included in the GenBank entry) [80].

While DDSR119 has no close relatives in GenBank, we compared it to Torfinnsbu, a P. poae phage that was recently reported to share 20% intergenomic similarity based on a VIRIDIC analysis of nucleotide sequence. The DNA similarities and shared proteins are largely clustered in two rations (Figure 6). In our analysis, these two phages share 26 phams, reflecting similar protein sequences for the terminase, some structural proteins, endolysin, spanin, DNA primase and DNA helicase; however, most of the shared genes are hypothetical (Figure 6).

3.4. Phage Lifestyle

Phage lifestyle prediction (virulent vs temperate) can be challenging. The presence of unambiguous integrase and/or repressor genes is reliably predictive of a temperate life history, even if a phage appears to undergo only lytic replication under a particular set of experimental conditions or carries other mutations that prevent lysogenic replication. However, bioinformatic identification of integrases, and especially repressors, is imperfect and therefore the apparent absence of these genes is not a reliable indicator of an obligately lytic lifestyle. In addition, some temperate phages are maintained as extrachromosomal elements in plasmidial or episomal form and do not encode integrases at all [81,82]. We therefore used multiple approaches to predict the lifestyles of these six phages: identification of specific lysogeny-associated genes such as integrases, and proteomic comparisons to well-characterized phages using ViPTree and Bacphlip.

Bacphlip predicted Dolphis and Jeremy to be temperate (p-values = 0.90, 0.837, respectively). We identified tyrosine integrases in both phages: gp42 in Dolphis and gp46 in Jeremy (Figure 5A,B, and Supplemental Figure S1). The two predicted tyrosine integrases are 99.3% similar over 98% of the protein suggesting they are the same integrase. We mapped Jeremy (gp46) to a Phage Integrase pFAM (PF00589; e-value 3.9 × 10⁻⁹), which also contains E. coli phage Lambda and Salmonella phage P22. The predicted protein also contains both the arm DNA binding domain (Jeremy 4–93aa, PFAM13356, e-value 2.7 × 10⁻¹¹) and the integrase catalytic core (Jeremy 222-411aa; e-value 7 × 10⁻²¹) [83]. We also identified highly related integrases in the distantly related phages PHB04 and Tuoem01 (~60% similar over 95% of the protein; Figure 5A,B).

Bacphlip predicted that our other 11 phages are obligately lytic (Zuri p = 1.0, Ebordelon p = 1.0, Holyagarpour p = 0.995, Meara p = 0.975, Waldo5 p = 0.873, PlaquesPlease p = 0.925, Lace98 p = 0.925, Stalingrad p = 0.997, Stamos p = 0.999, Bosely p = 0.997, DDSR119 p = 0.85) and no genes were annotated as integrases. The Bacphlip prediction for Zuri is both highly confident (p-value ≥ 0.95) [67] and consistent with the known lifestyle of other Schitoviridae phages. Bacphlip prediction confidences for Waldo5, PlaquesPlease, and Stalingrad were variable, but again the obligately lytic lifestyle is consistent with what is known about other Autotranscriptaviridae phages. DDSR119 does not have any close characterized relatives, but the Bacphlip results, and the lack of identifiable integrase genes support a tentative prediction that DDSR119 is obligately lytic.

3.5. Lysis Systems

Phages infecting Gram-negative hosts are known to encode three classes of lysis proteins: holins, endolysins, and spanins [84]. Each is responsible for subversion of one of the three layers of the cell envelope during host lysis: inner membrane (IM), peptidoglycan (PG) layer, and outer membrane (OM), respectively. We were able to identify all three lysis elements in our thirteen phages. These are identified in the whole genome maps in Figure 3, Figure 4, Figure 5 and Figure 6.

In Zuri-like phages (Zuri, Ebordelon, Holyagarpour, and Meara), all the lysis genes were grouped together as a lysis cassette (Figure 7A). Zuri encodes a class II holin (“N-in, C-in topology”, where both termini are cytoplasmic) with 2 predicted TMDs, followed by an endolysin with an N-terminal signal anchor release (SAR) domain and an embedded two-component spanin system, where the o-spanin gene is entirely embedded in the +1 reading frame of the i-spanin gene. The equivalent genes were easily identifiable based on sequence similarity in Ebordelon and Holyagarpour. Despite having lower overall sequence similarity, Meara appears to use the same embedded o-spanin arrangement, yielding a strong lipoprotein membrane anchor signal in the predicted protein. Like many embedded or overlapping o-spanins, the genes in these four phages use a non-canonical translation start codon.

Waldo5, PlaquesPlease, Bosely, Stamos, Laces98, and Stalingrad also encode class II holins with two TMDs and embedded two-component spanins, but their endolysins did not seem to contain any N-terminal SAR domains. Furthermore, the endolysin gene was located far upstream from the holin and spanin genes (Figure 7B). The lysis proteins in Waldo5, Laces98, and PlaquesPlease are 100% identical and share substantial sequence identity with those from Stalingrad: holin (45%), endolysin (61%), i-spanin (40%) and o-spanin (51%).

In Dolphis (Figure 7C), we identified gp37 as an endolysin based on sequence similarity. The upstream gene gp36 and downstream gene pair of gp39–gp40 have all the necessary features to be holin and the spanin pair, respectively, but the predicted holin has no sequence similarity to previously annotated holins. Based on our analyses, Jeremy appears to have essentially the same lysis cassette as its close relative, Dolphis. Similarly, in DDSR119 (Figure 7D), gp25 was annotated as the endolysin based on sequence homology. A potential i-spanin/o-spanin pair (gp26 and 26a) could be found immediately downstream with identity of 50% to other Rz-like spanins, but the best holin candidate (gp58) is far away from other lysis proteins and is not similar to other holin sequences.

3.6. Auxiliary Metabolic Genes

AMGs are important components of phage genomes useful for enhancing the host metabolism when integrated or altering the host metabolism when virulent to increase phage infection [85]. We used the standard workflow to predict AMGs in these thirteen phage genomes. We found AMGs in seven of the phages (Supplemental Figures S1 and S2). Dolphis (gp73) and Jeremy (gp76) contain a DNA (cytosine-5-)-methyltransferase which is likely used in epigenetic regulation of the phage genome to prevent host nuclease degradation [86]. Zuri (gp45), DDSR119 (gp31), Meara (gp37), Holyagarpour (gp42), and Ebordelon (gp43) all contain a dCTP deaminase which enhances DNA metabolism of the phage genome.

3.7. Podovirus Tail Spike Proteins

Tail spike proteins in podoviruses play several important roles in the early stages of infection, and extensive studies have been done with the spike proteins of E. coli phage T7 [87] and Salmonella phage P22 [88]. However, structural studies in P. putida are limited, and therefore we are limited to in silico analyses for this study [41]. Using Phamerator, we identified spike proteins in seven of the ten podoviruses in our new collection (the six Autotranscriptaviridae phages and Holyagarpour) and four previously isolated phages (gh-1 [38], phi15 [71], tf [89], and AF [90]) for comparison. The spike proteins from this dataset fall into 3 phams (41, 430, and 273) which are derived from amino acid comparisons and domain identification. Note that the pham numbers are derived within a given Phamerator dataset, so while they differentiate our tail spikes, they are otherwise arbitrary numbers that are not relevant beyond our dataset. We found that spike protein phams correlated with our phage taxonomic clusters, with one exception: Holyagarpour belongs to the Zurivirus genus, but its spike protein is more closely related to the spike proteins from the Waldovirus phages (Pham41), possibly indicating recombination among their ancestral phages. Also, while the spike proteins of singleton AF (protein: YP_007237194.1) [90] and tf (protein: YP_006382521.1) [89] are similar to the annotated spike proteins in our other podoviruses; no other proteins are shared among these phages.

In addition to the primary and secondary structure elements incorporated by Phamerator, we used SABLE to predict secondary structures in the N- and C- termini of several phage tail spike proteins and AlphaFold to predict tertiary structures (Figure 8; Supplemental Figures S4 and S5). In both datasets, we found that the first ~150 amino acids form 7–8 β-strands with an α-helix at the N-terminus. The exception to this was Holyagarpour, where the entire segment was primarily α-helices (and see below regarding tertiary structure). The last ~150 amino acids were not as conserved as the N-termini, but β-strands and β-sheets predominated. At the tertiary structure level, both the N-terminus and C-terminus of the spike proteins appear to be conserved among these P. putida phages regardless of the tail spike pham or the overall phage relatedness (Figure 8). The middle portions of the proteins show variability, but often contain an extended α-helix, a β-barrel, and a region of disorder. Using the overlay capability of MolStar, we conducted pairwise comparisons of tail spike proteins within phams and between phams (Figure 8). For each comparison, we prioritized aligning specific sections of the N- or C-terminus based on apparently conserved secondary structures. In most cases, there was good alignment of the N- and C-termini, even across phams. This is noteworthy because pham assignment is based on whole-protein similarity, not just conserved domains within them. When both termini of the pair could not be simultaneously aligned, this might be due to the flexibility of linker regions rather than some major difference in the structure itself. In addition, the N-terminus of the Holyagarpour tail spike was associated with a poor AlphaFold confidence score, so it is unclear how much of the apparent N-terminal differences in Figure 8D are due to a low-confidence prediction versus actual differences between the tail spike in this Zurivirus phage versus the tail spikes in the more closely related Waldovirus and Troedvirus phages. In Salmonella phage P22, and presumably in other podoviruses with related tail spike proteins, these proteins exist in a homotrimer in the virion (crystal structure shown in Supplemental Figure S4).

4. Discussion

The variety of P. putida phages described in this paper and in Brauer et al. [35] is unusual amongst other large phage collections in which a single bacterial strain has been used to isolate large numbers of phages [35]. Based on our proposed classifications, our thirteen phages include representatives of six different genera across at least four higher-order taxonomic groups, all isolated using a single P. putida strain. Similarly, Brauer et al. isolated 67 phages belonging to 22 groups, also on a single strain. Two examples show the opposite experience. From 2008 to 2011 a locally (Harrisonburg, VA, USA)-isolated Bacillus pumilus host was used for phage isolation. Some 120 phages were isolated, of which 27 were sequenced and all found to belong to one genetic cluster [91]. Also, Bacillus thuringiensis Kurstaki was used as a host in four different universities in Maryland and Virginia, in the United States, and several dozen phages were sequenced over the first five years and are continuing to be isolated. In this case, there is one cluster of B. thuringiensis phages with three highly related subclusters regardless of the year, the location, the type of sample, etc. [92]. While changing phage taxonomy over time makes it difficult to directly compare the number of genetic clusters reported by each study, it still seems clear that these examples from Bacillus phage discovery projects reflect a general theme of repeatedly isolating larger numbers of closely related phages that infect the same strain. This is in stark contrast to the findings of Brauer et al. [35] and to our findings in this work, where we see a wide diversity of phages infecting one strain of P. putida, even on the comparatively small scale of 13 phages. The reasons for this dramatic difference are unknown but could potentially reflect a greater diversity of P. putida phages in the sampled environments. Interestingly, our study used a host strain that is from a different clade of P. putida than the strain used by Brauer et al., so, at a minimum, it does not seem to be a strain-dependent phenomenon within P. putida [35]. In a whole genome BLASTp analysis, none of the Brauer study phages showed detectable similarities to any of the 13 we report here.

Two of the novel phages that we isolated, Dolphis and Jeremy, are predicted to be temperate. It is estimated that ~90% of phages are temperate or capable of undergoing lysogeny [93,94]. While prophages are evident in P. putida bacterial sequences, isolated temperate phages infecting P. putida are rare [95]. One of Dolphis and Jeremy’s distant relatives, PHB04, was previously predicted to be lytic [73]. However, we question this conclusion for several reasons. First, the authors identified a credible integrase gene in PHB04. Second, they concluded that PHB04 is not temperate based on a failure to amplify certain PHB04 genes from the genomes of 40 PHB04-resistant mutants of one particular host strain, which is insufficient to determine that PHB04 is not a temperate phage. Third, an intermediate relative, Touem01, was also predicted to be temperate based on multiple lines of evidence [72]. Therefore, while the creators of Bacphlip recommend that predictions with P-values lower than 95% be treated with caution, and future experimental work to verify them would be valuable, the preponderance of evidence points to Dolphis, Jeremy, and their three relatives (Touem01, ST4, and PHB04) being temperate phages.

The lysis cassettes of our phages have some interesting features. While the four basic types of endolysins (glycosylases, endopeptidases, amidases, and transglycosylases) are sufficiently conserved that they can usually be identified based on nucleotide sequence similarity [96], the holin and spanin sequences are much less well-conserved and are usually identified based on a combination of gene size, position relative to the endolysin, and predicted features such as transmembrane domains in the spanins [96]. In Zuri and its relatives, the lysis proteins are grouped together as a lysis cassette and manual annotation of the embedded i-spanin was straightforward. In all of our T7-like Waldovirus and Troedvirus phages, both the primary amino acid sequences and gene spacing (i.e., substantial genome distance between the endolysin gene and the other lysis genes) are similar to what is seen in phage T7. Interestingly, the T7 lysozyme has also been shown to play a role in regulating the balance between phage gene transcription and phage replication and packaging, which may explain its location in the genome among nucleic acid-processing enzymes, and its distance from the other lysis proteins [97]. In Dolphis and Jeremy, we proposed holins and spanins based on gene location, protein size, and some predicted secondary structures, but they have little or no sequence homology to known lysis proteins. If their function can be experimentally confirmed, their novel sequences should facilitate the rapid discovery of holins and spanins in other phages. In DDSR119, we note that the lysis gene organization is similar to that of some Pseudomonas pyocins. Specifically, when the DDSR119 genome is viewed in the circularly permuted form that we predict, the holin gene is located upstream of the endolysin and spanin genes, separated by the tail morphogenesis genes. This is analogous to the gene order of R-type pyocins in several Pseudomonas species, including P. putida [98]. The R-type pyocins are also called tailocins for their resemblance to myovirus tails, whereas DDSR119 has a siphovirus morphology. The R-type pyocins are also thought to be evolutionarily related to temperate phages [99] whereas we tentatively predict DDSR119 to be lytic. The similarity of gene order to tailocins could be coincidental, but with DDSR119 having no close known relatives among either lytic or temperate phages, it remains an intriguing observation.

When considering the fact that all thirteen of our phages infect the same host strain (and that they do not infect P. putida KT2440 or S12 strains) we wondered whether this might correlate with any obvious similarities in phage proteins that are typically associated with receptor binding. For example, if the bacterial receptor and the phage receptor binding proteins are conserved, this observation could be explained. The conserved lipopolysaccharide in Gram-negative bacteria has been shown to be the receptor for many phages [100], and specifically P. putida phages [35]. However, we could only identify tail spike proteins in some of our podoviruses (primarily among the T7-like phages) and even these were not highly conserved at the amino acid sequence level, nor did they cluster in a single pham, despite the fact that phams typically incorporate more predictive information than just primary amino acid sequence. Nevertheless, our AlphaFold predictions suggest that the conserved tertiary structure in P. putida spike proteins among even distantly related phages could play a role in the diversity of pairings in the genus and species of bacterial hosts. Both the N- and C-termini of spike proteins have attachment and enzymatic function, based on varying functional experimental data [101,102,103]. Bacteriophage hosts often have primary and secondary attachment sites, so both spike protein termini might be active simultaneously or in sequence,¹¹¹ but this would need to be tested in future work. It has also been hypothesized that the tertiary structure of these proteins would allow the N- and C-termini to come together physically to exhibit multiple functions [104]. The AlphaFold-predicted structure shows that the flexible and varied middle section of these proteins might allow such a conformational change.

All Pseudomonads are able to form substantial biofilms that protect them from antibiotics, environmental chemicals, and infection by bacteriophages. In the search for novel anti-biofilm agents that might be used in conjunction with antibiotics in multidrug-resistant infections, there is a rising interest in the biofilm-degrading properties of phages. A number of studies have suggested that spike proteins have such activity, using genetic manipulation and mutant analyses [22,41]. Building on the bioinformatic analysis of the spike proteins presented here, future molecular experiments may elucidate specific mechanisms that could be used in biofilm remediation studies.

The phages explored in this study were isolated and characterized in an undergraduate laboratory course, one of some 200 similar courses across the US and abroad (https://seaphages.org/, accessed on 3 February 2021). The extraordinary educational benefits of this combined educational and discovery-based model have been well characterized; see work by Hanauer et al. [105,106]. At James Madison University, a two-semester sequence, phage discovery and genomic analysis, is taken mostly by 1st or 2nd year students and open to any major, with no prerequisites. Our students have isolated upwards of thirteen hundred phages since 2009, using members of Mycobacterium, Microbacterium, Bacillus, and Pseudomonas genera as hosts. Other examples of how these research experiences are offered at different types of institutions and resources for getting started are included in the Supplemental Materials, “Resources and comments on using phage discovery and genomics in undergraduate settings.”

The sheer number of phages being collected has exploded and bacterial hosts expanded over the last 15 years, creating a large reservoir of phages with potential to address problems facing medicine, agriculture, and industry. In just the one study described herein, our students discovered over two dozen phages that infect the Pseudomonas genus, including the ones described here. Collectively, thousands of phages have been added to existing collections isolated as free phages, where the functional viral particles can be used for basic biology and biotechnology applications.