Influence of environmental conditions, bacterial activity and viability on the viral component in 10 Antarctic lakes

Authors


  • Editor: Riks Laanbroek

Correspondence: Christin Säwström, Department of Ecology and Environmental Science, Climate Impacts Research Centre (CIRC), Umeå University, SE-981 07, Abisko, Sweden. Tel.: +46 980 40162; fax: +46 980 40142; e-mail: christin.sawstrom@emg.umu.se

Abstract

The influence of biotic and environmental variables on the abundance of virus-like particles (VLP) and lysogeny was investigated by examining 10 Antarctic lakes in the Vestfold Hills, Antarctica, in the Austral Spring. Abundances of viruses and bacteria and bacterial metabolic activity were estimated using SYBR Gold (Molecular Probes), Baclight (Molecular Probes) and 6-carboxy fluorescein diacetate (6CFDA). Total bacterial abundances among the lakes ranged between 0.12 and 0.47 × 109 cells L−1. The proportion of intact bacteria (SYTO® 9-stained cells) ranged from 13.5% to 83.5% of the total while active (6CFDA-stained) bacteria ranged from 33% to 116%. Lysogeny, as determined with Mitomycin C, was only detected in one of the lakes surveyed, indicating that viral replication was occurring predominately via the lytic cycle. Principal component analysis and confirmatory correlation analysis of individual variables showed that high abundances of VLP occurred in lakes of high conductivity with high concentrations of soluble reactive phosphorus and dissolved organic carbon. These lakes supported high concentrations of chlorophyll a, intact bacteria, rates of bacterial production and virus to bacteria ratios. Thus, it was suggested that viral abundance in the Antarctic lakes was determined by the trophic status of the lake and the resultant abundance of intact bacterial hosts.

Introduction

Viruses exert a pervasive influence over aquatic ecosystems, mediating microbial abundance, production, respiration, diversity, genetic transfer, nutrient cycling and particle size distribution (Wommack & Colwell, 2000; Weinbauer, 2004; Suttle, 2005). It has been suggested that viruses can cause up to 30% of the bacterial mortality and 2–10% of the phytoplankton mortality in aquatic systems (Bratbak & Heldal, 2000). As bacterioplankton play an important role in the microbial loop, acting as recyclers of dissolved organic matter, viral infection and subsequent cell lysis have a significant impact on nutrient and carbon cycling in aquatic ecosystems (Bratbak et al., 1992; Murray & Eldridge, 1994; Gobler et al., 1997; Wilhelm & Suttle, 1999; Middelboe & Lyck, 2002; Suttle, 2005). A recent investigation by Säwström et al. (2007a) showed that over 60% of the carbon supplied to the dissolved organic carbon (DOC) pool in Antarctic lakes originated from viral lysis. Thus, the viral loop (the flow of carbon among bacteria, viruses and the DOC pool) can comprise much of the carbon flow in Antarctic lakes.

The interactions between viruses and bacteria in freshwaters have been investigated worldwide and it is evident that many factors can influence these interactions including humic matter content, temperature and nutrient levels, (Maranger & Bird, 1995; Anesio et al., 2004; Lymer & Vrede, 2006). However, little is known about the effect of the physiological state of bacteria on viral production in freshwaters. In most freshwater environments the bacterioplankton are living under suboptimal conditions, with low nutrient levels and temperatures restricting their abundance and production (Elser et al., 1995). Bacterioplankton in Antarctic lakes are no exception, particularly as phytoplankton productivity is low or nonexistent for much of the year due to light limitation (Delille, 2004). The resulting lack of bioavailable substrate for bacteria is likely to result in an inactive and/or dormant bacterial community, which is likely to have a negative effect on the production of viruses (Maranger et al., 1994; Davidson et al., 2004).

The metabolic activity of aquatic bacterioplankton determines the extent to which they contribute to nutrient cycling, the microbial loop and as potential hosts for virioplankton (Lenski, 1988; del Giorgio et al., 1996; Gasol & del Giorgio, 2000). It is now widely accepted that natural communities of aquatic bacteria contain cells that vary greatly in the metabolic state ranging from ‘active’ to ‘dormant’, ‘live’, ‘viable but nonculturable’ and ‘dead’, largely depending on their nutritional status (del Giorgio et al., 1996; Gasol & del Giorgio, 2000; Schumann et al., 2003; Smith & del Giorgio, 2003). Studies unanimously show that only a fraction of bacterioplankton cells are active (Smith & del Giorgio, 2003). The physiological state of the bacterioplankton in Antarctic lakes is largely unknown; however, studies have shown that bacterial production can occur throughout the year, often with the highest rates in winter (Laybourn-Parry et al., 2004). Although the microbial composition of some of the study lakes has been investigated previously, there is a lack of data on viral and bacterioplankton dynamics within these extreme lake environments (Laybourn-Parry et al., 1995, 1992, 2001, 2004, 2007; Bayliss et al., 1997; Laybourn-Parry, 2002; Madan et al., 2005; Säwström et al., 2007a, b). The aims of the present study were threefold: (1) examining the in situ abiotic and biotic factors that influence the viral component and replication pathway (lytic or lysogenic); (2) estimating the metabolic state of bacteria in Antarctic lakes using vital stains BacLight and 6CFDA and bacterial cell production; and (3) investigating any associations between the bacterial physiological state and the abundance of viruses.

Materials and methods

Study sites and sampling

This study was conducted in the Vestfold Hills situated in the Eastern Antarctica between latitudes 68°25′S 78°48′E and 68°41′S 78°36′E. Ten lakes were sampled, a range from freshwater lakes (Crooked, Druzhby, Bisernoye, Zvezda, Pauk and Lichen) to slightly brackish lakes (622–2330 μS cm−1) (Cowan, Depot, Nicholson and Watts) (Fig. 1). Each lake was sampled on one occasion in austral spring (September 2004). On each sampling occasion, water temperature and oxygen saturation (% DOsat) at 5 m was determined using a YSI 6600 sonde (YSI, Marion, MA). Water samples were obtained from a depth of 5 m using a 2.5-L Kemmerer bottle deployed through a hole drilled in the ice-cover with a Jiffy drill (Feldman Engineering, Sheboygan Falls, WI). Water was transferred to acid-washed bottles, returned to the laboratory at the Davis Station (Australian Antarctic Territory) in an insulated container within 6 h of collection. The container was then stored in a temperature-controlled cold room ±2 °C of ambient until subsampled for analyses. The conductivity and pH of each water sample was then determined using a WTW 197i conductivity meter (Weilheim, Germany) and a MeterLab pH meter (Copenhagen, Denmark), respectively.

Figure 1.

 Map showing the location of the 10 lakes in the Vestfold Hills, Eastern Antarctica.

Bacterial and viral concentrations

A known volume of water from each lake was stained to determine concentrations of viruses and intact, leaky and total bacteria within 8 h of sample collection and counts completed within 24 h. Counts of virus-like particles (VLP) and total bacteria were obtained using samples stained with SYBR Gold (Molecular Probes Inc., Eugene, OR). Immediately on return to the laboratory, 15 mL of water from each lake was fixed with 2% final concentration, 0.02 μm filtered, electron microscope grade glutaraldehyde. The fixed sample was then stained with SYBR Gold and filtered following the method of Chen et al. (2001).

Bacteria with ‘intact’ or ‘leaky’ membranes were stained by dual staining samples with Baclight (Molecular Probes Inc.) following the methods of Davidson et al. (2004). Baclight contains two nucleic acid-binding stains: SYTO®9, which stains all bacteria causing them to fluoresce green, and propidium iodidie (PI) that penetrates cells with leaky membranes, quenches the fluorescence of SYTO®9 and causes them to fluoresce red. Both 7 μL of 20 mM PI and 3 μL of 3.34 mM SYTO®9 were added per milliliter of unfixed lake sample and incubated in the dark for 30 min at ±2 °C of the ambient lake temperature. Abundances of intact and leaky bacteria were summed up to estimate the total bacteria in BacLight-stained samples. To stain active bacteria, 1 μL of a stock solution containing 1 g of 6CFDA (Molecular Probes Inc.) in 100 mL of AR grade acetone (BDH) was added per milliliter of the unfixed lake sample (Davidson et al., 2004). The sample was immediately mixed to avoid precipitation and incubated in the dark for 30 min at±2 °C of ambient temperature. Stained bacterial samples were then filtered to dryness onto a black 25 mm, 0.22 μm polycarbonate Nuclepore filter (Whatman, Middlesex, UK) over 0.8 μm Duropore (Millipore, Billerica, MA) backing filter.

Irrespective of the stain/s used, filters were mounted on a microscope slide, a drop of p-phenylenediamine antifade placed on the surface and a coverslip added (Noble & Fuhrman, 1998). Counts of viruses and, total, active, intact and leaky bacteria were performed at × 1000 magnification, using a Zeiss Axioscop (Oberkochen, Germany). Excitation of all stains was achieved under blue epifluorescent excitation, using filter set 487909 with a 450–490 nm exciter filter, a 510 nm chromatic beam splitter and a 520 nm barrier filter. Numbers of bacteria or VLP were counted in 10 randomly chosen microscope fields on duplicate filters from each lake. Abundances of intact and leaky bacteria in each sample were summed to estimate the total bacterial abundance and the abundance was compared with that obtained using SYBR Gold. The mean and SDs of bacterial and viral abundances in each lake were then computed.

Bacterial production (BP)

BP was estimated by the incorporation of [14C] leucine (306 mCi mmol−1) into the bacterial biomass by the microcentrifuge method as described by Kirchman (2001). Preliminary experiments on water samples from Lake Druzhby and Crooked Lake showed that [14C] leucine reaches saturating concentration at 57 nM (data not shown). Five replicate 1.7 mL samples were collected into 2 mL microcentrifuge tubes and [14C] leucine was added to a final concentration of 57 nM. Duplicate control samples were immediately inactivated with 90 μL of 100% trichloroacetic acid (TCA). All the samples were then incubated for 90 min to allow for the label to be incorporated into the bacterial biomass. After incubation, 90 μL of 100% TCA was added to all samples except for the duplicate inactivated controls. The tubes were then centrifuged at 16 000 g at 4 °C for 10 min, the supernatant removed by aspiration and 1.7 mL ice-cold 5% TCA added to each tube. Each tube was then vortex-mixed, centrifuged and the supernatant aspirated. Finally 1.7 mL of ice-cold 80% (v/v) ethanol in Milli-Q water was added, the centrifugation and aspiration steps repeated and 1 mL of scintillation cocktail (Ecoscint, National Diagnostics, Atlanta, GA) added. Samples were then counted by liquid scintillation in a Beckman LS6500 scintillation counter (Fullerton, CA). A conversion factor of 1.42 × 1017 cell mol−1 was applied to the incorporation rates of leucine into protein (Chin-Leo & Kirchman, 1988).

DOC and inorganic nutrients

Samples (∼50 mL) for DOC were filtered through GF/F filters (preashed for 12 h at 550 °C) and immediately transferred into acid-washed bottles and stored at −20 °C until analysed. Concentrations of DOC were determined using a Shimadzu TOC-5000 total carbon analyser equipped with an ASI-5000 auto sampler (Kyoto, Japan). Analyses were performed at high sensitivity using platinum catalyst and high-temperature combustion. The carbon analyser was calibrated using freshly prepared solutions of potassium hydrogen phthalate (0–5.0 mg C L−1) in Milli-Q water. Samples were thawed, 50 mL decanted into muffled sample vials, acidified to pH∼2 using HCl and sparged for 5 min with CO2-free air to remove dissolved inorganic carbon. Milli-Q blanks (unfiltered) and operational blanks (using Milli-Q filtered as above) were performed, both of which gave DOC concentrations near the limits of detection. A minimum of five replicate injections were made on the carbon analyser, resulting in a coefficient of variation of <2% for each analysis.

Concentrations of soluble reactive phosphorous (PO4) and ammonium (NH4) were assayed colorimetrically according to the methods of Mackereth et al. (1989).

Chlorophyll a (Chl a)

Chl a concentrations were determined by filtering a known volume (≤1 L) of water sample through a 13-mm-diameter GF/F filter which were then stored frozen at −80 °C at Davis Station. Filters were transferred to liquid nitrogen during transport to Australia and stored in an ultra-low freezer at –135oC until analysis. Pigments were extracted in 1.8 mL of MeOH containing 176 ng of apo-8′-β-carotenal [Fluka (Buchs, Switzerland)] in 25 μL MeOH as an internal standard. Following sonication, the extract was filtered through a 0.45 μm in-line filter and pigments were identified by HPLC using the methods of Zapata (Zapata et al., 2000). Hardware included a 626 LC pump (Waters, Milford, MA), an Alltima C18 column (250 mm × 4.6 mm, 5 μm bead size, Alltech Deerfield, IL), a Waters 996 photodiode array and F1000 fluorescence detectors (Hitachi, San Jose, CA). millenium 32 (ver. 3.05.01) and empower build 1154 software (Waters) were used in acquisiting and processing data. Chl a was identified against a pigment standard (Jeffrey & Wright, 1997) and quantified using the internal standard (Mantoura & Repeta, 1997).

Induction assay for lysogenic bacteria

Four replicate samples (15 mL) were either treated with Mitomycin C (a potent mutagen for prophage induction), (1 μg mL−1 final concentration), (Sigma, St Louis, MO), or left untreated (controls) (Paul & Jiang, 2001). The samples were incubated in the dark at in situ temperature (2–4 °C) for 24 h and then fixed with 0.02 μm filtered glutaraldehyde (final concentration 2%) and stored at 4 °C (storage <24 h). Bacterial and viral abundance were then determined using SYBR Gold staining (as described earlier). Significance of each induction event was determined by comparing of Mitomycin C treatment and control levels of viruses by an independent-samples t-test. A statistically significant increase in viral abundance in the Mitomycin C treatment relative to the control indicated the presence of lysogenic bacteria. The fraction of lysogenic bacteria (FLC) was then calculated as: % FLC=[(VtVc)/Bz]/Bc× 100, where Vt is the number of viruses enumerated in the Mitomycin C treatment at 24 h and Vc is the number of viruses enumerated in the control sample. Bc is the number of bacteria enumerated in the control sample at 24 h and Bz is the burst size. Two values were used for Bz, a value of 4 as presented by Säwström et al. (2007b) and a higher Bz of 26 calculated from 12 different freshwater environments (Säwström et al., 2007b).

Statistical analysis

Statistical analyses were performed in spss (version 11.0.0 for Windows). Following analysis of the data distribution (Kolmogorov–Smirnov test), data were ln-transformed to achieve normality. Variation in the measured parameters, both biological and physical, in the lakes was analysed using one-way anova and the Tukey post-hoc test was used to determine the significance of differences among lakes. Any associations between the measured biological and physical variables and their combined effect on viral concentrations were explored using principal component analysis (PCA) with a correlation matrix. PCA is an exploratory tool, useful for reducing the number of variables in a given data matrix by collapsing the dimensionality of the data. This is accomplished by projecting the original data onto new axes, or principal components (PCs) (Lebart et al., 1984). PC scores were derived from PC loadings and were used as variables to quantitatively assess whether there was any relationship between the extracted principal components and the viral concentrations. Furthermore, specific analyses of individual variables were carried out using Pearson's product–moment correlation.

Results

Physiochemical parameters and Chl a

Water temperatures in the lakes ranged from 0.42 to 2.85 °C with the lowest recorded in Watts Lake and the highest in Lake Druzhby (Table 1). Percentage saturation of dissolved oxygen ranged from 73% in Crooked Lake to 138% in Lichen Lake. Conductivity of Lakes Crooked, Druzhby, Pauk, Bisernoye, Lichen and Zvezda was low, ranging between 10 and 40 μS cm−1 while Lakes Depot, Cowan, Watts and Nicholson were comparatively high ranging between 622 and 2330 μS cm−1. The pH and concentrations of PO4 and DOC were low in lakes of low conductivity (≤40 μS cm−1) ranging from 7.0 to 7.4 pH units, 3.6–4.4 μg PO4 L−1 and 0.1–1.3 mg DOC L−1 but increased to 7.8–8.5 pH units, 4.6–5.6 μg PO4 L−1 and 2.8–25.0 mg DOC L−1 in lakes with higher conductivity (≥622 μS cm−1). In contrast, concentrations of ammonium were not related to lake conductivity and varied from 6.2 to 41.9 μg L−1 (Tables 1 and 2). Lowest concentrations of ammonium and PO4 were found in Lake Zvezda, while highest ammonium concentrations were found in Lake Nicholson and highest PO4 concentrations were found in Lakes Depot and Watts (Table 1). Concentrations of Chl a were ≤0.55 μg L−1 in all the investigated lakes, the exceptions being Lake Druzhby (1.22 μg L−1) and brackish Watts Lake (4.54 μg L−1) (Table 1).

Table 1.   Mean values of the physiochemical parameters in the 10 lakes studied
LakeDateTemp (C°)pHDOsat
(%)
Conductivity
(μS cm−1)
Soluble reactive
phosphorus
(PO4 μg L−1)
Ammonium
(NH4 μg L−1)
DOC
(mg L−1)
Chl a
(μg L−1)
  1. Temp, temperature; DOsat, dissolved oxygen saturation; DOC, dissolved organic carbon; Chl a, chlorophyll a.

Crooked21.09.041.337.073133.927.30.10.23
Druzhby21.09.042.857.0102193.936.90.41.22
Pauk22.09.041.997.0128103.615.30.40.30
Depot22.09.040.928.512414995.621.72.80.31
Bisernoye23.09.040.867.212223522.50.40.43
Lichen23.09.042.247.2138404.425.71.30.17
Cowan24.09.042.197.81116224.631.33.30.13
Zvezda24.09.042.157.4133313.16.20.50.37
Watts29.09.040.428.28723305.617.125.04.54
Nicholson27.09.042.687.7956784.741.94.70.55
Table 2.   Two-tailed Pearson product–moment correlation between abiotic and biotic variables in the 10 lakes studied
VariableVLPVBRChl aDOCpHTemp% DOsatCondPO4NH4BPTotal
bacteria
Intact
bacteria
Leaky
bacteria
  • ***

    P<0.01.

  • **

    P<0.05.

  • *

    P<0.1.

  • Neg, negative; VLP, virus-like particles; VBR, virus to bacteria ratio; Chl a, chlorophyll a; DOC, dissolved organic carbon; Temp, temperature; % DOsat, dissolved oxygen saturation in percentage; Cond, conductivity; BP, bacterial production.

VLP
 VBR***             
 Chl a*            
 DOC*******           
 pH****          
 Temp*Neg         
 % DOsat        
 Cond*********       
 PO4*****Neg**Neg***      
 NH4     
 BP*************    
 Total bacteria   
 Intact bacteria*****Neg**  
 Leaky bacteria**Neg**Neg 
 Active bacteria****

Bacterioplankton

Total bacterial abundance as determined with SYBR Gold ranged between 0.13 and 0.38 × cells 109 L−1 and between 0.12 and 0.47 × 109 cells L−1 with the Baclight viability kit (SYTO®9+PI-stained cells). Abundances of Baclight- and SYBR Gold-stained cells were significantly correlated (two-tailed Pearson product–moment correlation; r=0.841, P<0.01, N=10). Comparison of SYBR Gold and Baclight-stained bacterial cells by an independent t-test showed that there was no significant difference between the total bacterial abundances obtained using the two stains. Consequently Baclight-stained cells were used to estimate total bacterioplankton (Fig. 2a). Highest total bacterial abundances occurred in Lakes Bisernoye, Watts, Crooked and Druzhby while lowest abundances were found in Pauk and Depot Lake (Fig. 2a).

Figure 2.

 (a) Mean abundance of active, leaky, intact and total bacterial cells in the 10 lakes. Striped columns represent active bacterial cells, dark grey columns leaky bacterial cells, open columns represent intact bacterial cells and light grey columns total bacterial cells. Error bars represent±SD (N=20). (b) Mean bacterial cell production in the 10 lakes. Error bars represent ±SE (N=3).

Intact bacterial cells (SYTO®9-stained cells) varied between 0.04 and 0.33 × 109 cells L−1 (Fig. 2a) and were positively related to the abundance of VLP and total bacteria (Table 2). The proportion of bacteria that were intact ranged from 16.5% to 86.5% (mean of 40.2%). Abundance of intact bacteria were significantly lower than (one-way anova; F4,49=5.911, P<0.05), but positively correlated (Table 2) with, total bacterial abundance. Abundance of leaky (PI-stained cells) bacterial cells ranged from 0.05 to 0.39 × 109 cells L−1 (Fig. 2a) and was negatively related to VBR, pH, conductivity and DOC concentration (Table 2). In eight of the lakes studied, the fraction of leaky bacterial cells exceeded that of intact bacterial cells (Fig. 2a). The highest number of leaky bacteria was found in Lake Bisernoye. Active bacteria (6CFDA-stained cells) ranged from 0.06 to 0.35 × 109 cells L−1 or 33–116% of total bacteria (Fig. 2a) and were most abundant in Crooked Lake and least abundant in Pauk Lake and Depot Lake (Fig. 2a). Abundances of active bacteria were positively correlated with abundances of total and leaky bacteria (Table 2).

Rates of bacterial production ranged from 0.2 to 3.3 × 109 cells L−1 day−1 with the lowest value recorded in Crooked Lake and the highest value noted in Nicholson Lake (Fig. 2b). Bacterial production was positively correlated with viral concentration and VBR, but was not correlated with estimates of bacterial abundances using any of the stains (Table 2). DOC concentrations and conductivity were also positively correlated with bacterial production in the study lakes (Table 2).

Lysogenic bacteria were only detected in Lake Bisernoye and the calculated fraction of lysogenic bacteria was 23.9% (Bz=4) or 3.7% (Bz=26). The presence of lysogenic bacteria coincided with a high proportion of leaky bacteria (Fig. 2a).

Virioplankton

Abundance of VLP ranged between 0.7 and 12.6 × 109 L−1 and were significantly higher in Watts Lake (four to 18 times higher) than in the other lakes studied (Fig. 3) (one-way anova; F9,19=102.260, P<0.001). Virus to bacteria ratios (VBR) ranged between 3.5 and 32.6 (Fig. 3) and were positively correlated with concentrations of VLP (two-tailed Pearson product–moment correlation; r=0.863, P<0.001, N=10). VBR was positively correlated with DOC concentration and conductivity (two-tailed Pearson product–moment correlation; r=0.857, P<0.01, N=10 and r=0.663, P<0.05, N=10, respectively).

Figure 3.

 Mean abundance of virus-like particles (open columns) and VBR (black diamonds) in the 10 lakes. Error bars represent ±SD (N=20).

Relationships between all measured variables (not including viral concentrations), using PCA, showed that the first and second principal component explained most of the variance (44% and 23%, respectively) (Fig. 4) (Table 3). Inspection of the loadings of the first axis (PC1) showed that VBR, DOC, conductivity, Chl a, bacterial production, pH, PO4 and intact bacterial cells (SYTO®9-stained cells) had positive loadings (>0.5) whereas leaky bacterial cells (PI-stained cells) and temperature had negative loadings (<−0.5) (Fig. 4) (Table 3). Correlation analysis showed a significant relationship between abundances of VLP and PC1 scores (r=0.742, P<0.05, N=10), but not PC2 scores. Thus PC1 best explained variations in viral abundance, indicating that VBR, DOC, conductivity, Chl a, BP, pH, PO4, intact bacterial cells, leaky bacterial cells and temperature were the main determinants to the differences in abundance of VLP among the lakes studied. However, correlation analysis of the individual variables showed that the relationship between abundance of VLP and pH, temperature and abundance of leaky bacterial cells were not statistically significant (Table 2).

Figure 4.

 PCA correlation biplot of abiotic and biotic variables. The first two PCs account for 44% and 23% of the total variation, respectively. Variables with loadings >0.5 and <−0.5 on the PC1 axis are given in bold.

Table 3.   Factor loadings on the PC for each variable derived from PCA and per cent of variance explained by each PC
VariablePC1PC2
  1. VBR, virus to bacteria ratio; %DOsat, dissolved oxygen saturation in percentage; DOC, dissolved organic carbon; Chl a: Chlorophyll a. Loadings > 0.5 are given in bold. A negative interaction is indicated by –.

VBR0.810−0.045
Total bacteria0.1420.906
PO40.7730.125
Leaky bacteria0.7200.414
Intact bacteria0.6260.732
Active bacteria−0.4780.792
Conductivity0.913−0.169
% DOsat−0.2730.679
pH0.812−0.350
Temperature0.549−0.110
Bacterial production0.7610.124
NH40.1790.512
DOC0.953−0.134
Chl a0.5860.276
% of variance4423

Discussion

The Vestfold Hills has more than 300 lakes and ponds and the striking limnological feature of these ecosystems is the great range of salinities, extending from Crooked Lake, one of the largest surface freshwater lakes in the Antarctic, to Deep Lake, one of the most saline lakes in the world after the Dead Sea (Dartnall, 2000). The 10 lakes studied covered a range in conductivity from freshwater to brackish (10–2330 μS cm−1). These lakes can be separated based on their differing physiochemical characteristics with the brackish lakes (Depot, Cowan, Watts and Nicholson) having distinctly higher DOC concentrations and generally lower DO saturation, higher pH and higher PO4 relative to the freshwater lakes (Crooked, Druzhby, Pauk, Bisernoye, Lichen and Zvezda).

Bacterioplankton

Effectiveness of stains

Abundances of total bacteria obtained using SYBR Gold were strongly correlated and did not differ significantly from those obtained using Baclight. Thus, like Davidson et al. (2004), it was concluded that all bacterial cells were stained with Baclight at the concentrations used in this study. A range of techniques are now available to estimate the physiological state of bacterioplankton in aquatic ecosystems. Vital stains are attractive tools as they can rapidly provide an ecologically valuable measure of bacterial activity (e.g. del Giorgio et al., 1996; Sherr et al., 1999). Fluorescent probes such as 6CFDA and the Baclight dual stain used in this study have been successfully used to estimate abundances of bacterioplankton that are metabolically active, intact or have compromised (leaky) cell membranes (Naganuma & Miura, 1997;Decamp & Rajendran, 1998; Boulos et al., 1999; Gasol et al., 1999; Schumann et al., 2003; Davidson et al., 2004; Freese et al., 2006). However, different methods provide differing estimates of the proportion of viable or metabolically active bacteria and some studies find that the efficiency of vital staining can vary; some stains are toxic or inhibit bacterial metabolism; the abundance of metabolically active bacteria can vary with the stain used; and stains can correlate poorly with rRNA probes, fluorescent in situ hybridization, autoradiography, confocal microscopy and microbial culturing (Karner & Fuhrman, 1997; Smith & del Giorgio, 2003; Pirker et al., 2005). Smith & del Giorgio (2003) persuasively argued that, rather than invalidating the use of vital stains, the lower proportions of active bacteria obtained using vital stains was because they discriminated among bacteria at higher levels of metabolic activity. Thus, while not resolving all live, dead, active or intact bacteria (Schumann et al., 2003; Pirker et al., 2005; Freese et al., 2006), vital stains can rapidly provide an ecologically valuable measure of bacterial activity (e.g. Sherr et al., 1999; del Giorgio et al., 1996).

Active bacteria comprised a large proportion of the bacterial community among the lakes sampled. Such high proportions of active bacteria are characteristic of eutrophic waters rather than the oligotrophic lakes in this study (Porter et al., 1995; Yamaguchi & Nasu, 1997). In contrast to previous studies (Sherr et al., 1999; Davidson et al., 2004; Pearce et al., 2007), the abundance of active bacteria commonly exceeded that of intact cells and was not correlated with either intact bacterial abundance or rates of bacterial production. The results suggest a fundamental difference between lake and marine environments. The high abundance of 6CFDA-stained bacteria and absence of correlation with intact bacteria and bacterial production may reflect greater dependence by lake bacteria on esterase-based metabolism of polyester storage products during periods of substrate limitation or changes in the efficiency of staining due to salinity and/or bacterial community composition in the lakes (Jendrossek & Handrick 2002; Kadouri et al., 2005; Paoli et al., 2006). Underestimation of the viable bacterial population by Baclight (Pirker et al., 2005) may also have contributed to the lack of correlation between active and intact bacterial abundance among the lakes.

Although the proportion of total bacteria that were intact varied greatly among the lakes, the findings of this study are similar to the ranges reported from aquatic systems where 20–81% of bacteria were intact (Schumann et al., 2003; Davidson et al., 2004; Freese et al., 2006). No significant correlation was found between intact bacteria and rates of bacterial production. This may be due to environmental differences between lakes that expose bacteria to different physical, chemical and mechanical stressors and affect the permeability of bacteria to PI (Williams et al., 1998; Boulos et al., 1999) or poor staining of small starved bacteria by SYTO®9 (Boulos et al., 1999; Freese et al., 2006).

Abundance in lakes

Total bacterial abundances were comparable to those found by Laybourn-Parry et al. (2004) and Säwström et al. (2007a) in previous studies conducted in Lake Druzhby and Crooked Lake. Correlation analysis showed that total bacterial abundances were significantly correlated to abundances of active and intact bacteria. This suggests that total bacterial abundances were largely determined by the abundance of bacteria that were capable of metabolism and growth. Interestingly total bacterial abundances were not correlated with any other measure of biotic and abiotic variable, suggesting that factors determining their abundance were complex and insufficiently described by any single environmental variable.

Abundances of bacteria with compromised cell membranes (PI-permeable) were inversely correlated with concentrations of DOC. Like previous studies, these findings suggest that bacteria became leaky when nutrition was limiting (Yamaguchi & Nasu, 1997; Davidson et al., 2004; Pearce et al., 2007). The proportion of intact bacteria varied greatly among the lakes, comprising 16.5–86.5% of the bacterioplankton. Correlation analysis showed there was a significant relationship between abundances of VLP and intact, but not total bacterial cells. This indicates that intact (SYTO®9-stained, PI-impermeable) bacteria obtained using BacLight, which only comprised a small fraction of the total bacterial community in lakes other than Watts and Nicholson, supported most of the viral production. Furthermore, this study also found that the abundance of leaky (PI-permeable) bacteria was negatively correlated with VBR. This agrees with Maranger et al. (1994) who proposed that the viral infection of inactive bacteria may result in lysogeny and act as a sink for viruses. Thus, viral production appears to be enhanced by the availability of intact bacterial hosts but reduced by infection of leaky bacteria. Abundances of active (6CFDA-stained) bacteria were only correlated with that of total bacteria. Unlike BacLight, active bacterial abundances did not correlate with abundances of VLP or VBR, indicating that 6CFDA provided no useful index of viral host abundance. The possible reasons for the behaviour of this stain in this study are discussed above. However, our data of this study suggest that 6CFDA should be used with caution when comparing bacterial activity in lakes over a range of physical environments.

Virioplankton

Viral abundances were within previous reported values from temperate and Antarctic lakes (Kepner et al., 1998; Wommack & Colwell, 2000; Laybourn-Parry et al., 2001; Säwström et al., 2007a). PCA showed that both abiotic and biotic variables influenced viral abundances in the lakes and these included conductivity, nutrients, pH, temperature, intact bacterial cells and bacterial production. Further tests by correlation analysis showed a significant relationship between rates of bacterial production, VBR, DOC, conductivity, Chl a, PO4 and intact bacterial cells and abundances of VLP. Like Laybourn-Parry et al. (2001), viral abundance was found to have a tendency to increase with increasing trophy and conductivity in the lakes, as parameters such as PO4, DOC and Chl a were positively correlated with viral abundance. Maranger & Bird (1995) also found a positive relationship between Chl a and inorganic nutrients, especially phosphorus, and viral abundance in 22 lakes in Québec. Thus, these data reflect observations from lower latitudes where VLP abundance generally increases with trophic status (Maranger & Bird, 1995; Wommack & Colwell, 2000). Watts Lake, which also had the highest concentrations of PO4, DOC and Chl a, had the highest abundance of intact bacteria and viruses. This lake is distinctly different from the other lakes studied, as it has extensive algal mats (J. Laybourn-Parry, pers. commun.). It is likely that the resulting primary productivity and diffusion of mat-derived organic matter into the water column caused the very high concentration of DOC that was recorded in Watts Lake (Aiken et al., 1996).

VBRs were comparable to previous values reported by Laybourn-Parry et al. (2001) from lakes in the Vestfold Hills. Elevated VBR reportedly indicate high rates of viral infection of the bacterioplankton or a long persistence of viruses in the waters (Wommack & Colwell, 2000). Elevated VBRs (>10) were found in Lichen, Cowan, Watts and Nicholson Lakes, suggesting long viral persistence and that viral-infection might be responsible for a large fraction of bacterial mortality in these lakes. A positive relationship was found between VBRs and DOC concentration, conductivity and BP, which indicates that viral induced bacterial mortality is higher in the brackish lakes and that virus abundance increased with increasing bacterial production rates.

It has been proposed that lysogens have an advantage over their nonlysogenic counterparts in oligotrophic environments (Jiang and Paul 1998). As nearly all the lakes investigated could be classed as oligotrophic, a high proportion of the bacterial community was expected to be lysogenic. However, lysogeny appeared to be of little importance in the lakes of the Vestfold Hills. As found by Säwström et al. (2007a) in a seasonal study of Lakes Crooked and Druzhby, lysogeny was rare and only lysogenic bacteria were detected in one of the 10 lakes studied, Lake Bisernoye. Interestingly, Lake Bisernoye also had the highest proportion of leaky bacteria. Viral infection is known to impair the integrity of bacterial cell walls (Rienmann & Middelboe, 2002) and data of this study suggest there may be some connection between lysogeny and abundance of leaky bacteria. However, the prevalence of lytic infection in this study of the lakes may be due to the time of the year it was conducted. A recent study of the saline lakes of the Vestfold Hills showed that there was a clear seasonal pattern with higher rates of lysogeny during the winter (Laybourn-Parry et al., 2007).

In conclusion, these data indicate that the abundance of viruses in the investigated lakes was positively influenced by both abiotic and biotic factors including PO4, DOC, bacterial production and concentrations of intact bacteria and Chl a. The lytic viral replication pathway was found to be dominated at the sampling time when most bacteria were metabolically active. Correlation of intact bacteria and VLP suggests these cells were the principal viral hosts and supported viral production.

Acknowledgements

This work was funded by Australian Antarctic Science Advisory Committee grants to J.L.-P. and A.T.D. and a Marie Curie Scholarship held at the University of Nottingham (J.L.-P. and C.S.). The authors wish to thank the winter crew at Davis station, Antarctica 2004 for logistical support and field assistance.

Ancillary