We first describe the development of the method and its validation under controlled laboratory settings for both Synechococcus and Prochlorococcus host-phage systems. Then we apply the method to quantify the contribution of T4-like and T7-like cyanophage infection to the mortality of Prochlorococcus over the diel cycle in the upper mixed layer of the North Pacific Subtropical Gyre during the summer of 2015.
Development of the iPolony method for the detection of viral infection
Our goal was to develop a sensitive, high-throughput method amendable to analysis of field samples that can assess infection by virus families of interest. The polony method (named for PCR colonies that form during the reaction) is a direct PCR-based method that detects virus DNA inside infected cells. The original polony method was developed by Mitra and Church  and was subsequently adapted for the enumeration of free virus particles from the two major cyanophage lineages, the T4-like cyanomyoviruses  and the T7-like cyanopodoviruses . Here we describe further development of the method to simultaneously detect viral infection in thousands of individual cells embedded in a solid-phase gel in a high-throughput manner. We term this method for the quantification of infected cells ‘iPolony’.
The iPolony method has two steps (Fig. 1). In the first step, infection is arrested by fixation with glutaraldehyde, and target cells are isolated and concentrated (Fig. 1a). In its use here, Prochlorococcus and Synechococcus are sorted by flow cytometry based on forward angle light scattering, a proxy for cell size, and the autofluorescence of chlorophyll a and phycoerythrin, respectively. The concentration of cells is then quantified to calculate the number of cells analyzed in the downstream molecular analysis. In the second step, the cells are screened for the presence of intracellular virus DNA using the polony method (Fig. 1b). Polyacrylamide gels are cast with embedded sorted cells and a 5′-acrydite-modified primer to anchor the primer to the gel. PCR reagents are diffused into the gels. Degenerate PCR primers target a signature gene shared by the virus group of interest, in this case the DNA polymerase gene for the T7-like cyanopodoviruses  and the portal protein gene (g20) for the T4-like cyanomyoviruses . Prior to thermal cycling, cells are permeabilized to allow amplification of virus DNA with an in-gel heat lysis step (Supplementary text, Supplementary Fig. 3). During thermal cycling, a cell that contained virus DNA results in an anchored sphere of amplification or polony. After amplification, polonies in gels are hybridized with fluorescently labeled degenerate cyanophage group-specific probes, and gels are scanned with a microarray scanner. Each individual cell is counted as infected whether there is a single or multiple virus genome copies, which enables quantification of infection in a presence–absence manner per cell. Percent infection is calculated by dividing the number of polonies by the number of cells in the gel.
Validation of the iPolony method in controlled virus growth experiments
The method was first developed and tested on model lab systems under known conditions. We assessed the ability of the iPolony method to detect virus DNA inside cyanobacterial cells throughout the infection cycle. Two model systems were used: Synechococcus sp. strain WH8109 infected with the T7-like cyanophage, Syn5, and Prochlorococcus sp. strain MIT9515 infected with the T4-like cyanophage, S-TIM4. Virus DNA was detected inside cells at all stages of the infection cycle (Fig. 2). The efficiency of detection was lower prior to virus genome replication and was 43% for Syn5 and 11% for S-TIM4 (Fig. 2a–d). Detection of infected cells rose with the onset of DNA replication, reaching 86% of maximal infection levels based on host gDNA degradation, midway through genome replication for Syn5 and at the end of genome replication for S-TIM4 (Fig. 2a–d). Since more than a single phage can enter cells in these experiments, we verified that single gene copies can be detected inside cells for a single copy host gene, rbcL, in Synechococcus WH8109, as well as for E. coli carrying a single plasmid with a cyanophage g20 copy (see Supplementary results and discussion). These findings indicate the method is sensitive enough to detect infection throughout the entire infection process even when a single molecule of phage DNA is present prior to genome replication and reaches maximal levels after the onset of DNA replication.
Next, we assessed whether the method could accurately detect varying levels of infection by exposing Synechococcus sp. strain WH8109 to different numbers of the Syn5 phage, such that the infective virus-to-host ratios (multiplicities of infection, MOIs) ranged from 0.1 to 3, spanning infection percentages that theoretically range from 10 to 95% based on encounter theory estimates (Supplementary Table 3). The percent infection determined by the iPolony method was significantly and positively correlated with the MOI (Fig. 3a) (F = 143.1, R2 = 0.87, p < 0.001) and with values obtained using a lysis-based culture-dependent assay  (Fig. 3b) (F = 55.60, R2 = 0.70, p < 0.001). Therefore, the method reliably detects differences in the number of infected cells and provides estimates comparable to culture-based methods.
Single-cell based quantification of viral infection in the environment
We adapted the iPolony method to meet the challenges of the complexity of natural communities. The initial steps are similar to those described above for model systems. First, infection in field samples is halted by fixation and flash freezing, enabling simple and rapid sample collection in the field. Samples can be stored for at least a year without affecting infection levels allowing flexible analysis time (Supplementary text, Supplementary Fig. 1). Second, cell sorting with a flow cytometer removes cells from complex communities, from free viruses, and from fixative-containing seawater that inhibits the polony reaction . Furthermore, with our cytometer configuration sorting concentrates cells of interest to ~1000 ± 200 cells µl−1, which enables reactions to simultaneously screen up to ~10,000 cells for viral infection. Gels with fewer than 5–10 polonies per slide are considered to be below the limit of accurate quantification . Thus, the method has a sensitive detection limit of 0.05% infection in a reaction with 10,000 cells per gel. Third, infection values are corrected for the presence of free viruses either by empirically testing the filtrate of sorted cells for free viruses, as was done for this study, or through a correction factor employed when free virus abundances are above an empirically determined threshold of when co-sorting with cells occurs (Supplementary results and discussion, Supplementary Fig. 4). Finally, differences in detection efficiency before, during and after genome replication in the infection cycle (Fig. 2) are used to adjust and provide bounds for infection estimates (see Supplementary text) since microbial populations in the environment are likely at different infection stages (see below). The maximal difference in infection was 3.4-fold between the two most extreme scenarios where all infected cells were either prior to or after virus genome replication for both T4-like and T7-like cyanophage infection assays, respectively. Therefore, using latent period weighted efficiencies to calculate infection values in natural samples provides accuracy within a ~3-fold range. These procedures allow for sensitive and high-throughput detection of viral infection from environmental samples.
Low extent of viral infection of Prochlorococcus in the North Pacific Subtropical Gyre
Next we assessed whether viral infection exerted a significant control on Prochlorococcus abundances. We used the iPolony method to measure instantaneous levels of infection of Prochlorococcus at the high temporal resolution of every 4 h over 7 diel periods in the surface mixed layer in the North Pacific Subtropical Gyre (NPSG) in the summer of 2015. A Lagrangian sampling strategy was implemented to sample the same water mass during the sampling period by following drifters centered at 15-m depth .
Picocyanobacteria abundances were measured continuously with an underway flow cytometer  and varied between 1.2–2.3 × 105 and 0.4–1.6 × 103 cells ml−1 for Prochlorococcus and Synechococcus, respectively (Fig. 4a, Supplementary Fig. 5). Because of the low abundances of Synechococcus, we focused our attention on Prochlorococcus as the most abundant primary producer in these waters. Free cyanophages were measured using the polony method [33, 34] and were found to be 2–4-fold more abundant than Prochlorococcus in the surface mixed layer (Fig. 4b). T4-like cyanophages ranged between 2.7 and 4.1 × 105 viruses ml−1, and were an order of magnitude more abundant than T7-like cyanophages that numbered 1.7–3.8 × 104 viruses ml−1. T4-like and T7-like cyanophages constituted ~2.2% and 0.1% of total VLPs, respectively. These abundances are within the range of our previously reported measurements for cyanophages in the upper mixed layer of the NPSG . Furthermore, T4-like cyanophages, T7-like cyanophages, TIM5-like cyanophages, and cyanosiphoviruses were 89.0% (±1.2%), 3.55% (±0.96%), 0.35% (±9.5 × 10−5%), and 7.10% (±0.62%), respectively, of total cyanophage DNA sequence reads, normalized for genome size, in viromes taken in coordination with polony samples on the same cruise . This provides independent evidence that T4-like cyanophages were considerably more abundant than T7-like cyanophages during this cruise and that other known cyanophage groups were not abundant .
Metagenomes and viromes were also used to assess the extent to which our degenerate primers and probes captured the major cyanophage types in these waters. The degenerate primers and probes which target the T4-like cyanophage g20 gene captured the sequence variation in all 11 dominant T4-like cyanophage contigs  (Supplementary Fig. 6a). Sequence variation in the individual reads indicated that assemblies represented the dominant genotypes of cyanophages in the water and that primers and probes used in this study were compatible with the diversity of at least 93% of the individual reads (Supplementary Fig. 6b). Furthermore, sequenced amplicons from free cyanophages collected at 25- and 75-m depths during this cruise indicated that the primers captured a diverse set of T4-like cyanophage from within the population . Therefore, any underestimation for cyanophage abundances and infection is likely to be minor and within the threshold of detection, unless due to a presently unknown, nonetheless abundant, cyanophage genotype.
Although cyanophages were abundant in the water column, only a small percentage of the total Prochlorococcus population were infected in all samples analyzed over all 7 days (Fig. 4c). Infection averaged 0.79% of the total Prochlorococcus community spanning from 0.35 to 1.6% or 520 to 3100 virally infected cells ml−1 (Supplementary Fig. 7). T4-like cyanophages accounted for over 80% of the observed infections. Infection by T7-like cyanophages was at or below the limit of accurate quantification (<0.1% of the Prochlorococcus population after adjusting for cyanophage detection efficiency) for the majority of the sampling period (Fig. 4c). These data highlight the importance of understanding the nature of specific virus–host interactions as different virus lineages have significantly different impacts on cyanobacterial communities.
Our observation that 0.35–1.6% of Prochlorococcus is infected, with estimates of 0.35–4.8% mortality (see below), is striking given the current paradigm that 10–40% of all cells are lysed by viruses each day [10, 11, 13,14,15, 36, 47, 48]. How can our observations be reconciled with these previous studies? Viruses were estimated to contribute between −21 and 22% of Prochlorococcus mortality in oligotrophic surface waters using the dilution method , which indirectly assesses mortality and lacks the sensitivity to accurately measure low infection levels [27, 49, 50]. A wide range of mortality values from <0.1 to 46% have also been reported using contact rates [18, 19] or production-decay calculations  as a proxy for infection. These approaches require enumeration of cyanophages and have employed methods that systematically underestimate abundances [11, 33]. They also rely on assumptions that are poorly constrained for environmental populations including variable burst sizes and decay rates as well as high virus infectivity and host susceptibility, which introduce inaccuracy and lead to significant overestimations of mortality. For example, estimated infection frequencies would be lower by up to 10-fold if decay rates based on longer reported estimates  were used. Finally, the frequency of visibly infected cells method identifies cells containing viral particles (capsids) and uses debatable conversion factors for the period of time that particles are visible to estimate mortality [16, 52, 53]. For example, 30% of picocyanobacteria mortality was attributed to viruses assuming capsids are visible for 10% of the latent period based on a heterotrophic bacteriophage system , whereas had 50–60% been used based on cyanophage systems 5–6% mortality would have resulted given the same data . The iPolony approach improves on these most commonly used methods by directly measuring infection of tens of thousands of cells at all stages of the infection cycle and uses fewer and more well-constrained assumptions in converting infection to mortality estimates. This results in higher sensitivity with improved precision, reducing methodological variability and enabling a more accurate evaluation of viral mortality in natural systems.
Nonetheless, as a means to independently verify our estimates of 0.35–1.6% infection, we enumerated the number of single-cell genomes of Prochlorococcus collected from oligotrophic surface waters  that contain cyanophage contigs. While no single sample had enough cells sequenced to adequately determine infection, out of 298 sequenced Prochlorococcus cells from many geographically separate regions, 3 had cyanophage contigs in their genome assemblies , equivalent to ~1.0% infection. Thus, while not a robust means of quantification at low levels of infection, single-cell genomics provided independent support for our findings with the iPolony method.
Diel periodicity of infection
Prochlorococcus abundances displayed diel periodicity (p-value <0.001), with numbers increasing at night and decreasing during the day (Fig. 4a). Changes in Prochlorococcus cell size were also periodic (p < 0.001) with cell volume decreasing by nearly half during nighttime due to cell division and respiration (Fig. 4a). This suggests that cells were dividing almost once a day consistent with previous observations in the region [5, 7]. A daily doubling of the entire Prochlorococcus population would be expected to produce ~1–2 × 105 cells ml−1 each night if there was no mortality. However, the Prochlorococcus population only increased by 0.2–0.5 × 105 cells ml−1 after dusk indicating tightly coupled loss factors that maintained relatively constant Prochlorococcus abundances. Despite the diel periodicity in Prochlorococcus populations, standing stock abundances of VLPs and cyanophages in the water column showed no diel periodicity (Fig. 4b) (p-value = 0.62 for VLPs, p-value = 0.40 for T4-like cyanophage, p-value = 0.42 for T7-like cyanophage).
In contrast to cyanophage abundances, viral infection of Prochlorococcus displayed diel periodicity (Fig. 4c) (p-value <0.001). Viral infection was relatively stable and low during daylight hours (0600–1400 h) and increased ~2-fold toward dusk (1800 h), which coincided with the peak in Prochlorococcus cell division (Fig. 4a, c). At night (1800–0200 h) infected cell abundances were approximately two times greater compared to daylight hours (0600–1400 h) with maximal infection detected between 2200 and 0200 h. The observed increase in percent infection at night may be due to increased infection, increased detection of infection because infection had progressed into stages of phage genome replication, decreased lysis, or a combination of these factors. Finally, infected cell abundances dropped ~2-fold between 0200 and 0600 h indicating that most infected cells underwent an early morning lysis event. This punctuated lysis may be an important event for the surrounding heterotrophic bacterial community which are likely to be utilizing the cyanobacterial lysate and may be partially driving the observed diel patterns in heterotroph metabolism . It is important to note that a fraction of Prochlorococcus cells was always infected by T4-like cyanophages with between 450 and 1400 infected cells ml−1 found in morning hours (Supplementary Fig. 7). Thus, infection appeared to be strongly, but not entirely, phased to the light-dark cycle as there was some degree of asynchronous infection.
There is a clear decoupling between the low and cyclical infection dynamics and high and relatively invariable cyanophage abundances. We hypothesize that relatively slow cyanophage particle decay rates allow for the accumulation of a high standing stock of viruses which obscured the small periodic signal in daily changes in cyanophage abundances. We estimate that the turnover of the T4-like cyanophage standing stock of 3.3 × 105 phages ml−1 would occur every 8–23 days assuming an average burst size of 12 T4-like cyanophages cell−1 (Supplementary Table 4), an average of 1200 Prochlorococcus cells ml−1 are lysed by T4-like cyanophages during each infection cycle (Supplementary Fig. 7), and 1–3 infection cycles occur each day (see below). This viral turnover duration is at the slower end of a wide range of previously reported decay rates for cyanophages, ranging from undetectable in dark conditions to between 0.048 and 2.01 d−1 in surface waters (equivalent to turnover times of 0.5–21 days) [17, 51]. Furthermore, decay rates are typically based on the decline in infective titers [17, 19, 51]. We expect particle decay to be slower as it not only requires a loss of infectivity but also degradation of gDNA.
Transcriptional activity of cyanophages conducted during the same cruise from the same depth  provide qualitative support for the phasing of infection. Cyanophage transcriptional activity peaks in the late afternoon and is at a minimum at dawn, which is also consistent with other reports of cyanophage transcription in the NPSG [56, 57]. Notably, cyanophage DNA and RNA metabolism genes are maximally transcribed at dusk , suggesting that viruses are replicating their genomes during or just after the time of host cell division . Irrespective of viral infection, Prochlorococcus maximally expresses ribonucleotide reductases just prior to cell division . This creates a large pool of free nucleotides for host cell division that could be efficiently utilized by an infecting virus for its own genome synthesis. We hypothesize that the consistent synchrony of Prochlorococcus’ cell cycle to the light/dark cycle creates differences in the availability of intracellular resources throughout the day and may have selected for the phasing of cyanophage genome replication to coincide with host genome replication in order to maximize virus production.
The nighttime increase in cyanophage infection was unexpected as different lines of evidence suggest that darkness can prevent or reduce phage adsorption, reduce burst sizes, and lengthen or halt the infection cycle in laboratory experiments [59,60,61,62]. Furthermore, cyanophage genes are transcribed less  and cyanophage have decreased genomic DNA replication [35, 59, 61, 63] in the dark. However, the effects of darkness differ for different cyanophages even in a single family of T4-like cyanophages . Thus, various cyanophages in nature are also likely to respond differently to darkness, with some capable of adsorption at night, some capable of genome replication and some capable of neither. The decrease in cyanophage transcription observed near dawn is likely to be at least partially due to a 2-fold drop in infected cell abundances. This significant drop in infection prior to sunrise suggests that there may be a presently unknown mechanism that induces the early morning lysis event.
Estimating daily mortality and frequency of encounters that result in infection
We used the high frequency instantaneous measurements of viral infection to estimate the daily contribution of viruses to the mortality of Prochlorococcus (Supplementary text). Mortality is caused by lysis and is therefore relative to the number of infection cycles that can be completed in the population turnover time, which is approximately a day for Prochlorococcus in oligotrophic surface waters [5, 7]. Thus, the bounds of daily Prochlorococcus mortality fell between 0.35% (using the lower limit of instantaneous infection of 0.35% and 1 infection cycle a day) and 4.8% (using the upper limit of instantaneous infection of 1.6% and 3 infection cycles per day) (Fig. 4c). These bounds are based on the average latent period for numerous T4-like cyanophages (7.9 h, n = 9, Supplementary Table 4) determined in culture and assume all infections lead to lysis, which may be complicated by cells which overcome infections via intracellular resistance . Over the course of the last 3 days of the sampling period, T4-like cyanophages increased by 76,000 cyanophages ml−1 whereas T4-like infected cells increased by 500 cells ml−1. Had there been a single infection cycle per day, a burst size of ~150 T4-like viruses per infected cell would be required to produce 76,000 T4-like cyanophages, which is 3 to 13-fold greater than measured burst sizes of 12 and 20–50 viruses cell−1 for infection of Prochlorococcus and Synechococcus, respectively, under optimal culture conditions (Supplementary Table 4). This suggests that multiple infection cycles occurred each day and that different subsets of the Prochlorococcus population would be infected at different times over the diel cycle. Our observation that infection was always detected and not completely synchronized supports this hypothesis (see Fig. 4).
The estimated daily mortality values are surprisingly low considering the high abundances of Prochlorococcus and free cyanophages. In order to compare the number of expected encounters to the number of empirically observed infections, we used encounter rate theory , which does not consider viral infectivity and host susceptibility, to estimate the number of potential interactions between Prochlorococcus and T4-like cyanophages based on diffusion alone. An individual Prochlorococcus cell should encounter one cyanophage every 54 h for a total of 70,000 encounters per day in a population of 1.6 × 105 Prochlorococcus cells ml−1, which is the theoretical maximum rate given the average cell and virus concentrations observed at this time. If all encounters were to result in infection and lysis, viruses would be expected to kill 44% of Prochlorococcus daily. Yet our estimates of 0.35–4.8% daily mortality suggest that only 0.80–11% of the encounters resulted in infection.
Multiple mechanisms may explain the difference in predicted mortality based on encounter rates and the estimates of mortality, including the loss of infectivity, low adsorption efficiency, and host resistance. If each mechanism was solely responsible for mitigating infectious encounters, it would require that ~4.2% of cyanophages would be infectious, only 4.3% of encounters would result in adsorption, or 96% of Prochlorococcus (1.53 × 105 cells ml−1 out of 1.6 × 105 cells ml−1) would be resistant to their co-occurring viruses, to reconcile the observed infection rate with that expected from encounter theory. Instead, it is likely that all three mechanisms contribute to the observed infection frequencies. Previous reports in surface waters indicate that 20–92% of newly produced cyanophage lose infectivity per day [17, 65]. In cultured host-virus systems, 5–25% of encounters between infective phages and sensitive hosts result in adsorption and infection [66, 67]. While it is not known what fraction of wild Prochlorococcus populations is susceptible to any given virus, Prochlorococcus populations consist of hundreds of distinct types  with different sensitivities to co-occurring phages [69, 70]. Therefore, we suggest that a combination of these mechanisms, balanced at 10–35%, mitigates infections and thus sustains coexisting cyanobacteria and virus populations despite high Prochlorococcus cyanophage encounter rates (see Supplementary discussion, Supplementary Fig. 8).