Viral impacts on bacterial communities in Arctic cryoconite

The surfaces of glaciers are extreme ecosystems dominated by microbial communities. Viruses are found in abundance here, with a high frequency of bacteria displaying visible virus infection. In this study, viral and bacterial production was measured in Arctic cryoconite holes to address the control that viruses play in these highly truncated ecosystems. Mean bacterial carbon production in the sediments of cryoconite holes was found to be 57.8 ± 12.9 ng C g−1 dry wt. h−1, which predicted a mean of 1.89–5.41 × 106 cells g−1 dry wt. h−1 based on a range of conversion factors. Relative to this, virus production was found to be high, up to 8.98 × 107 virus like particles g−1 dry wt. h−1 were produced, which is comparable to virus production in sediments around the globe. The virus burst size was assessed by transmission electron microscopy and found to be amongst the lowest recorded in the literature (mean 2.4). Hence, to account for the measured virus production, the viral induced bacterial mortality was calculated to be more than capable of accounting for the mortality of all bacterial production. The data presented here, therefore, suggests that viral induced mortality is a dominant process for the release and recycling of carbon and nutrients in supraglacial ecosystems.


BDC
Bacterial direct counts BCP Bacterial carbon production B z Burst size CC Carbon content (cell −1 ) FVIC Frequency of visibly infected cells GrIS Greenland ice sheet NCP New cell production VBR Virus to bacteria ratio Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
Sedimentary ecosystems contain highly abundant virus communities which require high rates of new virus production to be maintained (Danovaro et al 2008a). In freshwater and marine sediments around the globe, virus production rates of 0.2-19.8 × 10 7 virus like particles (VLP) g −1 h −1 have been reported (Danovaro et al 2008a(Danovaro et al , 2008b. Such large scale virus production comes at a significant cost to heterotrophic bacterial hosts, as viral release from infected cells is generally via lysis. Lysis also releases cellular carbon and nutrients back into the dissolved and bioavailable particulate phase, effectively diverting the flow of carbon away from higher trophic levels, whist maintaining its availability for the remaining microbial community (Wilhelm and Suttle 1999). Viruses play a significant role in the carbon and nutrient transformations in sedimentary ecosystems. Between 10% and 100% of all bacterial carbon production is channelled through viral processes (Danovaro et al 2008a and references therein). The surfaces of glaciers are extreme environments, characterized by persistent low temperatures, yet they harbour abundant communities of microorganisms which cycle carbon and nutrients (Stibal et al 2012a). The most abundant and active microbial communities are found in the small, water filled depressions known as cryoconite holes (Hodson et al 2008). Cryoconite is dark material on glacial ice that may arrive from the surrounding moraines, tundra, bed of the glacier (Wharton et al 1985, or may even be produced in situ by carbon fixation (Anesio et al 2009). The dark colouration of the material causes it to melt into the ice maintaining a pool of water above. Cryoconite holes are typically tens of cm wide and deep, with water temperatures relatively constant at ∼0.1 • C (Säwström et al 2002). Cyanobacteria account for the majority of carbon fixation (Stibal et al 2006) and primary production is approximately balanced by respiration rates, depending on sediment depth (Telling et al 2012). However, bacterial production only accounts for <10% of the respiration values , Anesio et al 2010. In a previous study, bacterial production was not correlated to bacterial biomass, suggesting a strong top-down control on heterotrophic bacterial abundances in cryoconite holes (Anesio et al 2010). Grazers, such as ciliates and tardigrades, are present (Sømme 1996, Porazinska et al 2004, but their abundances are relatively low (Säwström et al 2002), suggesting an alternative control on bacterial abundances.
Viruses are likely to be important entities in cryoconite holes, responsible for a significant portion of heterotrophic bacterial mortality (Säwström et al 2007. The viral induced destruction of host cells and the release of cellular carbon and nutrients may be important for the continued production of the non-infected microbial community. Whether microbial communities and organic carbon build up on glaciers or are washed out to hydrologically connected ecosystems remains an important question. DOC in supraglacial runoff is believed to be derived, in part, from in situ microbial processes (Hood et al 2009). As a significant proportion of organic carbon may flow through viral mediated processes, there is a pressing need to investigate the impact of viruses on microbial communities and their potential to influence carbon cycling in these environments. In this study, the viral communities in the sediment of cryoconite holes are addressed. Here, virus abundances have been shown to be over an order of magnitude higher than the overlying waters  and the vast majority of microbial activity on a glacier takes place here. For the first time, rates of virus production are calculated and related to bacterial activity, allowing for the impact of viruses on the heterotrophic bacterial community to be estimated.

Study area and sample collection
This study addressed viral dynamics in cryoconite holes over two consecutive field seasons in Svalbard and Greenland. The Svalbard study centred on two valley glaciers, Midtre Lovénbreen (ML) and Austre Brøggerbreen (AB), near Ny-Ålesund (78 • 55 N 11 • 55 W, figure 1). Virus and bacterial production was measured in incubation experiments on the 25 July 2009 on sediment sampled from cryoconite holes at point ML, and on the 4 August 2009 from site AB (figure 1). Between June and August 2010, sampling and activity measurements were conducted on the western margin of the Greenland ice sheet (GrIS) around Kangerlussuaq. The study sites for activity measurements were approximately 1 km from the ice margin on the Russell Glacier (RG) at approximately 67 • 9 39.7 N, 50 • 0 52.7 W (figure 1). Viral production was measured in two independent cryoconite holes at RG1 on 15 June 2010 and three independent cryoconite holes on 27 July 2010 at RG2 (figure 1 and table 1). A transect was also conducted via helicopter on 2 August 2010 to measure virus and bacterial abundances throughout the entire ablation zone of the GrIS (sample points 5-42 km); (figure 1). These points represented elevations of ∼550 m above sea level at point RG1 to 1186 m above sea level at 42 km.

Virus and bacterial counts
For enumeration of viruses and bacteria, 1 ml of cryoconite was sampled with a large pipette and transferred into a sterile 15 ml centrifuge tube. The sample was mixed with 9 ml of 30 kDa glacial ultrafiltrate which had a virus concentration of at least four orders of magnitude lower than the sediment and is hence referred to as virus-free water. This was shaken vigorously for 2 min. Fifteen seconds after cessation of shaking, 100 µl of supernatant was sampled and fixed with glutaraldehyde (2% final conc.), which allowed us to detect viruses at concentrations of 10 6 VLP g −1 or greater in this instance. Samples were filtered onto 0.02 µm Anodisc filters (Whatman) within 24 h, stained with SYBR gold (Invitrogen) and counted using epifluorescence microscopy . Bacterial biovolume was also calculated by measuring approximately 2000 individual bacteria from six cryoconite holes (three from RG1, two from ML, one from AB) using an F-view II CCD (Olympus) and digitally measuring the length (L) and width (W) using Cell ∧ f software (Olympus) of bacteria in three fields of view per cryoconite. Cell biovolume (V) was calculated by using the formula V = (π/4)W 2 L − W 3 . The mean biovolume from the six cryoconites was used in further calculations. Biovolume (µm 3 ) was subsequently converted to carbon content (CC; fg cell −1 ) by using a range of suitable conversion factors in the literature.

Determination of burst size by TEM
1 ml of cryoconite from six cryoconite holes (three from RG1, two from ML, one from AB) was diluted with 9 ml of extraction buffer (2% glutaraldehyde, 10 mM sodium pyrophosphate and Milli-Q TM water to 9 ml). The mixture was subjected to three cycles of vortexing and sonication for one minute, before it was passed through a 3 µm mixed cellulose ester filter (SSWP Millipore). To further separate the bacteria from the sediment matrix, a 200 µl aliquot of the filtered sediment extract was centrifuged onto a 70% sucrose cushion at 10 000 × g for 10 min. The top layer of this mixture, plus 100 µl from the top of the sucrose cushion, were removed and made up to 2 ml with 2% glutaraldehyde. This mixture was then pelleted at 10 000 × g for 5 min. The supernatant was removed and discarded before the pellet was resuspended in 1 ml 2% glutaraldehyde. The dilution step and further pelleting served to reduce the background level of viruses which remained in the supernatant. 5 µl of each sample was placed onto a carbon coated 400-mesh Cu grid resting on filter paper. Grids were left overnight to dry, before being negatively stained for one minute with 2% uranyl acetate and then rinsed with deionized distilled water. Two grids per cryoconite were examined for visibly infected bacterial cells at 80 kV and 20 000-40 000× magnification. A cell was considered infected when two or more viruses were visualized inside a cell based on a dark staining and a clearly recognizable round or hexagonal capsid structure. Normally, counting cells with such low numbers of virus particles inside increases the risk of counting cells that are simply resting on top of viruses as infected bacteria. This is because ultrafugation is the usual method to pellet both virus and host on to TEM grids (Bettarel et al 2006). In this study, the low speed centrifugation used to pellet bacteria, but not viruses, greatly reduced the free viruses. This allowed use of a lower detection limit of two phages per cell in a similar fashion to Säwström et al (2007). The frequency of infected cells (FVIC) and burst size (B z ) were calculated from the mean and standard deviation of the six cryoconite holes.

Virus production
Virus production (VP) was determined in cryoconite sediment samples using the dilution method (Mei and Danovaro 2004). A 1:10 sediment: virus-free water dilution was achieved for each of three sub-replicates per cryoconite hole in three holes from Svalbard and five from Greenland. Virus-free water (13.5 ml) was added to 1.5 ml cryoconite sediment to significantly reduce encounter rates and new virus infections. Each incubation was conducted for 24 h in the dark at in situ temperatures. To determine the virus increase, subsamples of 100 µl were removed at approximately 4 h intervals and immediately fixed with 0.02 µm-filtered glutaraldehyde (2% final conc.) before slide preparation and viral counts by epifluorescence microscopy as above. VLP were counted at each time point. VP was calculated by plotting a first order regression of VLP against time during the first increase in VLP numbers in each of the three sub-replicates. To confirm viruses were not simply detaching from the sediment and influencing counts, a killed control incubation was performed by adding glutaraldehyde (2% final conc.) to an extra replicate from each sample location and treated as above. All counts were normalized to dry weight by desiccation at 105 • C for 24 h.
To determine the significance of lysogeny in cryoconite holes, the induction agent mitomycin C was added to triplicate  Calculation assumes a burst size of 2.4 from the TEM measurements and that all infections are lytic. sediment dilutions as above to a final concentration of 1 µg ml −1 . Treatments were incubated for 24 h before a 100 µl subsample was removed, fixed and made into a slide, as described above. Triplicate incubations in which mitomycin C was not added, served as controls. The percentage of lysogenic bacteria was calculated according to Weinbauer and Suttle (1996).

Bacterial production
Bacterial carbon production (BCP) was measured alongside VP in each cryoconite hole (n = 8) by 3 H-leucine incorporation, modified from the method of Smith and Azam (1992) (see supplementary information 1, available at stacks.iop.org/ERL/8/045021/mmedia). Briefly nine 1.5 ml subsamples from each cryoconite hole were removed and transferred to sterile 1.7 ml microcentrifuge tubes. Incorporation was started by addition of 3 H-leucine (100 nM final conc.). Three of the tubes were incubated as killed controls by the immediate addition of 100 µl glutaraldehyde. The tubes (six live, three killed) were incubated at in situ temperature for 3 h and incorporation was stopped by addition of 100 µl glutaraldehyde (2% final conc.). BCP was converted to new bacterial cell production (NCP) by dividing BCP by the conversion factors calculated from the biovolume (10.68-30.53 fg C cell −1 ) to produce a range of values.

Viral dynamics calculations
Viral turnover (VT) was calculated by dividing VP rates by initial viral abundances. The virus mediated mortality (VMM) of bacterial cells (cells lysed g −1 h −1 ) was calculated by dividing VP values by the viral burst size (B z ). VMM was further divided by the NCP rate (cells g −1 h −1 ), calculated from 3 H-leucine incorporation, to give a range of value for the virus induced bacterial mortality (VIBM) in each incubation expressed as a percentage of bacterial production.

Statistics
All data were tested for normality prior to statistical analysis using a Kolmogorov-Smirnov test conducted in SPSS (16.0) or SigmaPlot (12.0).

Virus and bacterial abundances
Virus abundances in independent cryoconite holes in both Greenland and Svalbard ranged from 5.62 × 10 8 to 24.5 × 10 8 VLP g −1 dry weight sediment (ML and GrIS 11 km respectively), with a mean across all sites of 14.8 ± 7.3 × 10 8 VLP g −1 (table 1). Bacterial abundances ranged from 2.68 × 10 8 to 19.6 × 10 8 cells g −1 (RG2 and GrIS 5.4 km respectively), with mean abundances of 8.62 ± 6.8 × 10 8 cells g −1 . We can report that the coefficient of variation of VLP and BDC within cryoconite holes was generally less than 0.3, which was similar to the variation between cryoconite holes from the same site. VLP and bacterial abundances were positively and significantly related across all sample sites (regression equation y = 0.988x + 6.121 × 10 8 , n = 12, R 2 = 0.76, P < 0.001) with a virus to bacteria ratio (VBR) between 1.2 and 4.4 (mean 2.1 ± 1). Over the 42 km transect of the ablation zone of the GrIS (figure 2), VLP abundances and BDC varied significantly (ANOVA, P < 0.001). Moving from the ice sheet margins to the interior, VLP abundances increased by a factor of three (ANOVA, P < 0.001 followed by post hoc Scheffé, P < 0.001) to a maximum at 11 km of 24.5 × 10 8 VLP g −1 . BDC also showed a seven fold increase between site RG (1 km) and their maximum at 5.4 km of 19.6 × 10 8 cells g −1 (ANOVA, P < 0.001, followed by post hoc Scheffé, P < 0.001). A planned 65 km site was also investigated but found to be above the snowline and insufficient cryoconite debris was obtained for counting.
For the purposes of comparing abundance variability, samples from Greenland were divided into outlet glacial locations and the ice sheet itself. Stibal et al (2012b) previously demonstrated that the marginal sites of the GrIS (<3.6 km from the margin) are physically and chemically different from the ice sheet interior because of their slope and proximity to deglaciated areas. Hence all cryoconite holes (n = 5) from sites RG1 and 2 (Greenland) (∼1 km from the ice sheet margin) were considered part of the Russell Glacier (figure 1). Abundance data from 5.4 to 42 km on the transect across the GrIS (n = 12) were considered together as representative of the main ablation zone of the ice sheet and are subsequently referred to as GrIS sites. Cryoconite holes from ML and AB represented Svalbard cryoconite holes (n = 3). Using the above criteria, there were significant differences in VLP abundances between sites (ANOVA P = 0.001). GrIS VLP abundances (sites 5.4-42 km) were approximately 2-4 times greater than the glacial sites in Svalbard and Greenland (post hoc Scheffé, P < 0.01). However, no significant differences were detected in VLP abundance between Svalbard and Greenland glacial sites (post hoc Scheffé, P > 0.05). Mean bacterial abundances were approximately 4 times greater on the main GrIS (Sites 5.4-42 km) than the Svalbard or Greenland glacial sites (ANOVA P < 0.001 followed by post hoc Scheffé, P < 0.05), but again no significant differences were found between Svalbard and Greenland glacial bacterial counts (post hoc Scheffé, P < 0.05).

Burst size
Visible virus infection in bacterial cells was clearly resolved using TEM (figure 3) by visualizing over 140 bacteria. Bacterial cells exhibited very low numbers of visible virus particles with a range of between 2 and 4 viruses per cell (mean 2.4 ± 0.5). No cells appeared replete with viruses. All viruses infecting bacteria were round or hexagonal in shape and were approximately 30-90 nm across. The frequency of visibly infected cells (FVIC) was calculated as 21 ± 6%.

Virus and bacterial production
Rates of virus production (VP) in independent cryoconite holes ranged from 4.41 × 10 7 VLP g −1 h −1 (RG2) to 8.98 × 10 7 VLP g −1 h −1 (ML) with a mean of 7.06 ± 1.8 × 10 7 VLP g −1 h −1 across all sites (table 1). No increase in VLP abundances was detected in the killed controls and lysogeny was also not detected in any of the mitomycin C incubations during the VP experiments. Viral turnover (VT) was calculated to range from 0.04 h −1 (RG2) to 0.16 h −1 (ML). Mean bacterial biovolume from cryoconite hole sediment was calculated to be 0.102 µm 3 , thus the mean bacterial carbon content (CC) for cryoconite heterotrophic bacteria was calculated to range from 10.68 to 30.53 fg C cell −1 . Bacterial carbon production (BCP) in cryoconite holes ranged from 35.4 to 73.3 ng C g −1 h −1 (ML and RG2 respectively), (mean 57.8 ± 12.9 ng C g −1 h −1 ). Using the CC conversion values BCP was equivalent to a mean new cell production (NCP) rate of 1.89-5.41 × 10 6 cells g −1 h −1 . Mean bacterial turnover, expressed as the ratio of NCP to bacterial abundance, ranged from 0.137 to 0.393 d −1 . VP was significantly related to bacterial abundance and could be predicted by the regression equation y = 0.298x − 3.325 × 10 7 (n = 8, R = 0.943, P < 0.005). There was no significant correlation between bacterial production and virus production (n = 8, R = 0.56, P > 0.05), nor any correlation between bacterial production and bacterial abundance (n = 8, R = 0.46, P > 0.05). By using the burst size derived from TEM measurements, viruses were calculated to account for the mean abatement of 598%-1710% of the heterotrophic bacterial production across all sites (table 1).

Discussion
This study provides evidence that viruses are highly abundant and active components of cryoconite hole ecosystems, with the potential to significantly impact upon bacterial mortality and the cycling of organic carbon. Virus dynamics in cryoconite holes appear to proceed at rates comparable to other sedimentary ecosystems, despite the extreme conditions they inhabit. Virus abundances and production are similar to median values reported from both freshwater and marine global sediments (table 2; Danovaro et al 2008aDanovaro et al , 2008b. The high rates of viral production measured in cryoconite holes results in a rapid mean viral turnover of 0.091 h −1 , which is also parallelled in marine sediments (mean 0.071 h −1 ; Danovaro et al 2008b). These high rates of virus production may be key to maintaining viral numbers in supraglacial ecosystems in response to the continual destruction of viruses in by such factors such as UV-B radiation (Suttle and Feng 1992), extracellular enzymatic degradation and adsorption to sediment particles (Noble and Fuhrman 1997). Assuming lytic interactions, virus production must come at the detriment of a large number of host organisms. Whilst both cyanobacteria and algae are also subjected to viral infection, heterotrophic bacteria are usually the main hosts for viruses in the environment because of their numerical dominance. Cyanobacteria made up 0.5%-18.3% (mean 10.5%) of the total cell counts over a similar transect of the GrIS (Stibal et al 2012b), and <0.8 ± 1.4% of the microbial community by abundance on four Svalbard glaciers (Stibal et al 2006). Hence, the most abundantly available hosts for viruses in cryoconite holes are heterotrophic bacteria. This is reinforced with TEM data which revealed that 21 ± 6% of the heterotrophic bacterial community displayed visible virus infection, meaning on average ∼18% of the total virions in cryoconite sediments were intracellular at any one time. As visible infection only occurs in the latter stage of infection, the real infection rates are likely to be much greater. Concurrently, viral and bacterial abundances over all activity sites and across our Greenland transect were positively correlated, and virus production was strongly correlated with bacterial abundance, suggesting heterotrophic bacteria are the main virion production factories in cryoconite hole ecosystems. Our transect data further revealed that the greatest virus and bacterial abundances are found further into the ablation zone of the Greenland ice sheet at 5-11 km. This implies that our viral and bacterial activity measurements likely do not represent values across the ablation zone, which may be significantly higher. Especially when considering virus production was correlated with bacterial abundance. Assuming all viruses produced are from lytic infection of heterotrophic bacteria, viruses were calculated to be responsible for the mean abatement of 598-1710% of all bacterial production in cryoconite holes (using a range of conversion factors for BCP from 10.68 to 30.53 fg C cell −1 ). High values, up to 1500%, have previously been reported from an aquatic hot spring (Breitbart et al 2004) which were attributed to Lotka-Volterra, predator-prey relationships between virus and host which may distort single time point measurements such as ours. However, we did not observe such a relationship between viral and bacterial abundances in this study. Bacterial numbers remained relatively unchanged or even slightly increased during the virus production incubations (supplementary figure 1, available at stacks.iop. org/ERL/8/045021/mmedia) which contradicts VIBM values over 100%. There are several factors which could account for these high VIBM values: (1) bacterial carbon production could have been underestimated, either by using incorrect isotopic dilution ratios of leucine in cryoconite bacteria (which is unknown), or the fact that high rates of lysis occurred in the incubations which would release bacterial bound 3 H-leucine to dissolved forms making it undetectable by the assay; (2) burst size could have been underestimated as the value calculated for cryoconite holes are the lowest ever reported for sedimentary ecosystems, however they are consistent with other studies in polar aquatic environments (Säwström et al 2007, Paterson andLaybourn-Parry 2012). A higher burst size would reduced VIBM estimates, however we also never observed an infected cell with greater than four viruses inside; (3) predator-prey dynamics could have occurred meaning lysis did occur faster than production. As bacterial counts were performed with SYBR gold we could not distinguish between live and dead cells, hence it is possible that recently lysed cells could still be counted and hide a decline in bacterial numbers; (4) other viral life strategies, such as chronic infections, could operate in addition to lysis and account for some of the viruses released. Chronically released virus could be produced without the lysis of the host. However we note the observed high infection frequencies were caused by non-filamentous virus forms. Because of these potential scenarios and the uncertainties in the calculations of virus induced bacterial mortality, we take this measurement to be an estimate of the maximum potential for viral lysis. Thus the production of viruses is a significant process in cryoconite holes, and one which can theoretically account for the lysis of bacterial cells at a rate equivalent to their production or greater. This large scale bacterial lysis and subsequent release of cellular contents has the potential to convert the majority of BCP back into dissolved or colloidal material, reducing its availability for uptake by grazers. This carbon may be less labile after viral lysis but would be at least partially available to the non-infected prokaryotic community. As heterotrophic bacteria are likely to consume a significant proportion of organic carbon from primary production, the high VIBM calculated here implies that the majority of carbon in cryoconite ecosystems flows through viral mediated processes. Hence it is probable that viruses strongly influence the transformation of carbon and nutrients in supraglacial ecosystems. A strong viral shunt is a typical occurrence in the cold, nutrient limited deep sea benthos (table 2 and references therein). Using similar methodology mean VIBM rates have been reported to account for the destruction of approximately 89% of bacterial cell production, decoupling bacterial production from secondary consumers (Danovaro et al 2008b). Cryoconite ecosystems are microbially abundant, truncated ecosystems, and whilst grazing rates are unknown, a strong viral shunt implies they are low relative to viral top-down control. A stronger viral shunt may also result in higher rates of bacterial respiration and lower bacterial growth efficiencies, as whilst DOM is recycled to the heterotrophic community, the overall efficiency of carbon uptake is reduced (Motegi et al 2009). Thus the likely dominance of the viral shunt in cryoconite holes may help to explain the high rates of respiration relative to bacterial production previously found in these habitats (Anesio et al 2010).
Given the potentially limiting nature of DOC on glacial surfaces, the continued viral mediated release of organic carbon may promote growth in the non-infected microbial community. However, supraglacial ecosystems may be subjected to significant flow of glacial meltwater, which could export DOC to downstream ecosystems. Over the Greenland ice sheet, debris bound organic carbon is significantly less at the marginal sites (sites < 5.4 km from the ice margin) and is negatively correlated with slope (Stibal et al 2012b). Recent evidence also suggests dissolved organic matter (DOM) in glacial run off is composed of highly labile carbon and is a significant contributor to bacterial productivity in downstream marine ecosystems (Hood et al 2009). Further to this, Hood et al (2009) found that this glacial DOM is microbial in origin. Therefore, the virally induced release of organic carbon from cryoconite bound bacteria could enhance its export from supraglacial environments and stimulate microbial communities in hydrologically connected ecosystems.