DNA polymerase diversity in Caudoviricetes
In the analyzed set of 18,382 Caudoviricetes genomes, we identified 6560 PolA, 2857 PolB, and 947 PolC proteins (Supplementary Table S1). The Caudoviricetes genome tree was split into subtrees at the depth of 0.15, roughly corresponding to a genus level (see an example in Supplementary Figure S1; https://ftp.ncbi.nih.gov/pub/yutinn/jumping_polymerases_2024/genome_tree/). Of the 1,514 subtrees that included more than one virus genome, 563 were found to encode a DNAP, and 8 encoded more than one DNAP. Analysis of the DNAP distribution pattern revealed two distinct types of DNAP heterogeneities within phage subtrees: (i) DNAPs of different families were encoded in closely related viruses, with a single copy in each genome, implying swapping of DNAP genes (6 subtrees), and (ii) the same viral genome encoded two DNAPs of different families (2 subtrees).
DNA polymerase swapping in Caudoviricetes
We identified 6 subtrees in the phage genome tree in which the phages encoded DNAPs of different families. Representative pairs of most closely related genomes with different family DNAPs from these subtrees are listed in Table 1. To investigate the provenance of these groups of phages encoding distinct DNAPs, we examined deeper clades in the phage genome tree that included each of the 6 subtrees (Table 1), apart from Crassvirales for which frequent PolA-PolB swaps have been reported previously [19]. Clades 1, 2, and 3 consisted of PolA- and PolB-encoding genomes, and clade 4 genomes encompassed PolA and PolC (Figs. 1, 2). The DNAPs from these clades were reciprocally mapped onto the corresponding PolA, PolB, and PolC trees (Supplementary Figure S2). Examination of this mapping suggests that, in addition to inter-family DNAP swaps, some intra-family DNAP swaps occurred in the selected clades, that is, PolA or PolB was apparently replaced by a distinct DNAP of the same family on several occasions. Below we discuss each of these clades in detail, in an attempt to reconstruct the evolutionary scenarios.
Table 1
Phage clades displaying DNAP swapping and examples of exchange between closely related genomes
DNAP
|
genome
|
Organism
|
genome length
|
ANI or AAI
|
clade
|
tree00180
|
PolA
|
MN184886.1
|
Erwinia phage pEp_SNUABM_08
|
62716
|
39% AAI
|
Clade1; subtrees 00180–00189
|
PolB
|
MN505213.1
|
Serratia phage JS26
|
63971
|
tree00183
|
PolA
|
KF626665.1
|
Phage Sano
|
56147
|
39% AAI
|
--“--
|
PolB
|
MT161385.1
|
Xanthomonas phage FoX4
|
60418
|
tree00184
|
PolA
|
MN536026.1
|
Pseudomonas phage vB_Pae-SS2019XI
|
57567
|
38% AAI
|
--“--
|
PolB
|
MF959998.1
|
Marinobacter phage PS6
|
58226
|
tree01348
|
PolA
|
MZ477002.1
|
Acinetobacter phage Phab24
|
93604
|
92.72% ANI
|
Clade2; subtrees 01336–01351
|
PolB
|
MN276049.1
|
Acinetobacter phage BS46
|
94068
|
tree01419
|
PolA
|
MG592602.1
|
Vibrio phage 1.237.B._10N.261.52.C5
|
60160
|
99.54% ANI
|
Clade3; subtrees 01419–01425
|
PolB
|
MG592464.1
|
Vibrio phage 1.089.O._10N.261.51.F9
|
59851
|
tree01472
|
PolA
|
MT601273.1
|
Bacillus phage vB_BsuS-Goe12
|
124287
|
99.74% ANI
|
Clade4; subtrees 01466–01481
|
PolC
|
MT601274.1
|
Bacillus phage vB_BsuS-Goe13
|
126848
|
We sought to validate the intra-clade cross-family swaps of DNAPs using the large subunit of the terminase (TerL) as the reference. The TerL sequences were collected from the genomes with identified DNAPs within each of the clades 1, 2, 3 and 4 and clade-specific TerL phylogenetic trees were constructed. Then, we constructed topologically constrained trees, separating the TerL from genomes encoding different DNAPs (for example, for clade 1, the constraint separated TerL from PolA- and PolB-bearing genomes). Such constrained topologies represent hypothetical phylogenies where DNAPs of different families are not intermixed within the clade histories. The optimal TerL trees satisfying these constraints were compared to the unconstrained trees, in an attempt to falsify the scenario with multiple DNAP swaps. For all clades, the Approximately Unbiased test decisively rejected the constrained topologies (Supplementary Table S2), suggesting that multiple DNAP swaps within each clade did occur.
Clade 1 included 126 genomes from several genera of the family Casjensviridae (genome size range 50–70 kb) of which 96 encoded PolA whereas the remaining 30 encoded PolB. The genome tree for this clade is dominated by a distinct group of PolA (A1 in Fig. 1) in which 3 disjointed PolB branches (B1-B3) and two branches from a separate group of PolA (A2) are embedded. Altogether, comparison of the phage genome tree with the phylogenetic trees of PolA and PolB suggests 8 independent DNAP swaps including both 6 inter-family (PolA to PolB) exchanges and 2 intra-family (A1 to A2) exchanges (Fig. 1). In most of the phage genomes in this clade, the swapped PolA and PolB genes share the same or similar genomic neighborhood (Fig. 1).
Clade 2 (subtrees 01336–01351 in the phage genome tree) unites phages from genera Plaisancevirus, Saclayvirus, Barbavirus, subfamily Ounavirinae, and several unclassified Caudoviricetes (genome size range 80–105 kb). The phages in subtree 01347 encode an additional protein with remote sequence and structural similarity to PolA, which we discuss in the next section. Here we address the apparent PolA to PolB swaps in this clade (Fig. 2). PolBs of Clade 2 are monophyletic (B4 in Fig. 2), but PolAs come from three distinct branches (A3, A4 and A5 in Fig. 2; see Supplementary Figure S2 for the DNAP trees). As in Clade 1, comparison of the phage genome tree with PolA and PolB phylogenetic trees suggests several independent swaps although the direction and the order of these events is difficult to establish.
Two Acinetobacter phages of Clade 2, Phab24 (MZ477002) and BS46 (MN276049), have average nucleotide identity (ANI) of 93%, and yet, encode DNAPs of different families, PolA and PolB, respectively (Table 1; Fig. 3a). Comparison of the genome organization in the vicinity of the DNAP genes (in which we additionally included the corresponding genome region of Acinetobacter phage TaPaz (MZ043613) because its PolB is most similar to PolB of MN276049 (Acinetobacter phage BS46). suggests that, in this case, the primase-helicase gene was replaced along with the DNAP.
In the upper part of each panel, nucleotide sequence similarity between the compared genomes is shown in green or yellow, on the scale from 0 to 100%. Red arrows denote the genomic regions containing the DNAP genes, visualized on the genome maps. The lower parts, show the blow up of the regions containing the DNAP genes. Functionally annotated homologous genes are shown in the same colors. DNAP genes are labeled with their GenBank protein IDs. Grey shading highlights genomic regions with high nucleotide sequence similarity (percent identity indicated). Pink lines connect genes with significant detectable amino acid sequence similarity detected with BLASTP but no significant nucleotide sequence similarity.
Clade 3 unites subtrees 01419–01425 (Fig. 2) and includes phages from genera Sashavirus, Nonanavirus, Gorganvirus, and unclassified Caudoviricetes, with genome size range of 42.5–62.5 kb. In this clade, a single PolB to PolA swap appears to have occurred within the subtree01419, whereas the basal PolB genes belong to distinct clades (Fig. 2). In this clade, we identified a pair of nearly identical genomes, Vibrio phage 1.237.B._10N.261.52.C5 (MG592602) and Vibrio phage 1.089.O._10N.261.51.F9 (MG592464), that encode different family DNAPs, PolA and PolB, respectively. Pairwise genome comparison shows that the DNAP neighborhood is the only large region with markedly lower similarity between the two genomes (Fig. 3b). The DNAP gene is the only one that obviously was replaced but additional rearrangements might have occurred in the adjacent genomic region containing genes encoding uncharacterized small proteins (Fig. 3b).
Clade 4 unites subtrees 01466–1481 including families Tybeckvirinae and Andrewesvirinae, genera Audreyjarvisvirus, Spbetavirus, Latrobevirus, Sextaecvirus, Slashvirus, and several unclassified Caudoviricetes (genome size range 61–185 kb). In this clade, the phages encode either PolC or PolA. PolC, specifically group C1, is likely to be ancestral in this assemblage of phages (Fig. 2 and Supplementary Figure S2). This ancestral PolC apparently was replaced with PolA on two independent occasions (Fig. 2), and furthermore, underwent several intra-family replacements involving PolC variants from groups C2 and C3. We also identified a pair of closely related genomes in this clade with over 99% ANI, Bacillus phage vB_BsuS-Goe12 (MT601273) and Bacillus phage vB_BsuS-Goe13 (MT601274), that encode DNAPs of different families, PolA and PolC, respectively. Comparative genome analysis of this pair of phages revealed an extended segment of dissimilarity suggesting that several replicative genes, including DnaB-like helicase, DnaG-like primase and RecJ-like exonuclease, traveled together with the DNAPs, but the replacement keeps the gene context unchanged, that is, the genes appear to have been replaced en bloc (Fig. 3c).
When the inter-family swaps were found to have occurred on shallow branches of the virus tree, these events typically involved phages that shared bacterial hosts (Supplementary Figure S3).
Two novel phage PolAs
In addition to the typical DNAPs, we identified two divergent variants of PolA encoded in phage genomes. One of these, denoted divPolA1, is present in a group of 14 phages (Flavobacterium phage vB_FspM_immuto_3-5A and related phages), with genomes in the range of 155–190 kb that also encode a ‘regular’ DNAP, either family A, B, or C (Fig. 4a). The conserved arrangement of genes implicated in replication downstream of the divPolA1 gene suggests that divPolA1 is involved in replication (Fig. 4b). Structural prediction for divPolA1 (Fig. 5) showed that it contains a Palm domain in which the main catalytic residues are conserved, whereas the Thumb domain is truncated and the 3’ exonuclease domain seems to be inactivated, with all three catalytic aspartates replaced (Fig. 5). Thus, divPolA1 most likely retains the DNAP activity whereas the exonuclease activity is lost.
Another diverged PolA variant, divPolA2, was initially identified in 110 genomes of barbaviruses (Rheinheimera phage vB_RspM_Barba18A and related viruses) with genomes of 80–85 kb, each also encoding a regular PolB. Additional PSI-BLAST searches against the Caudoviricetes database using barbavirus divPolA2s as queries revealed more divPolA2 proteins, sometimes with two or three paralogs per phage genome (Fig. 6).
Unlike the PolB of these phages, which is embedded within a typical context of replication-related genes, the gene encoding divPolA2 is located in variable gene neighborhood. Structural modeling suggests that divPolA2 contains an active DNAP (Palm) catalytic domain but lacks a Thumb domain homologous to those of any other DNAPs (Fig. 7).
Instead, this protein contains an N-terminal globular domain without detectable similarity to any other known domains that potentially might function as the Thumb. As in the case of divPolA1, the 3’ exonuclease domain is lacking. Most likely, divPolA2 is not the replicative enzyme of barbaviruses, a role that belongs to PolB. Instead, divPolA2 might be a DNAP involved in repair processes, or an RNA polymerase, given that PolA was co-opted for that function in T7 and related phages [5]. Of note, structural comparison did not only reveal DNAPs as the top hits for divPolA2, but also DNA-directed RNA polymerases (mitochondrial RNA polymerase (PDB ids: 7a8p, 6ymv, Dali z-score ~ 12) and, with lower z-score (~ 9), also a viral DNA-directed RNA polymerase from bacteriophage N4 (genus Enquatrovirus, class Caudoviricetes) (PDB id: 4ff3). These observations are compatible with the possibility that divPolA2 is actually an RNA polymerase although the N-terminal globular domain of divPolA2 is unrelated to the N-terminal domains of PolA-related RNA polymerases (Fig. 7).