Correspondence: George A. Kowalchuk, Netherlands Institute of Ecology, Centre for Terrestrial Ecology (NIOO-KNAW), P.O. Box 40, 6666 ZG, Heteren, The Netherlands. Tel.: +31 026 479 1314; fax: +31 026 472 3227; e-mail: email@example.com
The unusually harsh environmental conditions of terrestrial Antarctic habitats result in ecosystems with simplified trophic structures, where microbial processes are especially dominant as drivers of soil-borne nutrient cycling. We examined soil-borne Antarctic communities (bacteria, fungi and nematodes) at five locations along a southern latitudinal gradient from the Falkland Islands (51°S) to the base of the Antarctic Peninsula (72°S), and compared principally vegetated vs. fell-field locations at three of these sites. Results of molecular (denaturing gradient gel electrophoresis, real-time PCR), biochemical (ergosterol, phospholipid fatty acids) and traditional microbiological (temperature- and medium-related CFU) analyses were related to key soil and environmental properties. Microbial abundance generally showed a significant positive relationship with vegetation and vegetation-associated soil factors (e.g. water content, organic C, total N). Microbial community structure was mainly related to latitude or location and latitude-dependent factors (e.g. mean temperature, NO3, pH). Furthermore, strong interactions between vegetation cover and location were observed, with the effects of vegetation cover being most pronounced in more extreme sites. These results provide insight into the main drivers of microbial community size and structure across a range of terrestrial Antarctic and sub-Antarctic habitats, potentially serving as a useful baseline to study the impact of predicted global warming on these unique and pristine ecosystems.
Many factors are unfavorable to the majority of terrestrial life-forms in Antarctic regions, such as low thermal capacity of the substratum, frequent freeze–thaw and wet–dry cycles, low and transient precipitation, low humidity, rapid drainage, and limited organic nutrients (Wynn-Williams, 1990). These generally adverse conditions support relatively simple ecosystems with a noted reduction in the complexity of food webs, with highly simplified food web structures in the most extreme Antarctic habitats (Wall & Virginia, 1999). Annelids, mollusks, winged insects and mammals are effectively absent from these systems, and only two vascular plant species have been found to inhabit Antarctic terrestrial environments (Davis, 1981). Consequently, most of these soil environments are devoid of the root systems of vascular plants and larger animals which cause bioturbation. Although some complex trophic interactions have been identified in terrestrial Antarctic environments (Newsham et al., 2004), their less complex food-web structure provides a relatively simplified system in which to disentangle the drivers and consequences of soil microbial activities.
The Antarctic Peninsula is the most rapidly warming region in the world (Houghton et al., 2001). Predicted global warming will lead to longer growing seasons across this region, and extended plant distributions are anticipated (Frenot et al., 2005; Convey & Smith, 2006). Climate warming will not only affect Antarctic ecosystems directly, but associated changes in precipitation patterns and increased water availability due to melting are thought to be of perhaps even greater significance. Consequently, it has been hypothesized that direct temperature effects on soil-borne microorganisms will be less important than indirect effects, such as changes in vegetation density and other associated soil biophysical properties (Vishniac, 1993). Indeed, although decreases in bacterial abundances have been observed with increased latitude in terrestrial Antarctic systems, this is thought to be related to a concomitant decrease in vegetation density (thus carbon and inorganic nutrients) rather than to climate per se or latitude (Vishniac, 1993).
Although little is known about the structure and function of terrestrial microbial communities in southern polar regions, a number of important preliminary investigations have begun to shed some light on the ecology of these systems. For instance, it has been observed that culturable fungal communities are more diverse and more abundant in sub-Antarctic islands, where the climate is more humid and temperate, as compared to Antarctica proper (Smith, 1994; Azmi & Seppelt, 1998). Organic matter, soil water content, pH and total nitrogen have also been shown to be correlated with fungal abundance on the sub-Antarctic Signy Island (Bailey & Wynn-Williams, 1982). Also, several studies have reported that Antarctic fungal communities are dominated by cold-tolerant, as opposed to cold-adapted, fungi, suggesting that the superior tolerance of some fungal populations to the harsh habitats results in distinct Antarctic fungal assemblages (Kerry, 1990; Melick et al., 1994; Zucconi et al., 1996; Robinson, 2001). Interestingly, a recent molecular survey targeting all eukaryotes reported no decrease in diversity along a southern latitudinal gradient, but did discriminate between continental vs. maritime sites, with the former harboring lower eukaryotic diversity (Lawley et al., 2004).
Few detailed studies exist to date that provide in depth descriptions of bacterial communities in Antarctic soils. Nevertheless, it has been observed that bacterial counts, activity and community structure are related to soil type, nitrogen content, water abundance and type of plant cover (Christie, 1987; Tearle, 1987; Bölter, 1995; Bölter et al., 1997; Harris & Tibbles, 1997). Similarly, microbial activity was found to be controlled by not only short-term patterns of temperature and moisture, but also by the availability of organic matter and the supply of soluble carbohydrates and amino acids, but not N and P (Christie, 1987; Bölter, 1992). In contrast, another study found no relationship between moisture, soil particle size, salinity, pH and number of bacteria (Line, 1988). In one of the few molecular surveys of bacterial diversity in Antarctic terrestrial environments, it was recently reported that the extremely harsh environments of three different Antarctic cold desert mineral soils contained bacterial communities of relatively low diversity, with a high proportion of novel, potentially psychrotrophic taxa (Smith et al., 2006).
A number of studies have examined microfaunal diversity and distribution in Antarctic soils, revealing a patchy distribution of nematodes, collembola, acari, rotifers and tardigrades, hypothesized to follow patterns of vegetation, moisture retention or bird activity (Tilbrook, 1967; Spaull, 1973; Bölter et al., 1997; Sohlenius & Boström, 2005). Such studies of terrestrial invertebrates suggest distinct biogeographical regions within the Antarctic, although debate exists as to whether these adhere to the sub-Antarctic, maritime Antarctic and continental Antarctic regions delineated for vegetation patterns (Smith, 1984) or follow a discontinuity between the Antarctic Peninsula and continental Antarctica, along the newly coined ‘Gressitt Line’ (Chown & Convey, 2006). Despite the interest in soil microfauna, the relative importance of this group with respect to heterotrophic respiration appears relatively low. One study of relative respiration rates of soil organisms on Signy Island revealed that 81–89% of heterotrophic respiration could be attributed to bacteria and fungi, with a remaining 10–19% due to protozoan activity. Rotifers, tardigrades, nematodes, acari and collembola only accounted for 0.42–0.48% of total respiration (Davis, 1981). Nevertheless, soil microfauna may provide important clues into trophic interactions in Antarctic systems (Newsham et al., 2004) and may represent key indicators of change within such habitats.
The low number of recent studies about Antarctic soil ecosystems has hampered any attempts to predict or observe the possible effects of the rapid and ongoing warming of this region. The main goal of this study is to provide an in-depth assessment of soil-borne microbial communities across a range of Antarctic and sub-Antarctic terrestrial habitats. We further attempt to examine data on microbial abundance and community structure in relation to key various environmental factors to gain insight into the factors driving microbial communities in these unique environments.
Materials and methods
During the austral summer of 2003–2004, 2 × 2 m plots were established at the following sites (see Fig. 1 for a map): Falklands Islands (cool temperate zone; 51°S 59°W), Signy Island (South Orkney Islands, maritime Antarctic; 60°43′S 45°38′W) and Anchorage Island (near Rothera research station, Antarctic Peninsula; 67°34′S 68°08′W). At each location, two types of vegetation selected for sampling: (1) ‘vegetated’, where dense vegetation cover was present with retention of underlying soil; and (2) ‘fell-field’, represented as rocky or gravel terrain with scarce vegetation or cryptogam coverage. For the Falkland Islands, vegetated sites exhibited a dwarf shrub vegetation (Empetrum rubrum Vahl ex Willd.), and the fell-field site was rocky with sporadic grasses (Festuca magellanica Lam. and Poa annua L.). For the locations in the (maritime) Antarctic, vegetated sites were dominated by mosses (Chorisodontium aciphyllum Hook. F. & Wils on Signy Island and Sanionia uncinata Hedw. on Anchorage Island), and fell-field sites contained lichen cover (principally Usnea antarctica Du Rietz). Twelve plots were delineated per location with half of the plots positioned over each vegetation type. The Falkland Islands fell-field vegetation was not large enough to allow for such a design and nine of the 12 plots were therefore placed in the dwarf shrub vegetation. Two additional sites were chosen for sampling, but without delineation of permanent plots. Six frost polygons at two different sites were sampled near the Fossil Bluff (71°19′S 68°18′W) fuel depot, and five frost polygons were sampled from Coal Nunatak (72°03′S 68°31′W).
Environmental data collection
Automated weather stations and precipitation gauges (PLUVIO, OTT Hydrometrie, Hoofddrop, The Netherlands) were installed at the first three study locations. Temperature probes (copper/constantan thermocouple wires) were inserted in the plots 5 cm above the ground, at the soil surface and 5 cm below the soil surface. Soil moisture content was measured with a Water Content Reflectometer (CS616, Campbell Scientific, Shepshed, UK) to a depth of 30 cm. Each of these sensors recorded every hour for the duration of the study, with data being stored using a data logger (CR10X with a storage module of 16 Mb from Campbell Scientific). Soil microclimatic data retrieved from the automated weather stations were averaged over the whole year.
For molecular and cultivation analyses, five 1-cm diameter (from 2 to 3 cm to up to 15 cm deep) cores were sampled from each plot or polygon. They were frozen to −20°C as soon as possible (within 24 h) and maintained at that temperature until use. For soil analyses, one 10-cm diameter core was taken directly adjacent to the plots in order to minimize destructive sampling in the long term plots. Sampling took place on October 26–28, 2004 for the Falkland Islands, on January 2–3, 2005 for the Signy Island, on January 18–19, 2005 for Anchorage Island and on February 22–23, 2005 for Coal Nunatak and Fossil Bluff.
Soil biochemical and physical analyses
Soil analyses were carried out using standard protocols (Carter, 1993). Because this study represents the first characterization of these habitats, we assessed a wide range of soil parameters to allow full correlative comparison with measures of soil-borne community size and structure. phospholipid fatty acids (PLFA) analyses were carried out as outlined in Boschker (2004), using 1 g (Falkland, Signy, and Anchorage Islands) or 8 g (Fossil Bluff and Coal Nunatak) of soil (wet weight). i14:0, i15:0, a15:0, i16:0, C16:1ω7t, i17:1ω7, 10Me16:0, br17:0, a17:1ω7, i17:0, a17:0, C17:1ω8c, C17:1ω6/7, cy17:0, 10Me17:0, C18:1ω7c, 10Me18:0, cy19:0 PLFAs were used for determining bacterial biomass while C18:2ω6c was used to estimate fungal biomass. The whole peaks data set (except control peaks) was used for microbial community structure analyses (Table S1).
Soil subsamples originating from the same plot were pooled together and diluted in a basic salt solution (1% KH2PO4 and 5% NaCl). Two fungal and two bacterial media were chosen: 1/10 strength potato dextrose agar (PDA) with 100 mg L−1 of filter sterile streptomycin sulphate (for general fungi), water agar (WA) with 100 mg L−1 of filter sterile streptomycin sulphate (for oligotrophic fungi), 1/10 strength tryptic soy agar (TSA) with 50 mg L−1 of filter sterile cycloheximide (for general bacteria), and water yeast agar (WYA) with 50 mg L−1 of filter sterile cycloheximide (for oligotrophic bacteria). Following preliminary tests, fungal media were inoculated with 10−2 soil dilution and bacterial media with 10−3 soil dilution (10° is 1 g soil plus 9 mL basic salt solution). Inoculated agar plates were incubated in the dark at three different temperatures (4, 12 and 20°C). Colonies were counted after 9 or 17s day of incubation, depending on the type of medium and the incubation temperature.
Nucleic acid extractions
Soil DNA was extracted using the following protocol. Five hundred milligrams of soil was mixed with 250 mg of 0.1 and 0.5 mm (1 : 1) zirconia–silica beads, 500 μL of phenol–chloroform–isoamyl alcohol (25 : 24 : 1; Tris saturated, pH 8.0) and 500 μL of extraction buffer (12.2 mM KH2PO4, 112.8 mM K2HPO4, 5% w/v CTAB, 0.35 M NaCl; pH 8.0). Soils were then bead-beaten for 30 s at 50 m s−1, and centrifuged at 10 000 g for 5 min at 4°C. The supernatant was mixed with 500 μL of chloroform–isoamyl alcohol (24 : 1) and centrifuged again at 10 000 g for 5 min at 4°C. The supernatant was then precipitated at room temperature for 2 h with two volumes of a 30% w/v PEG 6000 and 1.6 M NaCl solution. The precipitated nucleic acids were then pelleted by centrifugation at 10 000 g for 10 min at 4°C. The nucleic acids pellets were then washed with 70% alcohol, dried, resuspended in 50 μL of deionized water and stored at −20°C until use.
PCR-denaturing gradient gel electrophoresis analyses
Table 1 summarizes the primers, thermocycling regimes and electrophoresis conditions used to analyze the different target communities examined in this study. All PCRs were carried out in 25-μL volumes containing 2.5 μL of 10x PCR buffer, 2.5 μL of bovine serum albumin (BSA; 4 mg mL−1), 0.75 μL of each primer (30 μM), 2.5 μL of dNTPs mix (8 mM), and 1.4 U of Expand high fidelity polymerase (Roche, Mannheim, Germany). All amplifications were carried out on a PTC-200 thermal cycler (MJ-Research, Waltham, MA). All thermocylcing programs were preceded by an initial denaturation step (95°C for 5 min) and followed by a final elongation step phase (72°C for 10 min). For each cycle of PCR, denaturation was at 95°C for 1 min, annealing at the specified temperature (Table 1) for 1 min and elongation at 72°C for 1 min. Touchdown protocols started with the highest annealing temperature, which was subsequently lowered by 2°C for each two cycles until the target annealing temperature was reached. Denaturing gradient gel electrophoreses (DGGEs) were carried using a D-Code Universal Mutation Detection System (Bio-Rad, Hercules, CA). All gradient gels were topped with 10 mL of acrylamide containing no denaturant and electrophoresis was carried at 60°C and 200 V for 10 min followed by an additional 16 h at 70 V. Gels were stained in ethidium bromide and digital images captured using an Imago apparatus (Gentaur, Brussels, Belgium) subsequent to UV transillumination. Banding patterns were normalized with respect to standards of known composition as well as samples loaded across multiple gels. The validity of intergel comparisons was tested by examining the grouping of like samples run across multiple gels, which revealed tight grouping of replicates and grouping according to gel (not shown).
Table 1. Primers, PCR and denaturing gradient gel electrophoresis (DGGE) conditions used in this paper
Real-time PCR was performed using the ABsolute QPCR SYBR green mix (AbGene, Epsom, UK) on a Rotor-Gene 3000 (Corbett Research, Sydney, Australia). All mixes were made using a CAS-1200 pipetting robot (Corbett Research, Sydney, Australia) to reduce variation caused by pipetting errors. Quantification of fungal and bacterial ribosomal genes in soil were carried as described elsewhere (Lueders et al., 2004a, b). For nematodes, the exact same amplification protocol was used as for PCR-DGGE analyses except that the ABsolute QPCR SYBR green mix was substituted for the normal PCR mix. Standards were made from full-length PCR-amplified 18S rRNA or 16S rRNA genes from pure fungal and bacterial isolates. To make the nematode standard, extracted soil DNA was PCR-amplified and cloned. One resulting clone that contained a proper insert of nematode origin was randomly chosen and used in a colony PCR procedure using plasmidic primers. PCR-amplified partial or full-length ribosomal genes of bacteria, fungi and nematodes were purified, quantified on a ND-1000 spectrophotometer (Nanodrop Technologies, Wilmington, DE) and the number of gene copies μL−1 was calculated using the molecular weight of ribosomal sequences as calculated from sequences deposited in GenBank. Using 10-fold increments, the standard concentrations were adjusted from 106 to 101 SSU rRNA gene copies μL−1 for bacteria and nematodes and from 105 to 101 SSU rRNA gene copies μL−1 for fungi. Most of the samples and all standards were assessed in at least two different runs to confirm the reproducibility of the quantification.
The banding patterns of DGGE gels (Fig. S1) were analyzed using the Image Master 1D program (Amersham Biosciences, Roosendaal, the Netherlands). The resulting binary matrices were exported and used in statistical analyses as ‘species’ presence–absence matrices. To test and have a graphical representation of the influences of environmental and soil variables on the microbial population structure, canonical correspondence analyses (CCAs) were carried in Canoco 4.5 for windows (ter Braak & Šmilauer, 2002). Location and vegetation cover were treated as ‘supplementary’ variables while soil and environmental data were included in the analysis as ‘environmental’ variables. Rare species were taken out of the analyses following an empirical method described by D. Borcard (http://biol10.biol.umontreal.ca/BIO6077/outliers.html). Variables to be included in the model were chosen by forward selection at a 0.05 baseline. Using only the chosen variables, the significance of each whole canonical model was tested with 999 permutations.
The effects of location, presence of vegetation and the interaction of these two factors on the community structure as analyzed by PCR-DGGE and PLFA were tested by distance-based redundancy analyses (db-RDA, Legendre & Anderson, 1999). Jaccard's coefficient of similarity (DGGE) or Bray–Curtis distance (PLFA) were first calculated between samples. The use of Jaccard's coefficient is recommended for binary species data, such as DGGE patterns scored for presence vs. absence, whereas Bray–Curtis is the distance of choice for species abundance data, such as PLFA patterns (Legendre & Legendre, 1998). The resulting similarity/distance matrices were then used for the computing of principal coordinates in the R package (Casgrain & Legendre, 2001). When necessary, eigenvectors were corrected for negative eigenvalues using the procedure of Lingoes (1971) and were then exported to Canoco as ‘species data’ for redundancy analyses (RDA). To test the effects of each of the two variables (vegetation and location), each was recoded using dummy binary-variables and one was used in Canoco as the only environmental variable in the model while the other variable was entered as a covariable. To test the interaction, the only variable entered in the model was the interaction between location and plant cover, while both individual factors were included (without interaction) as covariables. The significances of such models were tested with 999 permutations.
All anovas and correlation analyses were carried in Statistica 7.0 (StatSoft Inc., Tulsa, OK). For anova, data normality was tested with a Shapiro–Wilks test and variance homogeneity by Levene's test. When data failed to satisfy one of the tests, an appropriate transformation was applied (log or square root transformation). Tukey's honestly significant difference (HSD) method modified for unequal sample size (Unequal N HSD in Statistica) was used for post-hoc comparison with a 0.05 grouping baseline. For correlation analyses, CFU counts were averaged over all incubation temperatures to provide a simpler result table. Correlations were carried on the untransformed data using nonparametric Spearman rank–order correlations.
Soil and micro-climatic data
As expected, mean soil temperature (5 cm below surface) decreased with increasing latitude, while the vegetation cover did not have any significant effect (Table 2). Freeze–thaw cycles occurred more frequently at the Signy Island sites, whereas they hardly occurred at the Falkland Islands site (Table 2). Anchorage Island had a lower frequency of freeze–thaw cycles than Signy Island, but this difference was not significant. Average soil data and associated statistical tests are presented in Table 3. Some soil variables were clearly influenced by the vegetation cover, being generally higher in vegetated plots (Water content, Organic C, total N, K, Mg, Cl, conductivity and ergosterol). Some others soil variables were mostly influenced by location, decreasing (C : N ratio, pH, Mn) or increasing (NO3, P) with increasing latitude. The other variables measured showed a more complex pattern (Fe, NH4).
Table 2. Mean annual (2004–2005) micro-climatic characteristics at 5 cm depth at the Falkland Islands (FI), Signy Island (SI) and Anchorage Island (AI)
Soil temperature (°C)
Freeze–thaw cycles (per day)
Different letters within a column refer to significantly (P<0.05) different averages based upon an unequal N Tukey–HSD test.
Table 3. Mean soil characteristics for surface soil cores (0–5 cm depth) collected at the Falkland Islands (FI), Signy Island (SI), Anchorage Island (AI), Fossil Bluff (FB) and Coal Nunatak (CN)
Water content (%)
Organic C (%)
NH4 (mg kg−1)
NO3 (mg kg−1)
Total N (%)
C : N
P (mg kg−1)
K (mg kg−1)
Mn (mg kg−1)
Fe (mg kg−1)
Mg (mg kg−1)
Cl (mg kg−1)
Ergosterol (mg kg−1)
Different letters within a column refer to significantly (P<0.05) different averages based upon an unequal N Tukey–HSD test.
Effects of location and vegetation cover on microbial population structure
Preliminary microbial community analyses via the various PCR-DGGE strategies revealed from little to no detectable intraplot variation when five separate samples per plot were compared (data not shown). We therefore pooled five replicate individual nucleic acids extractions from each plot to produce one representative DNA template source for each experimental plot. PLFA analyses were also made on pooled soil samples. Coal Nunatak and Fossil Bluff samples were left out of the DGGE analyses because of insufficient PCR amplification for most of the samples. For cyanobacteria, only 14 samples provided sufficient amplification to be assessed by DGGE even with the use of a nested-PCR amplification approach. Location, plant cover and the interaction between these factors were tested by db-RDA for their influence on community structure assessed by DGGE and PLFA analyses (Table 4). These results taken together point out that the microbial communities have strongly dissimilar structures depending on the vegetation cover and sampling location.
Table 4. Distance-based redundancy analyses results for location and plant cover effects on different population structure assessed by PCR-DGGE and PLFA analyses at the Falkland Islands, Signy Island and Anchorage Island
Influence of environmental and soil factors on microbial community structure
Canonical correspondence analyses were used to determine the environmental factors that appeared to have the strongest influence on microbial community structure as assessed by the various PCR-DGGE strategies employed (Fig. 2). All the models produced when using the respective parameters represented in Fig. 2 were highly significant (test of significance of all canonical axes: P=0.0010). Latitude was the only factor that was chosen for all communities, indicating that community structure was at least partly dependent on latitude across a diverse range of soil-borne organisms.
Microbial abundance in soil
Although the different methods used to estimate microbial abundance in soil (real-time PCR, ergosterol, PLFA and CFU counts) were not always in complete agreement with each other, all showed a clear break in the data, with the two most southerly sites (Fossil Bluff and Coal Nunatak) as outliers. Due to this clear discontinuity in the data, and their lack of balanced sampling regime, these last two sites were excluded from anovas and associated post-hoc tests in our examination of trends from the Falkland Islands through Anchorage Island. The numbers observed for these samples were also typically several orders of magnitude lower than all the other samples, and that is concordant with the increased difficulty encountered in the amplification of certain SSU rDNA targets from these samples for PCR-DGGE analyses.
Real-time PCR results for bacteria and fungi are presented in Fig. 3 and associated anova tests in Table 5. Bacterial 16S rRNA gene abundance was influenced by location, plant cover and the interaction between these two factors in anova tests. Following post-hoc tests, fell-field sites at Signy Island were found to have lower 16S rRNA gene abundance than all other sites, except the fell-field Anchorage sites, while all other sites were similar (Fig. 3). There was also a trend toward decreasing bacterial 16S rRNA gene abundance with increasing latitude in fell-field plots. This trend was not evident in vegetated plots. Fungal 18S rRNA gene abundance in soil was significantly influenced by location and the interaction between location and plant cover, but plant cover by itself did not have any detectable effect in anova tests. Following post-hoc tests, fungal 18S rRNA gene abundance was found to be lower in the fell-field plots on Signy Island and, inversely, lower in the vegetated plots on Anchorage Island. Nematode 18S rRNA gene abundance was not influenced by any of the factors tested (Table 5) and averaged at 2.88 × 106 gene copies g−1 soil DW for the Falkland, Signy and Anchorage sites and at 1.23 × 102 copies g−1 soil DW for Fossil Bluff and Coal Nunatak.
Table 5. anova tests results for soil bacterial, fungal and nematode SSU rRNA abundance, bacterial and fungal PLFA abundance and bacterial and fungal CFU counts on PDA (nutrient-rich fungal media), water agar (WA; nutrient-poor fungal media), tryptic soy agar (TSA; nutrient-rich bacterial media) and water yeast agar (WYA; nutrient-poor bacterial media) at the Falkland Islands, Signy Island and Anchorage Island
For total bacterial PLFA, the only significant difference was between vegetated and fell-field plots on Signy Island (Fig. 3). Signy Island also exhibited a relatively low amount of bacterial PLFAs, especially for fell-field plots. On the other hand, total fungal PLFA amount was mainly influenced by location (Table 5), being significantly higher at Anchorage Island for most cases (Fig. 3). Ergosterol analyses revealed significant influences from plant cover (P<0.000001) and the interaction between plant cover and location (P<0.000001), but not from location by itself. No significant differences were found between vegetation types on the Falkland Islands, but at the two other locations, the amount of ergosterol was significantly higher in vegetated plots (Table 3).
Fungal/bacterial ratios were calculated using real-time PCR and PLFA data as a means of evaluating the relative dominance of these two main soil organisms in the different environment sampled. All the tested factors and their interaction terms were significant for both methods (Table 5). The main difference between the different ratios was that most of the ratios calculated using Real-time PCR results were approximately 10 times lower than the PLFA ratios (Fig. 3). However, the general trend was the same for both ratios: in the Falkland Island plots, the fungal/bacterial ratio was higher in the vegetated plots. The inverse was true for the Signy and Anchorage Island plots, where fell-field plots were significantly richer in fungi. The highest ratios (fungi relatively more abundant) were recorded for the Falkland Islands vegetated plots, in fell-field plots of Signy and Anchorage Islands, at Fossil Bluff and at Coal Nunatak, although the magnitude of the differences with other plots did vary in some cases depending on the method of abundance estimation used.
CFU counts for the different media and incubation temperature used are presented in Fig. 4 and the associated anova tests are presented in Table 5. For PDA (nutrient-rich fungal media), on the Falkland Islands, all plots had approximately the same number of CFU, for all incubation temperatures. In contrast, the number of CFU was consistently higher in vegetated plots on Signy Island. For CFU counts on WA (nutrient-poor fungal media), the only interaction significant was the one between plant cover and location. The effect of vegetation was not significant on the Falkland Islands, but was significant most of the time on Signy Island, as well as on Anchorage Island at the incubation temperature of 20°C.
For bacterial CFU on both TSA (nutrient-rich bacterial media) and WYA (nutrient-poor bacterial media), the pattern was somewhat more complex. The three second-order interaction terms were significant when analysed by anova (Table 5), indicating that the effect of vegetation cover differed depending on the incubation temperature and sampling location. Similarly, location effects depended on incubation temperature and were also different in fell-field vs. vegetated plots. Although not always significant, there was a consistent trend toward decreased bacterial CFU with increasing latitude in fell-field plots. Vegetated plots did not exhibit such a trend. Incubation temperature effects were always highly significant, both for bacterial and fungal media, and there was a general trend toward increased CFU with increasing incubation temperature.
Correlations between soil and environmental factors and microbial abundance
Following correlation analyses, two major groups of soil variables emerged as presented grouped in Table 6. The first group of factors was related to vegetation cover (see Table 3) and included water content, organic matter, total N, Cl, K, Mg and conductivity (rs with water content ranging from 0.56 to 0.95, P<0.05). The second was related to location or latitude (see Tables 2 and 3) and included soil mean temperature, pH-H2O, C : N ratio, P, Mn and NO3 content (rs with latitude in absolute value ranging from 0.62 to 0.95, P<0.05). Most of the abundance measures were significantly correlated with plant-related parameters (Table 6). Furthermore, the main factors influencing the fungal/bacterial ratios were also related to vegetation type. The different bacterial abundance measures were also significantly correlated most of the time: 16S rRNA gene abundance, bacterial PLFA abundance, CFU counts on TSA (nutrient-rich bacterial media) and on WYA (nutrient-poor bacterial media) were all significantly correlated with each other (rs all positive, ranging from 0.44 to 0.56, P<0.05), with the exception of the correlation between WYA counts and bacterial PLFA abundance. The picture was less coherent for fungal abundance measures: soil ergosterol content was positively correlated with CFU counts (rs=0.69 (PDA) and rs=0.52 (WA), P<0.05), with fungal PLFA abundance and fungal 18S rRNA gene abundance being correlated (rs=0.43, P<0.05). Fungal PLFA was also correlated to CFU counts on PDA (rs=0.46, P<0.05), but all other combinations were insignificant.
Table 6. Spearman rank order correlations between soil and micro-climatic parameters and diverse microbial abundance parameters measured at the Falkland Islands, Signy Island and Anchorage Island
Significant correlation (P<0.05) values are in bold. Mean temperature and the number of freeze–thaw cycles correlations were calculated using three plots per site per treatment (N=18) while the other correlations were calculated for all the plots (N=36).
Global warming is expected to have mainly indirect effects on microorganisms, especially via changes in macrophyte species composition, vegetation density, and litter quality and quantity, as well as associated changes in soil biochemical and biophysical characteristics (Panikov, 1999). This study therefore sought to provide a baseline of understanding regarding the drivers of microbial community structure across a gradient of Antarctic and sub-Antarctic environments with a special focus on the role of vegetation cover on the size and structure of associated soil-borne communities. Although spatial gradients have been used widely to predict long-term effects of global warming on ecosystems (Dunne et al., 2004), it should be recognized that the use of such a gradient along the Antarctic Peninsula region is not straightforward due to parallel variations in the severity of the thermal and hydric environments, differences in precipitation balance and disparate geological histories across the study range (Kennedy, 1993). Nevertheless, a number of useful general trends can be elucidated from the dataset examined here. For instance, the structure of the various subsets of the soil-borne communities examined assessed by several PCR-DGGE strategies was mostly coupled to factors related to latitude (mean temperature, pH, C : N ratio, etc.), whereas abundance data was mostly influenced by plant-related factors (organic C, soil humidity, total N, etc.). Thus, community structure appears to be determined to a large extent by the location and/or the specific location-dependent environmental conditions, whereas microbial abundance may be more associated with vegetation-related effects of nutrient input and climatic buffering. Different subsets of the total soil community also reacted differently to the presence of different vegetation and the range of environmental conditions encountered across the study area. Furthermore, conspicuous and complex interactions were apparent between location, vegetation cover and other variables, highlighting the fact that vegetation effects were highly dependent upon the environmental context in which they occurred.
Bacterial community size and structure
Antarctic environments are most well known for their severe climates. Bacterial processes are particularly sensitive to environmental conditions (Eriksson et al., 2001), yet bacteria are also highly adaptable to extreme and changing environments (Cavicchioli et al., 2000; Georlette et al., 2004; Thomas, 2005). Previous studies on bacteria in terrestrial Antarctic habitats have provided some general appreciation of such unique assemblages, but detailed community analyses across a range of systems were still lacking prior to this investigation. Previous reports have suggested that Antarctic bacteria are influenced by temperature patterns, plant cover, soil humidity and other soil characteristics (Christie, 1987; Tearle, 1987; Bölter, 1992; Bölter, 1995; Bölter et al., 1997; Harris & Tibbles, 1997). Our results support these suggestions, as we found that bacterial abundance and community structure to be influenced both by plant- and weather-related factors, with numerous complex interactions among these variables. Interestingly, bacterial abundance did not simply decrease with the coldness of the environment. For instance, the fell-field plots on Signy Island supported the lowest bacterial community densities (except for Fossil Bluff and Coal Nunatak). This Signy Island habitat is also subjected to a high frequency of freeze–thaw cycles, which may actually impose a greater stress level than conditions with a colder average temperature (Yanai et al., 2004). Plants are known to produce soil microhabitats (Kowalchuk et al., 2002), and even though the freeze–thaw frequency was unchanged by vegetation cover (Table 2), our results suggest that the dense vegetation in our experimental plots was able to counter the effects of the extreme environmental conditions to some extent. This influence of vegetation may explain the disparate effects of latitude on bacterial abundance in fell-field vs. vegetated sites. It is, however, not clear whether these effects are mediated by vegetation-induced protection of soil microhabitats, input of plant-derived substrates, or other mechanisms.
It is generally accepted that the harshness of Antarctic environments is not caused by the extreme climatic conditions per se, but perhaps more to the extreme range of conditions that are encountered. Previous reports have demonstrated that polar soils with no vegetation generally support fewer microorganisms than soils associated with mosses (Kaštovskáet al., 2005), and it was suggested that, for Antarctic soil, this may be caused by the combined effects of greater nutrient availability and more favorable physical conditions (Harris & Tibbles, 1997). These data, however, are confounded by the fact that mosses tend to occur at sheltered sites that already exhibit relative thermal- and hydro-stability. Nevertheless, the buffering action of mosses is likely to maintain soils beneath them at relatively constant water content and temperature which might strongly influence bacterial abundance. Indeed, all bacterial abundance measures were correlated to the soil water content (Table 6). Soils with dense vegetation cover also had a higher nutrient input than fell-field soils (Table 3), which might have helped to support a more abundant bacterial community. Such a separation of soils with high organic matter content from mineral soils was previously observed using cluster analysis of soil physical parameters and diverse microbial population descriptors (Bölter, 1990). This could also explain the lack of significance of vegetation cover at the Falkland Islands site for most data, as the environment was rather mild and all plots relatively rich in nutrients, which would decrease any buffering or nutrient effects conferred by increased vegetation cover.
Cyanobacterial community structure followed the same trends as seen for total bacterial communities, but showed a lower level of significance, probably due to the lower number of samples in the final analysis. Cyanobacterial community structure might be dictated to some degree by the presence of mosses, and a previous study demonstrated an association of specific Cyanobacterial assemblages with mosses (Solheim et al., 2004). In barren arctic and alpine environments, a significant portion of the bacterial community was related to the photosynthetic bacterial division Cyanobacteria (Kaštovskáet al., 2005; Nemergut et al., 2005). However, the difficulty we experienced in recovering Cyanobacterial-specific PCR products suggests that this division did not represent a significant proportion of the bacterial communities in these soils. Sequence data from bacterial 16S rRNA gene clone libraries from these sites also suggests that Cyanobacteria only make up a small minority of these bacterial communities (E. Yergeau & G.A. Kowalchuk, unpublished data).
Fungal community size and structure
Fungal community structure was not influenced by plant cover per se, but the interaction between location and vegetation cover was highly significant, indicating that the type of vegetation cover was of importance in how latitude affected fungal communities. This supports the previous finding that fungal communities can respond very differently to changes in organic input levels and quality depending on the environmental conditions (Tosi et al., 2005). Fungal quantification by real-time PCR showed a similar trend. A previous study using culture-based methods reported viable fungal propagules in a moss bank to be significantly influenced by the extent of the coverage of particular macrophyte species (Smith & Walton, 1985). This suggests that the species composition of the vegetation might also be of importance in influencing the response of the fungal community to latitude. This is partly supported in our dataset (DGGE and real-time PCR) by the fact that vegetation by itself did not have a significant influence. Instead, effects were more subtle, with fungal communities responding differently to vegetation cover depending on the environmental conditions present. This contrasted with data on bacterial community structure and abundance, where vegetation cover was a highly significant determinant across the range of latitudes examined.
In this study, we used several different approaches to estimate fungal biomass. In contrast to bacteria abundance measures, which all showed a similar picture, fungal abundance measures did not agree in all cases. Fungal abundance as estimated by 18S rRNA gene quantification via real-time PCR was correlated to values obtained using fungal PLFA estimates. However, these estimates showed no coherent picture in regard to correlations with the soil and weather factors measured. Using direct counts, it was reported that fungal abundance was similarly unaffected by environmental parameters (Bailey & Wynn-Williams, 1982). This contrasted with the results we obtained via fungal CFU counts and ergosterol measurement. As already reported for other environments (Widmer et al., 2001; Leckie et al., 2004), different soil microbial abundance estimators can give different, sometimes complementary results. However, CFU are known to provide a biased view of the abundance of microorganisms, as they only show the culturable part of the community (Staley & Konopka, 1985), and it is also impossible to translate propagation units into biomass. The high stress and high disturbance conditions of Antarctica are also thought to select fungal species that produce large numbers of small spores (Tosi et al., 2005), which could further bias fungal CFU counts. Additionally, it was reported that ergosterol content should be used with caution since it shows a high persistence in some soils (Mille-Lindblom et al., 2004), especially in low-temperature soils (Weinstein et al., 2000).
With these cautionary notes in mind, we still found some noteworthy trends in the culturable and ergosterol abundance data. At sites other than at the Falkland Islands, the fungal abundance was generally higher in vegetated plots, and correlation analyses indicated that the main factors influencing fungal abundance were plant-related (soil organic matter and water content; Table 6). These results are in line with previous culture-based studies carried out on Signy Island (Bailey & Wynn-Williams, 1982). Despite the shortcomings of ergosterol and CFU data for estimating fungal biomass, these data do lend some support to the notion that these fungal communities are shaped more by substrate quality and quantity, as well as other site-specific characteristics as opposed to pure weather-related parameters. CFU counts also showed a trend toward increased numbers with increasing incubation temperature, suggesting that some mesophilic strains that could not grow to the level of detection at low temperatures were present. This is in agreement with other studies that reported a prevalence of cold-tolerant fungal species rather than cold-adapted ones (Kerry, 1990; Melick et al., 1994; Zucconi et al., 1996; Robinson, 2001).
Fungal 18S rRNA gene/bacterial 16S rRNA gene and fungal PLFA/bacterial PLFA ratios followed the same trend, being the highest at sites with the harshest temperatures (fell-field sites at Signy, Anchorage and Fossil Bluff) or in the only plots having a dense cover of vascular plants (Falkland Islands vegetated plots). This could imply that fungi are less influenced by weather conditions than bacteria and can dominate more easily in harsh ecosystems, probably because of a better adaptation to lower temperatures or the presence of a cell wall. Fungi are also known to be able to degrade more complex organic matter than bacteria and that might explain their higher relative abundance in the Falkland Island vegetated plots. The only exception, where both methods did not completely agree, was in the case of the Coal Nunatak samples, probably because the biomarker amounts present in these soil samples were approaching the detection limits of the methods. The 10-fold difference between the two ratios is coherent with the differences in cell size and number of SSU rRNA gene copies per cell between fungi vs. bacteria. In support of other recent environmental studies (Malosso et al., 2004; Nemergut et al., 2005), it appeared that real-time PCR and PLFA analyses were the more consistent techniques for microbial biomass estimation in Antarctic soils.
Nematode community size and structure
Nematode community structure was strongly influenced by plant cover, location and their interaction term in our study, and although the abundance measured by real-time PCR was statistically similar for all samples, excluding Fossil Bluff and Coal Nunatak, we found nematode abundance to be highly correlated to soil organic matter and water content, both of which are vegetation-associated characteristics. In agreement with our study, nematode community structure, respiration and abundance were previously shown to be linked to the overlying vegetation on Signy Island (Caldwell, 1981). However, a slightly different pattern was observed for Mars Oasis, which is a coastal site close to Fossil Bluff, where nematode density, but not species richness, was considerably higher in naturally vegetated soil (Convey & Wynn-Williams, 2002). Antarctica nematodes were also found to be linked to organic matter, with higher numbers in the vicinity of bird colonies and moss patches (Sohlenius & Boström, 2005). In interpreting the differences between our results and those published previously, methodological differences also have to be considered. For instance, traditional nematode counts using migration extraction were reported to be difficult to adapt to some Antarctic soils (Freckman & Virginia, 1993). On the other hand, primer incompatibilities, as well as uncertainties in extraction and amplification efficiencies, are also potential sources of bias and error in the molecular estimators of nematode density. Nematode distribution in Antarctic habitats is also highly patchy (Sohlenius et al., 2004), making the use of small sample sizes for molecular analyses another potential source of error. Nevertheless, it was recently reported that the abundance of nematodes estimated by molecular methods could be related to biovolume, but not to the number of individuals in soil (Griffiths et al., 2006). The primers used here were also proven to be adapted to assess nematode communities in soil directly (Waite et al., 2003).
The analysis of soil-borne microbial communities described here for the first time examines the factors shaping microbial communities across a range of terrestrial Antarctic habitats. The latitudinal gradient examined was intended as a rough surrogate for long-term climate-change scenarios in soils, with our results forming an initial baseline to estimate the direct and indirect consequences of global warming in these extreme, pristine and rapidly changing environments. Given the rate of climate change, natural seasonal fluctuations and microbial abilities to adapt to environmental changes, it is hypothesized that the direct effects of climate change on soil-borne communities will be minor (Panikov, 1999). Accordingly, in light of the results presented here, we hypothesize that increases in bacterial, fungal and nematode abundance, and large changes in community structure, including changes in the relative abundance of fungi and bacteria, will only occur if climate change induces increases in nutrient inputs via increased vegetation density or productivity. Both in situ and laboratory experimental investigations into such hypotheses are clearly necessary to determine the functional consequences of Antarctic microbial community responses to global warming.
This study was supported by NWO grant 851.20.018 to R. Aerts and G.A. Kowalchuk. E. Yergeau was partly supported by a Fonds Québécois pour la Recherche sur la Nature et les Technologies postgraduate scholarship. The British Antarctic Survey, and especially Pete Convey, is gratefully acknowledged for supporting field operations. Merlijn Janssens and Kat Snel are acknowledged for sampling efforts. Wiecher Smant and Wietse de Boer are thanked for help with soil analyses. This is NIOO-KNAW publication 3892.