Abstract
Prochlorococcus is the numerically dominant phototroph in the tropical and subtropical oceans, accounting for half of the photosynthetic biomass in some areas1, 2. Here we report the isolation of cyanophages that infect Prochlorococcus, and show that although some are host-strain-specific, others cross-infect with closely related marine Synechococcus as well as between high-light- and low-light-adapted Prochlorococcus isolates, suggesting a mechanism for horizontal gene transfer. High-light-adapted Prochlorococcus hosts yielded Podoviridae exclusively, which were extremely host-specific, whereas low-light-adapted Prochlorococcus and all strains of Synechococcus yielded primarily Myoviridae, which has a broad host range. Finally, both Prochlorococcus and Synechococcus strain-specific cyanophage titres were low (< 103 ml-1) in stratified oligotrophic waters even where total cyanobacterial abundances were high (> 105 cells ml-1). These low titres in areas of high total host cell abundance seem to be a feature of open ocean ecosystems. We hypothesize that gradients in cyanobacterial population diversity, growth rates, and/or the incidence of lysogeny underlie these trends.
Phages are thought to evolve by the exchange of genes drawn from a common gene pool through differential access imposed by host range limitations3. Similarly, horizontal gene transfer, important in microbial evolution4, 5, can be mediated by phages6 and is probably responsible for many of the differences in the genomes of closely related microbes5. Recent detailed analyses of molecular phylogenies constructed for marine Prochlorococcus and Synechococcus7, 8 (Fig. 1) show that these genera form a single group within the marine picophytoplankton clade9 (> 96% identity in 16S ribosomal DNA sequences), yet display microdiversity in the form of ten well-defined subgroups8. We have used members of these two groups to study whether phage isolated on a particular host strain cross-infect other hosts, and if so, whether the probability of cross-infection is related to rDNA-based evolutionary distance between the hosts.
Figure 1: Host ranges of 44 clonal cyanophages exposed to marine Prochlorococcus and Synechococcus cultured isolates.
The evolutionary relationships between the 21 host strains are shown in the phylogenetic tree inferred using 16S–23S rDNA spacer regions8. For the cyanophages, red indicates Podoviridae, blue indicates Myoviridae and green indicates Siphoviridae. Filled circles, host strain used to isolate a particular cyanophage; open circles, cross-infection of cyanophage with another host; dash, no infection (that is, lysis). Symbols in parentheses indicate that the results do not match earlier studies10,13 with these phage and hosts (see Methods for details). HL Pros, high-light-adapted Prochlorococcus; LL Pros, low-light-adapted Prochlorococcus; Marine Syns, marine Synechococcus.
High resolution image and legend (78K)Analyses of host range were conducted (Fig. 1) with 44 cyanophages, isolated as previously described10 from a variety of water depths and locations (see Supplementary Information) using 20 different host strains chosen to represent the genetic diversity of Prochlorococcus and Synechococcus8. Although we did not examine how these patterns would change if phage were propagated on different hosts, this would undoubtedly add another layer of complexity due to host range modifications as a result of methylation of phage DNA6. Similar to those that infect other marine bacteria11 and Synechococcus10, 11, 12, 13, 14, our Prochlorococcus cyanophage isolates fell into three morphological families: Myoviridae, Siphoviridae and Podoviridae15.
As would be predicted10, 11, 12, 13, 14, Podoviridae were extremely host specific with only two cross-infections out of a possible 300 (Fig. 1). Similarly, the two Siphoviridae isolated were specific to their hosts. In instances of extreme host specificity, in situ host abundance would need to be high enough to facilitate phage–host contact. It is noteworthy in this regard that members of the high-light-adapted Prochlorococcus cluster, which yielded the most host-specific cyanophage, have high relative abundances in situ16. The Myoviridae exhibited much broader host ranges, with 102 cross-infections out of a possible 539. They not only cross-infected among and between Prochlorococcus ecotypes but also between Prochlorococcus and Synechococcus. Those isolated with Synechococcus host strains have broader host ranges and are more likely to cross-infect low-light-adapted than high-light-adapted Prochlorococcus strains. The low-light-adapted Prochlorococcus are less diverged from Synechococcus than high-light-adapted Prochlorococcus7, 8, suggesting a relationship, in this instance, between the probability of cross-infection and rDNA relatedness of hosts. Finally, we tested the Myoviridae for cross-infection against marine bacterial isolates closely related to Pseudoalteromonas, which are known to be broadly susceptible to diverse bacteriophages (bacterial strains HER1320, HER1321, HER1327, HER1328)11. None of the Myoviridae cyanophages infected these bacteria.
Phage morphotypes isolated were determined, to some degree, by the host used for isolation (Fig. 1). For example, ten of ten cyanophages isolated using high-light-adapted Prochlorococcus strains were Podoviridae. In contrast, all but two cyanophages isolated on Synechococcus were Myoviridae, a bias that has been reported by others14, and over half of those isolated on low-light-adapted Prochlorococcus belonged to this morphotype. We further substantiated these trends by examining lysates (as opposed to plaque-purified isolates) from a range of host strains, geographic locations and depths—of 58 Synechococcus lysates 93% contained Myoviridae, of 43 low-light-adapted Prochlorococcus lysates 65% contained Myoviridae, and of 107 high-light-adapted Prochlorococcus lysates 98% contained Podoviridae (see Supplementary Information).
Maximum cyanophage titres, using a variety of Synechococcus hosts, are usually found to be within an order of magnitude of the total Synechococcus abundance10, 14, 17, 18, and can be as high as 106 phage ml-1. One study17 has shown, for example, that along a transect in which total Synechococcus abundance decreased from 105 cells ml-1 to 250 cells ml-1, maximum cyanophage titres remained at least as high as the total number of Synechococcus. We wondered whether titres of Prochlorococcus cyanophage in the Sargasso Sea, where Prochlorococcus cells are abundant (105 cells ml-1), would be comparable to those measured in coastal oceans for Synechococcus where total Synechococcus host abundances are of similar magnitude. We assayed cyanophage titres in a depth profile in the Sargasso Sea at the end of seasonal stratification using 11 strains of Prochlorococcus (Fig. 2), choosing at least one host strain from each of the six phylogenetic clusters that span the rDNA-based genetic diversity of our culture collection8.
Figure 2: Cyanophage titres, measured using Prochlorococcus host strains, as a function of depth at the Bermuda Atlantic Time Series Station in the Sargasso Sea on 26 September 1999.
Nine of the cyanophages used in host range analyses were isolated from this depth profile (see Supplementary Information). a–c, Titres measured using high-light-adapted Prochlorococcus hosts (a) and low-light-adapted hosts (b), and total Prochlorococcus and Synechococcus cell abundances (Prochlorococcus cells 
104 ml-1; Synechococcus cells 103 ml-1) and 
t (a proxy for water density used to measure the depth of the mixed layer) (c). Titres were undetectable for low-light-adapted Prochlorococcus strains SS120, MIT9303 and MIT9313. Error bars represent the s.d. of assays.
Three Prochlorococcus host strains (MIT 9303, MIT 9313 and SS120) yielded low or no cyanophage. Other hosts yielded titres that reached a maximum at 70 m (NATL2A-phage) or 100 m (MIT 9302-, MIT 9515-, MED4-, MIT 9211-, NATL1A-phage) near the depth of maximum Prochlorococcus abundance (Fig. 2). All Prochlorococcus cyanophage titres were low (< 350 cyanophage ml-1) compared with those reported for Synechococcus in coastal regions (approximately 104–106 cyanophage ml-1) even though total host abundances were similar between these regions (approximately 105 cells ml-1)10, 14, 17, 18. Prochlorococcus cyanophage titres are comparable to those of Synechococcus from oligotrophic waters in the Gulf of Mexico—but in that instance the total Synechococcus abundance was also low (< 250 cells ml-1)17.
Cyanophage titres were also examined along a surface water transect from coastal (mesotrophic) to open ocean (oligotrophic) in the Atlantic Ocean to better understand the relationship between maximum phage titre and total host abundance along a trophic gradient. Titres were assayed with 12 strains of Synechococcus and Prochlorococcus that represented the known rDNA-based genetic diversity at the time that we began the study8 (but see also ref. 19). We found that Synechococcus cyanophage titres decreased by an order of magnitude or greater in surface waters between the coastal and open ocean (Sargasso) sites, whereas total Synechococcus abundance decreased from 3 104 to 7 103 cells ml-1 (Fig. 3). Prochlorococcus hosts did not yield cyanophage in coastal samples where there are no Prochlorococcus cells, and yielded relatively low titres (0 to 1.5 103 phage ml-1) at the shelf, slope and Sargasso stations where total Prochlorococcus abundance was between 4.5 104 and 1.4 105 cells ml-1. Even though total Prochlorococcus abundance at the Sargasso site was similar to that of Synechococcus at the coastal site (Fig. 3i, j), Prochlorococcus and Synechococcus cyanophage titres were significantly lower at the open ocean site (Fig. 3a–h). Moreover, regardless of the host used, titres never exceeded 3 103 cyanophage ml-1 at any depth throughout the photic zone even though total Prochlorococcus abundances exceeded 105 cells ml-1 (see Supplementary Information). Thus it seems that cyanophage titres at the end of summer stratification are relatively low in open ocean ecosystems, where the total possible host cell abundances are relatively high. Low titres lead to reduced contact rates and lowered mortality rates6.
Figure 3: Cyanophage titres measured in Synechococcus and Prochlorococcus host cells along a surface water transect from coastal (coast, Woods Hole, Massachusetts) to open ocean (Sargasso) conducted in September 2001.
Note that the magnitudes of the y axes are different for a–e and f–j. Ten of the cyanophages used in host range analyses were isolated from along this transect (see Supplementary Information). a–h, Cyanophage titres represent the averages and s.d. of triplicate plaque assays. i, j, Cell concentrations represent averages and s.d. of duplicate flow cytometry assays. Where no bar is shown, there were no plaques (a, c, e–h) or no cells (i). No plaques were observed at any of the surface samples along the transect for Synechococcus strain WH 8020 and Prochlorococcus strains MIT 9313, SS120 and MIT 9211 (data not shown).
High resolution image and legend (36K)Although it is difficult to draw definitive conclusions about causality from such trends because of the complexity of the phage–host interaction, there are some factors that might be implicated. If, for example, host strain microdiversity increased along the transect and cross-infection ability did not increase concurrently, this would lead to lower phage titres yielded by a suite of host strains20. Indeed, we know that the relative abundance of Synechococcus ecotypes changes from coastal to oligotrophic waters16. However, we observed a systematic decrease in cyanophage titres for all five Synechococcus hosts (note WH 8020 yielded no plaques) at the extremes of this transect (the coastal and Sargasso sites; Fig. 3a–e). If changing host abundance alone explained the change in titres, and if our suite of host strains is representative of natural diversity, then one might expect at least one host strain to yield increasing titres along the gradient (for example, for the 'open ocean strain' WH 8102); however, this was not observed.
Another possible explanation for decreasing phage titres as one goes from coastal to open ocean ecosystems is decreased nutrient availability along the transect21 resulting in suboptimal growth of host cells in the Sargasso Sea22 relative to Synechococcus at the coastal site23. Viral production is correlated with host growth rates in chemostats24 and in the field25, which could result from nutrient limitation causing physiological changes in the host that stall the lytic process of obligately lytic phage6, or favour lysogeny in temperate phage18, 26, 27. Although temperate phage have not been identified for marine Prochlorococcus or Synechococcus, INT family site-specific recombinases exist in the genomes of Prochlorococcus MED4 and MIT 9313, and Synechococcus WH 8102 (http://www.jgi.doe.gov/JGI_microbial/html/index.html), suggesting that prophages were once integrated into these host genomes28, 29.
The phage–host system described here should continue to be a useful framework for advancing our understanding of the ecology and evolution of phage–host interactions in marine ecosystems. We have known for some time that cyanophages must have a role in maintaining genetic diversity among hosts10, 17. The broad host ranges reported here indicate further their potential for mediating horizontal gene transfer, which may help explain the extensive microdiversity8, 19, 28, 29 seen in these two groups of marine cyanobacteria. The extent to which this potential is realized should become clear as more and more host and phage genomes are sequenced. Of significance also is the coupling between phage morphology and host type. Experiments designed to characterize phage resistance across variable hosts and phages (for example, identification of receptors and restriction and modification systems) should elucidate the underlying mechanisms responsible for these patterns. Finally, our analyses of cyanophage titres along a coastal–open ocean transect suggest that the underlying processes responsible for the production of free cyanophage differ along trophic gradients in the oceans. To fully explain these observations will require the development of approaches that allow one to determine which phage can infect which host(s) in a given community, and an understanding of the relative roles of lytic and lysogenic phases of the viral life cycle in aquatic systems.
Methods
Sample collection
Water samples for cyanophage titres, cyanophage isolations and cyanobacterial abundances were collected at the Bermuda Atlantic Time Series Station on 26 September 1999 and at four sites along a transect from Woods Hole to the Western Sargasso Sea on 5, 16, 17 and 22 September 2001 (see Supplementary Information). Water for cyanophage isolations was filtered (0.4 
m, Poretics number 13028 in 1999; 0.2 m, Osmonics number K02CP04700 in 2001) and stored at 4 °C in the dark in acid-washed polycarbonate (1999) or glass (2001) bottles until analysis (up to 15 months later). Cyanophage titres remain stable for at least one year27. Control experiments showed that titres were stable over a 15-month period (see Supplementary Information).
Culturing conditions
Prochlorococcus and Synechococcus strains were maintained in '75% Pro99' medium, a modification of the 'Pro2' medium30 with a 75% seawater base and the following final concentrations of N and P: 800 M NH4Cl, 50 M NaH2PO4. Cultures were grown at 19–21 °C under constant light 8–12 E m-2 s-1 for low-light-adapted Prochlorococcus and Synechococcus; 35–45 E m-2 s-1 for high-light-adapted Prochlorococcus.
Cyanophage isolations
Prochlorococcus cyanophage isolations were done initially using an axenic strain of Prochlorococcus (MED4ax; M. Saito and J.B.W., unpublished observations). Exponentially growing cells were transferred to fresh medium (1 ml:20 ml) and inoculated with 1 ml of 0.4-m-filtered sea water. The time course of auto-fluorescence (chlorophyll biomass) of these cultures was then followed with a Turner Designs 10-AU fluorometer. Cultures showing reduced fluorescence relative to controls were filtered and examined for phage particles as previously described10, 12. Lysates were stored at 4 °C in the dark. Subsequent isolations using 19 additional host strains were done using the same procedures scaled down to small volumes. Cyanophage isolates used in this study were plaque-purified twice before use, classified using morphology described by the ICTV15, and named according to suggestions made for cyanophage18.
Cyanophage host range
Host range analyses were conducted over a period of about 2 yr. Each interaction between a cyanophage and its potential host cell was performed with exponentially growing cells in triplicate on at least two different occasions. Marine bacterial strains were purchased from the Felix d'Herelle Reference Center for Bacterial Viruses (contact H. Ackermann). Several of the Synechococcus cyanophage used in this study ('Syn' phages; S-PM2 and S-WHM1) had been previously examined for host range cross-infectivity10, 13 and were maintained as lysates at 4 °C in the dark while host cyanobacterial cultures were serially transferred in late exponential and early stationary phase. A total of 103 of 108 cross-infections using these stored cyanophages yielded similar host range results in this study (Fig. 1). Of the differences observed, four of five were for one cyanophage isolate (Syn10), suggesting that it might have evolved an extended host range mutation. Host range can be altered through DNA modifications that can occur during propagation of a phage on an alternative host. Overall, these results suggest that cyanophage susceptibility of these host strains and the cross-infectivity of the cyanophage remained relatively stable throughout the 10 or more years of storage and culture maintenance.
Host cell and cyanophage quantification
Host cell abundance was measured using a modified Becton-Dickinson FACScan flow cytometer7. Cyanophage titres were quantified using most probable number (MPN) assays (1999) or plaque assay (2001). MPN assays were monitored for lysis relative to controls for 2–3 weeks depending on the host strain used. For the plaque assays, we plated the host strain in soft agarose (0.4% final concentration; GIBCO BRL, Life Technologies number 5517-014) along with the phage being titred; lawns of cells appeared 8–28 days after inoculation, depending on the strain (M. Saito, M.B.S. and J.B.W., unpublished data). Plaques were counted daily until they no longer appeared (3–14 days after the first plaques). Titres measured using both assays were not significantly different (t-test assuming equal variances, 
= 0.10).
Host dependency of measured titres
To see whether our standard suite of host cells used in our assays was giving us a representative picture of the maximum phage titre obtainable in a given sample, we used every cultured isolate of Prochlorococcus in our collection with a unique ITS rDNA sequence (23 isolates) to assay the cyanophage titre of a 50-m water sample from the Red Sea. The range of titres yielded was representative of the range we measured using our subset of Prochlorococcus isolates (see Supplementary Information).
