Coordinated transcriptional response to environmental stress by a Synechococcus virus

Abstract Viruses are a major control on populations of microbes. Often, their virulence is examined in controlled laboratory conditions. Yet, in nature, environmental conditions lead to changes in host physiology and fitness that may impart both costs and benefits on viral success. Phosphorus (P) is a major abiotic control on the marine cyanobacterium Synechococcus. Some viruses infecting Synechococcus have acquired, from their host, a gene encoding a P substrate binding protein (PstS), thought to improve virus replication under phosphate starvation. Yet, pstS is uncommon among cyanobacterial viruses. Thus, we asked how infections with viruses lacking PstS are affected by P scarcity. We show that the production of infectious virus particles of such viruses is reduced in low P conditions. However, this reduction in progeny is not caused by impaired phage genome replication, thought to be a major sink for cellular phosphate. Instead, transcriptomic analysis showed that under low P conditions, a PstS-lacking cyanophage increased the expression of a specific gene set that included mazG, hli2, and gp43 encoding a pyrophosphatase, a high-light inducible protein and DNA polymerase, respectively. Moreover, several of the upregulated genes were controlled by the host’s phoBR two-component system. We hypothesize that recycling and polymerization of nucleotides liberates free phosphate and thus allows viral morphogenesis, albeit at lower rates than when phosphate is replete or when phages encode pstS. Altogether, our data show how phage genomes, lacking obvious P-stress–related genes, have evolved to exploit their host’s environmental sensing mechanisms to coordinate their own gene expression in response to resource limitation.


Introduction
Viruses encounter their hosts under a range of environmental conditions.Classically, infection under sub-optimal conditions leads to lysogeny in temperate phages that have a choice between lytic and lysogenic life cycles [1,2].Additional studies have shown that the decision-making process leading to a lysogenic lifestyle is not necessarily as straightforward as previously thought and can be the result of other environmental triggers than low nutrient levels, such as high microbial density [3].In contrast, obligately lytic phages have no choice but to pursue a lytic cycle once they infect their host.While much of phage biology is based on laboratory experimentation under defined nutrient conditions, albeit usually using rich media, relatively little is known of phage infection dynamics under conditions of host nutrient limitation.
For many environmental bacteria, such as cyanobacteria, growth under nutrient-limiting conditions is the norm, especially those marine genera that occupy the vast oligotrophic ocean gyres that are particularly impoverished for the macronutrients N and P [4].While many of these marine cyanobacteria have optimized their physiology, biochemistry, and genomic footprint (see [5,6]) to optimize growth under such harsh surroundings, infection by a phage adds yet another burden on the cell.However, compared with host nutrient stress responses, relatively little is known about how such phage-host interactions play out under such suboptimal growth conditions for the host, particularly the molecular mechanisms that allow phages to lyse such a "sick" host.
Previous work on bacteriophage T4, infecting Escherichia coli cells in stationary phase, has shown that bacteriophages infecting cells deprived of nutrients can employ different strategies to deal with this deprivation, in this case altering life strategies from a regular infectious mode to one of "hibernation," where some phage proteins are produced yet there is no complete synthesis of mature virions until nutrients become available again [7].Another infection strategy is scavenging of the few nutrients available, together with the utilization of host cellular building blocks, which often get degraded in the course of infection [8,9], and the production of greatly reduced viral progeny from starved cells [7,10].
Similar to this E. coli work, cyanophage S-PM2d infection of the model marine cyanobacterium Synechococcus sp.WH7803 during growth under phosphate-deplete conditions also showed evidence of perturbed infection dynamics, noticeably an extended latent period [11].In contrast, no significant difference in the timing of cell lysis was observed with cyanophage P-SSM2 infecting the marine Prochlorococcus strain NATL2A [12] and cyanophage S-SM1 infecting Synechococcus sp.WH8102 [13], following P-deplete host growth.It was suggested that the expression of a gene encoding a phage version of the high affinity phosphate periplasmic binding protein PstS was a key factor here, potentially maintaining phosphate uptake at a time when host gene expression was being shut down [12].Indeed, expression of this P-related auxiliary metabolic gene (AMG) appears to be controlled by the host twocomponent PhoBR system, comprising a sensor (PhoR) and a DNA binding response regulator (PhoB), see [14] for a review, suggesting that phages can even hijack host regulatory networks in order to selectively overexpress specific AMGs under the relevant nutrient conditions [12,13].
Not all phages infecting marine cyanobacteria, which thrive in nutrient-deplete environments, possess obvious orthologues of these host metabolic genes, and, given that such types of suboptimal infections likely play out in great number in real-world environments, such as ocean gyres where nutrient impoverished sub-optimal host growth occurs, it is of general interest to investigate how obligately lytic phages optimize these infections under nutrient limiting conditions.
In this study, we used a cyanophage-Synechococcus model system to specifically address this question, focusing on a phage that lacks obvious AMGs that could potentially help it optimize the infection process under P-deplete conditions.Using wholecell transcriptomics and electrophoretic mobility shift assays, we describe a new network of genes up-regulated in a cyanophage genome during infection of a P-deplete host, but controlled by the host PhoBR system, that are widely distributed amongst cyanophages and potentially give a broad indication of how a variety of cyanophages undergo a lytic infection under suboptimal host growth conditions.

Culture growth conditions
Synechococcus sp.WH7803 (https://roscoff-culture-collection.org/ rcc-strain-details/752) cultures were grown in defined artificial seawater (ASW) medium [11], containing either 172-μM K 2 HPO 4 for phosphate-replete (ASW + P) growth or transferred at a 1:10 volume into ASW medium lacking phosphate (ASW-P), see below, under continuous illumination at a light intensity of ∼10-μmol photons m −2 s −1 with constant shaking at ∼220 rpm.Cyanophage S-PM2d [15,16] was propagated on Synechococcus sp.WH7803 grown in ASW + P medium.Upon lysis, the lysed culture was filtered through a 0.22-μm pore size vacuum filter (Corning, Corning, NY, USA) and the lysate concentrated with 10% w/v PEG 8000.Concentrated phages were resuspended in ASW + P or ASW-P, the PEG removed by centrifugation with chloroform and phages further concentrated in the appropriate medium, using Amicon filter columns (50 000 kDa MW cut-off, Merck, UK).The phage titre was measured via a plaque assay method [17].Infection parameters were measured either using plaque assays or via one-step growth experiments performed in 96well plates.To elicit phosphate-deplete infection dynamics daily infection assays were performed using 96-well plates as follows: Synechococcus sp.WH7803 (100 ml) was grown in ASW + P medium to an OD 750 = 0.35.The culture was then transferred into 1 L of fresh ASW + P or ASW-P medium.These freshly transferred cultures were grown under the same conditions as the initial starter culture until reaching an OD 750 = 0.25.Then, infection assays were performed daily, with Synechococcus cell numbers being enumerated by f low cytometry using a FACScan (Becton Dickinson, Franklin, NJ, USA).Each sample count was normalized to counts of multif luorescent beads (Polysciences, Warrington, PA, USA), the concentration of which was prior estimated via f luorescent microscopy.Red and orange f luorescence, accounting for chlorophyll/allophycocyanin and phycoerythrin, respectively, were measured through FL3 (650 nm) and FL2 (585/42 nm) filters, respectively.Cell and bead counts were collected using CellQuest software (Becton Dickinson, UK).P-replete and P-deplete Synechococcus sp.WH7803 cells (200 μl) were aliquoted into a 96-well plate in triplicate and cyanophage S-PM2d of known titre added to give a multiplicity of infection (MOI) of 10.A no virus control was also performed, adding ASW or ASW-P medium (100 μl) instead of phage, to wells containing the uninfected host grown under either P-replete or P-deplete conditions.The microtitre plate was then incubated under continuous illumination at an intensity of ∼10-μmol photons m −2 s −1 and the OD 750 measured every 2 h using an iMark Microplate reader (Bio-Rad, UK).A delayed infection phenotype was indicated when a lysis delay of more than 2 h in the P-deplete, compared with the P-replete culture, was observed.When cells were shown to be sufficiently P-stressed to exhibit a delayed infection, a large-scale infection experiment was performed, and samples were taken every 3 h for qPCR and RNAseq experiments (see below).

PCR assays to assess the presence of pstS in phage genomes
The presence of pstS in phage isolates was assessed via PCR amplification using degenerate primers (see Supplementary Table 1).Degenerate primers were designed using the HYDEN program [18].As an input for the HYDEN script, sequences of putative pstS genes from six sequenced cyanophages were used (see Supplementary Table 2).PCR amplifications were performed using MyTaq Mix (Meridian, Memphis, TN, USA) according to the manufacturer's instructions, over 35 amplification cycles with an annealing temperature of 55

qPCR assays
Upon onset of the lysis delay, as detected by the 96-well plate assay described above, Synechococcus sp.WH7803 cells growing in either P-replete or P-depleted ASW medium were enumerated using f low cytometry, as described above, and cultures diluted to a final concentration of 5 × 10 7 cells ml −1 in ASW/ASW-P medium, respectively.Diluted cultures (5 ml) were aliquoted into polycarbonate tubes and infected with cyanophage S-PM2d at an MOI of 10 in triplicate.Uninfected cultures were used as a control.Samples were taken at time points 0, 2, 4, 6, 8, 10, 12-, 14-, 16-, and 18 h post-infection, and 200 μl of each sample was diluted in 500-μl ASW medium and vacuum filtered through a 0.2-μm pore size polycarbonate Isopore filter (Merck Millipore, Burlington, MA, USA), mounted on a glass filter tower.While still on the filter, samples were washed three times with 1-ml preservation solution (100-mM EDTA, 500-mM NaCl, 10-mM Tris-HCl, pH 8.0).The filters were then inserted into a ribolyser Lysing Matrix E tube (MP Bioproducts, Irvine, CA, USA) and snap-frozen in liquid nitrogen.An additional 100-μl infected culture was fixed with paraformaldehyde (1% w/v final concentration) and stored at 4 • C for subsequent cell counting by f low cytometry.
To obtain cell lysates, 650-μl Tris-HCl pH 8.0 was added to each filter-containing ribolyser tube and cells lysed via 3 cycles of 30 s shaking at 30 Hz in a TissueLyser Qiagen (Retsch GmbH, Germany).The tubes were then subjected to centrifugation at 10 000 g and the supernatant snap-frozen in liquid nitrogen and stored at −80 • C. To calculate the intracellular phage DNA concentration, samples were diluted 1:10 (v/v) and qPCR performed as described elsewhere [19].

Whole-cell transcriptomics
A 1-l culture of Synechococcus sp.WH7803 was grown under Preplete and P-deplete conditions, until a delayed lysis phenotype was observed in the P-deplete compared with the P-replete culture (as detected via the 96-well plate assay, see above).Cells were then enumerated using f low cytometry, as described above, and cultures diluted to 5 × 10 7 cells ml −1 in ASW + P/ASW-P medium, respectively.The ASW + P and ASW-P cultures were then divided into six replicates of 100 ml each.Three replicates of each condition were infected with cyanophage S-PM2d at an MOI of 8, while the remaining three replicates were left uninfected as the nophage control.At each time point, 15-ml sample was taken from each replicate and filtered through a 0.2-μm pore size polycarbonate filter.Filters were washed three times using preservation buffer (100-mM EDTA, 500-mM NaCl, 10-mM Tris-HCl, pH 8.0), snap frozen in ribolyser tubes in liquid nitrogen, and stored at −80 • C.
RNA was extracted and purified according to a previous publication [20].The RNA concentration was quantified with a Qubit Fluorometer (Invitrogen, Waltham, MA, USA) using the Qubit RNA HS Assay and RNA integrity verified using Bioanalyzer RNA 6000 Pico chips (Agilent, Santa Clara, CA, USA).
Samples for RNA sequencing were sent to the Next Generation Sequencing Facility at the Leeds Institute of Biomedical and Clinical Sciences, St James University Hospital, University of Leeds.Libraries were prepared using the ScriptSeq Complete Kit (Bacteria) and ScriptSeq v2 RNA-seq Kit (Illumina, San Diego, CA, USA).Ribosomal RNA was removed using the RiboZero Kit -Bacteria (Illumina).Paired end sequencing was performed on the HiSeq 3000 platform, producing 150 base reads.
Paired reads were mapped to both the cyanophage S-PM2d and Synechococcus sp.WH7803 genomes using the BWA MEM v0.7.13 [21] with default parameters.The resulting SAM files were converted to BAM files and sorted using SAMtools v1.3.1 [22].To count the reads mapping to specific S-PM2d loci, sorted BAM files were first converted to BED files using the Bedtools bamtobed script [23].BED files were then used to count the reads mapping to different S-PM2d genes using the Bedtools intersect script with "-c -bed -s" options.Reads were normalized using the RPKM model, giving an estimate of relative gene expression calculated according to the following equation: where Read Count represents the number of reads mapping to a specific locus, Gene Length is the length of that locus, and Total Reads represents the number of total reads out of each sample mapping to the S-PM2d and Synechococcus sp.WH7803 genomes.
To establish changes in gene expression of S-PM2d genes between the corresponding time points in ASW + P and ASW-P infected cultures, differential gene expression values were calculated using the DESeq2 R package [24] in the R Studio environment.The gene expression fold change values, as well as statistical significance, were calculated using the DESeqDataSet-FromHTSeqCount function from the DESeq2 package.Only genes with a False Discover Rate (FDR) P-value <.05 were considered to be differentially expressed between the conditions.

Bioinformatics analysis of putative pho boxes in differentially regulated S-PM2d genes
Promoters of S-PM2d genes identified as differentially expressed under −P conditions were examined for the presence of putative PhoB-binding motifs.Promoter sequences were analyzed using the Pattern Locator script [25].The consensus sequence previously identified as a putative Pho box binding site in cyanobacteria -5 -PyTTAAPyPyT/A-3 - [26] was used to scan the upstream region of these S-PM2d genes and several potential binding sites were identified.

Over-expression of the Synechococcus sp. WH7803 PhoB protein in E. coli
The Synechococcus sp.WH7803 phoB gene (Syn_WH7803_01545) was codon optimized for expression in E. coli and cloned into the pMAL c4X expression vector (Genscript, Piscataway, NJ, USA) to produce a maltose binding protein (MBP)-PhoB fusion protein.
The construct was used to chemically transform NEBExpress competent E. coli and transformants selected using 100-μg ml −1 ampicillin.After reaching OD 600 ≈ 0.5 in LB medium, PhoB protein expression was induced using 0.3-mM IPTG for 4 h.Cells were then harvested by centrifugation, resuspended in BugBuster lysis buffer (Merck Millipore), and lysed using a French Press.Soluble and insoluble fractions were separated via centrifugation and the overexpressed MBP-PhoB fusion protein purified via an AKTA fast protein liquid chromatography (FPLC) system (GE Healthcare, Chicago, IL, USA) using an MBPTrap column (GE Healthcare).The amount of purified MBP-PhoB was quantified using a Bradford assay (Thermofisher, Waltham, MA, USA) according to the manufacturer's instructions.To purify the MBP protein alone, which was used as a control in Electrophoretic Mobility Shift Assays (EMSA), over-expressed MBP-PhoB fusion protein was eluted with 10-mM maltose in loading buffer (20-mM Tris, pH 8; 2-mM MgCl 2 , 200-mM NaCl) and Factor Xa protease (NEB, USA) added to a final concentration of 10 μg ml −1 in loading buffer containing 2-mM CaCl 2 and incubated at 4 • C overnight on a shaker.Factor Xa protease was removed using a HiTrap Q XL column (Cytiva, UK) washed with loading buffer.Separated MBP and PhoB proteins were eluted using a 0-1 M NaCl gradient.Purified MBP eluted at ∼300 mM NaCl and PhoB at ∼500-mM NaCl.Fractions were collected, concentrated, and further purified by size-exclusion chromatography in loading buffer using a Superdex 200 10/300 column (Cytiva, UK) and an AKTA FPLC system (GE Healthcare, Chicago, IL, USA).Note that purified PhoB alone was insoluble as evidenced by the formation of a precipitate in solution after removal of the MBP tag, elution of the protein as an unexpectedly high molecular weight peak on the AKTA column and protein stuck in the well in native PAGE.

Genome sequencing of cyanophages S-BM1 and S-BM3
An exponentially growing Synechococcus sp.WH7803 culture (100 ml) was infected with 100 μl of filtered cyanophage (S-BM1 or S-BM3) lysate and incubated for several days at 23 • C, under illumination at a light intensity of ∼10 μmol photons m −2 s −1 with constant shaking at ∼220 rpm.Lysates were subsequently centrifuged at 13 000 g to remove cell debris and filtered through 0.22-μm pore size syringe filters.Filtered lysates were then used to extract viral DNA, using a phenol-chloroform extraction method described in [27].

Phylogenetic analysis
Cyanophage and cyanobacterial PstS sequences were extracted from the NCBI nr database, Cyanorak (http://abims.sb-roscoff.fr/cyanorak/), and Cyanobase (http://genome.microbedb.jp/cyanobase/), using BLAST.Sequences were aligned using ClustalO [33].Trees were produced using FastTree 2.1 [34] with default settings, under the Jones-Taylor-Thorton ML model, visualized, and annotated using the iTol phylogenetic trees online annotation tool [35].Kyanoviridae PstS protein sequences were recovered using a uBLAST search against all published Kyanoviridae phage genome sequences [31], using the uSEARCH v11 script with the following parameters "-ublast -evalue 1e-20" [36].Sequences resulting from this query were added to the list of orthologous proteins, aligned, ordered in a phylogenetic tree, visualized, and annotated as described above.To examine congruency between the Kyanoviridae cyanophage core tree and the resulting PstS tree, pairwise leafleaf distances were extracted from each tree.Weighted and unweighted Robinson Foulds distances [37] were calculated using TreeDist in R.
The Kyanoviridae cyanophage core genome tree was generated using a list of 57 core genes (Supplementary Table 5), as defined in [38,39].Hidden Markov Model (hmm) files of each of the core genes were generated using HMMER 3.3.2,with hmmbuild with default settings, followed by hmmpress.Thus, generated hmm files were used to query a database of curated cyanophage genomes [31], and MAFTT v7.490 [40] was used to create separate alignments of each core gene homologue using a custom Perl script listed in Supplementary Data 2. The alignments were concatenated using catfasta2phyml.pl(https://github.com/nylander/catfasta2phyml), and a phylogenetic tree generated using FastTree 2.1 [34] with default settings, under the Jones-Taylor-Thorton ML model, visualized and annotated with the iTol phylogenetic trees online annotation tool [35].The cyanophage DNA polymerase tree was produced by performing a protein BLAST similarity search against a custom database containing protein sequences of cyanophage genomes using an e-value threshold of 1E−7.The list of 466 cyanophage genomes was generated by searching the INPHARED database [31] for words "cyano," "Prochlorococcus," "Synechococcus," and the names of cyanobacterial genera listed in Supplementary Table 2 of a previous publication [41].The sequences were aligned and the tree was produced in a manner described above.
A random representative list of cyanophages belonging to the Kyanoviridae (previously Myoviridae) and Autographiviridae (previously Podoviridae) families (roughly corresponding to myovirus and podovirus morphotypes, respectively) was generated, covering different clades of the DNA polymerase tree shown in Supplementary Fig. 7 (chosen genomes are marked by a star symbol).For each of those genomes, a Genbank file was generated, containing the DNA polymerase gene and 5 genes upstream and downstream from it, using a custom-made python script listed in Supplementary Data 2.Those GenBank files were used to demonstrate synteny between gene neighbours of the DNA polymerase genes, using Clinker [42].

Evolution of PstS within cultured Kyanoviridae cyanophages
To understand the distribution and evolution of the pstS gene in cyanophages, we analyzed 95 currently available complete genomes from the Kyanoviridae cyanophages.Of these, pstS was present in 30 isolates (∼1/3).Genomes containing pstS were distributed across the entire phylogeny of this group (Fig. 1A), suggesting either widespread loss or frequent acquisition of the gene.To understand which, we constructed a phylogenetic tree of cyanophage PstS proteins as well as freshwater and marine host proteins (Fig. 1B).Unlike E. coli, pstS in cyanobacteria is often found in multi-copy (see [43]).Here, we utilize the Cyanorak database [44] to assign a sphX cluster (CK_00001829) and two pstS clusters (CK_00000023 and CK_00043821), while taxonomically we also identify a discrete β-cyanobacterial, "freshwater-dominated" PstS cluster which encompasses both the pstS1 and pstS2 genes characterized experimentally by [43].Cyanophages possess only pstS cluster CK_00043821.The absence of sphX or pstS cluster CK_00000023 in cyanophages is notable, given that Synechococcus hosts frequently possess all three types (themselves often in multiple copy), whereas Prochlorococcus hosts only possess the pstS CK_00043821 cluster.The freshwater PstS cluster shown in Fig. 1B encompasses proteins encoded by the pstS1 and pstS2 genes in Synechocystis sp.PCC6803 which are known to have quite different transport kinetics for inorganic phosphate [43].Thus, it is difficult to predict functional properties of PstS proteins from sequence alone, which precludes understanding why cyanophages have only co-opted the CK_00043821 PstS cluster.
Our phylogeny suggests that pstS has been acquired twice in marine cyanophages, once from Synechococcus and once from Prochlorococcus (Fig. 1B).In most cases (24/30), the donor genus (Synechococcus or Prochlorococcus) corresponds to the isolation host of the phage, even though isolation host is not phylogenetically conserved (Fig. 1A).Thus, it is likely that these PstS variants facilitate specific interaction with the cognate PstABC membrane bound components of the transport system of their preferred host (see also Zhao et al. [45]).In the six cases where pstS has not been acquired from the host genus, it is not clear what the host ranges of these phages are, and therefore, the interaction of these "incorrect" PstS types with their host's PstABC requires further investigation.Within each acquisition event, we assessed the congruency between PstS subtrees and the core phylogeny.For those pstS genes acquired from Prochlorococcus, there was a significant but weak correlation between phylogenetic tree topologies.For Synechococcus, there was no significant correlation (Fig. 1C), suggesting that once acquired, there is frequent horizontal gene transfer of pstS between phages.This may explain their rather patchy distribution across the core phylogenetic tree ( Fig. 1A).The distribution of pstS in these phage genomes could ref lect niche adaptation in low P environments, as has been observed in marine metaviromes [46][47][48].Similar to previous reports [39], we observed that the phoA gene encoding a putative alkaline phosphatase is only present in phage genomes that also possess a pstS gene (Fig. 1A).

PstS-lacking cyanophages show delayed lysis in response to low P
To assess whether the absence of pstS in cyanophage genomes more generally correlated with a delayed lysis phenotype [11], we tested PstS-lacking cyanophages isolated from a variety of marine environments under +P/−P conditions (Supplementary Table 3).We used both plaque assays (Fig. 2A and B), measuring the yield and the potential delay in production of plaques, as well as onestep infection assays, measuring the OD 750 of phage-infected cells over the latent period of infection that indicates when cell lysis occurs (Fig. 2C and D).We further used degenerate primers targeting the phage pstS (Supplementary Table 1), to assess the presence of pstS in genomes of cyanophages isolated from a variety of marine environments (Supplementary Table 3).
We found a strong correlation between the absence of a pstS homologue in the phage genome and the presence of the delayed lysis phenotype when infecting under −P conditions.Of the 18 phages tested, 17 lacked pstS and showed a severe delay in lysis (see Fig. 2C for one example).In contrast, a newly sequenced phage, S-BM1, which encodes PstS, had identical latent periods under both P deplete and P replete conditions (Supplementary Table 3, Fig. 2D) in agreement with previous data for two other pstS-possessing cyanophages P-SSM2 and S-SM1 that also show no delayed lysis [12,13].Therefore, it appears that cyanophage-encoded PstS supports the phosphorus demand of an infected cell to avoid delayed lysis.

Is the DNA replication rate affected in pstS-lacking cyanophage infecting a P-deplete host?
Since phage PstS likely controls proper lysis time in response to P availability, we turned our attention to phages lacking pstS, in a bid to understand how they were limited.We hypothesized that phages lacking pstS would have a decreased DNA replication rate when infecting a P-deplete host.However, for the pstSlacking cyanophage S-PM2d, phage DNA replication rates were not statistically different 2-6 h post-infection between the P-replete and P-deplete conditions (2 Sample T-test values: t = 0.78, P = .48,Fig. 3A and C).The dramatic impact on phage infection kinetics with a lack of P in the medium on those cyanophages that do not possess pstS (Fig. 2A and C, Supplementary Table 3) points toward a distinct molecular mechanism, either controlled by the host or the phage, by which the infection process is delayed, and the burst size reduced, despite still maintaining similar DNA replication rates (Fig. 3A and C).From the host perspective, this mechanism would reduce the rate of infection of the P-depleted host by the phosphorus-demanding phages, allowing the persistence of the host population throughout periods of P-scarcity in the dynamic marine environment.From the cyanophage perspective, the altered kinetics would mean an alternative to undertaking an abortive infection pathway, and even though the infection is significantly affected and sub-optimal, the low number of viral progeny produced would still allow phage persistence, facilitating future host infection once the depleted nutrient(s), in this case P, become available again.
How then, does the cyanophage maintain a DNA replication rate similar to nutrient replete conditions when infecting a Pdeplete grown host?Or how does the phage (or indeed host) control the latent period or burst size?To determine this, we undertook a transcriptomics approach comparing gene expression under P-replete/P-deplete host growth.

Differential gene expression in response to P availability
Synechococcus sp.WH7803 grown in either P-replete or P-deplete conditions was infected with cyanophage S-PM2d at an MOI of 8, and samples taken for RNA analysis at t = 3, 6, and 9 h or additionally at t = 12 and 15 h post-infection under P-deplete host growth.Sequence coverage and mapping statistics of this RNAseq data are shown in Supplementary Table 4.We noted that the relative abundance of transcripts mapping to the Synechococcus host genome in the infected samples is significantly lower, compared with the uninfected control (Supplementary Fig. 1 and Supplementary Table 4), pointing to the immediate degradation of the host transcriptome, similar to what has been observed previously [8,19].This reiterates the importance of AMGs in phage genomes, with host transcription being abruptly halted and phage relying only on host genes expressed in the early stages of infection as well as its own genes to facilitate the metabolic needs of its infection cycle.
We did not observe any differentially expressed host genes in response to P conditions in infected cells (data not shown).We expect that this is due to the above-mentioned degradation of host mRNAs early in infection.In contrast, 14 S-PM2d genes were statistically significantly up-regulated 3 h postinfection under P-deplete conditions, compared with the Preplete control (Table 1; Supplementary Fig. 2).These included six S-PM2d genes located in close proximity on the genome, i.e.S-PM2d131 to S-PM2d136 of which S-PM2d133 encodes DNA polymerase (gp43), S-PM2d134-a UvsX RecA-like recombination protein, S-PM2d135-a DNA primase-helicase (gp41), and S-PM2d136-the MazG pyrophosphatase [49].Most of the other 8 significantly up-regulated genes are annotated as encoding hypothetical proteins apart from S-PM2d118 (gp46), encoding a recombination endonuclease subunit, and S-PM2d172, encoding a high-light inducible protein.There were an additional three genes that were identified as differentially expressed but are probably misannotated in this reference genome (Table 1).
Table 1.Differentially expressed cyanophage S-PM2d genes during infection of Synechococcus sp.WH7803 under P-deplete compared with P-replete conditions.

P-value
Presence in % of cyanophage genomes

S-PM2d004
CFW42138.The CDS is on the negative strand and contains no orthologs in cyanophage genomes, whilst there is a more probable overlapping ORF on the positive strand which is shared amongst cyanophages but is not differentially expressed.b These are very small ORFs (<50 aa's), not shared across cyanophage genomes, and are located within a tRNA operon.
Table 2.The sequence and position of putative Pho boxes found upstream of the differentially expressed cyanophage S-PM2d genes during infection of Synechococcus sp.WH7803 under P-deplete compared with P-replete conditions.Bold nucleotides represent those pertaining to the Pho box, while underlined nucleotides are conserved compared with the recognized consensus Pho box binding site (5'-PyTTAAPyPyT/A-3 ) [26 ].NA-no putative Pho box was identified.Position is the number of nucleotides upstream of the putative transcriptional start site of that gene.

Several of the differentially regulated S-PM2d genes under P-deplete host growth are controlled by the Synechococcus PhoBR system
Since these cyanophage S-PM2d genes were specifically upregulated following the infection of a P-deplete host, we sought to understand how these genes were up-regulated.Given it is known that the pstS gene in cyanophage P-SSM2 infecting marine Prochlorococcus is controlled by the host two-component system PhoBR [13].we investigated whether this was the case for these genes as well.Using a Pho box consensus motif previously described for marine Synechococcus as comparison [26], bioinformatics and visual inspection of promoter regions showed that 7 of the 14 up-regulated genes possessed a putative Pho box (see Table 2).To experimentally confirm that some of these putative Pho boxes were functional, the Synechococcus sp.WH7803 PhoB protein was over-expressed and purified in E. coli as a MBP-PhoB fusion protein and the purified protein used in electrophoretic mobility shift assays with DNA fragments of S-PM2d upstream regions either (i) containing a putative Pho box and differentially expressed (genes S-PM2d004, S-PM2d133, S-PM2d136), (ii) containing a putative pho-box, but not differentially expressed (gene S-PM2d130), or (iii) lacking a predicted Pho box but differentially expressed (S-PM2d134).
Of the upstream regions of the genes tested, the MBP-PhoB fusion protein bound to the promoters of S-PM2d004 (encoding a hypothetical protein), S-PM2d133 (encoding DNA polymerase), both with a predicted Pho box but also to SPM2d134 (encoding a putative UvsX, RecA-like protein), which lacks a predicted Pho box (Fig. 4).These three promoters showed no binding to the purified MBP alone (Supplementary Fig. 6).S-PM2d136 showed no binding to the PhoB-MBP fusion (Supplementary Fig. 3), despite being differentially expressed in response to low P and possessing a predicted Pho box.However, this gene may be co-transcribed with the upstream genes S-PM2d135 and S-PM2d134, the latter experimentally determined to bind PhoB.Alternatively, one or more of the over-expressed genes may be controlled by another transcription factor that is in turn under PhoB control.Gene S-PM2d130 had a predicted Pho box but neither bound PhoB (Supplementary Fig. 6) nor was differentially expressed in response to low P. Thus, bioinformatics prediction of Pho boxes in phage genomes is not reliable, and instead experimental validation of PhoB binding better predicts whether the gene was differentially expressed in response to low P. To confirm the specificity of PhoB binding, we performed competition assays in which f luorescently labelled upstream DNA fragments for S-PM2d004 and S-PM2d134 were competed by identical unlabelled DNA fragments, where the binding of the protein to the promoter was outcompeted by increasing the concentration of the unlabelled promoter-containing DNA fragment (Supplementary Fig. 4A), while competition with a non-binding fragment (S-PM2d136) showed no effect (Supplementary Fig. 4B).
In order to confirm that it is indeed the predicted Pho box to which the overexpressed PhoB protein is binding, we progressively reduced the size of these fragments to show that for both the S-PM2d004 (Supplementary Fig. 5A) and S-PM2d133 (Supplementary Fig. 5B) genes only those DNA fragments containing the predicted PhoB binding site resulted in a gel shift.

Discussion
Cyanophage infection represents a potentially important loss factor for picocyanobacterial populations throughout the global ocean via cell lysis (e.g.see [50]) but also has wide ranging implications for estimates of primary production given their ability to directly inhibit CO 2 fixation [51].Thus, given that cyanophage infections can theoretically occur under every environmental condition possible, it is important that we have a mechanistic understanding of how infection dynamics vary as a function of these in situ host growth conditions, given this may directly affect the latent period (the time before cell lysis, but also the time during which CO 2 fixation is inhibited) and burst size (the number of phages produced).
Phosphorus availability has previously been shown to affect cyanophage productivity in picocyanobacteria [12,13] but with no observed delayed lysis period.Here, we found a more general relationship between the presence of a pstS homologue in phage genomes and the lack of the delayed lysis phenotype under −P conditions (Fig. 1A; Supplementary Table 3).The acquisition of pstS may therefore enable cyanophages to overcome a bottleneck affecting lysis during phage morphogenesis.Phage pstS transcripts can be used to either produce additional copies of the periplasmic P-binding PstS protein, thus increasing the ability to acquire any available phosphate, or they can supplement the mRNA encoding the host pstS which is degraded together with the rest of the host transcriptome during the course of infection ( [8]; Supplementary Fig. 1), thus providing a continuous supply of the PstS protein in the periplasm of the infected cell.A similar strategy has been shown to be used by cyanophages carrying a photosynthetic gene psbA, which encodes a functional D1 protein and enables maximal energy production under high-light conditions in the infected cyanobacterial cell [19,52,53].
Phylogenetic examination of the cyanobacterial pstS shows strong clustering between the viral genes and the genes of the specific host these viruses infect (Fig. 1B).Thus, pstS sequences from Synechococcus phages form a sister clade to the Synechococcus pstS sequences, while the Prochlorococcus viral pstS sequences group together with the cyanobacterial Prochlorococcus pstS clade.This clear phylogenetic relationship becomes less obvious when looking at the totality of Kyanoviridae phages.Here (Fig. 1A), the P-stress-related gene pstS, as well as the putative alkaline phosphatase-encoding phoA, seem to have evolved numerous times in different phages and do not show clustering that is apparent when looking only at the cyanobacterial and cyanophage pstS sequences (Fig. 1B).Additionally, there seems to be no synteny conservation between pstS homologs found in different isolated cyanophage genomes, further supporting the multiple origin theory [54].This observation points to the possibility that adopting host pstS is not the only strategy available to phages infecting hosts under P-deplete conditions and that alternative infection strategies, one of which is described in this work, may be more prevalent than we previously thought.The existence of alternative strategies for optimizing infection under different P-conditions, the lack of synteny of pstS-surrounding genes between cyanophage isolates, and its apparent sporadic and repeated acquisition over ecological timescales, all suggest that, unlike their cyanobacterial hosts, cyanophages have not evolved into phosphorus-adapted ecotypes.
We have also shown that the delayed lysis phenotype of cyanophage S-PM2d during infection of a P-deplete host [11] appears to be a general phenotype of pstS lacking cyanomyoviruses (Supplementary Table 3).Given that S-PM2d genome replication rate was comparable during the infection of a P-replete or P-deplete host (Fig. 3A and C) excludes this as the underlying mechanism.Indeed, we show that under P-deplete conditions, phage genes for DNA replication are specifically upregulated and that phage genomes have evolved to exploit the host's regulatory system to control this.Since the infectious burst size is 5-fold lower under P-deplete conditions (Fig. 2; [11]), replicated phage genomes do not appear to get incorporated into infectious virions.How and why cyanophages maintain the synthesis of DNA, which is extremely phosphorus demanding (e.g.almost 50% of cellular phosphorus is in DNA [55]), is initially puzzling.However, cyanophages have an abundant source of nucleotides in the form of the host's chromosome which is immediately degraded upon infection [56], liberating free nucleotides.In E. coli phage T4, host chromosome degradation is catalyzed by a combination of the endonuclease II product of the denA gene and the gene product of d2a [57].While no homologues of these exist in Kyanoviridae cyanophages, it is likely that another endonuclease performs this function.In unrelated cyanophages, host mRNAs are degraded in a process thought be a result of the host's RNase E, with phages protecting their mRNAs by antisense transcription [8].A similar mechanism may exist in Kyanoviridae cyanophages where antisense transcription is observed [8].Free NTPs can be converted to dNTPs by the activity of the cyanophage-encoded ribonucleotide reductase [58,59].Altogether, it is likely that there is enough free or recycled nucleotides derived from host nucleic acids to meet the demands of phage chromosome replication.Moreover, DNA replication itself releases intracellular phosphate, via the pyrophosphatase activity of DNA polymerase [60].Since Synechococcus cells infected under P-deplete conditions produce 12 new cyanophages on average [11,61] and that each S-PM2d genome is ∼200 kb in length [16], while the addition of each nucleotide to the growing DNA chain produces 2 phosphate molecules, we estimate that during the course of DNA replication, ∼10 6 molecules of phosphate are released by the activity of DNA polymerase alone.Previous work has shown that Synechococcus cells growing at external phosphate concentrations of 1 μM accumulate a similar amount of phosphate per cell per hour [62].Thus, DNA replication may provide a rich source of phosphorus for high P demanding processes of phage morphogenesis.
We also observed that the phage mazG (S-PM2d136, Table 1) is up-regulated under P deplete conditions.S-PM2d mazG encodes a pyrophosphohydrolase which was initially hypothesized to play a role in suppressing the stringent response in infected cells [63].However, our previous work showed that the viral MazG does not hydrolyze the alarmone nucleotides ppGpp and pppGpp, but rather hydrolyses nucleotides, with increased affinity for dGTP and dCTP [49].Since cyanophage S-PM2d has a lower %GC (37.7%; [16]) compared with its Synechococcus sp.WH7803 host (60.2%; [64]), it was proposed that the viral MazG has a role in preferentially hydrolysing nucleotides for which it has lower demand, thus providing a potential additional source of intracellular phosphate produced by hydrolysis of less-required nucleotides.This observation further supports our model of amplification of nucleotide metabolism (including DNA replication) processes to provide a novel source of intracellular phosphate.In reality, there may even be an excess of nucleotides for DNA replication.Instead, cyanophages may scavenge phosphate from these to meet other phosphorus demands such as protein synthesis and packaging during the later stages of infection which have been implicated as the main energetic sink for phage production [19].Thus, it is possible that the role of the phosphate produced intracellularly during phage DNA replication is to provide the energy for the later stages of the latent phase of infection.
Since viral DNA polymerase seems to play an important role in the P-stress response of this group of cyanophages, we examined the phylogeny of DNA polymerases from published genomes of isolated cyanophages (Supplementary Fig. 7), as well as the preservation of synteny of DNA polymerase in these genomes (Supplementary Fig. 8).The phylogeny of DNA polymerase seems to align with the cyanophage family phylogeny with the main clades clustering along the family divide, with the Kyanoviridae family members possessing the Type B DNA polymerase, whereas the Autographiviridae contain the Type A DNA polymerase (Supplementary Fig. 7).There is no separate clustering between the Kyanoviridae DNA polymerases belonging to phages infecting Prochlorococcus and Synechococcus hosts, pointing toward a possibility of interchangeable acquisition of the gene from the common host of these cyanophages.Noteworthy is the split nature of the DNA polymerase gene found in most of the Autographiviridae strains (19/32).In these phages, DNA polymerase is divided into two or three ORFs, found next to each other, often partially overlapping, and encoded in a different reading frame to one another.This feature seems to be conserved among cyanophages from this family, as the second and third polymerase fragments all cluster together in a separate branch within the Autographiviridae clade.While this feature has been reported before [65], further work is required to establish whether this curious feature has any biological importance.
For both the Kyanoviridae and Autographviridae DNA polymerases, there is a high degree of synteny conservation within the two clades.In the case of Kyanoviridae polymerases, including the one belonging to cyanophage S-PM2d described in this work, there is the conserved presence of a helicase, UvsX-like recombinase, putative heat-shock protein, and MazG homologues surrounding the DNA polymerase (Supplementary Fig. 8A).Since we find all of these genes, together with the DNA polymerase, to be transcriptionally overexpressed during infection under P-deplete conditions, it is possible that this synteny is conserved as a part of a general cyanophage adaptation to infection under varying P-stress conditions.Certainly, this strategy appears to have been specifically selected for in S-PM2d given DNA polymerase (S-PM2d133) is under direct control of the host Pstress transcriptional regulation response (Fig. 4).Noteworthy here is that the host DNA polymerase is not under the same Pstress control [66,67].It is also notable that in the case of some cyanophages infecting both Synechococcus and Prochlorococcus, the immediate genomic context of the DNA polymerase also contains a pstS homologue.This proximity of pstS and DNA polymerase in some of these Kyanoviridae cyanophage genomes might explain the origin of the Pho-binding motifs in the promoters of these latter genes.The genomes of Autographiviridae cyanophages, although also showing a high degree of synteny around their DNA polymerases (Supplementary Fig. 8B), contain a somewhat different (compared with Kyanoviridae genomes) list of putative genes located in this genomic neighbourhood: ssDNA binding protein, RNaseH, nucleotide kinase, ribonucleotide kinase and RNA polymerase, with the conservation of mazG between these families, a gene which has previously been identified as a part of the core genome of Kyanoviridae cyanophages [38,39].
Overall, our results indicate that bacteriophages not only adapt to nutrient stress by incorporating host metabolic genes into their own genomes [13] but also acquire stress-specific promoters to exploit host transcription factors related to environmental stress.This mechanism presents an exciting new model of viral evolution which should be examined in other virus-host models.

Figure 1 .
Figure 1.The distribution and evolution of phosphate acquisition AMGs in Kyanoviridae cyanophages.(A) Core phylogenetic tree of Kyanoviridae cyanophages constructed from 57 gene markers.The tree is rooted on E. coli phage T4.The host genus, the presence/absence of pstS, phoA, and the delayed lysis phenotype is shown for each taxon.Circles on branch junctions indicate bootstrap values >80% (B).Phylogeny of PstS proteins from cyanophages and cyanobacteria.The tree is rooted with E. coli PstS.Black circles on branch junctions indicate bootstrap values >80%.(C) Comparison of the phylogenetic distance of cyanophage PstS proteins and their phylogenetic distance in the core tree.Grey dots indicate correlation between all cyanophage PstS proteins, while red and blue signify correlations within the red and blue groups highlighted in (B).Output linear regression statistical tests (Pearson's correlation) are shown next to the line.

Figure 2 .
Figure 2. Plaque assay results (A) and (B) and OD 750 values (C) and (D) of cyanophage infection of Synechococcus sp.WH7803 grown under P-replete and P-deplete conditions.Cyanophage used: (A) and (C) S-PM2d; (B) and (D) S-BM1.Error bars represent the standard error of the average of three replicates.T-tests between the number of plaques at the last time point show statistically significant differences between +P and −P-infected samples, both for S-PM2 (t-value: 3.511, P-value: .025)and S-BM1 (t-value: −3.024, P-value: .039).

Figure 3 .
Figure 3.DNA replication rate of cyanophage S-PM2d during infection of Synechococcus sp.WH7803 under P-replete and P-deplete conditions.(A) Percentage of the initial amount of intracellular S-PM2d phage DNA over the course of infection under P-replete and P-deplete conditions.(B) The percentage of initial Synechococcus sp.WH7803 cell abundance following infection with cyanophage S-PM2d under P-replete and P-deplete conditions.(C) The relative cyanophage DNA replication rate under P-replete and P-deplete conditions.The rate was estimated by calculating the slope of a curve representing the change in the amount of DNA per hour for each of the three replicates, between 2-and 6-hour time points post infection.Error bars represent the standard deviation of three replicates.

Figure 4 .
Figure 4. EMSA of purified MBP-PhoB from Synechococcus sp.WH7803 with specific cyanophage S-PM2d gene promoters.The concentration of MBP-PhoB protein used ranged from 0 to 1 μM, while 25-ng DNA fragment was used in each case.(−): Negative control, PhoB with an internal fragment of the Synechococcus sp.WH7803 phoB gene.(+): Positive control, PhoB with the promoter region of the Synechococcus sp.WH7803 phoB gene.