Single-cell and population level viral infection dynamics revealed by phageFISH, a method to visualize intracellular and free viruses

Phage–host dynamics during a one-step growth experiment

To evaluate the performance of phageFISH against standard metrics, we characterized the dynamics of podovirus PSA-HP1 infecting Pseudoalteromonas sp. strain H100 in a one-step growth experiment. It is not yet known conclusively whether PSA-HP1 is capable of an integrated mode; however, there are no genes indicative of a temperate lifestyle in the sequenced genome (M.B. Duhaime, unpubl. data).

The one-step growth experiment proceeded as follows: after physiological acclimation (SI Text and Fig. S2), Pseudoalteromonas cells were incubated 19 min with half as many phages (Multiplicity Of Infection, MOI = 0.5), diluted 100-fold to minimize further phage adsorption, and sampled immediately after dilution (T0; time points are denoted with a ‘T’ followed by the number of minutes post-dilution). Infected host cultures increased in cell abundances until T96, but their growth was compromised as compared with control cultures (Fig. 1A). Free phage numbers rose from T51 to T81, as detected by qPCR-based assays of extracellular phage genomic DNA (qEXT) and plaque-forming unit (PFU) assays of infective phage particles (Fig. 1A, controls see Fig. S3). Intracellular, mature phage particles were first detected at T36 using transmission electron microscopy (TEM), which marks the end of phage DNA replication at T36. The first visibly infected cells (defined as ≥5 phage particles per cell; Brum et al., 2005) were observed by TEM at T51, and the frequency of visibly infected cells (FVIC, see Experimental procedures) reached a maximum at T66, when 12.1 ± 0.02% (confidence interval, c. i.) of cells were infected, after which FVIC decreased to 3.0 ± 0.01% (c. i.) and remained low for the remainder of time points (Fig. 1B).

Open in a separate window

Figure 1

One-step growth infection dynamics of a marine virus–host system.A. Host abundance as total cell counts and extracellular virus abundance as virus abundance per cell by gene presence (qPCR) and plaque assays (PFU).B. Relative abundance of phage containing cells as FVIC by TEM count (black squares) and phageFISH count (coloured circles); pie charts indicate area size distribution among phage signals as determined by phageFISH.C. One-step growth model for phage PSA-HP1 and its host H100.Error bars indicate standard deviation except for FVIC data in which error bars indicate 95% confidence intervals. All data are based on measurements from two biological replicates. All data at T-19 are corrected by a 1/100 factor to be comparable to values measured after dilution of cultures.

Detailed phage–host dynamics observed with phageFISH

PhageFISH provided two metrics in this one-step growth experiment: (i) a quantitative metric – the fraction of infected cells, and (ii) a semi-quantitative metric – the relative extent of per-cell phage infection, measured as the area of the phage signal. The latter ranged from a fraction to the entire cell (see Fig. 1B, pie diagrams, and Fig. 2) and was grouped into three distinct size classes (for details, see Fig. S4): (I) < 0.4 μm², which was the majority of the infections at T0 and most likely represents new infections, (II) 0.4–1.4 μm², which most likely represents the viral DNA replication stage, and (III) 1.4–7 μm², which was the majority of the infections at T36, and most likely represents advanced infections (late replication and assembly).

Open in a separate window

Figure 2

Progression of infection over time: epifluorescence micrographs of virus-infected and uninfected host cells after phageFISH.A: T0–T51, B: T66–T96, C: T111–T46.Left: host only. Centre: virus only. Right column: overlay of host cells in green (Alexa₄₈₈) and virus in red (Alexa₅₉₄).Colour designation for arrows: white = cell lysis and release of free phages, cyan = new infection, blue = rRNA and phage localization, yellow = advanced infection, violet = infection by two phage. The scale bar indicates 5 μm.

The combination of the two metrics allowed us to discriminate between two waves of infection and to describe the infection stages within them (for details on how the two infection waves were segregated, see Fig. S4). The first wave had phage signals that grew from size class I at T0 to size class III (see Fig. 1B – blue circles and, Fig. 2A and B) up to and including cell lysis at T51 (but as late as T96, see below). The second wave signals appeared initially at T51 (as size class I) and grew (both in numbers and in signal area) until T146 (see Fig. 1B – green circles and Fig. 2B and C). Further, within each wave of infection we were able to document details of their relative infection stages. For example, first wave infections plateaued at T21 with 17.3 ± 1.4% (std. dev.) of the cells infected, but reached the largest phage signals (size class III) at T36 (pie charts in Figs 1B and ​and2A).2A). This indicates that few or no new infections were initiated post-T21, but that old infections were progressing via intracellular genome replication until T36. Starting with T51, the decline of the first wave and the onset of the second wave of infection are indicated by (i) the appearance of free phage particles due to cell lysis events (Fig. 2B, described further below), (ii) the decrease in the fraction of cells with advanced infections (size classes II and III), which was minimal at T81 (4.3 ± 1.3% st. dev.) and (iii) the shift back toward smaller signal size classes (Figs 1B and ​and2B),2B), indicative of new infections. The per-cell phage signal area allowed simultaneous tracking of these two waves of infection as they overlapped from T51 until T96, by which time even the latest of the first wave-infected cells had lysed. Notably, through the whole experiment, the phage signal showed distinct sub-cellular localization in the middle of the cell, while the rRNA signal localized at the periphery (Fig. 2), a feature further supported by TEM images of infected cells (Fig. S5).

In addition to these intracellular metrics, phageFISH also detected phage DNA extracellularly, in free phage particles. Simultaneous with the onset of the second wave of infection, we observed phageFISH-stained extracellular particles associated with lysed cells (Fig. 2B and C), suggesting that phageFISH can label target DNA encapsidated in free phage particles. While neither our original intention nor expectation, two follow-up experiments confirmed that phageFISH could indeed target free phage particles. First, a negative control gene probe (Moraru et al., 2010) applied to infected cells showed only ∼ 2% of the cells having false positive signals of the lowest size class, while no cell lysis-like events were visible (Fig. S6), indicating that the phageFISH-stained extracellular particles represented specific hybridizations of the phage probes to DNA. Second, phageFISH on a 0.2 μm-filtered lysate showed clear association between the phageFISH and phage DNA stain (SYBR Green) signals, while no hybridization signals were observed for the negative control gene probe (Fig. 3; for further details on the negative control gene probe, please see SI Text, phageFISH section).

Open in a separate window

Figure 3

Epifluorescence micrograph of free virus after phageFISH (red, Alexa₅₉₄) and SYBR staining (green).On the left: probe targeting a viral gene, on the right: negative control gene probe.Top: overlay of SYBR Green and phageFISH, centre: phageFISH only, bottom: SYBR Green only. Scale bar indicates 2 μm.

PhageFISH provides novel insights into phage–host dynamics

PhageFISH describes the phage infection in a higher degree of detail than any of the other methods used here for comparison. First, phageFISH allowed us to build a model of the previously uncharacterized infection dynamics between phage PSA-HP1 and Pseudoalteromonas sp. strain H100 (Fig. 1C), with the exception of the burst size, as explained above. Our observed phageFISH metrics (Fig. 1B) were corroborated by classical measurements, as the temporal dynamics of the phageFISH signal is consistent with qPCR- and PFU-based measurements of viral abundance (compare Fig. 1A and B) and FVIC counts (Fig. 1B). The latent period (51–66 min; Fig. 1) and phage generation time (70–85 min; Fig. 1) for PSA-HP1 on Pseudoalteromonas H100 are in the range of other marine phage–host systems studied in the lab (e.g. cyanopodovirus Syn5, Raytcheva et al., 2011, and phages infecting various marine heterotrophs, Jiang et al., 1998), while others, such as marine T7-like cyanopodovirus infecting Prochlorochoccus MED4, are known to require 8 h for phage production (Lindell et al., 2007). This variation in timing of lysis is thought to be a consequence of a trade-off between host quantity and quality (Wang et al., 1996), and thus is likely to be variable among environmental populations as well.

Second, phageFISH outperforms the other methods employed to monitor the phage infection dynamics. For example, while PFU and extracellular phage gene qPCR describe the infection stages moderately well (see Fig. 1A), they lack the ability to measure the fraction of infected host cells and fail to discriminate between the two waves of infection. Additionally, the more traditional FVIC metric successfully reports the fraction of visibly infected host cells, but is limited in that it can detect only late stage infection (mature viral particles), misses the second wave of infection, and, while only a problem for application to mixed communities, does not provide lineage-specific information. PhageFISH succeeds on all these fronts. Moreover, even where measurements can be similarly made, phageFISH is likely to be more sensitive than, e.g. FVIC due to its ability to detect phage replication through the entirety of the lytic infection. Specifically, we found that FVIC underestimated the maximum fraction of infected cells (at T66: 12.1 ± 0.02% c. i.) relative to phageFISH (at T21: 17.3 ± 1.4% std. dev.). Presumably, this is due to the fact that phageFISH detects intracellular viral DNA (both free and encapsidated) present through the infection, while FVIC is able to detect assembled virus particles present in the cells only in late-stage infection. In our experiment, by the time FVIC peaks at T66, part of the first-wave-infected cells have already lysed and released new phage (Fig. 1A), resulting in FVIC under-documenting the total fraction of infected cells from the first wave, as would be expected in a semi-synchronized population (Proctor and Fuhrman, 1990; Proctor et al., 1993). This is supported also by phageFISH, which, at T66, shows that 12.9 ± 0.6% std. dev. infected cells for the first wave. Additionally, the centrifugation step of the protocol for TEM samples could cause disruption of infected cells and thus lead to lower counts, as was contemplated by Weinbauer and Höfle (1998).

In addition to intracellular, single-cell detection capabilities, phageFISH also enables detection of free phages. This was most likely made possible due to various steps in the phageFISH protocol capable of denaturing proteins and nucleic acids, e.g. acid, SDS, and formamide and high temperatures (see SI Text, phageFISH section), thus rendering the virus particles accessible to probes. In our experiment, the detection of free phages was particularly valuable for careful documentation of the timing of cell lysis. For example, we observed free, presumably new, phages near cells that appeared to be lysed – i.e. cells with weaker 16S rRNA signal and/or broken cells (Fig. 2B). Almost simultaneously, the bacterial cells started to form aggregates, likely due to the secretion of extracellular polymeric substances. We suspect that these sticky extracellular secretions and cell debris of lysed cells caught a fraction of the free phage and prevented them from passing through the 0.2 μm pore-sized filters during sampling in late-stage infections (e.g. T51), while in early-stage infections (e.g. T0), neither cell debris, nor extracellular secretions, nor free phages were observed.

Current phageFISH applications

PhageFISH holds great promise for advancing viral ecology. First, phageFISH used in traditional model system experiments can advance the field one phage–host system at a time, as accomplished here. Beyond lytic infections, phageFISH offers a particularly valuable tool for studying temperate phage infections, which are now thought to be prevalent in marine systems as referenced by Paul (2008). Even while relatively quantitative, phageFISH allows a researcher to monitor when a temperate phage has switched from a single-copy, silent prophage state to a replicating lytic state.

Second, with improved sensitivity over geneFISH, we are confident phageFISH can be applied successfully to environmental samples, in a similar manner as geneFISH (Moraru et al., 2010; Petersen et al., 2011; Bernhard et al., 2012). Here, simultaneous visualization of phage genes and host rRNA offers a culture-independent means to answer, for targeted populations: how many microbes are infected at any given time, which phage groups are active, and whom do they infect? Such lineage-specific frequency-of-visibly-infected-cell measurements are pivotal steps towards linking virus and host community metrics (e.g. counts, production assays) with community sequence data, which have furthered our understanding of the identities and dynamics of key members of both communities already. Third, as FISH transformed environmental microbiology (sensu Amann et al., 1995), phageFISH enables sequence-based population ecology of free phage particles in the environment, as no current method allows.

While the application of phageFISH to the environment requires prior knowledge of phage and host gene sequence variation for probe design, such data are becoming routine as metagenomic sequencing improves and scales up. Even probe design for the large sequence diversity of environmental phages should prove surmountable by a combination of strategic target group choices and experience gleaned from polynucleotide probes targeting functional genes using geneFISH (Moraru et al., 2010 2011). Unlike cellular organisms, which share universal marker genes and several core genes, the most widely shared gene among viruses is found in only 37% of sequenced viral genomes (Kristensen et al., 2012). Investigation of orthologous genes in viruses has found that 21 of 57 tested viral taxa are represented by at least one signature gene (Kristensen et al., 2012) that (i) is present in all members of the taxon (a measure of signature gene ‘sensitivity’), (ii) does not exist outside the taxon (signature gene ‘specificity’), (iii) is virus-specific, and (iv) is single-copy in the viral genomes. These over 100 taxon-specific signature genes (Kristensen et al., 2012), plus alternatively determined core genes of T4-like myoviruses (Sullivan et al., 2010) and T7-like podoviruses (Labrie et al., 2013), can be further evaluated for phageFISH polynucleotide probes development (Moraru et al., 2011), to determine those which can tolerate up to 10–15% mismatches while still retaining their taxa specificity and sensitivity. Recent work examining T4-like virus genomic diversity suggests this mismatch tolerance to be relevant for both within-population (< 5% gene divergence) and between-population (10–15%) investigation ((L. Deng, J.C. Ignacio-Espinoza, A. Gregory, B.T. Poulos, P. Hugenholtz, M.B. Sullivan, submitted)). However, such family-specific gene marker studies are not without issue (reviewed in Duhaime and Sullivan, 2012). In particular, this approach suffers from limited and taxa-biased virus sequence databases (Kristensen et al., 2012), which will affect the power of phageFISH probe recall and precision and will require prudent consideration before application to viral populations in the environment.

Limitations and opportunities: the phageFISH crystal ball

The ability of phageFISH to quantify relative per-cell phage DNA copy number (e.g. by phage signal area) presents an opportunity for discovery of novel infection and cell biology features and phenomena. Here, it allowed us to document heterogeneity within the infected population and to visualize sub-cellular features of single cells, as we were able to resolve non-overlapping signals for phages and host ribosomes (Figs 2A and S5). This portends the study of cell biology of phage-infected microbes in ways not previously possible.

One current limitation to such absolute quantification is the CARD step (see SI Text). The amplified fluorescent CARD signal results from HRP activating labelled tyramides to bind to tyrosine residues in cellular proteins (Bobrow et al., 1992), with the number of HRP being related to the number of target-bound probes. Such a signal is likely to remain relatively linearly correlated to target copy number until the cellular environment becomes limited in tyramide binding sites (e.g. tyrosine residues in cellular proteins). Thus, broad size classes (e.g. pie charts in Fig. 1B) estimate phage infection progress at the single-cell level in model systems, but finer-scale infection dynamics and/or more complex community application will require more absolute quantification of this per-cell-area signal. Super-resolution microscopy (Schermelleh et al., 2010) might be the answer, as its higher sensitivity and resolution are likely to alleviate the need for CARD amplification altogether. Recently, super-resolution microscopy has allowed the quantification of ribosomes in E. coli based on detection of autofluorescent fusion proteins (Bakshi et al., 2012). Moreover, the first steps in combining FISH techniques and super-resolution microscopy in microorganisms have already been made and led to improved sub-cellular localization of rRNA, including the detection of sub-cellular areas of high versus low ribosomal content (Moraru and Amann, 2012). A CARD-free phageFISH protocol, combined with sub-cellular localization and quantification by super-resolution microscopy, would refine our understanding of the phage lytic process at stages where the numbers of per-cell phage DNA copies and encapsidated phages are informative and it could even allow differentiation of cells with actively replicating phages (concatenated DNA and unordered phage signal; Fig. 1C ‘replication’) versus cells with encapsidated phage DNA (structured localization of phage signal; Fig. 1C ‘assembly’). Both single-cell genome replication rates and burst sizes could be documented across a heterogeneously impacted host cell population, and, if performed on environmental samples, could be done in a lineage-specific manner. PhageFISH with absolute per-cell phageDNA copy quantification could further transform our understanding of lysogeny and its environmental significance (Weinbauer, 2004), e.g. by targeting well-conserved, known families of temperate phages to provide both quantitative estimates of the fraction of cells infected by single-copy (prophage state) and multiple-copy (lytic infection) phages. Further, such ecological descriptions coupled to phageFISH-enabled laboratory experimentation would undoubtedly advance our mechanistic understanding of a key feature of temperate phages – the factors controlling the lytic/lysogenic switch (Zeng et al., 2010; Ptashne, 2011).

Classical metrics

Phage numbers were quantified by SYBR Gold staining (see SI Text) prior to infection and by quantitative PCR (qPCR) and PFU during the experiment (see SI Text). For qPCR, primers targeted a single copy, non-coding sequence motif from the PSA-HP1 genome. For PFU, plaques were enumerated on a host lawn grown on agar plates (described in SI Text). Burst size, the number of phage progeny resulting from the infection of one host cell, was estimated by dividing the total burst size averaged over the last five time points of the one-step growth experiment by the number of infective phage. Since the observed phage numbers (Figs 1A and S3) were determined by different methods, i.e. plaque assay (for PFU) and qPCR (for gene copy number), two different approaches were followed to calculate the number of infective phages. The PFU-based number of infective phages was estimated by subtracting the number of extracellular phage present after dilution (PFU data, average of the first three time points) from the number of phage that were added initially as observed by PFU (i.e. burst size via PFU). The qPCR-based number of infective phages was estimated by subtracting the number of extracellular phage present after dilution (qPCR data, average of the first three time points) from the number of phage that were added initially as counted microscopically after SYBR staining (i.e. burst size via qPCR). Total bacterial cell numbers were microscopically evaluated after staining the immobilized host cells with 4′,6-diamidino-2-phenylindole (DAPI) (see SI Text). For a count of FVIC (Proctor et al., 1993), samples were fixed and prepared for TEM as described in SI Text. For each time point, the first 800 intact cells from one biological replicate were examined at 32 000–88 000 magnification. If the number of mature viruses in a cell was > 5, it was scored as infected (Brum et al., 2005).

We thank A. Wichels (AWI Helgoland) for providing the phage and host strain, members of the Tucson Marine Phage Lab for critical discussions, Gabriel Moraru (Heracle BioSoft SRL, Romania) for assistance developing the High Dynamic Range Imaging protocol for micrograph image processing, Phyllis Lam (MPI for Marine Microbiology) for critical reading of the manuscript, the University Spectroscopy and Imaging Facilities (USIF) in Tucson for providing TEM facility and services. Funding was provided by Biosphere2, BIO5 and a Gordon and Betty Moore Foundation grant to M. B. S. C. L. M. and R. A. acknowledge funding by the Max Planck Society and the Deutsche Forschungsgemeinschaft (Cluster of Excellence MARUM).