Prevalence and Evolution of Core Photosystem II Genes in Marine Cyanobacterial Viruses and Their Hosts

Cyanophages (cyanobacterial viruses) are important agents of horizontal gene transfer among marine cyanobacteria, the numerically dominant photosynthetic organisms in the oceans. Some cyanophage genomes carry and express host-like photosynthesis genes, presumably to augment the host photosynthetic machinery during infection. To study the prevalence and evolutionary dynamics of this phenomenon, 33 cultured cyanophages of known family and host range and viral DNA from field samples were screened for the presence of two core photosystem reaction center genes, psbA and psbD. Combining this expanded dataset with published data for nine other cyanophages, we found that 88% of the phage genomes contain psbA, and 50% contain both psbA and psbD. The psbA gene was found in all myoviruses and Prochlorococcus podoviruses, but could not be amplified from Prochlorococcus siphoviruses or Synechococcus podoviruses. Nearly all of the phages that encoded both psbA and psbD had broad host ranges. We speculate that the presence or absence of psbA in a phage genome may be determined by the length of the latent period of infection. Whether it also carries psbD may reflect constraints on coupling of viral- and host-encoded PsbA–PsbD in the photosynthetic reaction center across divergent hosts. Phylogenetic clustering patterns of these genes from cultured phages suggest that whole genes have been transferred from host to phage in a discrete number of events over the course of evolution (four for psbA, and two for psbD), followed by horizontal and vertical transfer between cyanophages. Clustering patterns of psbA and psbD from Synechococcus cells were inconsistent with other molecular phylogenetic markers, suggesting genetic exchanges involving Synechococcus lineages. Signatures of intragenic recombination, detected within the cyanophage gene pool as well as between hosts and phages in both directions, support this hypothesis. The analysis of cyanophage psbA and psbD genes from field populations revealed significant sequence diversity, much of which is represented in our cultured isolates. Collectively, these findings show that photosynthesis genes are common in cyanophages and that significant genetic exchanges occur from host to phage, phage to host, and within the phage gene pool. This generates genetic diversity among the phage, which serves as a reservoir for their hosts, and in turn influences photosystem evolution.

The movement of genes between organisms is an important mechanism in evolution.As agents of gene transfer, phages play a role in host evolution by supplying the host with new genetic material [11][12][13][14][15] and by displacing ''host'' genes with viral-encoded homologues [16][17][18].Phage evolution is in turn influenced by the acquisition of DNA from their hosts [13,[19][20][21][22] and by the swapping of genes within the phage gene pool [23,24].Recent evidence suggests that gene flow within the global phage gene pool extends across ecosystems [25][26][27].
Cyanophage genomes bearing key photosynthesis genes psbA and psbD provide a notable example of the co-option of ''host'' genes for phage purposes [13,22,[28][29][30].The psbA and psbD genes encode the two photosystem II core reaction center proteins, D1 and D2 (denoted here as PsbA and PsbD, respectively), found in all oxygenic photosynthetic organisms.It has recently been shown that the phage-encoded psbA gene is expressed during infection [31,32].Because maximal cyanophage production is dependent on photosynthesis [31,33], and the host PsbA protein turns over rapidly [34] and declines during infection [31], expression of these phageencoded genes likely enhances photosynthesis during infection, thus increasing cyanophage fitness.
If photosynthesis genes indeed provide a fitness advantage to cyanophages, one might expect them to be widespread among cyanophage genomes.Through whole or partial genome sequencing, psbA has been documented in three Prochlorococcus cyanophages (one podovirus and two myoviruses) and five Synechococcus myoviruses, whereas psbD was found in only some of these phages [13,29,35].Neither of these genes is found in the Synechococcus P60 podovirus genome [36].A survey of Synechococcus myovirus isolates revealed that at least 37 of them contained psbA [29], and this gene has also been found in cyanophage genome fragments in seawater samples [37].Thus, the presence of psbA is a common, but not universal, feature in the cyanophages examined to date, most of which have been Synechococcus cyanophages.
Using limited genomic sequence data from one Synechococcus and three Prochlorococcus cyanophages, we suggested that both psbA and psbD were transferred as whole genes from host to phage multiple times, but not from phage to host [13].Subsequently, Zeidner et al. [37] analyzed psbA data predominantly from field sequences and suggested that genetic exchanges of segments of the gene (intragenic recombination) may have occurred among host and phage copies in both directions [37].However, this novel and controversial hypothesis requires further investigation with sequences of known organismal origin and using methodology capable of identifying the recombination partners and the directionality of such potential exchanges.
To better describe and understand the phenomenon of photosynthesis genes in cyanophage, we looked for the psbA and psbD genes in 33 cultured cyanophage isolates that infect Synechococcus or Prochlorococcus (or both) and analyzed the sequences of these genes in the context of known host ranges of the phage.This dataset allowed us to address the following questions: (1) How prevalent are both psbA and psbD in cyanophages that infect Synechococcus and/or Prochlorococcus? and ( 2) To what extent have photosynthesis genes, or segments thereof, been moved between and among hosts and phages?

Results/Discussion
Prevalence of the psbA and psbD Genes in Cyanophages The psbA gene was amplified from 28 out of the 33 cyanophage isolates examined (Table 1).Combining these findings with published results (Table 1), we find that the psbA gene is present in 88% of cyanophage isolates examined, including all myoviruses (n ¼ 32) and all five Prochlorococcus podoviruses included in this study.However, this gene was not detected in Prochlorococcus siphoviruses (n ¼ 2) and Synechococcus podoviruses (n ¼ 3), suggesting that there are some combinations of phage family and host genus that do not lead to incorporation of the psbA gene into the phage genome.Six additional phages yielded ambiguous results and were excluded from these analyses (see Materials and Methods for details).
When present, the psbA gene is likely to be functional, as there is evidence for the conservation of amino acid sequences through purifying selection [13,37], and the gene is expressed during infection [31,32], implying that this gene confers a fitness advantage to the phages that carry it [13,22,29,31].Sustained photosynthesis is necessary for maximal phage production [31,33,38], and the long latent period of many freshwater and marine cyanophages (8 h or more; [9,31,33,38]) presumably results in energy-and/or carbon-limitation for phage replication.Thus, cyanophageencoded psbA likely serves to boost the photosynthetic performance of the host during infection, thereby increasing phage production.It is perhaps not coincidental that one of the phages that lacks psbA, Synechococcus podovirus P60 (Table 1), has a latent period of only 1 h (K.Wang and F. Chen, personal communication), which may be too short for psbA expression to be beneficial.Latent period information for marine cyanophages, however, is sparse.It is not known for the Prochlorococcus siphoviruses that lack psbA, and it has only been shown to be .8h for a single phage strain from each of the Synechococcus myoviruses [39] and Prochlorococcus podoviruses [31].Further, theory [40][41][42][43] and experiments [44] suggest that latent period length may be a transient property that rapidly evolves in response to changes in host cell densities.Thus, further exploration of this hypothesis requires analysis of the latent period of many more phage isolates under variable host cell concentrations.
The psbD gene was amplified from 15 out of the 33 cyanophage isolates examined (Table 1).Again, combining our data with published findings, we observe that psbD is found only in isolates that contain psbA and only in myoviruses, but not in all psbA-containing myoviruses.Only four of 12 Prochlorococcus myoviruses (as defined by original host strain of isolation; Table 1) contained psbD, whereas this was the case for 17 of 20 Synechococcus myoviruses.Although it is possible that differences in the photosystem II reaction center between Prochlorococcus and Synechococcus exist (such as differences in the rate of PsbD degradation) and could explain the biased distribution of the psbD gene among the myoviruses, there is no evidence that this is the case.The breadth of phage host ranges (as operationally defined in Table 1), however, appears to be a reasonably good predictor of whether a phage will contain psbD: 17 of 18 broad-hostrange phages encode it, whereas only one out of 21 narrowhost-range phages do so (Table 1).Perhaps broad-host-range phages have co-opted both psbA and psbD to better ensure the formation of a functional PsbA-PsbD protein complex in the host during infection.

Origins and Evolutionary History of psbA and psbD in Cyanophages
To investigate the origins of photosynthesis genes in phages and their hosts, we conducted phylogenetic analyses (using measures to minimize systematic errors; see Materials and Methods) of host and phage psbA and psbD sequences, including new sequence data for nine Synechococcus hosts (psbA), 19 Synechococcus and Prochlorococcus hosts (psbD), and 33 phages (both psbA and psbD).Phylogenetic reconstructions of host psbA and psbD genes in Prochlorococcus showed that well- Presence is indicated by þ, and absence by À.
Phages that contained identical sequences to other phages are noted as ''ID to X.'' a Cultured strain used for isolation of phage from natural seawater samples.Phages are defined as either Prochlorococcus or Synechococcus phages based on original host of isolation, but many of the myoviruses cross-infect both genera.Those phages that cross-infect both genera are marked ''D'', those that do not are left blank, and those that were not available for testing are marked ''na''.
b The number of host strains infected by each phage out of 21 strains tested [5].Operationally defined as narrow if a phage infects less than four hosts within a single cluster determined from 16S-23S rRNA internal transcribed spacer clustering [48] and broad if it infects more than four hosts within a cluster or at least two hosts that span more than one cluster.Small variations in this definition did not significantly affect the conclusions made.e Indicates whether the PCR results were corroborated by genome sequencing.In all cases where the genome sequence became available, it confirmed the PCR results.DOI: 10.1371/journal.pbio.0040234.t001supported sequence clusters contain only one organism type (Figures 1 and 2), with sequences from high-light adapted (HL) and low-light adapted (LL) Prochlorococcus [45] forming discrete clusters.These well-supported Prochlorococcus clusters are similar to those observed using other host genes such as rRNA, rpoC1, and ntcA [46][47][48][49], indicating that psbA and psbD have not been transferred between Prochlorococcus lineages.In contrast, the Synechococcus clusters for both psbA and psbD are poorly supported, a finding different to that obtained using other highly conserved genes [46][47][48][49] and thus may have resulted from genetic exchange between Synechococcus lineages.
The psbA sequences from Synechococcus myoviruses, Prochlorococcus myoviruses, and Prochlorococcus podoviruses generally formed discrete clusters consistent with their host ranges (Figure 1), suggesting that the transfer of photosynthesis genes from host to phage has been largely limited by host range (but see exceptions discussed below).Although many of these phages are capable of infecting both host genera (denoted as ''D'' in all figures), we designated each cyanophage isolate as a Prochlorococcus or Synechococcus cyanophage based upon its original host strain of isolation (as mentioned above and in Table 1).Given this designation scheme, it appears that transfers were predominantly from Prochlorococcus to their phages and from Synechococcus to their phages.This suggests host-range-limited host-to-phage transfer events, Phages are listed by their name, followed by their original host.Phages that are known to infect both Prochlorococcus and Synechococcus hosts are indicated with a ''D''; those that infect only one genus are labeled either P (infect only Prochlorococcus hosts) or S (infect only Synechococcus hosts), while those that are unknown are designated with a ''?''.Phages shown in italics and bracketed with ''**'' were isolated on hosts that do not belong to the same cluster and are thus exceptions to the general clustering pattern (see text).Taxa are color coded according to the following biological groupings: myoviruses (red), podoviruses (black), marine Synechococcus hosts (light blue), marine Prochlorococcus hosts (dark green, LL; light green, HL), freshwater cyanobacteria (dark blue).The tree topology was estimated by LogDet analysis of 1st and 2nd codon positions.Sequences where intragenic recombination was detected using other methods (see Materials and Methods) were not included in these phylogenetic analyses.Branch lengths were estimated by maximum likelihood under a model with nonstationary nucleotide frequencies.Numbers at the nodes represent neighbor-joining bootstrapping and maximum likelihood puzzling support.Anab, Anabaena; Gloe, Gleobacter; HL, high-light adapted; LL, low-light adapted; Syncy, Synechocystis; Thermo, Thermosynechococcus. DOI: 10.1371/journal.pbio.0040234.g001with subsequent horizontal and vertical transfers occurring among viral lineages.
Two isoforms of the PsbA protein are often found in cyanobacteria [50].The PsbA.1 (D1.1) isoform is constitutively expressed, whereas the PsbA.2 (D1.2) isoform is upregulated in response to high light and UV stress [51,52].Many of the differences between the isoforms are found in ten amino acids between position 121 and 312 [50].Based on which isoform the majority of these ten amino acids were identical to (including glutamine/glutamate at position 130), we determined that PsbA from both Prochlorocococus myoviruses and podoviruses are more similar to PsbA.1, the only isoform found in Prochlorocococus hosts so far [53] (unpublished data).Although Synechococcus hosts encode both isoforms (unpublished data), Synechococcus myoviruses encode the stressresponsive PsbA.2 isoform exclusively (unpublished data), which may be particularly beneficial during the stress of infection.These findings are consistent with the hypothesis of host-range-limited transfers of the psbA gene (but see exceptions below).
Host-to-phage transfers appear to have occurred at least four times for psbA and twice for psbD, as seen from the number of discrete clades containing phage-encoded genes in each case (Figures 1 and 2).The four psbA gene acquisitions by phage appear to include two transfer events for the Prochlorococcus myoviruses (Prochlorococcus myovirus group 1 and 2 in Figure 1) and a single event for Prochlorococcus podoviruses all from their Prochlorococcus hosts, as well as a single event for Synechococcus myoviruses from their hosts (Figure 1).The psbD gene appears to have been acquired once by both Synechococcus and Prochlorococcus myoviruses from their respective hosts (Figure 2).Interestingly, the three Prochlorococcus myoviruses that contain psbD all encode Prochlorococcus myovirus group 1 psbA sequences, suggesting that this gene was acquired only once by a subset of these myoviruses.Although the specific source is difficult to determine from phylogeny alone, the placement of the Prochlorococcus myovirus sequence clusters suggests that psbA was derived from either HL Prochlorococcus hosts or LL NATL2A-type hosts, while the psbD genes could have been acquired from any of the Prochlorococcus hosts other than MIT9313/9303.The placement of the Prochlorococcus podovirus (psbA only) and Synechococcus myovirus sequence clusters at the base of the host and virus clades provides little further information about the source of these phage genes.
We found three exceptions to the above host-constrained evolutionary scenario-i.e., cases where phage psbA and psbD genes did not cluster with those of their hosts (Figure 1 and Figures S1 and S2) and did not have PsbA isoforms consistent with that of their hosts (unpublished data).These include two narrow host-range Synechococcus myoviruses (S-ShM1, S-SSM1), which encode psbA sequences most similar to Prochlorococcus myoviruses (Figure 1) even to the extent that they encode the PsbA.1 isoform, as well as a Prochlorococcus myovirus (P-SSM1) with a psbA sequence that is most similar to those from Synechococcus myoviruses (Figure 1) and encodes the PsbA.2 isoform as expected for a Synechococcus myovirus.Although the latter can cross-infect across Prochlorococcus ecotypes, it has not been shown to infect Synechococcus [5].The P-SSM1 phage also encodes psbD, which, like its psbA gene, is more similar to Synechococcus psbD sequences than those of the Prochlorococcus host upon which it was isolated (Figure S2; note that this sequence does not appear in Figure 2 because it was a candidate for intragenic recombination; see Materials and Methods).It is likely that these exceptions to the rather consistent host-phage sequence clustering resulted from horizontal transfer events between a broad-host-range donor phage and a limited-host-range recipient phage during coinfection of a single host, i.e., swapping of genes within the phage gene pool [24].Whole gene transfers within the phage gene pool are likely to be more common than this, but undetectable when occurring within phages that form a discrete phylogenetic cluster.These observations call for caution when using clustering patterns of psbA and psbD sequences from uncultured phage (obtained from environmental genome data) to identify potential hosts.

Intragenic Recombination within Core Reaction Center Proteins
The lack of well-supported clade structure in phylogenetic reconstructions for Synechococcus host strains when using both psbA and psbD differs from those constructed using other genes [46][47][48][49], which led us to wonder about underlying mechanisms that could be responsible for such a blurred phylogenetic signal.In a recent study, Zeidner et al. [37] showed that Synechococcus-phage-like psbA sequences from the environment had a patchy %GþC distribution, which they suggest is due to intragenic recombination [37].Their analyses demonstrated that such recombination had occurred within the inferred-phage clusters and within clusters spanning both phage and host psbA sequences.They could not discern, however, whether the signal was caused only by phage-to-phage exchanges, or included phage-to-host exchanges, because the majority of their sequences were of unknown origin (i.e., they were derived from environment clone libraries), and the test employed does not assess the directionality of intragenic recombination events.Our cultured hosts and phages provide an opportunity to assess recombination partners without ambiguity regarding the source of the genes.In addition, the known host ranges of these phages [5] (Table 1), together with the types of recombination tests we have used (see Materials and Methods), allow us to assess the directionality and the pathways through phages and hosts that these recombination events are likely to have taken.
As a first assessment for potential intragenic recombination, we analyzed the %GþC patterns in all of the psbA and psbD genes (Figures 3 and 4, respectively).Prochlorococcus phage genes had similar average %GþC contents to those from their Prochlorococcus hosts (39%-46%), whereas those of Synechococcus phages had %GþC contents that were lower than those from their Synechococcus hosts (46%-51% versus 56%-62%), but not as low as those from Prochlorococcus hosts and phages.This intermediate %GþC could be the result of intragenic recombination between variants of the two host lineages.Alternatively, it may reflect the current state of mutational amelioration of the acquired gene from a high %GþC source towards the low genome-wide %GþC of the virus (Synechococcus myoviruses S-PM2 and Syn9 both have low genome-wide %GþC; [28]; P. Weigele, W. Pope, G. Hatfull, R. Hendrix, personal communication).If the latter is the case, we might expect such amelioration to be constant across the gene, resulting in an even %GþC distribution pattern.
To help differentiate between these hypotheses, we mapped the %GþC variation across the psbA and psbD genes using the methodology developed by Zeidner et al. [37].We detected patchiness of %GþC in Synechococcus myovirus psbA sequences dispersed along the length of the gene (Figure 3), confirming the findings reported by Zeidner et al. [37].We also detected %GþC patchiness among psbA from Prochlorococcus podoviruses, but not from Prochlorococcus myoviruses, despite overall similarity of their %GþC content with their Prochlorococcus hosts.This suggests that intragenic recombination has occurred among the podoviruses.In addition, patterns of %GþC were not uniform and even markedly clumped across the psbD gene from Synechococcus myoviruses (Figure 4), with the first segment resembling Synechococcus hosts and the last segment resembling Prochlorococcus hosts and their phages.Thus, intragenic recombination is likely to be at least partly responsible for the intermediate %GþC content in Synechococcus myovirus psbA and psbD sequences.
Statistical methods for detecting intragenic recombination (see Materials and Methods) revealed strong evidence for its presence in both the psbA and psbD sequence sets (Tables S1  and S2), but the relative frequency of recombination events was not equal for different groups of hosts and phages.Recombination appears most common among the cyanophages, and more so for Synechococcus than Prochlorococcus phages.Exchanges were detected between phages that infect both Synechococcus and Prochlorococcus as well as within myoviruses that infect a single genus (Synechococcus).Note that exchanges within a single phylogenetic phage cluster, such as within the Synechococcus myoviruses, were undetectable by our previous phylogenetic analyses.Interestingly, our analyses also revealed exchanges between Prochlorococcus-specific podoviruses and broad-host-range Synechococcus myoviruses, with the Prochlorococcus podoviruses serving as the donors (Table S1).Marine cyanobacterial podoviruses contain integrase genes and are thought to have the ability to integrate into the genomes of their hosts as prophages [30] (P.Weigele, W. Pope, G. Hatfull, R. Hendrix, personal communication).If true, genetic exchange could occur between the Prochlorococcus prophage and a Synechococcus lytic phage-a scenario well accepted in other phage-host systems for genetic exchange [14,15].
Intragenic recombination involving host genes appears less common than phage-to-phage recombination events (Tables S1 and S2).Exchanges between Synechococcus and their viruses are evident, however, and appear to have occurred both from host to phage and phage to host for both psbA and psbD.Although such events were not detected between Prochlorococcus and their phages, there were cases where Prochlorococcus myoviruses were the recipients of external DNA from an unknown source (i.e., recombination events possibly involving donors outside of our dataset).Thus, phages may be contributing to the intragenic recombination of portions of these genes in Synechococcus, perhaps explaining the lack of phylogenetic structure observed in psbA and psbD trees for Synechococcus clusters (but not for Prochlorococcus clusters) relative to those obtained when using other phylogenetic markers [46][47][48][49].Presumably, phage-host intragenic exchanges occur via homologous recombination during infection.Clearly, the transfer of DNA will be retained in host lineages only if infection fails to lyse the host (e.g., abortive infection [54]).
Finally, intragenic exchanges among hosts were also occasionally detected, particularly among Synechococcus (Tables S1  and S2).This may also play a role in the lack of clade structure among Synechococcus strains in the psbA and psbD trees.Although two possible intragenic recombination events between Synechococcus and Prochlorocococus were identified, they were resolved as small regions (15-16 bases) and may be false positives.Host-to-host transfers may have occurred through the uptake of DNA directly from the environment (e.g., via transformation) or through viral intermediates [37].Such host-to-host intragenic exchanges via viral intermediates presumably occur through generalized transduction [55].
In summary, our findings suggest that the shuffling of segments of psbA and psbD within the cyanophage gene pool has generated significant photosynthesis gene diversity and serves as an extended reservoir of genetic diversity for their hosts, influencing photosystem evolution.

psbA and psbD Gene Diversity in Cultured Isolates Captures Most of the Field Diversity
We next sought to determine how well psbA and psbD sequence diversity observed in culture collections represents that observed in wild phage populations, and whether additional whole-gene host-to-phage transfer events could be identified from these wild sequences from the phage gene pool.Zeidner et al. [37] had previously examined field diversity of the psbA gene sequence from environmental samples where Synechococcus strains were the dominant phototroph [37].Thus, we sought to examine genetic diversity of this gene, as well as that of psbD, from an environment where Prochlorococcus cells commonly outnumber Synechococcus cells by orders of magnitude [56].To this end, we amplified, cloned, and sequenced psbA and psbD gene sequences obtained from the viral-sized fraction (0.02-0.2 lm) of two seawater samples within (25 m) and below (75 m) the mixed layer in the Pacific Ocean off the coast of Hawaii (Figures 5  and 6, respectively).The psbA and psbD sequences from these viral-fraction samples clustered with cultured Prochlorococcus cyanophage isolates (with varying levels of support; Figures 5  and 6), but not with Synechococcus cyanophages.There was not a notable difference in the phylogenetic placement of the psbA or psbD clones obtained from within or below the mixed layer.Although this suggests a lack of vertical structure in diversity among the sequence types, we did not sequence these samples to saturation; thus, such conclusions are preliminary.
More than half of the wild psbA sequences (42 of 81) form a large cluster with cultured Prochlorococcus podoviruses (Figure 5).Within this group, all but one cluster of wild sequences  3. Note that the 21-nucleotide indel in Prochlorococcus hosts and their phages [13] (unpublished data) was excluded from the analysis at the position indicated by the ''//'' symbol to maximize the data that could be displayed using the sliding window approach.DOI: 10.1371/journal.pbio.0040234.g004contain cultured podovirus sequences (Figure 5).The extensive microdiversity in this cluster (labeled ''unrepresented 1'') was probably derived from within the podovirus gene pool, as evidenced by the presence of podovirus phage isolates in the more basal branches of the cluster.Other psbA sequences from the field samples form subclusters that contain cultured Prochlorococcus myoviruses and form a large group that also contains Prochlorococcus hosts (Figure 5).One cluster (''unrepresented 2'' in Figure 5) within this group also lacks sequences from cultured hosts or phages.The basal position of this cluster suggests that these sequences may belong to phages that infect as-yet uncultured Prochlorococcus hosts [57] and may represent an additional host-to-phage transfer event.Thus, our work here, together with that of Zeidner et al. [37], suggests that cyanophage culture collections represent much of the naturally occurring Prochlorococcus and Synechococcus cyanophage psbA gene sequence diversity [37].
All psbD sequences from wild phages fall into a single wellsupported cluster that includes a representative cultured Prochlorococcus cyanophage P-SSM4 (Figure 6).This cluster reveals significant microdiversity within the psbD Prochlorococcus phage gene pool in the viral-fraction from this Pacific Ocean site and suggests that phages that encode Prochlorococcus-phage-like psbD genes are perhaps not rare in this environment.The four Prochlorococcus cyanophages that  1. Synechococcus environmental ''viral'' sequences from [37].The tree topology was estimated by LogDet analysis of 1st and 2nd codon positions, with branch lengths estimated using stationary nucleotide frequencies.DOI: 10.1371/journal.pbio.0040234.g005contain the psbD gene in our culture collection originated from either the Sargasso Sea or the Red Sea; thus, it is perhaps not surprising that the viral-fraction microdiversity from the Pacific Ocean is largely unrepresented in this collection.

Conclusions
The phage genomic repertoire evolves through the exchange of genetic material from other phages [24] and by co-opting metabolic genes from their hosts [13,20,22].The prevalence of photosynthesis genes in cyanophages strongly suggests that the capture of these genes provides a significant fitness advantage among certain cyanophage types.Previously, we have shown that the horizontal transfer of hli genes from cyanophages to their hosts has likely played a role in driving host niche differentiation [13].More recently, cyanophages were hypothesized to be involved in partial gene exchanges even for the core photosystem gene psbA of their hosts [37].Here, we show that genetic exchanges involving cyanophages may have influenced the make-up of both of the core photosystem II genes (psbA and psbD) in Synechococcus, whereas this was less apparent for Prochlorococcus.Therefore, mounting evidence indicates that host-like genes acquired by phages undergo a period of diversification in phage genomes and serve as a genetic reservoir for their hosts.Thus, a complex picture of overlapping phage and host gene pools emerges, where genetic exchange across these pools leads to evolutionary change for host and phage.Fully understanding the mechanisms of microbial and phage coevolution clearly requires an improvement in our ability to quantify horizontal gene transfer at the whole and partial gene level and in our ability to accurately estimate the relative fluxes into and out of these pools.

Materials and Methods
DNA isolation from cultured hosts and phages and environmental samples.Eleven strains of Prochlorococcus, ten strains of Synechococcus, and 38 phages of Prochlorococcus and Synechococcus (seven podoviruses, 29 myoviruses, and two siphoviruses) were screened for psbA and psbD sequences for this study.We report here on new psbA sequences from nine Synechococcus hosts and new psbD sequences from 19 Prochlorococcus and Synechococcus hosts (including two from unpublished Synechococcus genomes for strains CC9605 and CC9902; available from http://genome.jgi-psf.org/mic_home.html.The 38 phages screened included seven phage templates for which genome sequences are now available (P-SSM2, P-SSM4, P-SSP7, S-PM2, S-WHM1, Syn5, Syn9), enabling us to validate our PCR amplification findings.Host genomic DNA was extracted using a DNeasy Tissue Kit (Qiagen, Valencia, California, United States).Filtered (0.2 lm, Acrodisc supor membrane syringe filter) phage lysates in Pro99 medium were used as DNA templates for subsequent PCR amplification experiments.
Environmental samples were collected from the Hawaii Ocean Time Series (HOT) on 15 October 2003 at 458N 1588W from depths of 25 m and 75 m.These samples were filtered through a 0.2-lm filter (Osmonics, Minnetonka, Minnesota, United States, Poretics polycarbonate 25-mm filter) to remove cellular material and substantially enrich for environmental phages.A 100-ml volume of 0.2-lm filtrate was then filtered onto a 0.02-lm filter (Whatman Anotop 25) to collect phage particles and resuspended in 7 ml of a modified SM storage buffer (600 mM NaCl, 8 mM MgSO 4 -7H 2 O, 50mM Tris [pH 7.5], 0.04% gelatin).
Overview of psbA and psbD screening strategy.PCR screening for psbA and psbD across a diverse set of samples presented several challenges.These included variable amplification efficiencies, uncertainty about whether amplicons derived from phage or host, and multiple gene copies in hosts.The amplification strategy was as follows: for each virus and host strain, four PCR reactions were carried out, pooled, and analyzed by gel electrophoresis; if the amplification product was not visible, it was diluted 10-fold and used as template for nested or semi-nested PCR and the resulting products analyzed; if still no product was visible, multiple phage stocks were rescreened.Multiple copies of psbA in Synechococcus strains were identified by sequencing many clones and were distinguished from sequencing errors as described below.We did not screen for multiple copies of psbA from Prochlorococcus or multiple copies of psbD from either Synechococcus or Prochlorococcus, as when present, they are generally indistinguishable from each other [58][59][60].
Amplification of psbA and psbD.PCR reactions were performed with Taq DNA polymerase and deoxyribonucleotide triphosphates from New England Biolabs (Beverly, Massachusetts, United States) or Invitrogen (Carlsbad, California, United States) and carried out with a PTC-100 or PTC-200 DNA Engine (MJ Research, Waltham, Massachusetts, United States) or a Robocycler Gradient 96 (Stratagene, La Jolla, California, United States).Template amounts were 10 ng of genomic DNA for Prochlorococcus and Synechococcus, 1 ll of lysate for cyanophages, and 2 ll of filtrate for environmental samples.PCR primers and amplification reaction conditions are shown in Tables S3  and S4.
The psbA gene from all sources was amplified using primer pair psbA-F/R [61] and PCR protocol A (Tables S3 and S4).Four reactions were conducted with each template, and the products were pooled and analyzed by agarose gel electrophoresis.Primer psbA-R falls on the intron region in S-PM2 [29].Therefore, for efficient amplification of phage psbA genes that may contain introns, and for increased sensitivity, we used the Pro-psbA-F/R primer set and protocol B in nested PCR reactions when no PCR product was visible from cyanophage lysates and environmental filtrates.To reduce the incidence of heteroduplex formation, amplification products from environmental samples were subjected to reconditioning PCR [62]: initial PCR products were diluted 1:10, then amplified using protocol A but for only three cycles.
The psbD gene from Prochlorococcus, Synechococcus, and cyanophages was amplified using primer pair psbD-54F/psbD-308R and protocol D. However, when product yield was low or absent, semi-nested PCR was carried out as follows.Amplification was first conducted using primer pair psbD-26F/psbD-308R and protocol C. Four reactions were conducted with each template, the products were pooled, diluted 1:10, and used as templates for a second round of amplification using primer pair psbD-54F/psbD-308R and protocol D. psbD from environmental samples was amplified using primer pair psbD-26F/psbD-308R and protocol C and subjected to reconditioning PCR as for psbA (see above).
In preparation for sequencing, PCR products were either purified directly using the QIAquick PCR Purification Kit (Qiagen) or separated on an agarose gel and then purified using the QIAquick Gel Extraction Kit (Qiagen).
To confirm that the absence of psbA or psbD PCR products from phage was not simply due to a lack of amplifiable phage DNA, we screened phage lysates for known phage genes: g20 (for myoviruses) and DNApol (for podoviruses).g20 was amplified using primer pair g20-F/R and protocol E, and DNApol using primer pair DNApol-F/R and protocol F, both with 1 ll of lysate.In all cases, a product was obtained, suggesting the phage template DNA was present and amplifiable by PCR (unpublished data).
Six phage lysates yielded PCR products with sequences identical to those of a known host.These six phage lysates include five cyanophages previously described (P-RSP1, P-SSP1, P-SSP2, P-ShM1, P-ShM2; [5]), as well as one cyanophage not previously reported in the literature (P-SSP9; M.B.S. and S.W.C., unpublished data).In these cases we could not eliminate the possibility that the amplicon resulted from host DNA, the amplification of which may be more likely to occur when there is no phage template for this gene.Thus, we excluded these phages from further analyses.In contrast, phages with amplicon sequences identical to those of other phages (indicated as ''ID to X'' in Table 1) were passed through multiple lysates, and a ''fingerprint'' phage gene (g20) was used to confirm that there was a single phage in the lysate.The psbA sequence was then reassayed, increasing our confidence in these results.Even with this precaution, we cannot rule out the possibility of PCR contamination for those few cases where identical sequences were amplified from different phage lysates.
Cloning and sequencing of PCR products.The psbA gene is often found in multiple distinct copies in marine Synechococcus [59], whereas in Prochlorococcus the psbA gene is either single copy per genome or encodes multiple copies that are nearly identical to each other [60,63,64].Among cyanophages, the psbA gene has only been found in a single copy per genome [28,30].To allow for the identification of multiple psbA gene copies in Synechococcus strains, PCR products from Synechococcus templates were cloned prior to sequencing.Cloning was performed using the TOPO TA Cloning Kit for Sequencing (Invitrogen) with the pCR4-TOPO vector.Ligation products were transformed into TOP10 competent cells.Plasmid purification and sequencing were conducted by Genaissance Pharmaceuticals (New Haven, Connecticut, United States).Inserts were sequenced from both forward and reverse directions, using the M13F and M13R primer binding sites in the pCR4-TOPO vector.
Approximately ten psbA clones were sequenced for each Synechococcus strain.The published genome of Synechococcus WH8102 provides an example of natural psbA diversity in a given strain, as it contains four copies of psbA: two copies that are 99.8% identical and a third and fourth copy that are 99.4% and 88% identical, respectively, to the above two psbA copies [59].Considering a Taq polymerase error rate of 3 3 10 À5 per nucleotide per duplication [65], at most one error could be expected in each psbA gene sequenced.Thus, sequences were considered identical, and removed from the analysis pool, if they were more than 99.8% identical, to avoid data issues stemming from possible PCR error (sequencing error should be nonexistent because consensus sequences were obtained from forward and reverse sequencing of the clones).Sequence identity levels for nonidentical clones from the remaining dataset ranged from about 60% to 99.0%.
PCR products from genes presumed not to have multiple distinct copies per genome (psbA from Prochlorococcus and cyanophage; psbD from all organisms) were generally sequenced directly (Harvard Medical School Biopolymers Facility [Boston, Massachusetts, United States], Davis Sequencing [Davis, California, United States], or Genaissance Pharmaceuticals).The absence of multiple significantheight peaks at single nucleotide positions in chromatograms from this direct sequencing (unpublished data) confirmed that single products were amplified during PCR.Each strain was sequenced in Table S5.Likelihood Ratio Tests for Variable Evolutionary Rates among Branches For both psbA and psbD, individual sequences exhibiting a signature for intragenic recombination (Tables S1 and S2) were excluded from analysis.Likelihood scores were obtained under a stationary HKY85 model combined with a gamma correction for among-sites rate variation.All model parameters, including nucleotide frequencies, were estimated by using maximum likelihood.Data analysis included all three codon positions.Models were employed as implemented in the baseml program of the PAML package [82].Tree 1 was obtained by neighbor-joining analysis of LogDet distances estimated from all three codon positions.Tree 2 was obtained by neighbor-joining analysis of LogDet distances estimated from 1st and 2nd codon positions.For both genes, Tree 1 grouped lineages along lines of similarity in evolutionary rates and compositional biases, and Tree 2 did not.Found at DOI: 10.1371/journal.pbio.0040234.st005(36 KB DOC).
Table S6.Likelihood Ratio Tests for Nonstationary Frequencies among Lineages H 0 denotes the null hypothesis of stationary nucleotide frequencies; this was modeled by specifying one set of nucleotide frequencies for all branches of the tree.H 1 denotes the alternative hypothesis of nonstationary nucleotide frequencies; this was modeled by assigning all branches of the tree topology to one of several independent sets of frequency parameters (six sets for psbA and five sets for psbD).Apart from nucleotide frequencies, H 0 and H 1 assumed a substitution process equivalent to an HKY85 model combined with a gamma model for among-sites rate variation.The transition/transversion ratio was assumed to be homogenous among branches.H 1 represents a user-defined version of the nonhomogenous models of Yang and Roberts [77].All model parameters, including nucleotide frequencies, were estimated by using maximum likelihood.Data analysis included all three codon positions.Models were employed as implemented in the baseml program of the PAML package [82].Tree 1 was obtained by neighbor-joining analysis of LogDet distances estimated from all three codon positions.Tree 2 was obtained by neighbor-joining analysis of LogDet distances estimated from 1st and 2nd codon positions.For both genes, Tree 1 grouped lineages along lines of similarity in evolutionary rates and compositional biases, and Tree 2 did not.User-defined sets of frequency parameters for H 1 were specified in the tree file (shown below) by using the ''branch label'' format described in the PAML manual.For both psbA and psbD, individual sequences exhibiting a signature for intragenic recombination (Tables S1 and S2) were excluded from analysis.Found at DOI: 10.1371/journal.pbio.0040234.st006(44 KB DOC).
Table S7.Likelihood-Based Statistical Comparison of Competing Evolutionary Hypotheses under a Model of Nonstationary Nucleotide Frequencies P KH denotes the p-value for the KH normal test of [83].P SH denotes the p-value for the SH test [84].P RELL denotes the RELL bootstrap proportion [83].Note that although Tree 1 and Tree 2 were not selected independently of the data, neither was selected according to its likelihood score.For both genes, Tree 1 grouped lineages along lines of similarity in evolutionary rates and compositional biases, and Tree 2 did not.For both psbA and psbD, individual sequences exhibiting a signature for intragenic recombination (Tables S1 and  S2) were excluded from analysis.Tree 1 was estimated by a neighborjoining analysis of LogDet distances from all sites, and Tree 2 was estimated by a neighbor-joining analysis of LogDet distances based on only 1st and 2nd codon positions.Likelihood scores were obtained under nonstationary models of nucleotide frequencies (see Table S5 for additional model details).

Figure 1 .
Figure 1.Phylogenetic Tree of psbA Gene Sequences from Cultured Cyanobacteria and CyanophagesPhages are listed by their name, followed by their original host.Phages that are known to infect both Prochlorococcus and Synechococcus hosts are indicated with a ''D''; those that infect only one genus are labeled either P (infect only Prochlorococcus hosts) or S (infect only Synechococcus hosts), while those that are unknown are designated with a ''?''.Phages shown in italics and bracketed with ''**'' were isolated on hosts that do not belong to the same cluster and are thus exceptions to the general clustering pattern (see text).Taxa are color coded according to the following biological groupings: myoviruses (red), podoviruses (black), marine Synechococcus hosts (light blue), marine Prochlorococcus hosts (dark green, LL; light green, HL), freshwater cyanobacteria (dark blue).The tree topology was estimated by LogDet analysis of 1st and 2nd codon positions.Sequences where intragenic recombination was detected using other methods (see Materials and Methods) were not included in these phylogenetic analyses.Branch lengths were estimated by maximum likelihood under a model with nonstationary nucleotide frequencies.Numbers at the nodes represent neighbor-joining bootstrapping and maximum likelihood puzzling support.Anab, Anabaena; Gloe, Gleobacter; HL, high-light adapted; LL, low-light adapted; Syncy, Synechocystis; Thermo, Thermosynechococcus. DOI: 10.1371/journal.pbio.0040234.g001

Figure 2 .
Figure 2. Phylogenetic Tree of psbD Gene Sequences from Cultured Cyanobacteria and Cyanophages Details as in Figure 1.Sequences where intragenic recombination was detected using other methods (e.g., P-SSM1) were not included in these phylogenetic analyses.DOI: 10.1371/journal.pbio.0040234.g002

Figure 3 .
Figure 3. Visualization of %GþC Content across the psbA Gene Colors represent the averaged %GþC in sliding windows along the length of the gene (20%-80%); white regions represent windows that included ambiguous bases in which %GþC could not be calculated for that region.The average %GþC content of the amplified sequence is tabulated on the right side of the figure.Phages are listed by phage name followed by their original host.Phages that are known to infect both Prochlorococcus and Synechococcus hosts are indicated with a ''D''; those that infect only one genus or the other have no marker, while those that are unknown are designated with a ''?''.Host names are prefaced with Syn or Pro for Synechococcus and Prochlorococcus hosts, respectively.Scale indicates nucleotide positions relative to the psbA gene sequence in Thermosynechococcus. DOI: 10.1371/journal.pbio.0040234.g003

Figure 4 .
Figure 4. Visualization of %GþC Content across the psbD Gene Details as in Figure3.Note that the 21-nucleotide indel in Prochlorococcus hosts and their phages[13] (unpublished data) was excluded from the analysis at the position indicated by the ''//'' symbol to maximize the data that could be displayed using the sliding window approach.DOI: 10.1371/journal.pbio.0040234.g004

Figure 5 .
Figure 5. Phylogenetic Tree of psbA Gene Sequences from Representative Cultured Cyanobacterial and Cyanophage Isolates and Cloned Environmental Sequences from the Hawaii Ocean Time Series Site in the Pacific Ocean Phylogenetic tree of psbA gene sequences and cloned environmental sequences were collected from above (25 m, black) and below (75 m, red) the surface mixed layer at the Hawaii Ocean Time Series site in the Pacific Ocean, a region where Prochlorococcus are the dominant phototrophs.Details for naming conventions are as in Figure1.Synechococcus environmental ''viral'' sequences from[37].The tree topology was estimated by LogDet analysis of 1st and 2nd codon positions, with branch lengths estimated using stationary nucleotide frequencies.DOI: 10.1371/journal.pbio.0040234.g005

Figure 6 .
Figure 6.Phylogenetic Tree of psbD Gene Sequences from Cultured Cyanobacterial and Cyanophage Representatives and Cloned Environmental Sequences from the Pacific Ocean Details for naming conventions are as in Figure 1, and phylogenetic analyses are as in Figure 5. DOI: 10.1371/journal.pbio.0040234.g006

Table 1 .
Presence or Absence of psbA and psbD among Prochlorococcus and Synechococcus Cyanophages