Virus Evolution, 2020, 6(1): veaa030

doi: 10.1093/ve/veaa030
Research article

UPU phages, a new group of filamentous phages found

in several members of Enterobacteriales

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

Jason W. Shapiro1,*,† and Catherine Putonti1,2,3,4
1
Department of Biology, Loyola University Chicago, 1032 W Sheridan Rd, Chicago, IL 60660, USA, 2Department
of Computer Science, Loyola University Chicago, 1052 W Loyola Ave, Chicago, IL, 60626, USA, 3Bioinformatics
Program, Loyola University Chicago, 1052 W Loyola Ave, Chicago, IL 60626, USA and 4Department of
Microbiology and Immunology, Stritch School of Medicine, Loyola University Chicago, 2160 S First Ave,
Maywood, IL 60153, USA
*Corresponding author: E-mail: [email protected]
†
https://orcid.org/0000-0002-1321-8418

Abstract
Filamentous phages establish chronic infections in their bacterial hosts, and new phages are secreted by infected bacteria
for multiple generations, typically without causing host death. Often, these viruses integrate in their host’s genome by
co-opting the host’s XerCD recombinase system. In several cases, these viruses also encode genes that increase bacterial
virulence in plants and animals. Here, we describe a new filamentous phage, UP/901, which we originally found inte-
grated in a clinical isolate of Escherichia coli from urine. UP/901 and closely related phages can be found in published
genomes of over 200 other bacteria, including strains of Citrobacter koseri, Salmonella enterica, Yersinia enterocolitica, and
Klebsiella pneumoniae. Its closest relatives are consistently found in urine or in the blood and feces of patients with urinary
tract infections. More distant relatives can be found in isolates from other environments, including sewage, water, soil,
and contaminated food. Each of these phages, which we collectively call ‘UP/ viruses’, also harbors two or more novel
genes of unknown function.

Key words: bacteriophage; inovirus; prophage; bladder.

1. Introduction Filamentous phages in the family Inoviridae go beyond the

standard dichotomy of viral lysis and lysogeny. Instead, the ma-
Phages are often described by their potential to kill their hosts.
jority of characterized inoviruses maintain productive infec-
Obligately lytic phages kill their hosts following infection, whereas
tions over multiple bacterial generations, without killing their
temperate phages may lie dormant as prophages within lysogen-
hosts. In many cases, these phages integrate as tandem repeats
ized bacteria for several generations before entering a lytic cycle.
(e.g. Derbise et al. 2007) into their hosts’ genomes at a locus
While killing by phages has immediate consequences for bacterial
called the dif site (Mai-Prochnow et al. 2015). The dif site is a
ecology and has led to the revival of phage therapy for treating
28 bp region at the bacterial terminus that includes two short
bacterial infections (Kortright et al. 2019), many phages also carry
palindromic regions recognized separately by the site-specific
genes that alter bacterial behavior (Mai-Prochnow et al. 2015;
recombinase pair XerC and XerD. In bacteria with this system,
Warwick-Dugdale et al. 2019). The effects of these phage-encoded
XerCD is responsible for resolving chromosome dimers during
genes range from modifying photosynthesis in cyanobacteria
(Sieradzki et al. 2019) to producing toxins in potential pathogens DNA replication and cell division (Carnoy and Roten 2009).
(Waldor and Mekalanos 1996). Filamentous phages co-opt this system by carrying their own
copy of the dif site within their genomes, causing XerCD to

C The Author(s) 2020. Published by Oxford University Press.

V
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/),
which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

1
 2 | Virus Evolution, 2020, Vol. 6, No. 1

confuse phage DNA for bacterial DNA. After integration, new Table 1. Bacterial strains in this study.
phages are produced at relatively low rates and may be main-
Strain ID Species Has UP/? WGS accession
tained indefinitely within the bacterial population (Val et al.
2005). Other filamentous phages that lack a dif site can also inte- UMB0731 E. coli No NZ_RRWP00000000a
grate elsewhere in the bacterial genome by encoding their own UMB0901 E. coli Yes NZ_PKHH00000000
integrases, as in the Pf bacteriophages of Pseudomonas aeruginosa UMB1220 E. coli Yes NZ_RRVZ00000000
(Mai-Prochnow et al. 2015). UMB1526 E. coli Yes NZ_RRVH00000000
Filamentous phages can have dramatic effects on their UMB5814 E. coli Yes NZ_RRUX00000000
hosts’ phenotypes. In Vibrio cholerae, the phage CTX/ encodes UMB6655 E. coli No NZ_RRUL00000000
the toxin genes responsible for bacterial virulence in humans UMB6890 E. coli No NZ_RRUI00000000
(Waldor and Mekalanos 1996). Similarly, filamentous phages UMB1389 C. koseri No Unsequencedb
have been associated with the virulence of other human patho- UMB7451 C. koseri Yes Unsequenced
gens including hemorrhagic Escherichia coli, Yersinia pestis (the UMB8248 C. koseri No Unsequenced
JE-1 E. coli No Unavailable
cause of plague), and P. aeruginosa in both patients with cystic fi-

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

brosis and skin wounds (Gonzalez et al. 2002; Derbise et al. 2007; a
UMB0901 was sequenced as ‘E75’ in its original BioProject (PRJNA316969).
Sweere et al. 2019). Inoviruses have also been correlated with b
Citrobacter strains not sequenced at time of writing. Phage identified by PCR.
the increased virulence of agricultural pests like tomato wilt
and potato blight (Kamiunten and Wakimoto 1982; Yamada
2013). Across these various infections, filamentous phages may sequenced. Escherichia coli JE-1 is an IncI plasmid-bearing strain
carry toxin genes, alter bacterial motility (Jian, Xiao, and Wang typically used to propagate phage I2-2 and was obtained from
2013) and biofilm formation (Rice et al. 2009; May, Tsuruta, and the Felix d’Herelle Reference Center for Bacterial Viruses
Okabe 2011), or indirectly interfere with the immune system’s (Université Laval, QC, Canada).
ability to clear the bacterial infection (Sweere et al. 2019).
Despite their myriad effects on a wide range of clinically and ag- 2.2 Identifying phage in bacterial genomes
riculturally important hosts, these phages remain under-
UP/901 was originally discovered as part of predicting phage
studied. As of this writing, there are only forty-six recognized
sequences using PHASTER (Arndt et al. 2016) in UMB0901 and
members of Inoviridae in RefSeq, and only thirty-three with
other E. coli isolates in a previous study (Miller-Ensminger et al.
established taxonomy by the International Committee on
2018; Garretto 2019). (The ‘UP’ in its name originated uncrea-
Taxonomy of Viruses (ICTV). Nonetheless, recent work has
tively from ‘urine project’, though not all related viruses are
shown that thousands of potential inoviruses can be found as
necessarily found in urine, as described in Section 3.) To identify
prophages in published bacterial genomes and metagenomes
UP/901 in other published genomes, we took advantage of its
(Roux et al. 2019).
integration at its hosts’ dif sites as a tandem repeat. We defined
Here, we describe a novel filamentous phage, UP/901, dis-
a preliminary BLAST (Altschul et al. 1990) query sequence for
covered as a prophage in a clinical E. coli isolate from a patient
UP/901 as the region starting with the dif locus in UMB0901 and
urine sample. UP/901 is most closely related to the non-
extending to the first phage repeat (7,560 bp). Throughout this
integrating phages I2-2 and IKe in the Lineavirus genus of the
paper, we will refer to the specific sequence (and homologs that
Inoviridae. Like many filamentous phages, UP/901 integrates as
are over 99% identical by amino acid sequence) as ‘UP/901’.
a tandem repeat at the dif site in its hosts. Using the phage se-
Remaining phages that share a common set of core genes (iden-
quence from our isolate as a query, we searched for close rela-
tified below) and with less than 99 per cent amino acid identity
tives of UP/901 in published bacterial genomes and assemblies.
are collectively referred to as ‘UP/ viruses’. The dif site itself is
We found homologous prophages in over 200 strains of E. coli,
not repeated in full within the tandem duplications. We then
Citrobacter koseri, Klebsiella pneumoniae, Salmonella enterica, and
used this single copy of the phage as a query for a blastn search
Yersinia enterocolitica. The most similar phages tended to come
using the NCBI web tool (https://blast.ncbi.nlm.nih.gov) with de-
from patient urine samples or from the blood or feces of
fault parameters. This BLAST search identified prophages with
patients with urinary tract infection (UTI); more distant rela-
over 99 per cent query coverage and over 98 per cent nucleotide
tives could be found in soil, water, animal feces, and contami-
identity to UP/901 in strains of C. koseri, E. coli, K. pneumoniae, S.
nated food. We refer to the collection of related filamentous
enterica, and Y. enterocolitica. Notably, neither PHASTER nor
phages as ‘UP/ viruses’. In several cases, identical UP/ strains
VirSorter (Roux et al. 2015) consistently predicted UP/901
were found infecting multiple bacterial genera, suggesting re-
homologs in bacteria that had significant BLAST hits. This high
cent host switching events. Last, we describe the prevalence of
false-negative rate for these two tools is likely due to UP/901’s
putative UP/ viruses in metagenomes.
short genome and the reliance of these tools on information
about known viruses. We have not tested the newer tool
2. Methods Inovirus_detector (Roux et al. 2019) with genomes carrying
2.1 Bacterial strains UP/901 homologs.
We next downloaded all available draft and complete as-
Clinical isolates of E. coli and C. koseri (see Table 1) were provided semblies on NCBI for the five bacterial species listed above (as
by the Wolfe Lab at Loyola University Chicago and were origi- of August 2018). Each assembly was queried locally with tblastx
nally isolated as part of separate IRB-approved studies on the with default parameters to identify contigs containing putative
urinary microbiota of women with and without symptoms of phage regions. We retained any contigs containing a hit cover-
UTIs or other urinary ailments (Price et al. 2016a,b; Garretto ing at least 1,000 nucleotides of the UP/901 query with over 75
2019; Penckofer et al. 2020). These strains and their GenBank per cent nucleotide identity. In no case did we observe a puta-
accessions (where available) are summarized in Table 1. tive UP/901 homolog on two contigs in the same assembly. The
Citrobacter koseri strains in this study have not been fully start positions of each phage were then confirmed by a separate
 J. W. Shapiro and C. Putonti | 3

local blastn search for the dif site (as identified in Carnoy and shared among each host that also contained a prophage in the
Roten 2009), and stop positions were identified by confirming virus phylogeny. These genes were aligned within Anvi’o using
the first repeat in the BLAST results. These repeats typically MUSCLE (Edgar 2004) and concatenated. We used IQ-TREE as
matched just the first fifteen bases in the dif site, and we ac- above to build the final phylogeny. For the host tree,
cepted putative hits that covered at least fourteen of the ModelFinder identified JTTþFþR10 (Jones–Taylor–Thornton
twenty-eight dif site nucleotides with at least 85 per cent iden- (Jones, Taylor, and Thornton 1992) with empirical state frequen-
tity. Due to varying assembly quality, not all assemblies in- cies and ten rate categories) as the best substitution model by
cluded phage repeats, and some sequences did not include a dif BIC.
repeat as an obvious stopping point. In these instances, we Trees were visualized using iTOL (Letunic and Bork 2019)
could only rely on the end position of the last significant hit in and the ape package (Paradis and Schliep 2019) in R (R Core
the initial tblastx query. Team 2013). Additional metadata for tree visualization were
To confirm the sensitivity of our UP/901 detection, we then obtained from NCBI, EBI, or from the literature. Key metadata
blasted each of the 229 putative UP/901-infected genomes against included the type of material sampled for isolating bacteria (e.g.
a query for the related phage, Ypf/ (from Y. pestis CO92, GenBank: soil, blood, urine, feces) and the more general source of that ma-

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

CP009973), which has been found at dif sites in other studies terial (e.g. environmental, animal, human, food). The full meta-
(Derbise et al. 2007). We liberally accepted all putative hits, regard- data associated with each sample is provided in the data
less of identity, and found a single case (GCA_001519645) with a repository (see Data availability). Nine strains had no available
cumulative query coverage over 1,000 bp. We queried both UP/901 metadata from a publication or database. Twenty-two other
and Ypf/ in SRA BLAST searches against the original reads for strains without available metadata were from the ‘100K
this assembly (SRX1528813). Pathogen Genomes Project’ (BioProject PRJNA186441; Weimer
2017) and are identified as ‘100K Project’ in figures.
2.3 Phylogenetic analyses
We used Anvi’o (Eren et al. 2015) to facilitate gene clustering 2.4 Phage in SRA metagenomes
and annotation for the putative phage regions found in We used searchsra (Levi et al. 2018; www.searchsra.org) to
GenBank assemblies. Standard annotation tools fail to identify check for UP/901 relatives in metagenomes. We used a trun-
inovirus genes accurately, so we then used blastp through the cated 6,703 bp UP/901 reference that included only the core UP/
NCBI web tool with individual phage IKe (GenBank: NC_002014) virus genes used in phylogenetic analyses. The results from the
and I2-2 (GenBank: NC_001332) genes to confirm each gene’s search contained over 100,000 potential hits, most of which are
correct annotation. These were identified by visually comparing false positives returned from a small number of erroneous reads
the aligned genes rather than relying on a significance thresh- aligning to the reference genome. Following the searchsra
old, as homologous sequences may share less than 30 per cent authors’ GitHub repository (https://github.com/linsalrob/
amino acid identity in these phages. Full-length UP/901 and SearchSRA), we first identified the depth of coverage across our
most of the putative phage regions have every gene found in reference genome for each potential hit. We used the pileup.sh
IKe and I2-2 (as in Fig. 2). function from BBMap (Bushnell 2014) to perform this step. We
We identified nine UP/901 core genes to use for phyloge- filtered results by both the query coverage and smoothness of
netic inference (genes I, II, III, IV, V, VI, VIII, h1, and h2). Genes VII coverage (Aziz et al. 2015) to identify true positives. A detailed
and IX were excluded from analyses, because they consist of description of these filtering steps with examples of true and
only about thirty amino acids and were not consistently identi- false-positive coverage plots is provided in the Supplementary
fied as genes. Gene X was also excluded as it is contained en- Methods. Metadata for the likeliest true-positive results
tirely within gene II. Next, we used MAFFT (Katoh and Standley are available in Supplementary File S1 and as part of the data re-
2013) to align each of the core genes. In reviewing individual pository. Statistical analysis related to these metadata was
gene alignments, we identified cases in genes I, II, III, and IV done in R.
where the original gene calls by Anvi’o split the open reading
frame (ORF) into two pieces because of early stop codons. It is
2.5 Phage presence in culture medium
unknown if these genes are truly pseudogenized or if they
might still encode functional proteins. For the purpose of build- UP/901-infected strains were grown in lysogeny broth (LB) at
ing the trees, we concatenated the two halves of these genes, 37  C with moderate shaking overnight. The following morning,
since their complete sequences reflect evolutionary relation- 1 ml of each culture was removed, centrifuged at 16,000 g for
ships, even if they might not be translated in full. Both the origi- 1 min, and the supernatants were filtered through 0.2-lm cellu-
nal and concatenated versions of these sequences are provided lose acetate syringe filters. An 80 ll sample of each filtrate was
in the data repository for this work. then treated with OPTIZYME DNase I (Fisher BioReagents
We then built a phylogeny of the 229 complete prophages with BP81071) for 30 min, followed by heat inactivation with EDTA at
IQ-TREE version 1.6.12 (Nguyen et al. 2015) using the concatenated 65  C for 10 min. This DNase step was included to remove any
alignment of each core UP/901 gene amino acid sequence with bacterial genomic DNA from the supernatant, which could re-
1,000 bootstraps. We took advantage of IQ-TREE’s integrated sult from cell death due to lysis from other prophages or shear-
ModelFinder (Kalyaanamoorthy et al. 2017) option to perform ing forces. DNase-treated samples were incubated at 95  C for
model selection. This step identified VTþFþR2 (variable time with 10 min to denature phage protein coats and expose the phage
empirical state frequencies and two rate categories) as the best ssDNA. We then amplified phage genomic DNA using UP/901-
substitution model according to Bayesian Information Criterion specific primers, UPphi_shortFW (GGGTTTATCAGAGGGGTCAG)
(BIC). Individual gene trees shown in Supplementary Fig. S1 were and UPphi_shortRV (AGGATGGCTCTAAGTCAACG). 16S PCR
built using FastTree (Price, Dehal, and Arkin 2010). (63 F/1387R primers) was used to confirm the absence of bacte-
We also generated a bacterial phylogeny for the phage hosts, rial genomic DNA in the DNase-treated filtrates. The UP/901
using Anvi’o to identify a set of 400 single-copy core genes and 16S PCRs were performed on 1 ll of unfiltered culture as
 4 | Virus Evolution, 2020, Vol. 6, No. 1

positive controls. Positive UP/ PCR products were verified by F-specific filamentous phage, M13. We also compared UP/901
Sanger sequencing using the UPphi_shortFW primer. to Ypf/, a dif-integrating filamentous phage that infects patho-
genic strains of E. coli and Y. pestis. Overall, Ypf/ shares less
2.6 Phage infection assays than 30 per cent amino acid identity with UP/901.
Figure 2 shows how UP/901’s genome compares with these
We tested the potential for UP/901 to infect new hosts using
other inoviruses, with each gene shaded according to its amino
standard plaque assays by spotting 5 ll of filtrate from UMB0901
acid similarity with the homolog in UP/901. UP/901 contains
(as described above) on 0.7 per cent agar LB overlays containing
each of the ‘core’ genes characteristic of filamentous coliphages
candidate bacterial hosts (UMB0731, UMB6655, UMB6890, JE-1).
with gene order preserved. UP/901 and Ypf/ are less than 20
We also tested for new infections using PCR. Colonies of pro-
per cent identical in all genes except for genes II and V, which
spective host strains were added to 5 ml of LB supplemented
are each over 85 per cent identical between these phages. Genes
with 50 ll of filtered supernatant from UMB0901. Uninfected
controls of each strain were also grown. These cultures were in- II and V are both involved in regulating phage DNA replication,
cubated for 18 h. Following growth, 2 ll of each culture were and this conserved module is less than 15 per cent similar in

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

used in a PCR designed to test for integration into the new IKe and M13.
host’s chromosome. The forward primer (IntCheck_FW: Two other differences set UP/901 apart from the non-
GTGTGTGGATGTGAATGGTG) in the assay was based on a integrating phages, I2-2, IKe, and M13: 1, UP/901 includes a
conserved sequence in all tested hosts just upstream of the dif copy of the E. coli dif site at the start of its genome, enabling it to
site, and the reverse primer (IntCheck_RV: CTGGCAGAACG integrate via the XerCD recombinase system; 2, UP/901 encodes
AACGATTAC) recognized a region early in the phage genome. two putative genes of unknown function within a region that is
This PCR can only amplify DNA if UP/901 integrates into a new intergenic in I2-2, IKe, and M13. These genes are not homolo-
genome. An overnight culture of UMB0901 was used as a posi- gous to any known protein, and all attempts to find annotated
tive control for each reaction, and the uninfected cultures were homologs of these genes in GenBank returned hits to hypotheti-
used as negative controls. cal proteins in prophages related to UP/901. In addition to these
hypothetical proteins, the UP/901 prophage region also con-
tains three ORFs in reverse orientation just downstream of the
3. Results
dif site. This genome organization is mirrored by Ypf/ and other
3.1 Initial identification integrating filamentous phages, including CTX/, where acces-
sory genes often surround the core functions (Derbise et al.
We originally identified UP/901 in a clinical isolate of E. coli
2007; Mai-Prochnow et al. 2015).
(UMB0901) as part of separate work examining prophages pre-
sent in the urinary microbiome (Garretto 2019). We then found
prophages with over 99 per cent nucleotide identity in three 3.3 Comparative genomics
other patient isolates in our lab collection (UMB1220, UMB1526, UP/901 strains share nearly 100 per cent amino acid identity
and UMB5814). In each case, we confirmed by PCR that the across infected clinical isolates of E. coli in our lab collection.
phage is shed from infected bacteria, as centrifuged culture su- The only distinction is that the attachment protein, g3p
pernatant treated with DNase is positive for UP/901 DNA
(encoded by gene III), in UP/901 has an additional repeat of a
(Fig. 1A) but negative for bacterial genomic DNA (Fig. 1B). Sanger
glycine-rich motif (‘GGGES’) than the viruses found in UMB1220,
sequencing confirmed that each PCR product was a true match
UMB1526, and UMB5814. As with many other filamentous
to the UP/901 genome.
phages, UP/901 is integrated as a tandem repeat at the dif site
in each of these genomes. In the case of UMB0901, we re-
3.2 Comparison to characterized inoviruses assessed our own assembly by aligning the raw reads to the
UP/901 is approximately 50 per cent identical to phages IKe and contig with UP/901 and estimating the depth of coverage
I2-2 (70% coverage and 70% amino acid identity for covered (Supplementary Fig. S2). On average, the UP/901 prophage re-
genes), each a non-integrating member of the Lineavirus genus gion has three times the coverage of the surrounding loci, indi-
of filamentous phages. These phages are closely related to the cating a tandem triple in UMB0901.

Figure 1. Phage DNA confirmation. PCR results using (A) UP/ or (B) 16S primers after DNase treatment to remove bacterial genomic DNA from filtrates of UMB0901,
UMB1220, UMB1526, and UMB5814 cultures. The þ control lane was from UMB0901 colony PCR.
 J. W. Shapiro and C. Putonti | 5

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

Figure 2. Comparison of UP/901 to related inoviruses. Roman numerals are the names for core inovirus genes common in M13 and the Lineavirus members. Ypf/ genes
labeled as in Derbise et al. (2007). Genes shaded gray are darker if they are more similar to the sequence from the UP/901 prophage in UMB0901.

Table 2. Prevalence of UP/ in GenBank assemblies. We then built a phylogeny for the 229 complete UP/ pro-
phages. Overall, the phylogeny reflects low genetic variation
Host species GenBank assemblies Assemblies with UP/ hit
across the UP/ viruses, and most clades are separated by short
E. coli 12,401 189 branches and include polytomies. Even the more distant phage
S. enterica 9,144 95 relatives found in Y. enterocolitica are 98 per cent identical in
K. pneumoniae 4,647 29 amino acid sequence to the original UP/901 core genes. As a re-
Y. enterocolitica 177 10 sult, the tree has multiple splits with bootstrap support below
C. koseri 28 8 50 per cent. For this reason, we show the tree as a cladogram in
Fig. 3A and will not attempt to over-interpret the tree topology.
(A high-resolution phylogeny with complete bootstrap supports
Using the UP/901 sequence as a BLAST query, we searched is shown in Supplementary Fig. S3.) For each of the genomes in
GenBank for homologous prophages in other bacteria. Prophages the phylogeny, we identified metadata, where available, for the
with over 98 per cent nucleotide identity across the UP/901 query sample source material (e.g. urine, feces, blood) and source en-
sequence could be found in published genomes of E. coli, S. enter- vironment (e.g. human, animal, environmental) and added
ica, C. koseri, K. pneumoniae, and Y. enterocolitica. We then down- these data to the tree visualization.
loaded all available assemblies from GenBank of these five species UP/901 is indicated by an arrow in Fig. 3A and is part of a
(26,397 in total) and used tblastx to identify viruses integrated at clade of sixty-five phages. Fifty-five of these prophages come
dif sites. In all, 331 genomes harbored a potential UP/901 pro- from E. coli, of which thirty-six came from urine (including
phage (see Table 2 for a summary by host). Of these 331 putative fourteen UTI samples). Two large polytomies make up the ma-
phages, 229 contained each gene found in UP/901, excluding the jority of this group, and at least one sample (assembly
three ORFs neighboring the dif site. The third of these ORFs (la- GCA_000457405) was from the blood of a patient with UTI-
beled h3 in Fig. 2) is present in 221 of these 229 putative prophages, induced bacteremia. In addition, each polytomy includes cases
whereas h4 and h5 are found in only 56 and 57 genomes, respec- where identical phages (i.e. 100% amino acid identity across
tively. For the remainder of this paper, we will refer to this collec- the genes used in the phylogeny) were also found within C.
tion of 229 strains as ‘UP/ viruses’. The 102 remaining genomes koseri and K. pneumoniae. This group contains three samples
contained a significant hit to UP/901, but assembly or sequencing from animals. The rest of the tree includes seventy prophages
quality was inadequate to predict full phage genomes reliably. We found in S. enterica (from both animals and humans), though
also checked each of the 229 UP/-infected genomes for Ypf/ to phages infecting each of the other host species are also pre-
rule out possible false-positive results. We found one instance (E. sent. None of these S. enterica samples were isolated from
coli assembly GCA_001519645) of coinfection between a dif-inte- urine, and only ten of the remaining ninety-four sequences
grated UP/ prophage and a putative Ypf/ prophage integrated at came from urine or UTI samples. The total counts for each
a different locus. None of the remaining 228 genomes harbored a host species, sample material, and source are indicated in the
prophage resembling Ypf/. legend for Fig. 3.
 6 | Virus Evolution, 2020, Vol. 6, No. 1

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

Figure 3. Cladograms of UP/ prophages in GenBank assemblies (A) and their hosts (B). The outer ring is the host species, the middle ring is the sample material, and
the inner ring is the source environment. Black arrows indicate (A) UP/901 and (B) UMB0901. Samples sizes shown in parentheses in the legend. Trees are based on
amino acid sequences of core genes. Bootstrap support is shown by the size of black dots with at least 51 per cent support. High-resolution trees with scale bars and
bootstrap supports are in Supplementary Fig. S3.

We next constructed a phylogeny for the host bacteria, using breakdown by host species, with additional structure corre-
a set of conserved single-copy genes identified with Anvi’o sponding to sample material and whether the strain was from
(Eren et al. 2015). The bacterial tree shows the expected human or animal sources. Branch supports are generally higher
 J. W. Shapiro and C. Putonti | 7

twenty-four of the sixty-four unique gut samples came from

nine different studies of the infant microbiome and included
children from China, Estonia, Finland, Singapore, and the USA.
One hypothesis for this high number of infant samples carrying
UP/-infected bacteria is that UP/ may be more common in
hospital-acquired Proteobacteria.
Several filamentous phages have been associated with path-
ogen virulence (Mai-Prochnow et al. 2015), including one study
identifying their potential involvement in cases of neonatal
meningitis caused by E. coli O18:K1:H7 (Gonzalez et al. 2002). We
were interested to see if any of the infant microbiome studies
identified through searchsra might shed light on potential con-
nections between UP/ viruses and bacterial virulence.

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

In one study (Ward et al. 2016), the authors sequenced the
gut microbiomes of 144 pre-term and twenty-two full-term
infants and identified cases where necrotizing enterocolitis
(NEC) was associated with strains of uropathogenic E. coli.
Furthermore, the anonymized patient metadata were available
with the publication and could be matched to sample metadata
from the SRA. This patient data included indicator variables for
high (over one or ten) per cent abundance of E. coli and K. pneu-
Figure 4. Number of unique amino acid sequence variants for each UP/ gene
used to build trees.
moniae, as well as information on patient outcomes. We identi-
fied nine patients in this study that carried bacteria infected by
a putative UP/ prophage (patient IDs: 11211, 11961, 11962,
in the bacterial tree except for clades with very short branches.
12321, 21461, 30031, 30081, 30251, 30252). Of these, six contained
Due to the low branch support of the phage tree, it is difficult to
E. coli and five contained K. pneumoniae; two patients had both
compare the phage and host tree topologies.
bacteria present. The overall rate of UP/ presence (9/166) is
As expected from the short branches of the phage tree, indi-
about four times higher than across GenBank assemblies of E.
vidual phage gene trees (Supplementary Fig. S1) show little vari-
coli and K. pneumoniae (218/17048). This difference represents a
ation in amino acid sequence across the phages, with most
significant enrichment (v2 ¼ 170.44, P < 0.001) in this study.
genes composed of a small number of unique variants (Fig. 4).
Moreover, two of the nine patients were ultimately diagnosed
Gene III stands out as having the most variation, with ninety-
with NEC. This is not significantly different (v2 ¼ 0.0011,
five distinct amino acid sequences among the 229 genomes.
Examining the alignment for gene III, most of the variation is in P ¼ 0.973) from the rate of NEC without UP/ in the study (25/
the number of glycine-rich repeats in the region that also varied 157). In addition, four of the patients (IDs starting with ‘30’)
in our lab isolates. were full-term and not treated in the NICU. The rate of UP/ pro-
phages among the full-term infants (4/22) is significantly
greater than among the pre-term patients (5/144) (v2 ¼ 5.44,
3.4 Phages in metagenomes
P ¼ 0.020). Thus, while it is possible that UP/ is more common in
We next used searchsra (Levi et al. 2018) to find relatives of hospital-acquired bacteria, there is no evidence to suggest that
UP/901 in metagenomes. The initial results included over it is associated with patient outcomes in this study.
100,000 sequencing runs with at least one read aligned to
UP/901. After a series of filtering steps to remove erroneous
3.5 Identifying additional infections in the lab
data and false-positive results (see Supplementary Methods) we
identified 257 SRA runs that were most likely to have true- After identifying UP/ viruses in other hosts, we returned to the
positive results. We then used SRA BLAST (Camacho et al. 2009) lab to test additional strains. Though only twenty-eight C. koseri
to validate these putative hits and identified any instances of genomes were available in GenBank, nearly one-third carried a
redundant BioSample numbers (multiple sequencing runs from UP/ virus. In two cases, these C. koseri phages were identical to
the same sample) or redundant sample source (e.g. multiple prophage sequences found in urine E. colis. Given this high fre-
samples from the same individual). This validation step identi- quency of infection and similarity to UP/901, we tested three
fied 78 true positive, non-redundant samples that covered over unsequenced C. koseri isolates (UMB1389, UMB7451, UMB8248)
75 per cent of the UP/901 reference genome with over 90 per using UP/901 PCR primers. The PCR results identified an inte-
cent sequence identity. SRA BLAST is sensitive enough to grated phage in UMB7451, but it was not actively produced, in
distinguish between UP/ and Ypf/ prophages (Supplementary contrast to the infected E. coli isolates (Supplementary Fig. S5).
Fig. S4). We also attempted to establish new infections using filtrate
These positive metagenomic results include sixty-four from from UMB0901 cultures. Filamentous phages typically rely on
gut, five from urine, and four from environmental samples. Two conjugative pili as a primary receptor for infection (Mai-
additional hits came from reference isolates of C. koseri and Prochnow et al. 2015), and each E. coli carrying UP/901 in our lab
from mock communities used in validating microbiome proj- collection also harbored an IncI conjugative plasmid. We identi-
ects. (Full metadata are provided in Supplementary File S1.) Five fied three uninfected E. coli strains (UMB0731, UMB6655, and
of the positive gut samples were from non-human mammals, UMB6890) as candidates for testing UP/901’s ability to establish
including three from cats and two from pandas. Environmental new infections. These strains all clustered together with
samples included two from the New York subway, one from UP/901-infected strains (UMB1220, UMB1526, UMB5814) in a
wastewater, and one from river sediment. Most notably, previous gene presence–absence analysis of strains in our
 8 | Virus Evolution, 2020, Vol. 6, No. 1

collection of urinary E.colis (Garretto 2019), contained an IncI Genes II and V are both involved in regulating phage DNA repli-
plasmid, and were not already infected by UP/901. cation, and this modular split in homology between the dispa-
Despite their potential as candidate hosts, UMB0901 filtrates rate viruses likely reflects a deeper history of recombination
that were confirmed to contain the phage did not produce pla- among the phages.
ques on any of these strains. We were also unable to observe Because they infect related hosts and might have a history
plaques on E. coli JE-1, an IncI-bearing strain that is the standard of recombination, we also checked for any instances of co-
host for UP/901’s closest known relative, I2-2 (Bradley, Coetzee, infection between UP/ and Ypf/ in our data. We found one
and Hedges 1983). Last, we used PCR to test for phage integra- such case (GCA_001519645), corresponding to E. coli bacteremia
tion in these candidate hosts but were again unable to confirm in a patient with sepsis. This assembly, unfortunately, consisted
new infections. of 206 contigs, and Ypf/ appears to be split across at least four
different contigs. Notably, Ypf/ is not included on a large contig
4. Discussion containing both dif and UP/, suggesting this Ypf/ prophage in-
tegrated at a different locus using another mechanism or is pre-
We have introduced a new filamentous phage, UP/901, found in

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

sent as a plasmid in this host. SRA BLAST (Supplementary Fig.
multiple environments, with the largest clade found predomi-
S6) confirmed that each phage is fully covered by the reads. It is
nantly in urine E. colis. In addition to infecting E. coli, related vi-
unknown if the prophage genomes interact. Sachs and Bull
ruses can be found in S. enterica, K. pneumoniae, C. koseri, and Y.
(2005) demonstrated that co-infecting filamentous phages (in
enterocolitica. In several cases, identical phage genomes were
found in multiple host genera. These phages all share two puta- their case, f1 and IKe) could coevolve to package their genomes
tive genes of unknown function. We have tentatively named together using shared coat proteins. It is feasible that UP/ and
this collection of phages carrying these genes as ‘UP/ viruses’. Ypf/ could also produce chimaeric virions. In fact, the UP/ re-
Three additional novel ORFs were found in UP/901 and fifty- gion in GCA_001519645 has an early stop codon in its gene III
five other UP/ viruses, and most UP/ prophages lacking these and would likely be unviable unless it is able to incorporate
genes harbored other ORFs in these positions. Ypf/’s version of the attachment protein during assembly.
These UP/ viruses are related to phages IKe and I2-2 in the
Lineavirus genus of Inoviridae, but IKe and I2-2 cannot integrate
into the host genome and lack the accessory genes that appear 4.2 Tandem repeats and prophage diversity
to be unique to UP/ phages (Fig. 2). Given the prevalence of dif dif site integration typically results in tandem duplication of the
site integration among filamentous phages and their high se-
phage genome (Mai-Prochnow et al. 2015). It has been observed
quence similarity to much of the UP/901 genome, it is likely
in CTX/ and Ypf/ that hosts harboring a dif-integrated phage
that IKe and I2-2 evolved from an integrating ancestor.
can be super-infected by other such viruses, with the new phage
Recent work has called into question the current taxonomy
potentially supplanting the original (Midonet et al. 2019) or inte-
of filamentous phages and has suggested that the Inoviridae
grating in the same region (Derbise and Carniel 2014). In the lat-
may require substantial revision into multiple new families of
ter cases, the phage can integrate between the original copies or
viruses (Roux et al. 2019). We are, therefore, hesitant to claim
downstream (Chouikha et al. 2010). It is, therefore, possible to
that UP/ viruses deserve to be identified as a new phage genus.
For the time being, we propose that the UP/ viruses should be observe tandem triples or even quadruples of inovirus pro-
considered members of the existing Lineavirus genus, with the phages at the dif site of infected bacteria, and these downstream
current members (IKe and I2-2) forming a subgenus of phages homologs are not always identical. Furthermore, this series of
that have lost the dif site and nearby accessory genes. We ac- integration events can promote hybridization between co-
knowledge, though, that this taxonomy could change as the infecting inoviruses when new phages are produced by infected
Inoviridae are revised and may be complicated by recombination hosts (Davis and Waldor 2000; Derbise and Carniel 2014), as well
between co-infecting phages. as co-packaging of phages into shared coats as described above
(Sachs and Bull 2005). In the case of UP/ prophages, we did not
4.1 Relationship to CUSf and Ypff always know the exact number of tandem repeats within each
host genome, as most sequencing efforts relied on only short
Prior work identified the filamentous prophages CUS/ and Ypf/
Illumina reads that were not resolved into tandem copies dur-
in strains of extra-intestinal E. coli (ExPEC) and Y. pestis
ing assembly by the original data providers. For UMB0901, we
(Gonzalez et al. 2002; Derbise et al. 2007). While these phages
used depth of coverage to infer that UP/901 is integrated as a
have variable accessory gene regions (the same regions where
tandem triple.
UP/ has its own novel genes), CUS/ and Ypf/ otherwise share
Given the propensity for tandem prophage recombination,
99 per cent nucleotide identity with each other (Derbise et al.
there is additional uncertainty in interpreting and comparing
2007). These phages are found in similar hosts as UP/ viruses
the genomes of UP/ viruses that presents an additional chal-
and rely on dif site integration. In these prior studies, the
lenge for taxonomy. The E. coli strain ECONIH2 (GenBank:
authors demonstrated both the potential for these phages to in-
crease the virulence of ExPEC (Gonzalez et al. 2002), and also the CP014667) was originally sequenced with both Illumina and
likely association between Ypfu and the modern strains of Y. PacBio technologies and offers a valuable case study. ECONIH2
pestis pv orientalis associated with the third plague pandemic contains a tandem triple of UP/ prophages. The first two copies
(Derbise et al. 2007). of the phage are identical by nucleotide sequence and contain
As part of the exploration of UP/, we compared the individ- each of the genes also found in UP/901, as well as three addi-
ual genes from our UP/901 reference sequence to the Ypf/ pro- tional ORFs. The third copy of the phage in ECONIH2, however,
phage region in its type host, Y. pestis CO92. We found that only contains only the first of these additional genes followed by a
genes II and V have significant homology (Fig. 2), whereas the fourth novel ORF. This one bacterial strain demonstrates the po-
rest of UP/901 is more closely related to phages I2-2 and IKe. tential heterogeneity among co-infecting prophages.
 J. W. Shapiro and C. Putonti | 9

4.3 Future work Data availability

Additional work remains to understand the prevalence and role Data from this work is available at figshare (https://figshare.
of UP/ viruses in different microbial communities. First, future com/s/235048b35ae0617dac68).
research will need to characterize the two novel genes, h1 and
h2. These genes are found in a region of the genome that often
includes genes that alter host behavior or virulence (Mai- Supplementary data
Prochnow et al. 2015). In the prototypical integrating inovirus,
Supplementary data are available at Virus Evolution online.
CTX/, this region encodes the cholera toxin genes and their reg-
ulators (Waldor and Mekalanos 1996). It appears likely that
these two new genes interact with one another, but they share Acknowledgments
no homology to any known gene, and there is little to hint at
We are grateful to the Wolfe Lab for generously providing
their possible functions.
While we have identified putative phages from genome as- strains of E. coli and C. koseri used in this work and to Putonti

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

semblies, it was not always possible to predict phage end posi- Lab members for helpful comments on the manuscript. We
tions or the number of tandem repeats. These issues make it also thank O. Pybus, M. Marston, and two anonymous
difficult to assess the pan-genome of UP/ viruses, as some reviewers for constructive feedback and suggesting a deeper
might carry additional accessory genes not found in UP/901. analysis of the metagenomic data and comparison to Ypf/.
Future research that includes long-read resequencing of
infected hosts and phages released into culture medium would
help to resolve these ambiguities. Funding
We also attempted to determine the host requirements for This work is supported by NSF (1661357 to C.P.).
establishing new infections of UP/901. We were able to confirm
by PCR that new virions are released into the culture medium Conflict of interest: None declared.
by infected bacteria, but we were unsuccessful in our attempts
to infect new hosts. Our failure to observe new infections may References
be the result of lacking appropriate host strains for testing or
Altschul, S. F. et al. (1990) ‘Basic Local Alignment Search Tool’,
due to low efficiency of establishing new infections. At the
Journal of Molecular Biology, 215: 403–10.
same time, UP/901’s closest relatives, I2-2 and IKe, each infect
Arndt, D. et al. (2016) ‘PHASTER: A Better, Faster Version of the
E. coli harboring IncI conjugative plasmids, and each E. coli strain
PHAST Phage Search Tool’, Nucleic Acids Research, 44: W16–21.
infected by UP/901 in our collection also contained an IncI plas-
Aziz, R. K. et al. (2015) ‘Multidimensional Metrics for Estimating
mid. It, therefore, appears likely that at least some of these
Phage Abundance, Distribution, Gene Density, and Sequence
phages rely on IncI-encoded pili, and future work with addi-
Coverage in Metagenomes’, Frontiers in Microbiology, 6: 381.
tional strains and different growth conditions will test this
Bradley, D. E., Coetzee, J. N., and Hedges, R. W. (1983) ‘IncI2
possibility.
Plasmids Specify Sensitivity to Filamentous Bacteriophage
It is also possible that different UP/ viruses rely on different
IKe’, Journal of Bacteriology, 154: 505–7.
conjugative pili to initiate new infections. The phage attach-
Bruno, R., and Bradbury, A. (1997) ‘A Natural Longer Glycine-Rich
ment protein, g3p, showed the greatest variation across strains,
Region in IKe Filamentous Phage Confers No Selective
and each of these variants might correspond to different host
Advantage’, Gene, 184: 121–3.
specificity. Most of this variation was observed in a glycine-rich
Bushnell, B. (2014) BBMap: A Fast, Accurate, Splice-Aware Aligner
repeat region that links N- and C-terminal domains of g3p.
(No. LBNL-7065E). Berkeley, CA: Lawrence Berkeley National
Previous work showed that this region can diversify in phage
Lab.
IKe in a few generations in the lab, but no effects on phage fit- Camacho, C. et al. (2009) ‘BLASTþ: Architecture and
ness were observed (Bruno and Bradbury 1997). Those experi- Applications’, BMC Bioinformatics, 10: 421.
ments, however, did not test for changes in host range. Carnoy, C., and Roten, C. A. (2009) ‘The Dif/Xer Recombination
Systems in Proteobacteria’, PLoS One, 4: e6531.
Chouikha, I. et al. (2010) ‘Insights into the Infective Properties of
YpfU, the Yersinia pestis Filamentous Phage’, Virology, 407:
5. Conclusion
43–52.
UP/ viruses represent an exciting new group of filamentous Davis, B. M., and Waldor, M. K. (2000) ‘CTXu Contains a Hybrid
phages. Many inoviruses play important roles in modifying bac- Genome Derived from Tandemly Integrated Elements’,
terial pathogens within eukaryotic hosts. It remains to be seen Proceedings of the National Academy of Sciences, 97: 8572–7.
if UP/ viruses, though frequently found as prophages in bacteria Derbise, A. et al. (2007) ‘A Horizontally Acquired Filamentous
from urine, affect the frequency of UTI or other aspects of uri- Phage Contributes to the Pathogenicity of the Plague Bacillus’,
nary health. In the case of pre-term infant microbiomes, we ob- Molecular Microbiology, 63: 1145–57.
served no correlation between UP/ prophage presence and , and Carniel, E. (2014) ‘YpfU: A Filamentous Phage Acquired
patient outcomes. It is possible that UP/ phages affect bacterial by Yersinia pestis’, Frontiers in Microbiology, 5: 701.
ecology within the urinary environment without changing bac- Edgar, R. C. (2004) ‘MUSCLE: Multiple Sequence Alignment with
terial virulence. Perhaps more important will be understanding High Accuracy and High Throughput’, Nucleic Acids Research,
how frequently the phages shift hosts within different environ- 32: 1792–7.
ments and whether these new infections alter the phenotypes Eren, A. M. et al. (2015) ‘Anvi’o: An Advanced Analysis and
of multiple genera within a community. Visualization Platform for ‘Omics Data’, PeerJ, 3: e1319.
 10 | Virus Evolution, 2020, Vol. 6, No. 1

Garretto, A. (2019) ‘Exploring Associations between Lysogeny Penckofer, S. et al. (2020) ‘Characteristics of the Microbiota in the
and Host Abundance’, Master’s thesis, Loyola University Urine of Women with Type 2 Diabetes’, Journal of Diabetes and
Chicago, Chicago. Its Complications, 107561.
Gonzalez, M. D. et al. (2002) ‘Conserved Filamentous Prophage in Price, M. N., Dehal, P. S., and Arkin, A. P. (2010) ‘FastTree
Escherichia coli O18: K1: H7 and Yersinia pestis Biovar Orientalis’, 2—Approximately Maximum-Likelihood Trees for Large
Journal of Bacteriology, 184: 6050–5. Alignments’, PLoS One, 5: e9490.
Jian, H., Xiao, X., and Wang, F. (2013) ‘Role of Filamentous Phage Price, T. K. et al. (2016a) ‘Genome Sequences and Annotation of
SW1 in Regulating the Lateral Flagella of Shewanella piezotoler- Two Urinary Isolates of E.coli’, Standards in Genomic Sciences, 11:
ans Strain WP3 at Low Temperatures’, Applied and 79.
Environmental Microbiology, 79: 7101–9. et al. (2016b) ‘The Clinical Urine Culture: Enhanced
Jones, D. T., Taylor, W. R., and Thornton, J. M. (1992) ‘The Rapid Techniques Improve Detection of Clinically Relevant
Generation of Mutation Data Matrices from Protein Microorganisms’, Journal of Clinical Microbiology, 54: 1216–22.
Sequences’, Bioinformatics, 8: 275–82. R Core Team. (2013) R: A Language for and Environment for
Kalyaanamoorthy, S. et al. (2017) ‘ModelFinder: Fast Model

Downloaded from https://academic.oup.com/ve/article/6/1/veaa030/5860829 by guest on 12 November 2024

Statistical Computing.
Selection for Accurate Phylogenetic Estimates’, Nature Methods, Rice, S. A. et al. (2009) ‘The Biofilm Life Cycle and Virulence of
14: 587–9. Pseudomonas aeruginosa are Dependent on a Filamentous
Kamiunten, H., and Wakimoto, S. (1982) ‘Effect of Infection with Prophage’, The ISME Journal, 3: 271–82.
Filamentous Phage Xf on the Growth, Ultrastructure and Roux, S. et al. (2015) ‘VirSorter: Mining Viral Signal from
Virulence of Xanthomonas campestris pv. oryzae N5850’, Japanese
Microbial Genomic Data’, PeerJ, 3: e985.
Journal of Phytopathology, 48: 642–7.
et al. (2019) ‘Cryptic Inoviruses Revealed as Pervasive in
Katoh, K., and Standley, D. M. (2013) ‘MAFFT Multiple
Bacteria and Archaea across Earth’s Biomes’, Nature
Sequence Alignment Software Version 7: Improvements in
Microbiology, 4: 1–12.
Performance and Usability’, Molecular Biology and Evolution, 30:
Sachs, J. L., and Bull, J. J. (2005) ‘Experimental Evolution of
772–80.
Conflict Mediation between Genomes’, Proceedings of the
Kortright, K. E. et al. (2019) ‘Phage Therapy: A Renewed Approach
National Academy of Sciences, 102: 390–5.
to Combat Antibiotic-Resistant Bacteria’, Cell Host & Microbe,
Sieradzki, E. T. et al. (2019) ‘Dynamic Marine Viral Infections and
25: 219–32.
Major Contribution to Photosynthetic Processes Shown by
Letunic, I., and Bork, P. (2019) ‘Interactive Tree of Life (iTOL) v4:
Spatiotemporal Picoplankton Metatranscriptomes’, Nature
Recent Updates and New Developments’, Nucleic Acids
Communications, 10: 1–9.
Research, 47: W256–9.
Sweere, J. M. et al. (2019) ‘Bacteriophage Trigger Antiviral
Levi, K. et al. (2018) Searching the Sequence Read Archive Using
Immunity and Prevent Clearance of Bacterial Infection’,
Jetstream and Wrangler. Proceedings of the Practice and Experience
Science, 363: eaat9691.
on Advanced Research Computing, p. 50.
Val, M. E. et al. (2005) ‘The Single-Stranded Genome of Phage
Mai-Prochnow, A. et al. (2015) ‘Big Things in Small Packages: The
Genetics of Filamentous Phage and Effects on Fitness of Their CTX is the Form Used for Integration into the Genome of Vibrio
Host’, FEMS Microbiology Reviews, 39: 465–87. cholerae’, Molecular Cell, 19: 559–66.
May, T., Tsuruta, K., and Okabe, S. (2011) ‘Exposure Waldor, M. K., and Mekalanos, J. J. (1996) ‘Lysogenic Conversion
of Conjugative Plasmid Carrying Escherichia coli Biofilms by a Filamentous Phage Encoding Cholera Toxin’, Science, 272:
to Male-Specific Bacteriophages’, The ISME Journal, 5: 1910–4.
771–5. Ward, D. V. et al. (2016) ‘Metagenomic Sequencing with
Midonet, C. et al. (2019) ‘The TLCU Satellite Phage Harbors a Xer Strain-Level Resolution Implicates Uropathogenic E.coli in
Recombination Activation Factor’, Proceedings of the National Necrotizing Enterocolitis and Mortality in Preterm Infants’, Cell
Academy of Sciences, 116: 18391–6. Reports, 14: 2912–24.
Miller-Ensminger, T. et al. (2018) ‘Bacteriophages of the Urinary Warwick-Dugdale, J. et al. (2019) ‘Host-Hijacking and Planktonic
Microbiome’, Journal of Bacteriology, 200: e00738–17. Piracy: How Phages Command the Microbial High Seas’,
Nguyen, L. T. et al. (2015) ‘IQ-TREE: A Fast and Effective Stochastic Virology Journal, 16: 15.
Algorithm for Estimating Maximum-Likelihood Phylogenies’, Weimer, B. C. (2017) ‘100K Pathogen Genome Project’, Genome
Molecular Biology and Evolution, 32: 268–74. Announcements, 5: e00594–17.
Paradis, E., and Schliep, K. (2019) ‘ape 5.0: An Environment for Yamada, T. (2013) ‘Filamentous Phages of Ralstonia solanacearum:
Modern Phylogenetics and Evolutionary Analyses in R’, Double-Edged Swords for Pathogenic Bacteria’, Frontiers in
Bioinformatics, 35: 526–8. Microbiology, 4: 325.