Global morphological analysis of marine viruses shows minimal regional variation and dominance of non-tailed viruses

Abstract

Introduction

Viruses are key players in the Earth's ecosystem not only because they are the most abundant and diverse biological entities in marine environments (reviewed by Wommack and Colwell, 2000; Breitbart et al., 2007) but also because they have considerable influence on ecological, biogeochemical and evolutionary processes in the ocean (reviewed by Fuhrman, 1999; Weinbauer, 2004; Suttle, 2007; Breitbart, 2012). Viral-induced mortality of microorganisms in the ocean can affect microbial species composition (Thingstad, 2000) and alter the flux of nutrients and organic matter by increasing recycling of these materials through the microbial loop (reviewed by Fuhrman, 1999). Expression of viral auxiliary metabolic genes (sensu Breitbart et al., 2007), such as core photosystem genes, during infection may also substantially impact oceanic productivity (Lindell et al., 2005; Clokie et al., 2006; Lindell et al., 2007; Sharon et al., 2007; Dammeyer et al., 2008; Thompson et al., 2011). In addition, viral-mediated horizontal gene transfer can profoundly alter the evolution of oceanic microorganisms as has been demonstrated in marine cyanobacteria (for example, Lindell et al., 2004; Sullivan et al., 2006; Ignacio-Espinoza and Sullivan, 2012).

With these significant roles in oceanic ecosystems, it is important to understand the characteristics of marine viruses and their distribution in the oceans. The majority of marine viruses are thought to infect bacteria (Wommack and Colwell, 2000) and taxonomic surveys based on the bacterial 16S rRNA gene have shown that bacterial assemblages vary between oceanic regions (Schattenhofer et al., 2009; Barberan et al., 2012). Thus, one would expect viral assemblages to vary between oceanic regions as well. Viruses do not have a universal marker gene so assessing their diversity across spatial scales is challenging and has resulted in the use of metagenomics to compare viral assemblages from different environments (Breitbart et al., 2004b; Angly et al., 2006; Dinsdale et al., 2008; Hurwitz and Sullivan, 2013). The first study to compare marine water column viral metagenomes showed that viral assemblage genetic distance not only increases with geographical distance but also that there is considerable overlap in viral assemblages across sites even though constituent viral abundances vary (Angly et al., 2006). In fact, one particular podovirus DNA polymerase sequence is present in several aquatic and terrestrial environments (Breitbart et al., 2004a). A much larger-scale Pacific Ocean viral metagenomic data set (Hurwitz and Sullivan, 2013) employing quantitative methodologies (John et al., 2011; Duhaime and Sullivan, 2012; Duhaime et al., 2012; Hurwitz et al., 2013; Solonenko et al., in press) is now available to examine biogeography, but such studies have not yet been conducted. This is because the database representation for sequence comparisons are so poor that most ocean viruses are not yet identifiable (for example, Angly et al., 2006; Hurwitz and Sullivan, 2013). Thus simple questions such as how viral assemblages vary across oceanic regions remain unanswered.

An alternative to metagenomics is comparing viral assemblages throughout the world's oceans using morphology. Viral morphology is central to modern viral taxonomy (King et al., 2012) and commonly correlates with whole-genome-derived taxonomy (Rohwer and Edwards, 2002) and aspects of their biology (reviewed by Ackermann, 2001). Thus, morphological metrics have applications ranging from medical diagnostics (Doane, 1980) to environmental virology (for example, Bratbak et al., 1990; Weinbauer and Peduzzi, 1994). In aquatic environments, morphological metrics documented spatiotemporal changes in viral assemblages, revealing aquatic viruses as dynamic and varied across large environmental gradients (Bratbak et al., 1990; Auguet et al., 2009; Brum and Steward, 2010; Bettarel et al., 2011a, 2011b). Environmental morphological studies also aid viral discovery, finding novel morphologies, including large viruses (Bratbak et al., 1992; Gowing, 1993; Sommaruga et al., 1995), spindle-shaped viruses (Oren et al., 1997) and filamentous viruses (Hofer and Sommaruga, 2001). Finally, morphological analyses are not plagued by the database bias issues (Edwards and Rohwer, 2005) that undermine quantitative viral taxonomic analyses in metagenomic studies.

Sample preparation, however, has only recently been resolved for quantitative viral metagenomic studies (reviewed in Duhaime and Sullivan, 2012) and remains an obstacle to being quantitative in environmental viral morphological studies. Transmission electron microscopy (TEM) sample preparation generally includes one of the two approaches: either viruses are concentrated and then adsorbed to TEM grids (for example, Sommaruga et al., 1995; Stopar et al., 2003) or they are directly deposited onto TEM grids using traditional (for example, Bergh et al., 1989) or air-driven ultracentrifugation (Maranger et al., 1994; Brum and Steward, 2010). Here, we use an air-driven ultracentrifuge with a rotor designed to quantitatively deposit viruses onto TEM grids (Hammond et al., 1981), resulting in high recovery of viruses (Maranger et al., 1994). We evaluate this quantitative TEM (qTEM) method to determine the best conditions for sample collection and processing, as well as its biases when applied to marine samples. Using qTEM, we then document viral morphological diversity in the upper water column at 14 stations in six global ocean regions using highly contextualized samples collected on the Tara Oceans Expedition (Karsenti et al., 2011).

Materials and methods

Results

Morphological characteristics of oceanic viral assemblages by depth and oceanic region

The Tara Oceans samples were collected from the surface and DCM of 14 stations in six oceanic regions with a range of environmental conditions (Supplementary Table S1). Across 2600 viruses and 26 samples examined, only four viral morphotypes were observed: myoviruses, podoviruses, siphoviruses, and non-tailed viruses (Figure 1). Overall, viral morphotype composition and capsid diameter were remarkably consistent with depth and oceanic region (Figure 2; details for each sample in Supplementary Figures S4–S9). Non-tailed viruses dominated in each depth and oceanic region (average 66–85%), while myoviruses, podoviruses and siphoviruses were the next most abundant morphotypes, in that order, except in the Mediterranean Sea where podoviruses exceeded myoviruses (Figure 2a). Regionally, non-tailed viruses were negatively correlated with salinity and podoviruses were positively correlated with salinity (Supplementary Table S2, Supplementary Figure S10). For correlations among individual samples, non-tailed viruses and podoviruses were correlated with salinity while myoviruses and podoviruses were correlated with temperature (Supplementary Table S2). However, these relationships reflected changes in the range of the relative percentage of these morphotypes and were often driven by only 3–4 samples (Supplementary Figure S10). No morphotype was significantly correlated with oxygen or chlorophyll concentration (Supplementary Table S2).

Open in a separate window

Figure 1

Examples of the four viral morphotypes observed in this study ((a), myovirus; (b), podovirus; (c), siphovirus; (d), non-tailed virus).

Open in a separate window

Figure 2

(a) Percentage of viral morphotypes in all the surface samples combined, all the DCM samples combined and each oceanic region. Error bars represent s.ds. of the means of all the samples. Letters indicate significant differences between depths or oceanic regions while numbers indicate significant differences within depths or oceanic regions (ANOVA with Tukey's post-hoc test, P<0.001 for all). (b) Box and whisker plots of viral capsid diameters in all the surface samples combined, all the DCM samples combined and each oceanic region. Top, middle and bottom lines of each box correspond to the 75th, 50th (median) and 25th percentiles, respectively. Whiskers extending from the top and bottom of each box correspond to the 90th and 10th percentiles, respectively. Circles represent capsid diameters that are outside of the 90th and 10th percentiles (outliers). Letters indicate significant differences between depths or oceanic regions (ANOVA with Tukey's post-hoc test, P<0.001 for all). The number of viruses used for each data set is given in parentheses.

With respect to capsid diameters, there was no significant difference between pooled surface and DCM samples (Figure 2b). Regionally, viral capsid diameters in the Mediterranean, Red and Arabian Seas were significantly larger than those in the Indian, Atlantic and Pacific Oceans (Figure 2b). These larger overall capsid sizes occurred in the highest salinity oceanic regions (Supplementary Table S1) with average capsid diameter positively correlated with salinity for individual samples (Supplementary Table S2, Supplementary Figure S10). There were no significant relationships between average capsid diameter and environmental parameters when considering pooled data for oceanic regions (Supplementary Table S2).

CA to compare sample capsid diameter distributions, as well as capsid diameter bins (Figures 3a and b), was then used to more deeply explore biogeography and the influence of environmental variables on viral assemblage morphological characteristics. Differences between surface and DCM samples were highly variable (Figure 3a), with some surface samples more similar to the DCM sample at the same station (for example, station 41) and others much more divergent (for example, station 34). Further, there was no significant correlation between depth of the DCM and distance between surface and DCM samples at each station on either the CA1 or CA2 axes of the ordination plot (Pearson's correlations; P>0.3 for both). Biogeographical differences in viral assemblages were also not well supported, with considerable overlap between samples from each ocean and sea. In fact, the distance between samples on the CA1 or CA2 axis of the plot was not significantly correlated with geographical distance between samples considering either all samples or only surface or DCM samples separately (Pearson's correlations, P>0.4 for all).

Open in a separate window

Figure 3

Ordination of Tara Oceans samples (a) and capsid diameter bins in nm (b) using CA based on distribution of viral capsid diameters with 7 nm bins (s, surface sample; d, DCM sample; surface sample from station 36 is omitted due to missing oxygen data; percentage of total inertia explained by CA1 and CA2 is reported on the axes). Lengths of vectors overlaid on the sample ordination plot correspond to the strength of influence for each environmental variable, with r² and P-values reported for each vector (c–f). Response surfaces for each environmental variable are also overlaid on the sample ordination plot to assess linearity of the relationship, with r² (adjusted), P-values and the percentage of deviance explained reported for each response surface (c–f). CA1 was negatively correlated with salinity (Pearson's correlation, r=−0.486, P=0.014) while CA2 was negatively correlated with temperature (Pearson's correlation, r=−0.623, P<0.001) and positively correlated with oxygen (Pearson's correlation, r=0.646, P<0.001).

Environmental variables were more explanatory than geography or depth in evaluating viral assemblage morphology in the global oceans. Salinity was the most important environmental variable explaining capsid diameter distributions (CA1 was negatively correlated with salinity and explained the most inertia in the ordination plot; Figures 3a). Vectors and response surfaces of environmental variables showed that, while the relationship with temperature was non-linear, temperature, salinity and oxygen, but not chlorophyll a, significantly influenced capsid diameter distributions (Figures 3c–f). For example, samples from the surface at station 23 and the DCM at stations 23 and 30 in the Mediterranean Sea grouped together (Figure 3a), sharing both narrow capsid diameter peaks (49–63 nm; Supplementary Figure S4) and similar environmental conditions (low temperature plus higher salinity and oxygen; Supplementary Table S1). By contrast, samples from the DCM at station 41 and surface at stations 34 and 41 from the Red and Arabian Seas grouped together (Figure 3a), sharing wider capsid diameter peaks (49–91 nm; Supplementary Figures S5 and S6) and similar environmental conditions (higher salinity and temperature, lower oxygen; Supplementary Table S1). However, most samples were closer to the CA plot origin, suggesting weaker influences from environmental variables (Figure 3a).

Ordination of capsid diameter bins was also influenced by environmental variables (Figure 3b). However, bins furthest from the origin tended to have the fewest viruses, although this relationship was only significant for the CA2 axis (Pearson's correlation, r=−0.659, P=0.004), suggesting that bins with the most viruses were least influenced by the environmental extremes observed, resulting in relatively consistent abundances across samples. To evaluate the influence of low abundance bins (<5 viruses), the CA was repeated without them and did not significantly change the analysis results (Supplementary Table S3).

Similarly, the ordination analyses were relatively insensitive to capsid diameter bin size. Analyses using each of the minimum (5 nm) and maximum (10 nm) optimal bin sizes determined for the samples provided similar results for the influence of environmental parameters on capsid diameters of viral assemblages (Supplementary Table S3). Exceptions include reduced significance of the temperature vector and oxygen response surface with 10 nm bins, most likely because this larger bin size insufficiently resolved capsid diameter distributions in most samples.

Tailed virus sample size was relatively low, reducing statistical power to evaluate spatial differences in their morphological characteristics. With this caveat, morphotype-specific tail lengths were not different between the surface and DCM samples, except for siphovirus tails which were longer in surface samples (Figure 4, but note that only six siphoviruses were detected in DCM samples). Among oceanic regions, myovirus tails were longer in the Arabian Sea than Mediterranean Sea, Red Sea and Atlantic Ocean; siphovirus tails were longer in the Red Sea than Mediterranean Sea; and podovirus tail lengths were not significantly different among the oceanic regions (Figure 4). Correlation analyses between tail lengths and environmental variables were not attempted owing to low sample sizes.

Open in a separate window

Figure 4

Box and whisker plots of myovirus, siphovirus and podovirus tail lengths in all the surface samples combined, all the DCM samples combined and each oceanic region. Refer to Figure 2 for a description of box and whisker plot construction. The number of viruses used for each data set is given in parentheses. Letters indicate significant differences between depths (t-test, P=0.001) or oceanic regions (ANOVA on ranks with Dunn's post-hoc test, P<0.05 for all).

Global marine viral morphological characteristics

Pooling all the sample data allowed examination of overall characteristics of upper water column viruses. Again, non-tailed viruses dominated (averaging 79% of all the viruses), followed by myoviruses, podoviruses and siphoviruses, in that order (Figure 5a). Myoviruses had the largest overall capsid diameters followed by siphoviruses, podoviruses and non-tailed viruses, with combined tailed viruses having significantly larger capsids than non-tailed viruses (Figure 5b). Also, tail lengths statistically differed with siphoviruses having the longest tails, followed by myoviruses, then podoviruses (Figure 5c). In addition, 48% of the 27 observed siphoviruses had prolate capsids and 3% of all the observed myoviruses had both capsid diameters and tail lengths either within or smaller than the dimensions described for dwarf myoviruses (Comeau et al., 2012).

Open in a separate window

Figure 5

Morphological results of all the viruses in this study, including the percentage of each morphotype (a), as well as capsid diameters (b) and tail lengths (c) of all the viruses and each morphotype. The average and s.d. are given for each set of viruses, with ranges reported in parentheses, and the number of viruses analyzed (N) is given for capsid diameters and tail lengths. Refer to Figure 2 for a description of box and whisker plot construction. Letters indicate significant differences between morphotypes (ANOVA on ranks, P<0.001 for all) and numbers indicate significant differences between capsid diameters of non-tailed and all tailed viruses combined (b; Mann–Whitney rank sum test, P<0.001).

Discussion

Global ocean qTEM analyses showed that while viral assemblage morphological attributes vary between samples, there is little evidence for consistent variation with depth or oceanic region. The proportion of observed morphotypes (myoviruses, podoviruses, siphoviruses and non-tailed viruses) was highly similar in each oceanographic region, suggesting that there are controlling factors maintaining their relative abundances in the world's oceans. Average capsid diameter was significantly greater in the Mediterranean, Red and Arabian Seas, but neither depth nor inter-sample geographical distance explained variations in sample capsid diameter distributions. Thus, viral morphological attributes in the upper global oceans were not explained by depth or biogeography.

Instead, environmental conditions appear to influence viral morphological characteristics. Although no strong relationships between viral morphotype percentages and environmental variables emerged, larger average viral capsid diameters were significantly associated with higher salinity in individual samples. Using capsid diameter distributions as a more refined metric for viral morphology resulted in temperature, salinity and oxygen concentration, but not chlorophyll a concentration, having significant influences on viral assemblages, with salinity as the most explanatory. However, this effect was most evident at relative extremes of environmental conditions examined, and most samples lacked such evident environmental influence. This is probably explained by limited variations in surface ocean physico-chemical variables compared with previous studies in which freshwater to saline (Bettarel et al., 2011b) or oxic to anoxic gradients (Brum and Steward, 2010) resulted in very strong changes in viral assemblage morphological characteristics. Linking these global viral morphology data to viral genomic and bacterial taxonomic data will be the next logical step in refining our understanding of marine viral biogeography.

Only four morphotypes were observed in this study, indicating that other morphotypes (for example, lemon-shaped or filamentous) comprised <1% of these marine viral assemblages (with 100 viruses examined per sample). Additionally, while 100 viruses per sample sufficiently characterized viral assemblages, this resulted in insufficient data to fully investigate spatial variability of tailed viral morphological attributes (for example, tail length). We estimate that 5–100-fold more viruses per sample (depending upon morphotype) are required to investigate the possible presence of other morphotypes and more robustly evaluate effects of geography and environmental variables on morphological characteristics of tailed virus subgroups.

With the assumption that most marine viruses are phages (viruses that infect bacteria; Wommack and Colwell, 2000) and the knowledge that ca. 96% of all isolated phage are tailed (Ackermann, 2007), one would expect most marine viruses to be tailed. Instead we found that non-tailed icosahedral viruses dominate the upper water column of the global oceans, comprising 51–92% of viral assemblages. This corroborates two previous marine studies and contrasts three in freshwater systems (Table 1). Commonly, however, this high proportion of non-tailed viruses in marine environments is attributed to tail loss during sample preparation (reviewed by Proctor, 1997). The only empirical test of this assertion showed substantial viral tail loss from marine sediment samples (Williamson et al., 2012) but used harsher preparation methods (sonication and/or vortexing) than was used for qTEM in this study. By contrast, qTEM sample preservation and preparation does not cause tail loss or substantially alter other community viral morphological characteristics for water column samples. In addition, not once, in 2600 viruses documented in Tara Oceans samples, were viral tails observed separated from capsids.

It is possible that small podovirus tails may be obscured if these viruses landed directly on their tails when deposited onto the grid and the g-force used (118 000 × g) was insufficient to force them to a prone position. This would result in erroneous documentation of podoviruses as non-tailed viruses but would not change our major conclusions. Specifically, even if 50% of podoviruses were recorded as non-tailed, podovirus fractional abundances would double (to 12%) and non-tailed fractional abundances would only decrease to 73% (refer to Figure 5), leaving our concluded relative order of viral morphotypes intact. Further, for non-tailed viruses to actually be rotated, podoviruses would require this scenario to occur at much higher frequency in seawater than freshwater, as non-tailed viruses only comprise 0–30% of investigated freshwater viral assemblages (Table 1).

Marine viruses may lose their tails before sample collection through natural decay. In this scenario, one would expect similar capsid diameter distributions for tailed and non-tailed viruses if the ‘non-tailed' viruses had lost their tails; instead, tailed viruses had significantly larger capsids than non-tailed viruses. Further, the much lower portion of non-tailed viruses observed in freshwater environments (Table 1) would require vastly different viral decay processes in fresh versus saltwater, which seems unlikely.

The observation that upper ocean viruses are predominantly non-tailed raises questions regarding what organisms these viruses infect, and whether they contain double-stranded DNA (dsDNA), single-stranded DNA (ssDNA) or RNA genomes. The most abundant potential hosts for viruses in the surface ocean are bacteria (reviewed by Pomeroy et al., 2007), but there are few marine non-tailed phage isolates (Table 2). Early marine phage isolations yielded one non-tailed dsDNA phage in 1968 and one non-tailed RNA phage in 1976, and more recent efforts have added nine ssDNA phages and a phage of unknown nucleic acid type (Table 2). Notably, two of these non-tailed phages were isolated using the cyanobacterium Synechococcus sp. WH7803 (McDaniel et al., 2006; Kuznetsov et al., 2012) from which a decade of viral isolations had previously resulted in only tailed phages (Waterbury and Valois, 1993; Wilson et al., 1993; Fuller et al., 1998; Lu et al., 2001; Chen and Lu, 2002; Marston and Sallee, 2003; Sullivan et al., 2003). Collectively, this suggests that the relative dearth of non-tailed phage isolates (Ackermann, 2007) may result from ascertainment bias derived from a combination of limited host diversity and non-tailed phages being less easily propagated or recognized than their tailed counterparts.

The upper ocean, although dominated by bacteria, contains other potential microbial hosts for viruses, including archaea and eukaryotes. Marine archaea numerically dominate the mesopelagic oceans (Karner et al., 2001), with increased abundance in some surface waters (for example, the Southern Ocean; DeLong et al., 1994), yet their viruses are represented by a single isolate—a lemon-shaped virus from a hydrothermal deep-sea environment that infects Pyrococcus abyssi (Geslin et al., 2007). We observed no lemon-shaped viruses nor any of the myriad ‘exceptional' morphotypes isolated from archaeal extremophiles (reviewed by Prangishvili et al., 2006). This is likely because physico-chemical variables in the oceanic samples did not approach the ‘extreme' conditions from which these exceptional morphotypes have been isolated. However, there are non-marine archaeal viral isolates with icosahedral non-tailed morphology (Bamford et al., 2005; Atanasova et al., 2012; Jaakkola et al., 2012) and further exploration of marine archaeal virus-host systems may yield more examples.

To date, the majority of isolated marine non-tailed viruses are derived from eukaryotes, including 28 dsDNA viruses isolated from marine algae; three ssDNA viruses isolated from marine diatoms; and six RNA viruses isolated from diatoms, a fungoid protist and picophytoplankton (Table 2). Although less abundant than prokaryotes, the relatively high number of viruses released per eukaryotic cell (reviewed by Lang et al., 2009) may increase representation of their viruses in the oceans (Steward et al., 2013) such that they could comprise a significant portion of non-tailed viruses.

Capsid diameters of marine non-tailed viral isolates (Table 2), while admittedly limited, may be useful in hypothesizing potential hosts for the observed non-tailed viruses. The range of capsid diameters for isolated eukaryotic dsDNA viruses (115–220 nm), smaller eukaryotic RNA viruses (22–32 nm), larger eukaryotic RNA viruses (90–95 nm) and smaller ssDNA phages (30–32 nm) each comprised <1% of non-tailed viruses in the Tara Oceans samples, while eukaryotic ssDNA viruses (30–38 nm) and larger ssDNA phages (72–77 nm) only comprised 3 and 5%, respectively. However, the lone dsDNA and RNA non-tailed phages isolated from marine bacteria had 60 nm capsids, which most closely represented the mean capsid diameter for Tara Oceans non-tailed viruses (54±12 nm). Assuming that these trends from so few cultivated non-tailed viruses are robust, this suggests that most non-tailed marine viruses may infect the numerically dominant bacteria. However, the primary conclusion from comparing capsid diameters is that most observed non-tailed viruses have no cultivated representatives.

Cultivation-independent approaches also provide information about marine non-tailed viruses. First, marine viral metagenomes have yielded assembled genomes with similarity to non-tailed ssDNA Microviridae phages (Tucker et al., 2011; Roux et al., 2012), and to several families of eukaryotic non-tailed RNA viruses (Culley et al., 2006), providing genomic information about uncultured groups. Second, recent work suggests that RNA viruses are nearly as abundant as dsDNA viruses, comprising 15–77% of total viruses at one coastal Hawaii location (Steward et al., 2013). Extrapolating this to the global oceans where 51–92% of viruses were non-tailed, and assuming all the RNA viruses are non-tailed, suggests that RNA viruses could comprise 16–100% of the non-tailed viruses observed.

Finally, 65–93% (reviewed by Hurwitz and Sullivan, 2013) and 41–81% (Culley et al., 2006; Steward et al., 2013) of sequences in marine DNA and RNA viral metagenomes, respectively, are not represented in existing genomic databases. Given that observed non-tailed virus capsid diameters were largely inconsistent with those from cultivated marine non-tailed viruses, we posit that non-tailed viruses may comprise the majority of this vast ‘unknown' marine viral metagenomic sequence space. Several existing and emerging approaches will likely help identify and characterize non-tailed marine viruses. These include culture-based approaches (for example, targeted isolations with existing and new marine bacterial, archaeal and eukaryotic cultures), as well as new methods that either require only the host to be in culture (for example, viral tagging; Deng et al., 2012) or are completely cultivation-independent (for example, physical fractionation of viral assemblages; Bergeron et al., 2007; Steward and Rappé, 2007; Brum and Steward, 2011; Brum et al., 2013). The abundance and distribution of genetically characterized, non-tailed viruses could also be explored using phageFISH (Allers et al., 2013). Also, viruses with particular nucleic acid types can be examined by enriching for ssDNA (Kim and Bae, 2011) or specifically targeting dsDNA, ssDNA and RNA pools (Andrews-Pfannkoch et al., 2010).

In summary, morphological analysis was fundamental to the origin of modern aquatic viral research (for example, Bergh et al., 1989; Borsheim et al., 1990; Bratbak et al., 1990; Borsheim, 1993) and, with careful methodological evaluation, it continues to be a valuable tool to understand the ecology and diversity of aquatic viruses. This use of qTEM to assess marine viruses across six ocean regions shifts the paradigm to non-tailed viruses as dominant, which should guide future work towards characterizing these abundant and nearly unexplored viruses.

Acknowledgments

Notes

The authors declare no conflict of interest.

Footnotes

References