Frequent freeze–thaw cycles yield diminished yet resistant and responsive microbial communities in two temperate soils: a laboratory experiment


  • Editor: Gary King

Correspondence: Blaž Stres, Department of Animal Science, Biotechnical Faculty, Chair for Microbiology and Microbial Biotechnology, University of Ljubljana, Groblje 3, 1230 Domžale, Slovenia. Tel.: +386 1 7217 869; fax: +386 1 7241 005; e-mail:


Few studies have been conducted on adaptations of microbial communities to low and fluctuating temperatures using environmentally relevant conditions. In this study, six Himalayan and two temperate soils were selected as candidates for low-temperature/freeze–thaw (FT)-adapted and susceptible soils, respectively. Redundancy analysis with forward selection was used to create a model of environmental parameters explaining variability in the initial microbial abundance and 4 °C activities. The best predictor was soil carbon, explaining more than 74% of data variability (P=0.002), despite significant differences in the soil characteristics and environmental history. We tested the hypothesis that the reproduced Himalayan FT fluctuations select physiologically similar communities in distinct soils. Microcosms were experimentally subjected to two separate 50 and 60 FT cycle (FTC) experiments. A significant decrease in abundance, 4 °C basal respiration and drastic rearrangements in community-level physiological profiles (CLPP) were observed in microcosms with temperate soils until 40 FTC. CLPP remained distinct from those of the Himalayan soils. Minor changes were observed in the Himalayan soils, confirming that microbial populations with physiological traits consistent with the noncontinuous permafrost conditions reside in the Himalayan soils, whereas the surviving temperate soil microorganisms actively adjusted to novel environmental conditions.


Most of our current knowledge on microbial adaptations to cold or freeze–thaw (FT) is based on cultivated isolates obtained from cold environments that grow under laboratory conditions (Walker et al., 2006; Visnivetskaya et al., 2007; Rodrigues & Tiedje, 2008). Only recently have cultivation-independent studies described major microbial groups residing in Arctic and Antarctic soil environments (Yergeau & Kowalchuk, 2008; Männistöet al., 2009). However, few studies have been conducted at the microbial community level, rendering their physiologies and ecologies virtually unexplored (Ley et al., 2001; Rodrigues & Tiedje, 2008). In addition, most studies on microbial adaptations to temperature changes in temperate, polar or alpine regions on Earth have focused on the effects of global warming (Bardgett et al., 2008), whereas only recently Hartley et al. (2008) reported a lack of evidence of compensatory thermal acclimation of temperate microbial respiration to lowered temperatures. Studies using distinct soils and approaches have reported conflicting effects of FT on microbial biomass and activities (Henry, 2007). In some cases, a significant decrease in soil microbial abundance or activity was reported (Schimel & Clein, 1996; Larsen et al., 2002), whereas others reported insignificant effects (Lipson & Monson, 1998; Sharma et al., 2006). In addition, based on the total plate counts of one soil, FT cycles (FTC) were suggested to select an FT-tolerant community (Walker et al., 2006). Although informative, the relevance of these studies to metabolic adaptations at the community level to cold is not known due to the low number of soils tested within one study, the large divergence in the number (up to 20 FTC day−1), the rates (up to 48 °C h−1) and amplitude (from −18 to +20 °C) of FTC used in these studies, frequently generating environmentally irrelevant FT conditions thus making it hard to reconcile the observed effects.

In this study, six Himalayan and two temperate Central European soils were adopted as candidates for cold adapted and susceptible soils, respectively. Temperature measurements [Ishikawa et al., 2001; T. Watanabe, pers. commun.; field observations (B. Stres, unpublished data)] show that the high-altitude soils of the Kanchenjunga region are encountering frequent FT events (approximately 100 per year concentrated mostly during spring and autumn) and are characterized as noncontinuous permafrost (Ishikawa et al., 2001).

Because one of the important characteristics of the Himalayas is slope steepness and altitude, we hypothesized that a 1000 m difference in altitude at elevation from 5000 to 6000 m on the south-facing slope of Drohmo Peak (6980 m), Kanchenjunga region, Nepal Himalaya, would help disentangle the factors driving the quantitative and physiological aspects of high-altitude microbial communities and contrast them with those of temperate soils. In a laboratory, 50 and 60 cycle FT microcosm experiments reflecting the Himalayan in situ temperature fluctuations were tested for the general FT tolerance of microbial abundance and activities in all eight soils in an attempt to disentangle environmental and FT effects. Our final aim was to test whether the increasing number of FTC selected for functionally resilient microbial communities in susceptible soils capable of fine-scale adjustments of community-level metabolic profiles at low temperatures. We believe this is the first investigation of the influences of FT cycling mimicking natural conditions on the abundance, microbial activity and community-level metabolic properties all measured at 4 °C in eight soils of distinct origin.

Materials and methods

Sampling sites

Topsoil samples were collected on the south-facing slope of a high-alpine ridge descending from Drohmo peak (6980 m), Nepal (27°48′00″N and 88°07′02″E) (Supporting Information, Fig. S1). Sharp boulders formed small wind-protected patches containing barren shallow soil material at 6000 m. The 5800 m site was gravel swept with no vegetation cover (Fig. S2). Patches of moss (<1 m2, 1–3 cm thick) were found at 5600 m among larger boulders with no vegetation except lichens covering the limited boulder surface. Mixed grass–moss community was present on up to a 5-cm-thick layer of soil at 5400 m, whereas glacial moraine supported up to 10 cm layer of soil, giving rise to the polar tundra-like plant community dominated by Kobresia spp. and Poa spp. at 5200 m. Small portions of freshly exposed gravel and sand material brought from upstream-glaciated regions were sampled from the surface of the Kanchenjunga glacier at 5000 m.

The mean annual temperature (1998–2001) measured in the close proximity (27°48′16″N and 88°06′03″E) was −7.1 °C (Ishikawa et al., 2001; T. Watanabe, pers. commun.).

The mean annual precipitation in the form of snow was 2500–3000 mm (Hidy, 2003; Barros et al., 2004). Satellite image analysis (Kellenberg et al., 2007) and field explorations (Thomas & Rai, 2005) show an increase in the alpine grass-covered area at the expense of the snow-covered area, characterizing the soils as noncontinuous permafrost (Ishikawa et al., 2001). Similar permafrost degradation was also reported in the nearby Qinghai–Tibet Plateau, with the permafrost estimated to disappear within the next c. 150 years (Wang & French, 1995).

Two distinct, but closely located arable soils from Central Europe, mineral (46°02′54″N, 14°28′13″E) and marsh (45°58′35.5″N, 14°28′11″E) soil, used previously to characterize the seasonal effects of soil bacterial communities (Kraigher et al., 2006; Stres et al., 2008), were sampled. The mean annual temperature and precipitation of both soils are +10 °C and 1400 mm, respectively (Stres et al., 2008). The two sampling sites are located in close proximity of Ljubljana (Fig. S3) and experience at most two mild FT events per year under snow cover or do not freeze at all.

Soil sampling

The topsoil samples were collected at the end of October 2002. The Himalayan soils were collected in 200 m horizontal transects and 200 m vertical increments starting from the Kanchenjunga glacier (5000 m), resulting in 24 sampled locations. Each transect consisted of four equally spaced locations, where 20 cores of loosely structured soil material were collected down to the bedrock or the frozen surface (from 2–3 cm to up to 7 cm deep), using a soil corer (l=10 cm, d=3 cm), stored at 4 °C and airlifted. This resulted in 480 cores packed separately for 24 locations. Soil aliquots of 0.5 g were frozen (−20 °C) for molecular analyses. Experiments were initiated within 2 weeks after sampling and portions of soil intended for activity measurements were stored in a cold room at 4 °C. The top 5 cm of mineral and marsh soils from Ljubljana were sampled in the same manner, resulting in 80 cores each (4 sites × 20 cores per site). Stones and roots were immediately removed by hand after sampling. During the second sampling in October 2005, the 6000, 5200 m locations, mineral and marsh soils were processed again as described above.

Soil analyses

The soil samples were analyzed for a variety of physical and chemical characteristics to determine the gradients in total soil carbon (C) and nitrogen (N) (Hedges & Stern, 1984), pH, ammonia (NH4+) and nitrate (NO3) (Kraigher et al., 2006), water-holding capacity (WHC), soil texture and moisture (Kassem & Nannipieri, 1995). The quotient of wavelength absorption at 365 and 250 nm was used as a measure of the molecular weight of aromatic compounds in dissolved organic matter (Osburn et al., 2001). The soil content of the reductive sugar glucose equivalents was determined in a 1 : 1 cold water extract (Lever, 1977). To avoid chemical contaminations, particularly in soil solutions from the samples from a high altitude, only double-rinsed glass ware was used in the entire procedure. Aggregate size classes were determined by wet sieving at the start and after the completion of the FT experiment (Gale et al., 2000). The data were corrected for water content and expressed per gram of soil dry weight. Analyses performed on each soil were replicated at least three times in the experiments. In addition, each soil served as an experimental replicate of the soil environment (giving n=8) to provide a conservative test of our hypotheses and also ensuring that the study design remained feasible.

Microbial abundance

The DNA extraction and quantitative PCR targeting bacterial 16S rRNA genes were conducted in triplicate as described before (Henry et al., 2006) using two universal primers specific for members of Eubacteria 341f (5′-CCT ACG GGA GGC AGC AG-3′) and 515r (5′-ATT CCG CGG CTG GCA-3′) to amplify a 174-bp region. Real-time PCRs were carried out in a Smart Cycler (ChypheidR). The 25 μL reaction mixtures contained 0.3 μM each primer for 16S rRNA gene amplification, 12.5 μL of SYBR Green PCR master mix, including HotStar Taq DNA Polymerase, Quanti Tec SYBR Green PCR Buffer, dNTP mix with dUTP, SYBR Green I, ROX and 5 mM MgCl2 (QuantiTectk SYBRR Green PCR Kit, Qiagen, France), 2.5 μL of DNA diluted template corresponding to 25 ng of total DNA and RNase-free water to complete the 25 μL volume.

The overall abundance of microorganisms was determined in quadruplicate by direct counting using a Nikkon Exclipse TE 300 fluorescence microscope equipped with a DXM 1200 digital camera. Soil suspensions were prepared as described before (Bloem, 1995). Briefly, soil aliquots were homogenized in a 0.05 M Na2HPO4 in a blender and fixed with formaldehyde (3.7% final concentration). After settling for 2 min, aliquots were stained with acridine orange (1 mg mL−1 final concentration). The SD obtained from>25 microscopic fields was estimated empirically at the onset and at the end of the FTC experiment to be within ± 10% and ± 15% of the mean, respectively.

Microbial activities

The degradation of aromatic compounds was determined by decolorization of triphenylmethane dye crystal violet (Sigma) (Sun-Young et al., 2002) and conducted in 50-mL tubes containing 3 g of each sample in 10 mL dye solution at a concentration of 3 mg L−1 in triplicate. The discoloration was followed spectrophotometrically at a constant pH at 550 nm for 24 h (4 °C, 150 r.p.m.). The blanks incubated at 4 °C consisted of autoclaved soils and were subtracted. The rates of activity measurements were expressed per gram of soil.

To determine the glutamate-induced respiration at 4 °C (Lipson et al., 2000; Ley et al., 2001, 2004), three replicates of 3 g portions of each soil were amended with 2.25 mg C g−1 as the glutamate concentration exhibiting the highest respiration, incubated at 4 °C for 10 h and the linear increase in the CO2 concentration was used for calculations. A six times higher concentration of glutamate was used for marsh soil to promote the maximum respiration rates in these soils as determined before the start of the experiments. Glutamate is involved in cryoprotection and also functions as an osmolyte (D'Souza-Ault et al., 1993; Rodrigues & Tiedje, 2008) and is an important constituent of decaying cell material. In addition, glutamate typically resulted in the highest biomass estimates in cold tundra soils (Lipson et al., 2000; Ley et al., 2001).

The basal respiration rate was used as an estimate of the gross activity of all heterotrophic microorganisms (West & Sparling, 1996; Ley et al., 2004; Bradford et al., 2008). Three replicates of 10 g of each soil sample were placed in experimental flasks. Gas samples of 0.5 mL was taken from the headspace 4–8 h after closure at 4 °C for gas-phase analysis, and a Beckman gas chromatograph equipped with a thermal conductivity detector was used (Kraigher et al., 2006). Before each measurement, the experimental flask was placed in an insulated container to limit temperature change when removed from a 4 °C incubator or sand bed (see FT experiment). Gas concentrations were corrected for dissolution at 4 °C (Tiedje, 1994). The rates were calculated from the increase in the CO2 concentration in the headspace over each incubation period.


FT experiment

The gross activity of soil microorganisms in the FT experiment was measured through the basal respiration rate at 4 °C as described above. Diurnal FTC (n=50) with an amplitude of +4 to −4 °C and FT rates of 2.4 °C h−1 were applied to realistically reproduce the conditions in the high-altitude soils of the Himalaya (Ishikawa et al., 2001; B. Stres, unpublished data). The constant 4 °C incubation served as controls. In order to model the formation of a temperature gradient resembling those encountered in the field, i.e. the top-down direction of soil freezing and thawing, all microcosms were placed in a sand bed with a particle size distribution similar to that of soils during FT experiments. The moisture of the sand bed and soil samples was maintained at field capacity throughout the experiments by spraying with sterile bidistilled water at 4 °C. The temperature of the incubated microcosms was directly controlled by a temperature-calibrated tensiometer (Düwi 7975®, Germany) immersed into parallel sample soils. In order to determine the increase in the gas-phase CO2 concentration in the respiration rate measurements, the microcosms were gas tight closed before the onset of the FTC 1, 3, 7, 10, 25 and 50. Gas samples were taken 24 h later at the end of the 9-h FTC period at 4 °C and the microcosms were then uncapped to allow gas equilibration before the next cycle of freezing. The starting headspace contained atmospheric gas composition. Replicate microcosms were sacrificed for direct counts and soil analyses.

After 10 FTC, a subexperiment monitoring of DNA degradation and DC for 6 days at 4 °C was conducted. DNA was extracted from triplicate 0.5 g subsamples (0.25 for marsh soil) on days 0, 2 and 6 after the completion of 10 FTC. The Ultraclean Soil DNA kit (MO BIO Laboratories, CA) was used after an extended cell disruption protocol according to the manufacturer's instructions for the maximum DNA yield. The DNA concentration was measured using a spectrophotometer (Nanodrop 1000, Thermo Scientific) and expressed per gram of dry soil. Direct counts were performed as described above.

MicroResp physiological profiling experiment

The second FTC experiment was conducted on the four soil samples collected in October 2005 (5200, 6000 m, mineral and marsh), essentially as described above. The microcosms were gas tight capped only before the onset of FTC 1, 10, 25, 40, 50 and 60. The controls again consisted of soil microcosms incubated at 4 °C throughout the experiments (n=60 diurnal FTC).

Using the MicroResp approach of community-level physiological profiling (Campbell et al., 2003), substrates were prepared so that 2.25 mg C g−1 soil was supplied. The following 16 substrates were used: six carbohydrates (glucose, mannose, fructose, galactose, arabinose, trehalose), five amino acids (glutamate, histidine, serine, alanine, arginine), five organic acids (α-ketobutyrate, malate, citrate, ascorbate, panthotheic acid) and water, based on the work of Stevenson et al. (2004). Soil samples were removed from FTC or 4 °C incubation at regular intervals at days 0, 10, 25, 40, 50 and 60. Soil (250 μL total volume) was added to 96-well microtiter deep-well plates after the solutions of each substrate had been added first. Three wells per substrate were prepared for each soil and independently replicated twice. Measurements were conducted at 4 °C for up to 24 h to avoid the substrate use change due to fluctuations in temperature and free water availability in FTC (Mikan et al., 2002; Schimel & Mikan, 2005). The concentration of the reactants in the Cresol Red gel detector plate was the same as described before (Campbell et al., 2003). A calibration curve of absorbance vs. headspace equilibrium of the CO2 concentration in the well headspace was fitted to an exponential model (R2=0.969) as follows: OD540 nm=inline image. This was determined by equilibrating 96-well detector microtiter plates with different concentrations of CO2 prepared from a commercial gas mixture (2% CO2) in a chamber for 4 h at 4 °C. After equilibration, plates were read on a plate reader (Dynex Technologies, Microplate reader) using the dynex revelation 3.04 program at 540 nm.


Soil physical and chemical parameters were log transformed and analyzed for differences using repeated anova procedures using statistica6 (StatSoft Inc.) with ‘region/geology’ as a factor. A pairwise comparison of means was performed using Tukey's honestly significance test. Individual MicroResp substrate CO2 responses were standardized by the sum of CO2 responses of all three substrate groups (amino acids, carbohydrates, carboxylic acids) for each soil and arcsine square root transformed.

Nonmetric multidimensional scaling (NM-MDS) using the Bray-Curtis distance, 250 runs with real and randomized data, was applied for visualizing the differences in the soil characteristics of sampling sites and MicroResp responses using PC-ORD V4.32 (mjm Software Design). The Monte-Carlo (MC) technique was used to assess the significance of correlations, stress decomposition (MC Scree plot) and the number of dimension axes to retain (McCune et al., 2002). The proportion of variance represented by each axis based on the r2 distance in the ordination space and in the original space next to correlations of variables with axes are reported in the text.

A linear-constrained ordination, redundancy analysis (RDA) with forward selection and line transect restrictions on the permutations where random shifts of a mirror image were disabled were used to create a model of environmental parameters (Table 1) explaining the variability in biological variables (Table 2). The MC permutation test (499 permutations), with environmental parameters used as predictors, was applied to compute the significance of hypothetical relations. Patchiness and vegetation cover were used as qualitative explanatory variables. All analyses were carried out using multivariate data analysis software canoco for Windows V4.5 (Leps & Smilauer, 2003) in conjunction with canodraw 4.0 and canopost programs. The response of microbial parameters to the gradient in the statistically significant environmental factors is reported in the text.

Table 1.   Environmental parameters of the six Himalayan high-altitude soils and two temperate Central-European soils
 Himalayan high-altitude soilsCentral-European temperate soils
6000 m5800 m5600 m5400 m5200 m5000 mMarshMineral
  • Mean values are reported. Values in parentheses denote 1 SD.

  • *

    Significantly different parameters after the completion of the FT experiment are reported.

  • MWI, molecular weight index of dissolved organic carbon.

Depth (cm)4 (± 2)5 (± 3)3 (± 2)5 (± 4)10 (± 4)1 (± 0.5)275 (± 45)35 (± 9)
WHC (g water g−1)0.33 (± 0.08)0.19 (± 0.07)0.26 (± 0.06)0.33 (± 0.06)0.4 (± 0.12)0.18 (± 0.03)1.67 (± 0.24)0.344 (± 0.11)
Moisture (g water g−1)0.046 (± 0.01)0.019 (± 0.01)0.16 (± 0.01)0.167 (± 0.06)0.128 (± 0.04)0.010.45 (± 0.09)0.098 (± 0.02)
Slope (°)704535255500
Patchiness (<m2)PatchesPatchesPatches
VegetationMossMoss, grassGrassGrasslandGrassland
Soil classificationSandy loamLoamy sandLoamy sandLoamy sandSandy loamLoamy sandHistosolSilt loam
Clay (%)6.85 (± 1.42)1.62 (± 0.89)1.93 (± 0.23)1.6 (± 0.44)6.85 (± 2.49)2.66 (± 1.11)55.6 (± 9.15)24 (± 3.91)
Silt (%)20.94 (± 4.34)19.78 (± 1.52)20.62 (± 0.89)27.22 (± 2.19)28.27 (± 5.21)25.12 (± 3.71)31.6 (± 2.11)53 (± 2.17)
Sand (%)72.2 (± 6.98)78.59 (± 2.33)77.44 (± 5.81)71.16 (± 6.41)64.88 (± 3.31)72.21 (± 3.19)12.8 (± 3.19)23 (± 4.15)
Bedrock materialGraniteGraniteGraniteGraniteGraniteGraniteCarbonateCarbonate
pH5.4 (± 0.22)5.6 (± 0.31)5.9 (± 0.18)5.9 (± 0.3)5.1 (± 0.2)6.4 (± 0.3)7.3 (± 0.16)6.9 (± 0.2)
Corg (%)1.53 (± 0.3)0.62 (± 0.17)1.97 (± 0.47)1.91 (± 0.33)2.07 (± 0.63)0.51 (± 0.14)15 (± 1.8)2.5 (± 0.49)
Ntot (%)0.11 (± 0.03)0.018 (± 0.01)0.12 (± 0.028)0.17 (± 0.07)0.11 (± 0.06)0.02 (± 0.01)1.2 (± 0.69)0.26 (± 0.35)
NH4+ (μg g−1)15.56 (± 9.43)11.2 (± 5.08)22.6 (± 10.83)10.15 (± 4.24)32.4 (± 7.31)0.77 (± 0.61)15.6 (± 3.56)10.1 (± 3.68)
NO3 (μg g−1)18.6 (± 13.58)15.9 (± 10.87)16.65 (± 9.18)24.0 (± 15.28)62.5 (± 17.93)0.9 (± 1.03)20.7 (± 3.39)92.7 (± 5.91)
MWI (DOC)2.33 (± 0.45)3.22 (± 0.23)2.74 (± 0.34)2.51 (± 0.51)2.78 (± 0.12)2.91 (± 0.45)5.97 (± 0.23)4.23 ± 0.56)
Sugars (μg g−1)13.84 (± 1.32)10.44 (± 4.86)10.29 (± 2.36)13.57 (± 3.31)6.01 (± 1.71)0.96 (± 0.71)45.89 (± 4.46)29.13 (± 3.81)
>150 μm aggregates (% of the initial texture content)*99.1 (± 3.1)98.1 (± 2.4)97.1 (± 3.2)98.1 (± 2.5)99.1 (± 1.3)99.1 (± 1.1)19.5 (± 4.1)21 (± 3.5)
Table 2.   Microbial parameters of the six Himalayan high-altitude soils and two temperate Central-European soils before the start of the FT experiment (n=50 diurnal cycles)
 Himalayan high-altitude soilsCentral-European temperate soils
6000 m5800 m5600 m5400 m5200 m5000 mMarshMineral
  1. Basal 4°C respiration rate measurements during the course of the FT experiment are presented in Fig. 2. The data are presented in the form: mean (1 SD).

  2. DC, direct cell counts; glutamate-IR, glutamate-induced respiration.

 Bacterial 16S rRNA genes copies (108 g−1)6.82 (1.15)2.07 (0.95)7.36 (2.71)4.7 (1.30)7.5 (2.00)0.5 (0.19)9.23 (3.53)16.5 (4.84)
 DC (108 g−1)4.2 (0.35)1.26 (0.09)5.15 (0.45)3.63 (0.19)5.75 (0.09)0.39 (0.04)8.6 (0.23)4.9 (0.31)
 DNA (ng g−1)1393.3 (145)523.2 (97)1460.2 (88.5)1073.3 (45.5)1373.3 (49.9)147.3 (98.1)1873.3 (105.8)1493.3 (99.1)
 4°C respiration (μg CO2- C g−1 h−1)9.40 (0.18)6.04 (0.24)15.91 (0.22)11.22 (0.53)7.10 (0.66)3.25 (0.63)105.45 (2.5)12.6 (0.55)
 4°C glutamate-IR (μg CO2- C g−1 h−1)14.92 (0.49)8.87 (0.44)19.61 (0.31)15.44 (0.25)15.85 (0.76)4.39 (0.45)187 (4.89)16 (2.24)
 4°C crystal violet degradation (μg h−1)14 (0.48)8 (1.51)11.25 (0.75)12.25 (1.26)14.25 (1.44)6.25 (2.06)12.87 (0.89)14.75 (0.39)
 4°C respiration (μg CO2- C g−1 h−1)9.20 (0.27)5.97 (0.47)16.37 (0.46)11.89 (0.67)7.60 (0.49)3.01 (0.71)101.10 (4.5)12.10 (0.41)
 4°C DC (108 g−1)4.04 (0.25)1.21 (0.11)5.21 (0.32)3.58 (0.35)5.84 (0.49)0.45 (0.06)8.1 (0.35)5.0 (0.21)
 4°C DNA (ng g−1)1360 (101)490 (56)1590 (109)1040 (75)1290 (78)143 (104)875.6 (87)531.7 (114)
 DC (108 g−1)4.31 (0.35)1.11 (0.15)5.01 (0.14)3.41 (0.23)5.63 (0.41)0.34 (0.05)1.2 (0.13)0.6 (0.02)

Results and discussion

Microbial environments, abundance and 4 °C activities

The physical and chemical characteristics of tested soils (Table 1) showed a strong region-geology dependence (Fig. 1), mirroring the differences in the soil properties between the Himalayan and the temperate soils (for statistics, see Table 3). The NM-MDS analysis of soil characteristics where the final stress, instability and the number of iterations were 5.36, 0.00001 and 83, respectively (Fig. 1) (P<0.02), indicated reliable ordination (McCune et al., 2002). The proportions of variance represented by each axis, based on the r2 between distance in the ordination space and distance in the original n-dimensional space, were 0.579 and 0.402. The first ordination axis was highly correlated (Pearson's r) with WHC, clay, Corg and the sugar content (r>0.942), while the second axis was highly correlated to sand and the NO3 content (r>0.814) (P<0.02). Both axes clearly separated the Himalayan soils from the two temperate soils, with the exception of the 5200 m soil, which was, in some aspects, similar to temperate mineral soil. Exclusion of marsh soil with the highest soil carbon from analysis showed that clay, silt, sand, Ntot, NO3, MWI and sugar content separated soils with comparable carbon content, confirming the observed differences in soil characteristics (Table 1) between the Himalayan and the temperate soils. Subsoils could not be used as alternative controls lacking FT history as the Himalayan sampling sites were shallow (Table 1). In addition, surface and subsurface horizons generally differ in pedological, physicochemical features and interactions with the atmosphere, which are all likely to affect the mechanisms and biological actors involved (Bullock et al., 1988; Salome et al., 2010).

Figure 1.

 NM-MDS ordination of soil physical and chemical parameters: (◆) 6000 m, (▪) 5800 m, (▴) 5600 m, (×) 5400 m, (▵) 5200 m, (•) 5000 m, (○) mineral, (□) marsh. Similarity index: Bray-Curtis. Final stress, instability and the number of iterations in NM-MDS were 5.36, 0.00001 and 83, respectively, and indicated a stable 2D solution (P<0.02). The proportions of variance represented by each axis, based on the r2 between distance in the ordination space and distance in the original n-dimensional space, were 0.579 and 0.402 for axis 1 and axis 2, respectively.

Table 3.   Repeated-measure anova table on F-values and degrees of freedom (d.f.) on the effect of ‘region-geology’ [Himalayan (granite); Central-European (carbonate)] for soil parameters
 Between-subject effects
All soils, F (d.f. 1)Without marsh
soil, F (d.f. 1)
  • The asterisks describe the significance of the relationship:

  • ***


  • **


  • *


Soil parameters
 Clay (%)53.43***69.27***
 Silt (%)22.93***55.79***
 Sand (%)181.50***85.88***
 Corg (%)19.14***3.47
 Ntot (%)27.07***10.21**
 NH4+ (μg g−1)0.24610.5
 NO3 (μg g−1)6.5477*26.42***
 MWI (DOC)51.4379***25.93***
 Sugars (μg g−1)57.5296***23.94***

RDA showed that neither the altitude nor the slope alone could explain significantly the variability of the measured soil parameters of the Himalayan soils (Table 1). The habitat types alone were significantly correlated to the Ntot and NH4+ content, explaining only 36.87% of data variability (P=0.002), but were not correlated with any of the other soil variables measured. It is conceivable that the effects of the 1000 m altitude gradient were deflected by the environmental disturbances present in this region such as frost-heaving, solifluction, pervection, avalanches, eolian deposition (unpublished data) and translocation of particles and snow, each exerting a variable impact (Ono et al., 1999; Ishikawa et al., 2001; Hidy, 2003; Barros et al., 2004; Carson et al., 2009; Schütte et al., 2009). Thus, considering the environmental conditions of the Himalayan soils, the effects of the altitude gradient measured on bulked (this study) or separately measured transect samples (unpublished data) were not linear and predictable, indicating a complex site history that could not be regarded as a simple succession after glacier or snow retreat in the High Arctic (Schütte et al., 2009).

Incorporation of the biological data (Table 2) into RDA analyses in relation to the sampling localities, habitat types and soil physical–chemical parameters (Table 1) showed that the 1000 m altitude gradient did not explain any of the measured biological variables significantly. The best predictor of variability in the high-altitude microbial abundance and microbial activity at 4 °C was soil organic carbon, explaining 88% and 74% of the data variability (P=0.002), respectively. The other environmental parameters tested in this study (Table 1) did not have significant effects. When mineral and marsh soils were included in a separate RDA (random shifts enabled), the variability in the abundance and activity of microbial communities in the Himalayan and temperate soils was most strongly explained again by soil carbon and C/N ratio (68% and 61% of the data variability, P=0.002 and 0.028, respectively). The tight correlation of soil carbon content and biomass was also observed in a large number (n>150) of distinct Central European soil habitats (Weigand et al., 1995; Lentzsch et al., 2005) and also in previous studies on European and North American alpine soils (Ley et al., 2001; Sigler et al., 2002; Tscherko et al., 2003, 2004). This contrasts with the findings on the Arctic and Antarctic soils and deglaciated sediments that appear to be governed by distinct environmental parameters: soil texture and pH (Kaštovskáet al., 2005; Stibal et al., 2007), moisture and temperature (Bekku et al., 2004) were found to be significantly more important than soil carbon (Bekku et al., 1999; Beyer et al., 2000).

Resistance to FTC

The high frequencies of FTC with low cooling and warming rates such as the reproduced Himalayan FTC used here (0.04 °C min−1) are particularly damaging for microorganisms (Walker et al., 2006) due to extracellular ice formation, leading to the concentration of soil solutes and resulting in protein denaturation, membrane damage, cell dehydration and lower metabolic rates (Nedwell, 1999; Rodrigues & Tiedje, 2008). However, the microbial abundance in the noncontinuous permafrost soil microcosms was not significantly affected after 50 reproduced Himalayan FTC (Table 2). In contrast, more than a sevenfold decrease in abundance was observed in the two temperate soils (Table 2), indicating that a significant fraction of at least 45–48% (P<0.05) was affected by FTC. This is in line with previous reports on the effects of FT conditions that were beyond the range of natural FTC (Soulides & Allison, 1961; Henry, 2007; Elliot & Henry, 2009). The actual fraction of affected microbial community could be even higher as we cannot rule out the possibility that some of the cells stained after FT were dead.

Unfavorable conditions such as FT generally impair active metabolism; therefore, basal respiration at 4 °C was taken as a measure of general microbial activity (Hartley et al., 2008). Following the FT experiment, basal respiration observed in the Himalayan microcosms was 15% lower (Fig. 2; Table 3) (not significant at P=0.05) in contrast to more than a 90% decrease (significant at P=0.05) in both temperate marsh and mineral soils. The control 4 °C respiration rates of all soils used in this study did not differ significantly from the initial respiration rates throughout the experiment (Table 2; Stres et al., 2008). Sufficient carbon accessible at 4 °C was therefore available despite the generally reduced affinities for substrates below the optimum growth temperature (Nedwell, 1999; Gonzalez-Quinonnes et al., 2009). Recent studies have also shown that soil microbial respiration in various soils does not acclimate to low temperatures (Hartley et al., 2008; Bárcenas-Moreno et al., 2009), not even at the time scale of months. Thus, the remaining respiration observed after the completion of FT experiment in mineral and marsh microcosms (5.4% and 6.1% of the initial 4 °C respiration, respectively) (Table 2, Fig. 2) must have derived from the FT surviving and cold active portion of the original temperate microbial communities. The resistance of the Himalayan communities was congruent with the noncontinuous permafrost conditions and the long-term positive selection for species with higher affinities for substrates at low temperatures and frequent FT (Nedwell, 1999; Ishikawa et al., 2001; Schimel et al., 2007).

Figure 2.

 Relative basal respiration rates obtained from the Himalayan and temperate soils sampled in 2002 as a function of the reproduced Himalayan FTC. Time zero values preceding the FT experiment are reported in Table 2. (◆) 6000 m, (▪) 5800 m, (▴) 5600 m, (×) 5400 m, (▵) 5200 m, (•) 5000 m, (○) mineral, (□) marsh. The error bars indicate 1 SD.

Differences in the soil characteristics could generate unique differences in the FT conditions experienced at the scale of microbial cells like the phase changes from liquid to solid (and vice versa), freezing point depression, the thickness of adsorbed water films and the distribution of nucleation agents (Rodrigues & Tiedje, 2008). However, the overall temperature fluctuations in experimental microcosms determined in the temperature range of −4 to +4 °C were virtually indistinguishable (Fig. S4), indicating similar transfer of heat to different soils studied at the laboratory scale. Further, the amplitude, frequency and the top-down direction of FTC closely reproduced the environmental conditions of the high-altitude Himalayan soils, but were fully out of range for the two temperate soils. This was also mirrored in the lack of detectable aggregated structure in the Himalayan soils, whereas 69% and 81% decreases in soil macroaggregates (d>150 μm) was observed in mineral and marsh microcosms after the FT experiment, respectively. The disrupting forces inside the aggregates were most probably a result of ice crystals expanding in pores between particles and breaking the particle–particle bonds, resulting in aggregates small enough to pass through a 150-μm screen.

Some of the biological findings presented here (Fig. 3; Fig. S5) are intriguing in their own right, although a detailed follow-up is beyond the scope of the present paper. We note that the 10 Himalayan FTC resulted in decreased cell abundance, but an unaltered soil DNA content in temperate soils. The apparent slow degradation of soil DNA content (Levy-Booth et al., 2007; Nielsen et al., 2007) during the subsequent 6-day incubation at 4 °C could represent degradation of DNA from damaged cells (Henry, 2007). Moreover, coextraction of DNA from annihilated biomass could conceal the loss of a numerically significant portion of the microbial community in response to more extreme reproduced FT. Indeed, in cases where authors used more extreme FT than observed in nature, the lack of an observed community change could in fact be due to the presence of slow to degrade DNA from annihilated biomass or due to true community resistance. The currently available approaches do not allow for effective removal of the free DNA from complex environmental samples due to interferences between chemicals, DNA, particles and organic matter (Wagner et al., 2008). However, this is critical for deciphering whether or not the physiological properties measured in experiments (as presented in our study) are significantly linked to the measured community change (or the lack of it) as was described before (Yergeau & Kowalchuk, 2008; Männistöet al., 2009).

Figure 3.

 DNA content of the mineral (○) and marsh (□) soils following 10 reproduced Himalayan FTC and subsequent 4°C incubations at days 2 and 6. The error bars indicate 1 SD. Only the data for 5000 m (•) and 6000 m (◆) soils are presented for clarity (see Fig. S5 for details).

MicroResp physiological responses to FTC

In order to obtain detailed information on the initial community-level physiological responses and their differences in response to FT in different soils, the four soils (5200, 6000 m, mineral and marsh) sampled in 2005 were exposed to a 60-day FT experiment. More than a 90% decrease in the 4 °C respiration was reproduced again for the two temperate soils (Fig. 4), consistent with the first FT experiment (Fig. 2). The NM-MDS analysis of community-level metabolic properties in response to FT (Bray-Curtis distance, multiple random starts and the MC permutation test) resulted in two-dimensional ordination, with final stress (4.05651), instability (0.00001) and the number of iterations (107) (Fig. 5) (P<0.02) indicating reliable ordination (McCune et al., 2002). The proportions of variance represented by each axis, based on the r2 between distance in the ordination space and distance in the original n-dimensional space, were 0.61 and 0.373 for axis 1 and axis 2, respectively. The first ordination axis was highly correlated (Pearson's r) with arabinose, pantotheic acid and serine (r>0.923), while the second axis was highly correlated to the amino acids [glutamate, alanine, arginine, histidine (r>0.914)] and carboxylic acid citrate (r=0.918).

Figure 4.

 Relative basal respiration rates obtained from the Himalayan and temperate soils sampled in 2005 as a function of the number of the reproduced Himalayan FTC. Time zero points represent the starting values preceding FT experiments that did not differ significantly from those reported in Table 3. (◆) 6000 m, (▵) 5200 m, (○) mineral, (□) marsh. The error bars indicate 1 SD.

Figure 5.

 NM-MDS ordination of the mean of the standardized metabolic responses of the ecosystems sampled in 2005: (▵) 5200 m, (◆) 6000 m, (○) mineral and (□) marsh soils over 60 days of a diurnal FT cycling experiment. Similarity index: Bray-Curtis. Final stress (4.05651), instability (0.00001) and the number of iterations in NM-MDS (107) indicated a stable 2D solution (P<0.02). The proportions of variance represented by each axis, based on the r2 between distance in the ordination space and distance in the original n-dimensional space, were 0.61 and 0.373 for axis 1 and axis 2, respectively. The numbers next to symbols designate the number of FTC.

The initial metabolic profiles of mineral and marsh soils were almost identical (Fig. 5) and were clearly separated from the 6000 and 5200 m profiles by the first and second ordination axes. The FT experiment caused the metabolic profiles of mineral and marsh soils to change significantly in the first 40 FTC. On the other hand, the Himalayan soil microcosms exhibited only minor adjustments in their metabolic profiles. After roughly 40 FTC, the profiles of mineral and marsh soils stabilized (Fig. 5), but remained distinct, and did not converge with those of the Himalayan soils. This suggests that the surviving and cold-active microorganisms that were initially masked by the more abundant, but FT-sensitive species actively adjusted their metabolic profiles in the two temperate soils in response to novel environmental conditions (Nedwell, 1999; Schimel et al., 2007). Temperate soils initially had a significantly higher response to organic acids and carbohydrates than the Himalayan soils, but shifted toward the use of amino acids during the FT experiment (Fig. 5). This is surprising as the carbon and sugar content in addition to MWI of temperate soils indicated a higher availability of light carbon (Table 1). The generally decreased uptake of the carbohydrate–organic acid pool and stimulated uptake of amino acids are indicative of the stimulated and active microbial physiological response to FT according to Lipson & Monson (1998), who linked the reduced concentration of water-extractable amino acids in alpine tundra in response to FT to the activity of the FT-resistant microbial communities. Jones et al. (2004) also observed rapid amino acid cycling (i.e. the rate of turnover of the amino acid pool between the constituents of the soil microbiota) in Arctic and Antarctic soils upon thaw, while Schmitt et al. (2008) observed the disappearance of plant and microbial sugars upon soil freezing due to soluble organic matter stabilization through various chemical alterations. The latter could have already taken place in the Himalayan soils. The observed FT-related metabolic adjustments in the two temperate communities are thus most probably a consequence of aggregate decomposition and cell disruption-derived carbon availability exacerbated by low-temperature cold starvation (Nedwell, 1999). The lack of metabolic adjustments in response to FT of the Himalayan soils suggests the resistance of the Himalayan soil microcosms to the effects of the reproduced Himalayan FT. Adjustments in the metabolic profiles observed in the cycled temperate microcosms after 40 FTC were also of a magnitude similar to those of the Himalayan soils (Fig. 5). This suggests that the environmentally relevant amplitude (−4/+4 °C), the rate (0.04 °C min−1) and the number (n=50–60) of FTC yielded microbial communities significantly less susceptible to the reproduced Himalayan FT in the two temperate soils in the laboratory experiment. On the other hand, differences in the metabolic profiles observed in the control temperate and Himalayan soils incubated at 4 °C (not shown) were consistent with the recently described minor effects of temperate soils' storage at 4 °C on their metabolic properties using the MicroResp approach (Gonzalez-Quinonnes et al., 2009).


Our study shows that the carbon content in the high-altitude and temperate soils explains significantly the largest fraction of variability in microbial abundance and activities at a low temperature (4 °C) despite the profound differences in other soil characteristics, site history and geology. The soils originating from a temperate region and from the Himalaya responded differently to the diurnal FTC, indicating that the initially observed responses to FT cycling are not common for both ecosystem types, most probably due to the different environmental histories. The microbial communities from the high-altitude Himalayan noncontinuous permafrost soils were largely FT resistant and may thus include microbial populations active at the ambient Himalayan temperature fluctuations. In contrast, the two temperate soils subjected to FT showed an initial decrease in microbial abundance and activity. However, the remaining community was capable of functional reorganization at low temperatures that resulted in its tolerance to additional FT, activity at low temperature and modified, but rather stable community-level physiological traits.


We are indebted to late Ivan Mahne, head of the Chair for microbiology, for the ideas and ongoing support during the 2002–2006 period. We acknowledge the efforts of Andrej Štremfelj and late Miha Valič, who provided visual material, additional soil temperature and eolian deposition measurements in the years 2005–2007. We thank Fiona Moore and Stephen Chapman (Macaulay Institute) for discussions during the initial adaptation of MicroResp and Teiji Watanabe (Hokkaido University) for releasing the complete 1998–2002 temperature measurements. This research was supported by the Slovenian Research Agency (ARRS) Agreement No. P0-0562-0481, ESF No. 3311-05-837013 COST-856 and STSM to B.S.