Cyanophages from a less virulent clade dominate over their sister clade in global oceans

Infection properties of clade A and clade B T7-like cyanophages

We set out to test the hypothesis that the phylogenetic separation of T7-like cyanophages into two major clades reflects differences in their infection physiology. To do this we investigated a suite of infection properties of three pairs of clade A and B phages, each pair infecting the same Synechococcus host (Table 1) to allow us to control for variability in host genetics and physiology. These six cyanophages are representatives of 3 clade A and 2 clade B cyanophage subclades (SI Appendix, Table S1).

We began by investigating adsorption kinetics and the length of time taken to produce new phages in the infection cycle, the latent period, from phage growth curve experiments. In all three pairs of phages, adsorption was 7–15-fold more rapid in the clade A phage versus the clade B phage (Fig. 1, Table 1). Furthermore, the clade A phage had a faster infection cycle with a latent period that was 3-5-fold shorter than the clade B phage on the same host (Fig. 1a–c) (Table 1). To determine how representative these findings are for a greater diversity of T7-like cyanophages we report the latent period of nine additional non-paired phages that infect a variety of hosts and span the diversity of this cyanophage genus, measured here and taken from the literature (SI Appendix, Table S1). These phages showed the same pattern as observed between phage pairs, although one clade A phage had a relatively long latent period (see SI Appendix, Table S1). Overall, the 5 clade A phages representative of 5 subclades had a significantly shorter latent period (3.3 ± 3.6 h, n = 5 phages (mean ± SD) than the 10 clade B phages from 7 subclades (7.7 ± 2.0 h, n = 10 phages) (Kruskal-Wallis: χ² = 4.72, df = 1; p = 0.029, n = 15). No significant differences in the length of the latent period were found for clade B phages that infected Synechococcus and Prochlorococcus (Kruskal-Wallis: χ2 = 1.13, df = 1; p = 0.29, n = 10).

We determined the number of infective phage progeny produced per cell, the burst size, using a single cell approach [30]. In this assay, single infected cells are separated by flow cytometry into individual wells at maximal adsorption (SI Appendix, Table S2), allowed to lyse, and the number of infective phages produced is determined by the plaque assay (see Methods). The three clade A phages had significantly larger burst sizes (135.5 ± 49.6 phages·cell⁻¹) than the three clade B phages (55.0 ± 48.2 phages·cell⁻¹) (paired t-test: df = 2, t = 5.28, p = 0.03, n = 6) by 2-8-fold for each phage pair, as determined from 42–270 individual cells in five independent experiments (Fig. 1d–f, Table 1). Thus, clade A phages had higher burst sizes relative to clade B phages despite having a shorter infection cycle. This finding challenges current thinking that the evolution of shorter latent periods necessarily results in a tradeoff of smaller burst sizes for all phage types [31, 32].

The combined effects of all stages of infection, including adsorption kinetics, the length of the latent period and the burst size can be seen from the timing of plaque formation and their size [33]. Clade A plaques became visible in less than 24 h post infection while clade B plaques took 3-4 days to appear. Furthermore, 2.4–4.3-fold larger plaques were produced by clade A phages (11–16 mm) than clade B phages (3–7 mm) over the same period of time (paired t-test: df = 2, t = 13.714, p = 0.0052, n = 6) (Fig. 1m–o) (SI Appendix, Table S3). These findings show that, under these laboratory conditions, clade A phages have significantly greater fitness than clade B phages on the same hosts.

We then quantified virulence which we define here as the probability that a phage kills and lyses a cell after adsorption [30, 34]. This was determined from the percentage of individual cells in a population that were lysed by each phage using a single cell approach [30]. The host was challenged with the same number of infective phages at a multiplicity of infection (MOI) of 2, and each phage was allowed to adsorb to the host until maximal adsorption was achieved (see Methods and SI Appendix, Table S2). The clade A phage lysed between a 2–3-fold higher proportion of cells than the clade B phage when comparing two of the phage pairs, whereas no significant difference was found for the third phage pair (Fig. 1g–i, Table 1).

The viability of a phage is affected by its extracellular decay rate which influences the period of time it has to encounter and infect a new host. To assess whether rates of decay differ between clade A and clade B phages we determined the loss of infectious phages over time when incubated under host growth conditions. No consistent pattern was observed across phage pairs (Fig. 1j–l, Table 1) and there was no significant difference in mean decay rates between clade A and clade B phages (t-test: df = 0.13, t = 2.5308, p = 0.9, n = 6).

These combined findings show that for most infection properties, clade A phages are more aggressive than clade B phages as they complete their infection cycle more rapidly, produce more progeny and kill more host cells. Thus, phylogenetic differences within a diverse, ecologically important phage genus have clear manifestations in infection properties at the clade level and are thus likely to be a result of adaptation and selection. These differences in infection properties are present even though members of both clades have the same genomic backbone of replication and morphology genes and infect the same cyanobacterial host taxa. The genomic underpinnings of the observed clade-level differences in infection properties are currently unknown and could be due to allelic differences in core genes, the result of distinct gene repertoires, or a combination of both.

Annual population dynamics of T7-like cyanophages in the Gulf of Aqaba

Differences in infection properties are expected to influence the abundance and distribution patterns of cyanophages. We assessed the population dynamics of clade A and clade B cyanophages over the annual cycle in the Gulf of Aqaba, Red Sea. To do so, we collected samples from monthly depth profiles over a 1-year period, focusing on the upper 140 m photic zone where their cyanobacterial hosts reside. To put our findings into their environmental context, we first describe the seasonal dynamics of the water column and cyanobacterial distributions in these waters.

Physicochemical conditions of the water column and cyanobacterial population dynamics

The Gulf of Aqaba has a characteristic seasonal cycle in water column stability (see Methods), affecting nutrient availability and phytoplankton abundances in the photic zone [20, 35] that was also observed over our period of sampling (Figs. 2, 3 and SI Appendix, Fig. S1). Winter mixing reached a depth of ~300 m by the end of February 2013 (Fig. 2a) and injected nutrients into the photic zone (Fig. 2b, c). Stratification began upon warming of the upper surface layers in March. Maximal stratification was observed by August. During the stratification period nutrients in the photic zone were utilized by the phytoplankton and dropped below limits of detection by mid-spring (Fig. 2b, c). Mixing commenced again in October as the upper layers cooled. The mixed layer extended below the photic zone by the middle of December and reached its maximal depth again in February 2014 (Fig. 2a). For simplicity we refer to two periods that differ in their water column stability: the stratification period from March to September and the mixing period from October to February.

Chlorophyll a, present in all primary producers, is often used as an approximate proxy for phytoplankton biomass. Chlorophyll a concentrations and phytoplankton group abundances were uniformly distributed throughout the mixed layer during periods of mixing (Fig. 3, and SI Appendix, Fig. S1a). A shallow subsurface peak in chlorophyll a concentration (the deep chlorophyll maximum, DCM) developed early during stratification in the spring and deepened to 100 m by the summer (Fig. 3a). The DCM coincided with maximal abundances of small photosynthetic eukaryotes and Synechococcus during the short-lived spring bloom and with Prochlorococcus and photosynthetic eukaryotes in summer (Fig. 3, and SI Appendix, Fig. S1a). As found previously, eukaryotic phytoplankton, Synechococcus and Prochlorococcus were most abundant in winter, spring and summer, respectively [20].

T7-like cyanophage annual population dynamics

Virus abundances were determined from a total of 107 samples from 12 depth profiles collected from March 2013 to February 2014. First we quantified virioplankton from virus-like particles (VLPs) which are generally considered to reflect abundances of dsDNA viruses [36]. VLPs were most abundant in transition periods as the water column changed from mixing to stratification (March to May) and from stratification to mixing (October to December) (Fig. 4a). Maximal abundances of 5–7 × 10⁷ VLPs·ml⁻¹ were observed in the upper 60 m of the water column in April and October. Abundances were lowest during stable stratification from June to September, but were still observed at densities in excess of 10⁷ VLPs·ml⁻¹ (Fig. 4a). VLP abundances were significantly correlated with trophic status of the water column, represented by chlorophyll a concentration (ρ = 0.62, p = 1.1 × 10⁻⁸; n = 68) as well as with Synechococcus (ρ = 0.68 p = 1.6 × 10⁻¹⁰; n = 69) and heterotrophic bacteria (ρ = 0.28, p = 0.014; n = 75).

We quantified clade A and clade B T7-like cyanophages over the annual cycle in the Gulf of Aqaba using the polony method, a solid-phase single-molecule PCR method [37]. T7-like cyanophage population dynamics were quite different to those of total VLPs (Fig. 4a, b). Maximal abundances of T7-like cyanophages were observed during stable stratification when VLPs were at their annual minimum. Thus, while T7-like cyanophages made up between 0.3–12% of the VLPs over the annual cycle, they contributed most to the virioplankton pool between June-September, with a maximum contribution of 12.1 ± 6.8% of VLPs at 100 m in August (SI Appendix, Fig. S2). These findings show that T7-like cyanophages have different population dynamics compared to the dsDNA virus community as a whole.

T7-like cyanophage populations were dominated by clade B cyanophages (Fig. 4b–d, and SI Appendix, Fig. S3). Their maximal monthly abundances typically ranged from 0.6–1.5 × 10⁶ phage·ml⁻¹. In contrast, clade A cyanophage abundances were never higher than 6.0 × 10⁴ phages·ml⁻¹ and were below the limit of accurate quantification (1 × 10⁴ phages·ml⁻¹) in 75% of the samples (Fig. 4d, and SI Appendix, Fig. S3). As such, clade B cyanophages were more abundant than clade A cyanophages at all depths and in all seasons at ratios that ranged from 2.8-fold to over a 1000-fold. In fact, clade B phages were at least an order of magnitude more abundant than clade A cyanophages in 97% of all samples collected in the photic zone (n = 84).

We then assessed whether differences in environmental cyanophage abundances translated into differences in the extent of infection. We assessed the percent of infected Synechococcus and Prochlorococcus cells by clade A and clade B phages during March and September in 2014 using the iPolony method [38]. The less aggressive clade B cyanophages infected significantly more cyanobacteria than clade A phages in all but one sample (paired Wilcoxon test: V = 44, p = 0.0078, n = 18 for Synechococcus and V = 78, p = 0.0004, n = 24 for Prochlorococcus) (Fig. 5). Moreover, in 85% of the samples clade B phages infected at least 10-fold more cyanobacteria than clade A phages.

Distribution patterns of clade A and clade B cyanophages changed with seasonal shifts in water column conditions and cyanobacterial abundances. At the beginning of the stratification period in March 2013, clade B cyanophage abundances were highest in the upper 60 m (3.3 × 10⁵–5.5 × 10⁵ phages·ml⁻¹), coincident with the Synechococcus bloom (Fig. 4b and SI Appendix, Fig. S3). Their numbers increased and the maximum deepened as stratification intensified during spring-summer, coinciding with the peak in Prochlorococcus. Annual maxima in clade B cyanophage abundances were observed in the summer with highest numbers in August at 100 m (1.58 × 10⁶ ± 0.63 × 10⁶ phages·ml⁻¹) (mean ± ci95%). Abundances remained relatively high through the beginning of the autumn mixing period in October-November. As mixing progressed, abundances became uniformly distributed over the mixed layer and dropped down to 1.8 × 10⁵ phages·ml⁻¹ (Fig. 4c, and SI Appendix, Fig. S3). Overall, clade B cyanophage abundances correlated with Prochlorococcus (assessed for 60 to 140 m depth, see Methods) (ρ = 0.83, S = 2378, p < 2.2 × 10⁻¹⁶, n = 44), especially during the stratification period (ρ = 0.93, S = 262, p < 2.2 × 10⁻¹⁶, n = 28). Since members of clade B cyanophages can infect either a Synechococcus or a Prochlorococcus host [10], the greater correlation with Prochlorococcus may be explained by their higher abundances relative to Synechococcus during stable stratification, supporting an overall larger population of clade B cyanophages.

Clade A cyanophages had somewhat similar seasonal dynamics to those of clade B cyanophages. They were most abundant during the stable stratification period (June-September) and during early mixing (November) (Fig. 4d). However, clade A cyanophages were present at notably shallower depths than clade B cyanophages, being most abundant in the upper 60 m of the water column throughout the stratification period (Fig. 4d, and SI Appendix, Fig. S3). Similar to clade B phages, maximal annual abundances of clade A cyanophages were found in August but at 20 m (5.8 × 10⁴ ± 2.0 × 10⁴ phages·ml⁻¹) with similarly high abundances also found in November at 60 m (5.1 × 10⁴ ± 2.0 × 10⁴ phages·ml⁻¹). Since clade A cyanophages were often close to or below the limit of accurate quantification, we concentrated samples from four depth profiles in different seasons (SI Appendix, Fig. S3). Correlation analysis with data from these profiles showed that clade A phages were highly correlated with Synechococcus during months of stratification (ρ = 0.93–0.97, p < 0.005, n = 7 per profile) but not during mixing (ρ = 0.12, p = 0.65, n = 7) (SI Appendix, Fig. S4). The correlation with Synechococcus rather than Prochlorococcus is expected given that cyanophages belonging to this clade primarily infect Synechococcus [10], (Fig. 5).

T7-like cyanophage dynamics correlated to seasonal changes in cyanobacterial populations. Such linked dynamics might be expected at the population level since phages are obligate intracellular parasites that require their hosts to replicate and because T7-like cyanophage dynamics were measured on monthly time scales that integrate dynamics across many cyanophage infection cycles (hours) and cyanobacterial divisions (on the order of a day). Previous findings investigating cyanophages over seasonal cycles at a similar temporal resolution, but at different taxonomic levels, show similar correlations between cyanobacterial and cyanophage abundances. These include studies that used infective assays measuring cyanophages that infect a specific cyanobacterium [39,40,41] and an amplicon study investigating single cyanobacterial and T4-cyanophage genotypes [42].

This study of T7-like cyanophage populations revealed the dominance of clade B over clade A cyanophages at all depths and in all seasons over the annual cycle in the Gulf of Aqaba, Red Sea. This dominance was apparent both when Prochlorococcus was the more abundant cyanobacterium in late spring-summer and when Synechococcus was most abundant in winter-early spring. The dominance of clade B phages is not restricted to the Red Sea. Recently, we found that clade B phages were significantly more abundant and infected more cyanobacterial cells than clade A phages in 97% of samples from surface transects across vast regions in the North Pacific Ocean, including samples where Synechococcus was more abundant than Prochlorococcus by more than 5–10-fold [43]. These patterns are also consistent with metagenomic comparisons of relative read numbers, from both the viral fraction and cellular metagenomes, sampled sporadically from surface waters at various oceanic sites [29, 44].

Intriguingly, clade B phages have significantly higher abundances and infect more cyanobacteria in the environment even though their infection properties show lower fitness than clade A phages. This phenomenon of dominance and more infections by the less virulent virus is not likely to be unique to the T7-like cyanophages, as a slower, less-virulent virus was recently suggested to infect more coccolithophore cells in the Atlantic Ocean [45]. These findings indicate that greater fitness, determined as the greater number of viral progeny produced per unit time in single-host infection settings, does not necessarily predict the dominance of populations in complex communities in the environment.

The dominance of clade B over clade A phages in seasons and in regions with large Prochlorococcus populations is likely to be largely due to the ability of many clade B phages but only a minority of clade A phages to infect Prochlorococcus [10] (Fig. 5). Since, the dominance of clade B phages was also observed at times (Fig. 4) and in regions [43] where Synechococcus was the dominant cyanobacterium, other explanations are required for understanding their high abundances at those times and regions. It is feasible that the greater diversity of clade B phages allows them to infect more Synechococcus genotypes than clade A phages. However, it is also possible that the differences in infection properties play a direct role in this phenomenon. These possibilities are not mutually exclusive.

Modeling abundances based on the infection properties of clade A and clade B phages

Here we address the possibility that the dominance of clade B phages is directly related to their infection properties. This is particularly relevant for when Synechococcus is the dominant cyanobacterium since many clade A and clade B phages infect members of this genus. For this, we developed a mathematical model of host-phage population dynamics suitable for narrow host-range phages, in which each phage infects a single susceptible host, and assessed host and phage abundances in steady-state environmental conditions [46] (see Methods). We used the average latent period, burst size and virulence based on our empirical results for clade A and clade B phages and assumed equal decay and contact rates for both phages.

We considered highly specific interactions, in which distinct cyanobacterial genotypes (H) were each infected by either a distinct clade A or a distinct clade B phage (V): H_A infected by V_A1 and H_B infected by V_B1. We assumed the same growth rates and carrying capacity for the two hosts. At steady-state, the clade A phage significantly drove down the population size of its host, while the clade B phage reduced its host to a much lesser extent (Figs. 6a, c and 7b). This subsequently resulted in a larger mean population size for the clade B phage relative to the clade A phage (Figs. 6a, d and 7a). Moreover, this model predicts that clade B phages have a greater ecological impact, both infecting more cyanobacteria and causing considerably more cyanobacterial mortality than clade A phages (Fig. 7c, d, SI Appendix, Fig. S5). This is in line with our observations that more Synechococcus and Prochlorococcus cells are infected by clade B than by clade A cyanophages in the Red Sea (Fig. 5) and in the North Pacific Ocean [43].

Our model indicates that virulence has a strong nonlinear effect on host-phage interactions resulting in non-monotonic outcomes with peak phage abundances, infected cells and virus-induced mortality occurring at intermediate virulence values (Figs. 6d and S5). Towards the lower end of the virulence scale, clade A phages are predicted to be more abundant than clade B phages (Fig. 6d). At higher virulence values, closer to those found empirically in our study, the more virulent clade A phages are less abundant than clade B phages (Figs. 6d, e and 7a), presumably because clade A phages draw down their host populations to such an extent that they do not support large phage populations (Figs. 6c and 7b). In this model formulation, burst size and virulence have equivalent impacts on drawing-down the steady-state host population from the phage-free steady state (Fig. 7b, Eqs. 7 and 10 in Supplementary Methods). However, at infection properties relevant for T7-like cyanophages, virulence has a much stronger effect on phage abundances than does burst size since reduction in virulence of clade A phages would lead to a substantial increase in their abundance (Fig. 7a, Eqs. 9 and 12 in Supplementary Methods).

We also addressed the situation where a distinct clade A phage (V_A2) and a distinct clade B phage (V_B2) infect a single host genotype (H_AB). In this model, the aggressive clade A phage outcompetes the clade B phage (Fig. 6b), as expected due to its superior infection properties (including burst size [46]). When in direct competition, the clade A phage drives down the host population to levels below those that support replication of the clade B phage (see Methods Eqs. 19–23). This competitive exclusion suggests that, in order for clade A and clade B cyanophages to be found in the same body of water, host separation likely occurs spatially or temporally under direct competition, with a particular host genotype being infected by either a clade A or a clade B phage but not both. Indeed, local patches of microbes and interactions on the microscale between microorganisms are likely in planktonic environments [47,48,49]. We note that multiple phages with similar infection properties could, in principle, infect and coexist on the same host genotype.

Host separation could also result from evolutionary processes. Host evolution through selection for resistance to a phage from either clade would lead to local host separation. The selection pressure for resistance against a clade A phage is likely to be greater than against a clade B phage since resistance to the former would result in the greater increase in host population size. Furthermore, mutations in cyanophages can lead to a change in the hosts they are able to infect [50, 51] and thus allow them to avoid extinction when exposed to direct competition. An example of host separation is apparent in this system for Prochlorococcus since many clade B phages can infect Prochlorococcus genotypes whereas few clade A phages can. Furthermore, the presence of hundreds of diverse cyanobacterial genotypes [17, 18] with different sensitivities to co-occurring cyanophages [25, 26] in the oceans also supports the possibility of host separation. Similar support for host separation in phage-host interaction networks has also been reported for a variety of other taxa [52,53,54]. Irrespective of whether host separation is due to ecological and/or evolutionary processes, larger clade B phage populations are predicted to persist when distinct clade A and clade B phage genotypes infect different host genotypes, as described in the first model formulation (Figs. 6a and 7). As such, the lower fitness and virulence of clade B phages can be reconciled with substantially higher abundances of this clade of phages even when Synechococcus is the dominant cyanobacterium.

Infection properties may also influence phage population diversity and host range. We hypothesize that intermediate virulence allows clade B phages to infect members of the slower growing Prochlorococcus genus [compare [55] and [56]], and perhaps more cyanobacterial types in general under a variety of suboptimal conditions, since clade B phages would reduce their host populations to a lesser extent than the more virulent clade A phages (Figs. 6, 7, SI Appendix, Fig. S6). Having more host types and maintaining larger host populations would result in more overall infections (Fig. 7c). Thus, clade B phages with more overall DNA replication cycles would have greater chances for mutation resulting in increased phage diversity and a greater pool of viral variants available for genetic drift or natural selection. Irrespective of whether the greater diversity of phages and larger repertoire of hosts for clade B phage populations is a consequence of their infection properties or not, the combination of both higher numbers of host types, and intermediate virulence leading to larger sustainable host populations, can explain the greater abundance of clade B T7-like cyanophages over clade A T7-like cyanophages in the ocean.

Our findings raise the possibility that two opposing processes are driving the evolution of virulence in the T7-like cyanophages: Direct phage competition for the same host may lead to the evolution of higher virulence and spatial or temporal host separation. At the community level, however, phage-host separation may select for intermediate virulence which can lead to more sustainable host populations that in turn support larger phage populations. These ideas support the evolution of intermediate virulence in parasites [57, 58], and expand them to include viruses that infect single-celled organisms in complex ecological settings. It will be important for future research to attempt to disentangle the combined effects of multi-scale selection processes [59] in the context of community-level diversity.

Source: Ecology - nature.com