Closely related Vibrio alginolyticus strains encode an identical repertoire of prophages and filamentous phages

Filamentous vibriophages represent a massive repertoire of virulence factors which can be transferred across species boundaries, leading to the emergence of deadly pathogens. All filamentous vibriophages that were characterized until today were isolated from human pathogens. Considering frequent horizontal gene transfer among vibrios, we predict that other environmental isolates, including non-human pathogens also carry filamentous phages, of which some may encode virulence factors. The aim of this study was to characterize the phage repertoire, consisting of prophages and filamentous phages, of a marine pathogen, Vibrio alginolyticus. To do so, we sequenced eight different V. alginolyticus strains, isolated from different pipefish and characterised their phage repertoire using a combination of morphological analyses and comparative genomics. We were able to identify a total of five novel phage regions (three different Caudovirales and two different Inoviridae), whereby only those two loci predicted to correspond to filamentous phages (family Inoviridae) represent actively replicating phages. Unique for this study was that all eight host strains, which were isolated from different eukaryotic hosts have identical bacteriophages, suggesting a clonal expansion of this strain after the phages had been acquired by a common ancestor. We further found that co-occurrence of two different filamentous phages leads to within-host competition resulting in reduced phage replication by one of the two phages. One of the two filamentous phages encoded two virulence genes (Ace and Zot), homologous to those encoded on the V. cholerae phage CTXΦ. The coverage of these zot-encoding phages correlated positively with virulence (measured in controlled infection experiments on the eukaryotic host), suggesting that this phages is an important virulence determinant. Impact statement Many bacteria of the genus Vibrio, such as V. cholerae or V. parahaemolyticus impose a strong threat to human health. Often, small viruses, known as filamentous phages encode virulence genes. Upon infecting a bacterial cell, these phages can transform a previously harmless bacterium into a deadly pathogen. While filamentous phages and their virulence factors are well-characterized for human pathogenic vibrios, filamentous phages of marine vibrios, pathogenic for a wide range of marine organisms, are predicted to carry virulence factors, but have so far not been characterized in depth. Using whole genome sequencing and comparative genomics of phages isolated from a marine fish pathogen V. alginolyticus, we show that also environmental strains harbour filamentous phages that carry virulence genes. These phages were most likely acquired from other vibrios by a process known as horizontal gene transfer. We found that these phages are identical across eight different pathogenic V. alginolyticus strains, suggesting that they have been acquired by a common ancestor before a clonal expansion of this ecotype took place. The phages characterized in this study have not been described before and are unique for the Kiel V. alginolyticus ecotype. Data Summary The GenBank accession numbers for all genomic sequence data analysed in the present study can be found in Table S1. All phage regions identified by PHASTER analysis of each chromosome and the respective coverage of active phage loci are listed in Table S2. GenBank files were deposited at NCBI for the two actively replicating filamentous phages VALGΦ6 (Accession number: MN719123) and VALGΦ8 (Accession number: MN690600) The virulence data from the infection experiments have been deposited at PANGAEA: Accession number will be provided upon acceptance of the manuscript. Data statement All supporting data have been provided within the article or through supplementary data files. Four supplementary tables and six supplementary figures are available with the online version of this article.

process was repeated for 10 6 generations and sampled every 5000 generations. The first 2 3 8 2000 trees were deleted as burn-in processes and the consensus tree was constructed from the 2 3 9 remaining trees. Convergence was assured via the standard deviation of split frequencies 2 4 0 (<0.01) and the potential scale reduction factor (PSRF~1). The resulting phylogenetic tree 2 4 1 and associated posterior probabilities were illustrated using FigTree version 1.4.2 2 4 2 (http://tree.bio.ed.ac.uk/software/figtree/). We used Propionibacterium phage B5 which is a 2 4 3 phage preying on a gram-positive bacterium as an outgroup.

4 4
We additionally compared the predicted phage regions from the present study with 2 4 5 potential phage regions from all other so far published fully closed V. alginolyticus genomes 2 4 6 ( Table S4). To do so, we identified potential phage regions on each chromosome using 2 4 7 PHASTER (33) and compared those with the phage regions from the present study using 2 4 8 Easyfig (34). 2 4 9 2 5 0 Analysis of virulence factors: We found that one of the active filamentous phages (i.e.

5 1
Vibrio phage VALGΦ6) contains the virulence cassette comprising the Zot and the Ace 2 5 2 proteins, which is frequently found in vibriophages and responsible for severe gastro-2 5 3 intestinal diseases (47,48). To compare these two proteins with other Zot and Ace proteins 2 5 4 isolated from various vibriophages we generated protein alignments using AliView (40) and 2 5 5 examined the presence of Walker A and Walker B motifs in Vibrio phage VALGΦ6 Zot 2 5 6 proteins. We further used the TMHMM Server (http://www.cbs.dtu.dk/services/TMHMM/) 2 5 7 to confirm the presence of a transmembrane domain typically found in the Zot protein. We performed a controlled infection experiment to estimate the virulence of the eight 2 6 3 sequenced strains on juvenile pipefish (a detailed description of methods and the statistical 2 6 4 analysis can be found in (28). Briefly, we fed 9-12 juvenile pipefish per tank in using 2 6 5 triplicate tanks with Artemia nauplii which were previously exposed to ~10 9 CFU/ml or 2 6 6 seawater as control. Twenty-four hours post infection, each fish was killed and bacterial load 2 6 7 was determined as colony forming units (CFU/ml) as in (28) Figure S1). This indicates that induced and non-induced cultures produce 2 8 6 comparable amounts of particles encoded by the same filamentous phage. As a further 2 8 7 control total DNA without DNase A treatment resulted in a coverage increased by the factor 2 8 8 of 100-100,000 at the loci encoding filamentous phages compared to the average 2 8 9 chromosomal coverage. We thus conclude that the cultures produced a permanent amount of 2 9 0 phage particle protected ssDNA independent of the induction from mitomycin C. 2 9 1 2 9 2 2. Caudovirales 2 9 3 Whole genome comparison between the eight sequenced V. alginolyticus strains revealed 2 9 4 the presence of three different prophage regions belonging to the family Caudovirales, none 2 9 5 of which generated phage particles nor protein protected DNA in the experimental settings 2 9 6 used in this study ( Figure 1). Thus, a more thorough classification based on morphological 2 9 7 characterization was not possible. We further did not find regions of increased coverage for 2 9 8 these three Caudovirales regions (Supplementary material Figure S1) on the bacterial 2 9 9 chromosomes indicating that these phages were neither actively replicating in uninduced 3 0 0 bacterial cultures nor able to switch to the lytic cycle upon induction with mitomycin C. We 3 0 1 could not identify sequence similarities between these three different Caudovirales phages, 3 0 2 suggesting that they are genetically distinct phages. However, each of the three Caudovirales (Replication, Assembly, Structural proteins, Integration, Lysis) and 12 ORFs to hypothetical 3 1 0 proteins ( Figure 1). All ORFs were orientated in the same direction. Even though Vibrio 3 1 1 phage VALGΦ1 could not be found in induced and uninduced supernatants, it is predicted to 3 1 2 be intact according to PHASTER. Vibrio phage VALGΦ1 is exclusively found on 3 1 3 Chromosome 1, where it has a unique integration site, which is identical across all eight 3 1 4 sequenced strains. The phage genome as well as the flanking regions (five genes upstream 3 1 5 and five genes downstream of the integrated phage) showed 100% sequence similarity across 3 1 6 all eight sequenced strains, suggesting that the phage is highly conserved across host-strains. 3 1 7 Comparative genomic analysis between Vibrio phage VALGΦ1 and ten closest hits on NCBI 3 1 8 reveals that the two closest related phages are FDAARGOS_105 integrated on chromosome 1 3 1 9 but only when Vibrio phage VALGΦ8 was integrated on chromosome 2, not when it existed 3 8 0 exclusively extrachromosomal or had an additional copy on chromosome 1. genomes suggest that Vibrio phage VALGΦ6 and Vibrio phage VALGΦ8 are different from 3 9 0 other bacteriophages described until today. For Vibrio phage VALGΦ6 the two closest hits 3 9 1 were the two V. parahaemolyticus phages VfO4k68 and VfO3k6 with query covers of 66% 3 9 2 and 75% and similarity values of 94.65% for each phage. The closest hits for Vibrio phage 3 9 3 VALGΦ8 were the two V. parahaemolyticus phages Vf12 and Vf13 with a query cover of 3 9 4 88% and a similarity of 94.65%. To compare all non-lytic phages between strains from the present study and other V. 3 9 7 alginolyticus isolates we used PHASTER to predict prophages from all available closed non-3 9 8 Kiel V. alginolyticus genomes and found a total of 14 predicted prophage regions (Table S4).

9 9
Comparisons between those uncharacterized vibriophages and phages from the present study 4 0 0 revealed that Vibrio phage VALGΦ6 and the Caudovirales cassette consisting of Vibrio 4 0 1 phage VALGΦ2 and 2b is unique to the Kiel alginolyticus system. In contrast, we found 4 0 2 integrated Inoviridae with high similarity to Vibrio phage VALGΦ8 in four of the six non-4 0 3 Kiel V. alginolyticus strains and one integrated Caudovirales that shared high similarity with 4 0 4 Vibrio phage VALGΦ1 and had the same attL/ attR sequence, i.e. CGTTATTGGCTAAGT Multi-phage-cassettes 4 1 8 We found multi-phage cassettes on chromosome 2 in two strains, i.e. K04M3 and K04M5. 4 1 9 While the cassette in K04M5 consists of Vibrio phage VALGΦ6 followed by Vibrio phage 4 2 0 VALGΦ8, K04M3 has two multi-phage cassettes. The first cassette consists of Vibrio phage 4 2 1 VALGΦ6, followed by a tandem repeat of two identical Vibrio phage VALGΦ8 regions, the 4 2 2 second cassette, which is located 10389 bp downstream of the first cassette is identical to the 4 2 3 one identified in K04M5. Even though Vibrio phage VALGΦ6 is identical across all strains, 4 2 4 the second replicate in K04M3, which represents the start of the second multi-phage cassette, 4 2 5 misses the transcription regulator and has major deletions, particularly affecting assembly 4 2 6 and structural proteins (Supplementary material, Figure S4). 4 2 7 4 2 8 Virulence of Kiel V. alginolyticus ecotypes 4 2 9 Comparative genomic analysis between virulence factors commonly encoded on 4 3 0 filamentous phages revealed that only Vibrio phage VALGΦ6 contains the virulence cluster 4 3 1 containing Ace and Zot. In contrast no known virulence factors could be found on Vibrio 4 3 2 phage VALGΦ8 and the described Caudovirales. Sequence comparisons of Zot proteins 4 3 3 encoded on different vibrios revealed that the Vibrio phage VALGΦ6 encoded Zot is highly 4 3 4 similar to Zot genes encoded on other closely related Vibrio species from the harveyi clade 4 3 5 (such as V. parahaemolyticus or V. campbellii, Figure S5). Even though we found less 4 3 6 similarity between the Vibrio phage VALGΦ6 encoded Zot protein and CTXΦ-encoded Zot 4 3 7 proteins, we found two conserved motifs (Walker A and B, common among human 4 3 8 pathogens), which were at the N-terminal side of the Zot proteins ( Figure S5). In addition, we 4 3 9 found a transmembrane domain in the Vibrio phage VALGΦ6 encoded Zot protein ( Figure  4 4 0 S6), suggesting that similar to the CTXΦ-encoded Zot, the Vibrio phage VALGΦ6 encoded 4 4 1 Zot is also a transmembrane protein.

2
Controlled infection experiments on juvenile pipefish revealed differences in total 4 4 3 bacterial load among strains, a proxy for virulence. We found that those strains, that only integration sites (possible recombination with both chromosomes). All eight strains have no 4 7 5 core genomic variation and sequence variation is mainly attributable to differences in mobile 4 7 6 genetic elements (MGEs), such as plasmids and presence/ absence of Vibrio phage VALGΦ8 4 7 7 (49). Comparative genomic analyses across a wider range of V. alginolyticus isolates 4 7 8 indicated that the phage repertoire of the Kiel alginolyticus ecotype is unique and cannot be 4 7 9 found elsewhere. Thus, we hypothesize that the identical prophage composition in this 4 8 0 ecotype together with the identical integration sites and flanking regions suggests that these 4 8 1 phages may have been acquired from a common ancestor before a clonal expansion of the 4 8 2 Kiel alginolyticus ecotype took place. Under this scenario we predict that these five 4 8 3 prophages are increasing the fitness of this ecotype in the present habitat and are thus 4 8 4 maintained by selection. 4 8 5 The sampling design, spanning two different organs (gills or gut) from six different 4 8 6 pipefish allows us not only to look at the phage composition of closely related bacteria across 4 8 7 eukaryotic hosts but also within eukaryotic hosts. We found more similarity within pipefish 4 8 8 Nr. 4 (strains K04M1, K04M3 and K04M5) than across all six pipefish: First, all three strains 4 8 9 contained Vibrio phage VALGΦ8, and second, the only two multi-phage cassettes were 4 9 0 found in strains K04M3 and K04M5, both isolated from pipefish Nr. 4. It is tempting to 4 9 1 speculate that the high prevalence of Vibrio phage VALGΦ8 relative to all eight sequenced 4 9 2 strains is a result of the close proximity between strains inside the gut, which favours the 4 9 3 rapid horizontal spread of Vibrio phage VALGΦ8. Future experiments would be needed to 4 9 4 study the likelihood for Vibrio phage VALGΦ8 to establish successful chronic infections and 4 9 5 the circumstances which favour the different life-styles (extra-or intra-chromosomal) and 4 9 6 integration sites (chromosome 1 or chromosome 2). 4 9 7 4 9 8 Filamentous phages differ in their life-style 4 9 9 While Vibrio phage VALGΦ6 was exclusively found at one integration site across all 5 0 0 eight sequenced strains (exceptions: the multi-phage cassettes in strains K04M3 and 5 0 1 K04M5), Vibrio phage VALGΦ8 had different integration sites on both chromosomes and 5 0 2 existed intra-and extrachromosomal. We identified one, respectively two extrachromosomal 5 0 3 closed circular contigs within the assembly of strains K04M1 and K05K4 representing 5 0 4 multimers of Vibrio phage VALGΦ8 (Figure 4). This indicates the presence of 5 0 5 extrachromosomal phage replicons in two out of the eight sequenced V. alginolyticus 5 0 6 genomes. Filamentous phages typically multiply via the rolling circle replication (RCR) 5 0 7 mechanism (41). Considering that K05K4 contains another copy of Vibrio phage VALGΦ8 5 0 8 integrated on chromosome 1 and that the extrachromosomal contigs contain two, respectively 5 0 9 three copies of Vibrio phage VALGΦ8 (Figure 4), we hypothesise that the K05K4 5 1 0 extrachromosomal contigs represent RCR intermediates of the integrated Vibrio phage 5 1 1 VALGΦ8. However, to confirm or falsify this hypothesis experiments using knock-out 5 1 2 versions of the intrachromosomal copy of Vibrio phage VALGΦ8 in strain K05K4 have to be 5 1 3 performed which are beyond the scope of this study. In contrast, K04M1 does not contain an 5 1 4 intrachromosomal version of Vibrio phage VALGΦ8 and the extrachromosomal contig of 5 1 5 K04M1 only consists of one phage replicon (Figure 4). This suggests that Vibrio phage 5 1 6 VALGΦ8 is able to establish a chronic extra-chromosomal infection without the need of an 5 1 7 intrachromosomal copy. 5 1 8

1 9
Within-host competition can lead to the reduction of phage producing particles 5 2 0 Vibrio phage VALGΦ1 has been predicted to be complete, but we did not find phage 5 2 1 particles of head-tail phages in the supernatant nor did we detect any DNA-sequences in phage 5 2 2 particles that map to its region in the chromosome under lab-conditions. Without a proof-of 5 2 3 principle which would again require knock-out versions of the filamentous phages, we can 5 2 4 only speculate that the Caudovirales Vibrio phage VALGΦ1 is supressed by the two actively 5 2 5 replicating filamentous phages. Indeed, within-host competition between different 5 2 6 Caudovirales has been found in other systems, for instance in Bacillus licheniformis (29). 5 2 7 Alternatively, Vibrio phage VALGΦ1 might be not induced within the conditions of our 5 2 8 experimental set up or could have been wrongly predicted to be complete by the software but 5 2 9 is, however not able to actively replicate suggesting prophage domestication, which has also 5 3 0 been predicted for Vibrio phage VALGΦ2 and 2b. When head-tail phages switch from the 5 3 1 lysogenic to the lytic cycle they always kill their host. Selection for a strict repression of the 5 3 2 lytic life cycle of prophage inactivation should thus be strong (50). Indeed, bacterial genomes 5 3 3 have numerous defective prophages and prophage-derived elements (51, 52), which 5 3 4 presumably originate from pervasive prophage domestication (50). By domesticating 5 3 5 prophages, bacteria can evade the risk of getting lysed but are still able to maintain beneficial 5 3 6 accessory genes, in the present case for instance the Mar family proteins encoded on the 5 3 7 defective Vibrio phage VALGΦ2b, which encode transcriptional regulators involved in the 5 3 8 expression of virulence, stress response and multi-drug resistance (53, 54). Another case of 5 3 9 within-host competition between phages is the 10-x reduced coverage of Vibrio phage 5 4 0 VALGΦ6 in strains where both filamentous phages were present. Again, we hypothesise, that 5 4 1 one phage, in this case Vibrio phage VALGΦ8, negatively affects the replication of another 5 4 2 phage, here Vibrio phage VALGΦ6. As above, to verify or falsify this hypothesis, knock-out 5 4 3 versions of strains containing both filamentous phages would be required, as have been used 5 4 4 in (29). Within-host competition is common among different head-tail prophages within the 5 4 5 same host leading to strong selection for short lysis time (55) The accession numbers of the eight Vibrio algionolyticus genomes analysed in the present 6 2 9 study are provided in Supplementary information, Table S1.
The accession numbers of the two newly discovered filamentous phages are provided in the 6 3 2 manusrcipt, see Data statement. 6 3 3 6 3 4 The accession numbers of all other filamentous phage genomes used for comparative 6 3 5 genomics in the present study are provided in Supplementary information, Table S3 and 6 3 6 Figure 5.