A novel lineage of large aquatic bacteriophages identified through metagenomics

We constructed phylogenies of the moraphages together with divergent relatives, using three viral marker genes: the major capsid protein (MCP: VOG00292), terminase large subunit (TerL: VOG00012), and family B DNA polymerase (PolB: VOG00275). For each of these trees, reference sequences in the NCBI database were identified using BLASTp, with representative moraphage proteins as the query. Amino acid sequences were aligned using Clustal Omega with default parameters [31], and trees were constructed using IQ-TREE multicore v. 2.3.4 [32, 33]. The.treefile file of the output was visualized in iTOL [23]. This same approach was used to make trees of the viral DnaK homologs identified in some moraphage genomes (see below). Average amino acid identity was calculated as described previously [34, 35], and figures were made using the pheatmap function in R.

Proteins were predicted using Prodigal v. 2.6.3 with default parameters [36]. Functional annotation was done using the eggNOG-mapper v2 (emapper-2.1.12) [37] tool based on eggNOG v. 5.0 [38]. In addition, we compared all of the proteins to the VOG database v. 214, using hmmsearch in HMMER3 (e-value cutoff, 1e-5 [39]).

We used OrthoFinder v. 2.5.4 [40] to identify orthologous groups (OGs) in the moraphages (-og parameter). We selected an 80% cutoff, meaning that the core genome would consist of the OGs that are shared with 80% or more of the phages. The three ERM phages were counted as one occurrence, and this percentage was therefore calculated out of 26 rather than 28 phages. The core genome was identified based on occurrence rather than the number of hits to an OG a phage may have. The predicted functions of the core genome OGs were predicted manually using BLASTp and HHpred. From the OrthoFinder results, which compiled amino acid sequences into their respective OGs, sequences were input into each database, and the best annotation was chosen based on percent identity and e-values. Pan-genomic statistics were compiled using the micropan utility in R.

The predicted proteins encoded by each phage genome were searched against the RefSeq v. 227 database using LAST (parameters -m 500), and the best hits were recorded. The phylum of each best match was identified, and the bacterial phylum Bacteroidota received the most best hits for almost every phage. To assess the confidence that members of the Bacteroidota were the hosts of the phages, we calculated a metric by dividing the number of best hits to the Bacteroidota by the number of hits to the next-most-prevalent phylum and expressed this value as a ratio. If the ratio exceeded 1.5, the most prevalent host phylum was chosen as the likely host.

A synteny plot of the complete phage genomes was made using GenoPlotR [41] in the R programming environment. The genomes and gene predictions were used as input, together with pairwise BLASTp outputs (BLASTp version 2.16.0+, e-value threshold of 1e-10 [42]).

Here, we examined a novel clade of large phages found in metagenomes generated from a wide range of environments. The genome length of the moraphages ranged in size from 200 to 307 kbp, and four contained DTRs indicative of potentially complete genomes. Moraphages were identified primarily from freshwater sources such as lakes and rivers, but some variety in their environment was found (Table 1). For example, moraphages were also found in human-made facilities, as one phage was obtained from reclaimed water used to irrigate crops, and another was obtained from sediment near a natural gas well. Three distinct natural environments apart from freshwater are also represented: a marine water column, a coastal salt marsh, and an estuary. Out of the four moraphages with complete sequences, three were of freshwater origin, and one was isolated from an estuary. Throughout our genomic and phylogenetic analyses, we observed that the moraphages are related to three “ERM phages” that were recently discovered in marine sediment metagenomes [26]. We therefore also included these genome sequences in most of our comparative analyses. Other reference phages that were included in our phylogenetic analysis were found by comparing moraphage marker genes to entries in the NCBI databases, which showed similarities to smaller phages derived from various sources, including aquatic systems as well as the human gut and mouth.

The moraphage sequences were compared with those of large phages of other lineages using VirTree to determine whether they share similarities with members of these groups (Fig. 1). Also included in this analysis were members of the T4 group, specifically the families Straboviridae, Kyanoviridae, and Ackermannviridae, as well as the Lak phages, which have been found in human and animal gut metagenomes [23, 43]. The resulting phylogeny showed that moraphages are a distinct lineage that does not fall within any of these established groups of large phages.

Bild vergrößern

To identify the possible hosts of moraphages, we performed a homology search of all proteins in the phage genomes against the NCBI RefSeq database to identify hits to host bacterial lineages as described previously [26]. This analysis was performed on the grounds that phages often acquire genes from their host and that homology searches often provide clues regarding the identity of the host. From this analysis, we determined that the likely hosts of most moraphages are members of the class Flavobacteriia, with the plurality of best hits for 23 phages falling within the phylum Bacteroidota, while the plurality of hits for two were to members of the phylum Pseudomonadota (Fig. 1). The prediction of a host within the Bacteroidota is also consistent with phylogenetic analysis of the heat shock proteins identified in the moraphage genomes (see details below).

To determine the phylogenetic placement of this novel clade of jumbo phages relative to other phages, phylogenetic trees were constructed based on the terminase large subunit (TerL: VOG00012), major capsid protein (MCP: VOG00292), and family B DNA polymerase (PolB: VOG00275) sequences. When constructing these trees, we included a set of the best BLASTp hits to these proteins (Fig. 2, panels A-C). Some of the moraphage sequences did not show any similarity to MCP, TerL, or PolB VOGs and were therefore excluded from further analysis. In all of these trees, the moraphages formed a distinct clade that was divergent from any reference strains found in the NCBI database, confirming that they form a unique lineage of large phages. It should be noted that not all of the reference phages were jumbo phages, despite encoding proteins with sequence similarity to jumbo phage proteins. A common trend can be seen in all three trees, as the majority of the reference phages did not form a clade with the moraphages, resulting in two major nodes. A second trend was observed regarding the environments of the phages included in each tree. The reference phages, with a few exceptions, had been identified in metagenomic samples from the human gut or mouth. The trees suggest that the moraphages are a diverse lineage of phages that are distinct from previously characterized groups, and we therefore sought to quantify the genomic diversity within this lineage.

Bild vergrößern

We also compared the moraphages to each other by examining their pairwise average amino acid identity (AAI), which is an index used to quantify whole-genome relatedness. We visualized these results using a heat map and a clustering dendrogram (Fig. 3). The values are mostly in the low percent identity range, with the majority of phages falling between 39 and 65%, suggesting that this lineage has broad diversity. There are some clear outliers from this average, namely three pairs of phages that have a high degree of amino acid sequence similarity. The pair with the second-highest identity value of 84.48% (SAMN12796234:SAMN08331091) and the pair with the highest identity value of 95.21% (SAME6649935:SAME8660007) originated from freshwater samples. Interestingly, the pair with the third-highest identity value, at 77.44% (SAMN09092241: SAMN08778078), are from different environments, from a coastal sea marsh and a marine water column, respectively. The phage SAMN09092241 contains a PolB gene but appears to lack an MCP or TerL VOG gene and, this phage was judged to have the most divergent PolB due to the length of its branch (Fig. 2 A-C). In sharp contrast, its paired phage SAMN08778078 had hits to all three hallmark proteins. The pair with the highest identity value were isolated from different freshwater locations but clustered together in all three trees. However, it should also be noted that the only phage genome out of these pairs that has DTRs was that of SAMN0833109, suggesting that, despite these phages being highly similar to each other, the AAI values of these three pairs may have been based on AAI measurements of incomplete genome comparisons.

Bild vergrößern

The majority of the moraphages have low AAI percentages to each other, and we investigated if there were any genes that they shared as a group by assembling a pan-genome. The core genome consists of genes that are shared between all or the majority of the members of the clade, which offers a more focused view into the proteins that define the phage lineage. The pan-genome analysis also assesses the degree of gene diversity within the moraphage genomes by plotting a rarefaction curve (Fig. 4). While the moraphages may share a set of essential genes, the majority of their DNA encodes divergent and novel genes.

Bild vergrößern

As all three ERM phages shared the same OGs, they were counted only once rather than three times for the core genome analysis. Many phages also had multiple hits to the same OG, but we identified the core genome based on occurrence. Our analysis showed that there are only 10 OGs that are shared between all moraphages, while 98 OGs are shared between 80% of them or more. To gain insight into the function of these core OGs, we manually investigated their predicted functions using BLASTp and HHpred (see Methods). There are some notable results from both BLASTp and HHpred analysis, namely that four separate OGs (0000011, 0000025, 0000026, and 0000028) are putative glycosyltransferases. The function of these proteins in phages has been reported to be glucosylation of phage DNA, which protects it from host restriction endonucleases as part of the phage replisome. Glycosyltransferases have also been observed in temperate phages, modifying host serotypes during lysogeny [44]. Genetic material is not the only target of glycosylation, the virion itself can be altered as well. Glycans have been observed at the C-termini of both the capsid and tail tube subunit of Mycobacterium phages, acting as a shield against anti-phage responses [45]. Due to their diverse roles, the acquisition of diverse glycosyltransferases has occurred many times independently in different phage lineages [46], and there is great diversity in their structure and function. Apart from this, structural proteins such as baseplate proteins, contractile tail proteins, and capsid assembly proteins also make up the core genome, which can be expected, as all of these components are necessary for the assembly of a virion. However, a gene encoding a tape measure protein (TMP), an essential tail assembly protein, was not identified within the genomes. The length of the TMP gene has been shown to be correlated with the length of the phage tail [47], but it might also have other important functions, such as peptidoglycan-degrading activity [48]. As phage tails are used for infection of the host and injection of the phage genome, the TMP is believed to contribute to this process [49]. Because of its importance for large tailed phages, we believe that the moraphages have a TMP gene within their genome that was not detected because of its divergence from its homologs in other phages.

Also of note, OG0000024 and OG0000031 are associated with RNA polymerase, serving as a sigma factor and a mediator of transcription, respectively, and OG0000043 was identified as a polymerase B subunit (see Data availability for raw data files). These results indicate that proteins associated with DNA replication, transcription, and virion structure are the most prevalent of the shared genes in the moraphage core genome. We also examined a set of 26 VOGs that were found in >80% of the moraphage genomes, which once again includes genes for a range of protein families predicted to be involved in DNA replication, transcription, and virion structure (Table 2).

We also constructed a synteny plot to visualize genome-level variability between the moraphages (Fig. 5). For this, we used the four moraphages that were predicted to be complete based on the presence of DTRs, as well as one representative ERM phage. This analysis confirmed the presence of conserved blocks of homologous proteins encoded by the phages, which in many cases correspond to genes involved in virion morphogenesis and genome replication. Moreover, it is apparent that several inversions and other genomic rearrangements have taken place in the evolution of this group, suggesting that these may be important events that shape their genomic organization. Lastly, the synteny plots revealed several genomic regions with no or little homology to related phages, suggesting that they may represent genomic islands with variable composition across this lineage. If this is correct, an important avenue of future research would be to examine the functional repertoires of these islands and the dynamics of their gain and loss among the moraphages.

Bild vergrößern

While the moraphages encode many unique proteins (discussed below), there are many that are missing that have been identified within other large phage lineages. For example, in a study by Al-Shayeb et al. [26], some clades of jumbo phage from many environments were found to encode tRNA synthetases that are distinct from those of their host, but we did not find any of these in the moraphage genomes. Another significant discovery was the “phage nucleus” that forms in the host during infection by phages of the family Chimalliviridae [16, 50]. Members of this family were found to encode a multi-subunit RNA polymerase for the transcription of viral genes within the “phage nucleus”, limiting the phage’s dependence on the host. We did not find any RNA polymerase gene homologs in the moraphages, suggesting that this clade of viruses is dependent on host transcriptional machinery for gene expression. This also suggests that the moraphages have a distinct infection program that differs from those of other lineages of large phages that have been characterized previously.

Of the 25 novel moraphage sequences, only four contained DTRs and are therefore presumed to represent complete genomes. We therefore performed detailed functional analysis using these four genome sequences. For this, we performed manual curation of protein annotations using BLASTp and HHpred, as well as cross-referencing with our automated annotations using EggNOG (see Methods). Using this approach, we found that these four moraphages were predicted to encode a histone H1-like protein (PF07432). Also referred to as a “linker histone”, this protein helps to stabilize the nucleosome by linking adjacent nucleosomes together [51], and homologs have been identified in the genomes of giant eukaryotic viruses such as medusavirus [52]. Medusavirus has been shown to encode a full set of histone homologs, but this H1-like protein is distinct in the timing of its expression. In contrast to the other histones, H1 is expressed early in the infection cycle and may perform independent functions such as shutting down the host’s genome or regulating the virus’s genome for transcription [53]. A homolog of H1 has also been identified in Chlamydia trachomatis as Hc1, which binds to and densely compacts chromatin, regulating chlamydial gene expression [54, 55]. Hc1 has been expressed in E. coli to observe its effect on DNA structure and transcription/translation, and it was found that protein and DNA synthesis in the Hc1-expressing strain was considerably reduced [56, 57] Although the function of these histone-like proteins in moraphages is still unknown, it is intriguing to consider that the phage may use it to compact its own genome during packaging in a manner analogous to that described for giant viruses [58, 59].

Looking at the moraphages as a group, nine of them encode AMG glutamine synthetase (GS), whose function in bacteria is associated with nitrogen metabolism and ammonium reassimilation [60, 61]. This protein has been identified in oceanic viruses, suggesting that these viruses may manipulate nitrogen cycling in their environment [62]. GS components can also be used to activate alternative nitrogen-producing pathways under nitrogen-starved conditions [63]. Since the moraphages are of aquatic origin, nitrogen usage and potentially limited access to this element may have contributed to the phage’s acquisition of a GS,

We also identified multiple antitoxin proteins in the four genomes with complete sequences. Antitoxins are typically components of toxin-antitoxin (TA) systems and are a common genetic module found in both bacterial and archaeal chromosomes and plasmids, with some species encoding multiple distinct systems [64]. TA systems have also been commonly identified in prophage genomes [65, 66]. These modules typically comprise a growth-inhibiting toxin and its antitoxin, allowing for reversible changes to the bacterial cell, including some in response to phage infection [67]. Recently, there has been renewed interest in the anti-phage activity of TA systems. Bacteria use the modules as a broad method of defense against phage infection, with various biochemical activities that can target different phage processes [68]. Some phages are known to use antitoxin mimics to abrogate restriction by anti-phage TA systems. This has led to speculation that freestanding antitoxins in phage genomes may act as a counter to host anti-phage immune systems. The abundance of antitoxin genes in moraphage genomes may reflect an investment in countering host cell defenses.

Intriguingly, we observed an association between CcdA/HicB-like antitoxins and a group of Pac-I-like putative homing endonucleases present in multiple copies in some of the moraphage genomes. Homing endonucleases are selfish genetic elements that can spread within phage genomes and facilitate allelic conversion during phage coinfection [69]. In addition to neutralizing toxin activity, CcdA and HicB are known to transcriptionally repress toxin expression [70]. The T4-encoded homing endonuclease Itev-I has been shown to repress its own promoter [71]. Pairing an antitoxin gene with a homing endonuclease in a single module may be an alternative strategy for regulating homing endonuclease activity. Alternatively, if the antitoxin provides a fitness benefit to the phage, pairing this gene with a homing endonuclease may be a way to offset any fitness cost imposed by the selfish endonuclease.

One of the most striking features of the moraphages is the presence of several heat shock proteins (HSPs), molecular chaperones that are widespread in plants, animals, and prokaryotes, regulating the folding of proteins and protection from improper folding [72]. Similar to antitoxin mimics, phage-encoded HSPs may be a means of overcoming anti-phage defenses. HSPs are known to promote antiviral responses in eukaryotes [73], and recent work has shown that DnaJ is required for the anti-phage activity of an NLR-like protein in E. coli [74]. Additionally, another bacterial chaperone, SecB, is required for the function of some antitoxins [75], and some HSPs are able to phenocopy SecB’s canonical function for protein secretion [76]. As an alternative to a counterdefense function, phage-encoded DnaJ chaperones might recruit the host’s DnaK and facilitate the assembly of viral proteins, as seen in eukaryotic viruses such as simian virus 40 [77] and crocodilepox virus [78]. In a 2012 study, a T4-related enterobacteriophage named RB43 was found to encode a putative functional J-domain protein with 63% amino acid sequence identity to known Escherichia coli DnaJs [79]. DnaJs also seem to be more common in the moraphage genomes, as can be seen in Figure 1. While these proteins are highly conserved across the domains of life, their presence in phage genomes and their overall role in phages is not well understood.

In addition to DnaJ and DnaK, some moraphages encode the small heat shock protein HSP20. A subclass of HSPs, small heat shock proteins (sHSPs), with a few exceptions, are also ubiquitous in all forms of life but differ from HSPs. These proteins have a low molecular weight, ranging from 16 kDa to 42 kDa, and exhibit lower nucleotide and amino acid sequence conservation across species groups [80, 81]. Like other HSPs, these proteins have been found to participate in host protein folding and mediated disaggregation of proteins [82] and to respond to environmental changes [83]. The number of sHSPs in different organisms is also quite variable, with humans having 10 [84], Drosophila having four [85, 86], and some pathogenic bacteria such as Helicobacter pylori not possessing any [87].

Using HSP70 (DnaK) sequences from the eight moraphages encoding the protein and related sequences from bacteria indentified by BLASTp searches of the NCBI RefSeq database, a phylogenetic tree was constructed. One particular moraphage (SAMN19228552) displayed a relatively long branch length and did not cluster with the other seven DnaK-encoding moraphages, suggesting that it might have acquired a Dnak gene independently from the others. The other seven moraphage DnaK homologs formed a distinct clade within the family Flavobacteriaceae of the phylum Bacteroidota, consistent with the view that these phages likely infect these hosts and acquired HSP70 from a member of this bacterial lineage (Fig. 6).

Bild vergrößern