This study investigated the effects of environmental variables on the bacterial and fungal communities of the Beilu River (on the Tibetan Plateau) permafrost soils with different vegetation types.
This study investigated the effects of environmental variables on the bacterial and fungal communities of the Beilu River (on the Tibetan Plateau) permafrost soils with different vegetation types.
Microbial communities were sampled from meadow, steppe and desert steppe permafrost soils during May, June, August and November, and they were analysed by both pyrosequencing and the use of Biolog EcoPlates. The dominant bacterial and fungal phyla in meadow and steppe soils were Proteobacteria and Ascomycota, whereas Actinobacteria and Basidiomycota predominated in desert steppe soils. The bacterial communities in meadow soils degraded amines and amino acids very rapidly, while polymers were degraded rapidly by steppe communities. The RDA patterns showed that the microbial communities differed greatly between meadow, steppe and desert steppe, and they were related to variations in the soil moisture, C/N ratio and pH. A UniFrac analysis detected clear differences between the desert steppe bacterial community and others, and seasonal shifts were observed. The fungal UniFrac patterns differed significantly between meadow and steppe soils. There were significant correlations between the bacterial diversity (H′) and soil moisture (r = 0·506) and C/N (r = 0·527). The fungal diversity (Hf′) was significantly correlated with the soil pH (r = 0·541).
The soil moisture, C/N ratio and pH were important determinants of the microbial community structure in Beilu River permafrost soils.
These results may provide a useful baseline for predicting the variation in microbial communities in response to climate changes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Permafrost underlies ~26% of terrestrial ecosystems and is estimated to contain around 15% of the world's soil carbon (Post et al. 1982). However, there is a growing risk of losing this soil carbon due to the acceleration in decomposition of organic soil matter under a warming climate (ACIA 2005). The micro-organisms found in permafrost are important drivers of the decomposition and mineralization processes in permafrost soil carbon stocks, and they have recently become a focus of attention (Mannisto et al. 2007; Chu et al. 2010; Chu et al. 2011; Mackelprang et al. 2011). However, we understand little about how climate changes might affect the microbial community composition of permafrost soils. Thus, it is particularly important to collect baseline information on the structural composition of these communities and to determine the factors that influence their structure and function before changes occur.
Studies in the Arctic and Antarctic have shown that the microbial communities in permafrost soils respond to variations in the soil's environment. For example, the vegetation type has a strong effect on soil microbial community structure through the combined effects of nutrient availability and the physical condition of the soil (Neufeld and Mohn 2005; Yergeau et al. 2007; Chong et al. 2010; Chu et al. 2011). The bacterial communities found in Arctic shrub soils differed from those in tussock soils due to variations in the bioavailable C fraction (Wallenstein et al. 2007). Changes in plant cover and morphology may have an impact on the transmission of biologically damaging UV-B to the soil surface, which may affect microbial diversity (Zepp et al. 2003; Hughes et al. 2006). Further studies have suggested that vegetation enhances the moisture and thermal retention of soil microhabitats (Harris and Tibbles 1997; Yergeau et al. 2007). In addition, the bacterial and fungal community composition of Alaskan tundra soil has shown significant seasonal changes at the species level (Wallenstein et al. 2007). Shifts within the bacterial community structure and functions in both carbon utilization and turnover rates were observed with the variations in the redox potential determined by the water content (Wagner et al. 2009). In addition, soil moisture, pH, bioavailable C and soil C/N have also been shown to be correlated with variations in bacterial communities (Mannisto et al. 2007; Wallenstein et al. 2007; Yergeau et al. 2007; Chu et al. 2011). These observations suggested a close association between the microbial community and soil characteristics in response to changes in climate.
The Qinghai–Tibet Plateau has an area of about 2·5 million square kilometres and an average elevation of >4000 m. It is one of the main low-latitude permafrost regions in the world (Cheng 1998). The permafrost found on the Qinghai–Tibet Plateau is warm and thin compared with the high-latitude permafrosts of North America and Russia, which makes it more sensitive to changes in climate (Cheng 1998). The soil organic carbon (SOC) reservoirs found in this area comprise 30–40 × 109 tonnes, accounting for more than 20% of the SOC storage in China and 2–3% of global soil carbon reservoirs (Wang et al. 2002; Wu et al. 2012). These SOC reservoirs are at risk due to the effects of global warming. These soils are also alkaline, in contrast to acid soils in the Arctic (Neufeld and Mohn 2005; Mannisto et al. 2007; Chu et al. 2010). Recent investigations of the Tibetan permafrost tundra have shown that the moisture of the soil is correlated with the vegetation cover and the SOC (Wang et al. 2008; Wu et al. 2012). However, the micro-organism in this unique permafrost soil remains relatively unexplored except for a few culturable diversity studies (Bai et al. 2006; Zhang et al. 2007). Thus, the objectives of this study were to determine the microbial community distributions among different vegetation types and to analyse their correlations with soil characteristics in the Beilu River permafrost soils of the Tibetan Plateau, using pyrosequencing and Biolog EcoPlate analysis.
On 10 May, 10 June, 10 August and 10 November 2009, 0–20 cm soil samples were collected from four well-separated sites (1000–3000 m apart) with different types of vegetation at the Beilu River Long-term Permafrost Research Station (34°49·602′N, 92°55·689′E, 4885 m above sea level) (Chen and Lin 2012) (Fig. 1). This region is underlain by continuous permafrost and an active layer develops during the thawing season, reaching a maximum depth of 3 m (Niu et al. 2005). Two alpine meadow sites (meadow: M1 and M2), one alpine steppe (steppe: S) and one alpine desert steppe (desert steppe: DS) sites were sampled. The meadow sites were dominated by Kobresia pygmaea, Kobresia humilis, Kobresia capilifolia and Polygonum viviparum, covering about 90% of the sites. Steppe sites were dominated by Stipa glareosa, Christolea crassifolia and Oxytropis glacialis, covering about 70% of the site. Desert steppe sites were dominated by Stipa purpurea and Festuca sinensis, covering about 40% of the site. On each sampling date, three soil cores (20 cm deep, 5 cm in diameter) were collected within a 2 m2 area of each site and composited together as a single replicate sample (Yergeau et al. 2007; Chong et al. 2010; Chu et al. 2011). Samples for each season were stored in a cooler with ice packs during transfer to the laboratory. The DNA was extracted within 24 h of processing. The physicochemical properties of the soil samples were determined by standard methods, as described previously (Chu et al. 2011).
DNA was extracted from 2·0 g soil using a MoBio PowerSoil DNA Isolation Kit (MoBio Laboratories, Carlsbad, CA, USA) and MP Bio FastDNA® SPIN Kit (MP Biomedicals, Santa Ana, CA, USA) and purified using a Sepharose 4B column (Sigma-Aldrich, St Louis, MO, USA). Four extracted DNA samples were pooled to obtain a diversity of micro-organisms (Wallenstein et al. 2007; Mareckova et al. 2008; Delmont et al. 2011). Partial sequences of the 16S rRNA gene including the variable V3 region, and the 18S rRNA gene including the variable V9 region, were amplified from the DNA using two primers: V3F (5′-nnnnnnnn CCAGACTCCTACGGGAGGCAG-3′) and V3R (5′-nnnnnnnnCGTATTACCGCGGCTGCTG-3′) (Muyzer et al. 1993), and V9F(1380f: 5′-nnnnnnnn CCCTGCCHTTTGTACACAC-3′) and V9R (1510r: 5′-nnnnnnnn CCTTCYGCAGGTTCACCTAC-3′) (Amaral-Zettler et al. 2009). The eight lowercase letters in these sequences were designed as a unique tag for sample identification when multiple samples were analysed in parallel. The PCR amplification was performed as described previously (Delmont et al. 2011). The PCR products were examined by electrophoresis in a 1·2% (w/v) agarose gel, stained with ethidium bromide in TAE buffer and purified using a Gel/PCR DNA Fragment Extraction Kit (Geneaid, CA, USA). The concentrations of the PCR amplicons were measured using a Fluoroskan Ascent with Quant-iT PicoGreen dsDNA reagent (Invitrogen, Carlsbad, CA, USA) as the fluorescent dye. The sample was mixed with other PCR products and submitted to the Chinese National Human Genome Center in Shanghai for pyrosequencing using the GS FLX platform (Roche).
All of the sequences from the pyrosequencing reads were preprocessed to remove ambiguous sequences with lengths of <100 nucleotides and >1000 nucleotides, excluding the primers. This filtering step also eliminated sequences with more than two ambiguities. Reads with the corresponding barcodes were extracted and used in the analyses. All of the sequences were verified by the Ribosomal Database Project II (RDP Release 10) using Chimera Check (http://rdp.cme.msu.edu/index.jsp), and all chimeric sequences were discarded. The remaining pretreated sequences obtained by pyrosequencing were aligned using Nast (http://greengenes.lbl.gov/ NAST), according to a criterion of 75% identity for >50% of the sequence length (DeSantis et al. 2006). The distance matrix was computed using Arb (Ludwig et al. 2004). Sequences with a similarity >97% were classified into a single operational taxonomic unit (OTU) based on the distance matrix (Schloss and Handelsman 2005) using Dotur (http://schloss.micro.umass.edu/software). Taxonomy classification was performed using BlastN with an e-score cut-off of 0·01 against version 10.18 of the RDP database and the Greengenes web service (http://greengenes.lbl.gov/cgi-bin/nph-classify.cgi). The diversity indices were calculated based on the OTU analysis.
Biolog Eco-microplates (Biolog Inc., Hayward, CA, USA) were used to determine the spectrum of carbon sources utilized by the bacterial community in different seasons (May, August and November) at the four study sites (M1, M2, S and DS) in aerobic conditions. The EcoPlates allowed intraplate replication because each 96-well plates contained 31 different carbon substrates and an additional blank in a triplicate array, according to the methods of Insam (1997). Micro-organisms were extracted from 10 g soil samples using 90 ml of Ringer's solution (0·25%). The soil suspensions were shaken for 30 min on a reciprocal shaker. After extracting the micro-organisms, the cell densities were calculated by direct cell counts of dichlorotriazinyl aminofluorescein (DTAF)-stained samples, as described previously (Kobabe et al. 2004). The Ecoplate wells were then inoculated with 150 μl of a suspension containing a cell density of 5 × 107 cell ml−1. The plates were incubated in the dark at 10°C, and colour development was measured at a wavelength of 590 nm every 12 h for 180 h using an automated microtitre plate reader (Biolog).
The seasonal and spatial differences in the soil characteristics and frequency of the major microbial groups determined by pyrosequencing, which were expressed as the percentage of each group in the libraries of each soil replicate, were analysed using a one-way analysis of variance (anova) with Spss (ver. 17.0; SPSS Inc., Chicago, IL, USA). Correlations between the microbial community structures and soil variables were analysed by distance-based redundancy analysis (db-RDA)in Canoco (ver. 4·5). The bacterial and fungal models produced using the parameters shown in Fig. 4 were highly significant with 499 permutations test (test of the significance of the first and all canonical axes: bacteria, P < 0·01; fungi, P < 0·05). A Monte Carlo test (499 permutations) based on the db-RDA was used to assess the effects of each variable. The bacterial and fungal community comparison was statistically analysed at the OTU level using UniFrac (Lozupone and Knight 2005).
Based on the Biolog data, the absorbance measurements of six major substance classes (amines, amino acids, carbonic acids, carbohydrates, phenolic compounds and polymers) were compared among M1, M2, S and DS by anova. The absorbance value of each well at 72 h was selected to minimize the effects of different inoculum densities in the plates.
Across all soil samples, 13019 high quality 16S rRNA gene sequences (621–1214 sequences per sample) and 24359 18S rRNA gene sequences (479–5014 sequences per sample) were obtained, and 92·3 and 76·2% of those sequences could be classified, respectively. The 16S rRNA gene sequences were clustered into 5386 OTUs, which belonged to 19 phyla and 30 classes. Eight phyla (Proteobacteria, Actinobacteria, Acidobacteria, Bacteroidetes, Firmicutes, Gemmatimonadetes, Chloroflexi and TM7) were considered abundant with sequence frequencies of >1% (Fig. 2a), whereas 11 phyla (Nitrospirae, Chlorobi, Verrucomicrobia, WCHB1-60, Cyanobacteria, Armatimonadetes, Candidate Division WS3, JL-ETNP-Z39, Planctomycetes, TA06 and SM2F11) were considered low abundance with sequence frequencies of <1% (Fig. 2b). The classified fungal sequences were clustered into 832 OTUs, which represented seven phyla and 19 classes. The dominant fungal phyla were Ascomycota (38·66%), Basidiomycota (21·76%) and Mucoromycotina (7·67%). Chytridiomycota, Kickxellomycotina, Ascomycota and Fungi incertae sedis were also identified (Fig. 3). In addition, each sample contained a large number of bacterial and fungal sequences that could not be classified, even at the phylum level (Figs 2 and 3).
At the phylum level, the RDA patterns indicated that there were distinct differences in bacterial and fungal community compositions between vegetation types for M, S and DS (Fig. 4a,b). Although most of the more numerous microbial groups were present in all samples, community structure clustered according to the vegetation type because of differences in the relative abundance and differences in the composition of the less numerous phylotypes. The communities of bacteria and fungi in DS were clearly different from those in S, and those from M which were relatively similar and clustered together, explaining ~65 and 54% of the total variation along the first axis, respectively. The S and DS bacterial communities were different from those in M along the second axis, which accounted for ~22·4 and 29% of the variation, respectively (Fig. 4a,b).
The relative abundances of the dominant bacterial taxa varied among the M, S and DS vegetation types. The dominant taxon in M was Proteobacteria (Fig. 2a), which was followed in relative abundance by Acidobacteria, Actinobateria and Bacteriodetes. In S, Proteobacteria and Acidobacteria abundance decreased, whereas that of Bacteriodetes increased greatly compared with M. Nevertheless, the DS soils were dominated by Bacteriodetes and Acidobacteria, followed in relative abundance by Proteobacteria and Actinobateria. In general, Proteobacteria abundance decreased gradually from M through S to DS soils, that is, from 42·06 to 33·6% and 18·43%, respectively. By contrast, the Bacteroidetes, abundance increased from 4·59% in M, to 14·16% in S and 18·84% in DS. The anova test indicated that there were significantly higher abundances of Proteobacteria in M and S compared with DS soils (P < 0·01), and a significantly lower abundance of Bacteroidetes in M than S (M vs S, DS; P < 0·05). However, there were few major differences in the bacterial community compositions at the phyla level between sampling dates.
In contrast to the RDA patterns, the bacterial UniFrac analysis suggested that there were significant seasonal community shifts, that is, on the first axis, which explained ~42% of the total variation (Fig. 5). The May and June communities were located very close together in the same quadrant, suggesting a similar bacterial community composition during these 2 months. The August and winter bacterial communities were located in different quadrants. Among sites, however, the DS bacterial communities were located separately in another quadrant, explaining ~17·4% of the variation along the second axis.
The most abundant fungal division in M was Ascomycota, followed in relative abundance by Basidiomycota, Blastocladiomycota, Mucoromycotina and Chytridiomycota. By contrast, the DS soils were dominated by Basidiomycota, followed by Ascomycota, Mucoromycotina, Chytridiomycota and Kickxellomycotina. The S fungal communities were dominated by Ascomycota and Mucoromycotina, followed by Basidiomycota, Chytridiomycota and Fungi incertae sedis (Fig. 3). The anova test indicated that Basidiomycota differed significantly between M, S and DS (P < 0·01) and that Mucoromycotina were significantly more abundant in S soils (P < 0·01). No significant differences in fungal community compositions at the phyla level were detected in different months. The UniFrac analysis showed that there were clear fungal community differences between sites, with M vegetation-type clusters differing from the S and DS vegetation types (Fig. 6).
Analysis of the physicochemical properties of the Beilu River permafrost soils showed that the soil was alkaline, ranging from pH 8·21 to 9·58 (mean = 8·62; SD = 0·33; n = 16) (Table 1). The soil moisture, pH, total nitrogen and cell counts differed significantly among the M, S and DS vegetation types. In particular, the soil moisture was negatively correlated with pH (r = −0·508, P < 0·05). The highest average soil moisture and lowest average pH were detected in M soils, followed by S and DS soils. No significant differences were detected between the different months (Table 1). In addition, the Shannon Indices (H′) calculated for the 16S rDNA sequences ranged from 4·01 to 5·87, which were lowest for DS samples and highest for M samples. The H′ was significantly correlated with C/N (r = 0·527, P < 0·05) and water content (r = 0·506, P < 0·05). The Shannon Indices (H′f) calculated for the 18S rDNA sequences ranged from 1·4 to 3·87, which were lowest for M and the highest for DS samples. The H′f was significantly correlated with soil pH (r = 0·541, P < 0·05) (Tables 1 and 2).
|Site and seasons||Total C (%, w/w)||Total N (%, w/w)||H2O (w/w)||pH||CFU (×105 g−1 dw)||Cell counts (×108 cell g−1 dw)||H′||H′(F)||C/N|
|Meadow||2·51 (0·19)||0·0805 (0·007)a||23·10 (1·92) a||8·51 (0·24) a||4·65 (3·09)||4·12 (0·35) a||5·65 (0·33)||2·33||31·28 (3·09)|
|Meadow||1·81 (0·03)||0·0602 (0·0008)b||28·45 (2·91) c||8·44 (0·13) a||3·56 (2·61)||3·65 (0·22) ab||5·41 (0·65)||2·12||30·78 (0·85)|
|Steppe||2·09 (0·13)||0·0663 (0·009)b||20·25 (2·69) a||8·59 (0·12) a||9·20 (5·43)||3·00 (0·37) c||5·23 (0·26)||3·05||31·52 (2·95)|
|Desert||2·13 (0·22)||0·0705 (0·016)a||11·57 (2·08) b||8·99 (0·29) b||1·51 (0·79)||3·12 (0·32) bc||4·77 (0·52)||3·33||30·21 (2·52)|
|May||1·88 (0·37)||0·0613 (0·014)||19·00 (4·86)||8·61 (0·17)||3·05 (3·09)||3·25 (0·79)||5·18 (0·56)||2·53||31·24 (4·56)|
|June||1·95 (0·34)||0·0679 (0·023)||19·87 (6·32)||8·62 (0·11)||4·27 (2·92)||3·48 (0·49)||5·33 (0·72)||2·68||31·05 (7·02)|
|August||1·97 (0·43)||0·0700 (0·026)||21·00 (5·29)||8·41 (0·21)||9·45 (5·18)||3·62 (0·15)||5·09 (0·58)||2·35||29·87 (5·29)|
|November||2·13 (0·31)||0·0708 (0·021)||23·50 (5·52)||8·89 (0·38)||2·15 (1·14)||3·22 (0·59)||5·45 (0·20)||3·27||31·56 (5·52)|
|Soil characteristics||H′||H′ (F)||H2O||pH||Total C (%,w/w)||Total N (%,w/w)||C/N||Cell Counts||CFU|
|Total C (%,w/w)||−0·047||−0·401||0·357||0·079||1|
|Total N (%,w/w)||−0·350||−0·324||0·233||0·020||0·868b||1|
|C : N||−0·527a||0·013||0·075||−0·061||−0·380||−0·776b||1|
The Monte Carlo test showed that the total percentage variance explained by the measured variables was 64·2 and 53·2% in bacterial and fungal community patterns, respectively (Fig. 8). A breakdown of the proportions of explained variance attributed to each soil variable is shown in Fig. 8. Across all samples, soil moisture was the most important soil parameter underlying the variations in the bacterial and fungal community structures. The soil C/N ratio was the second most important limiting factor for bacterial communities, while pH was the second most important soil factor affecting the fungal community structure. Overall, the M soil microbial community cluster was associated with relatively high soil moisture and C/N levels, the DS soil microbial community cluster was associated with relatively high soil pH levels, and the S community cluster was associated with a relatively high C/N ratio. Only the relative abundance of Proteobacteria increased significantly with the soil moisture in the soils analysed (P < 0·05) (Table 2).
The substrate turnover in the M, S and DS vegetation types for the six major substance classes (amines, amino acids, carbonic acids, carbohydrates, phenolic compounds and polymers; Insam 1997) showed that amines, amino acids and carbonic acids/carbohydrates were significantly more rapidly degraded by M communities, whereas phenolic compounds and/or polymers were significantly more rapidly degraded by S/DS communities (P < 0·05) (Fig. 7).
Our study of soil microbial communities in the Tibetan Plateau permafrost tundra detected a distinct clustering of bacterial and fungal communities according to M, S and DS vegetation types, while the soil moisture, pH and C/N, which were concomitant with vegetation types, were found to be important determinants of microbial community profiles in Beilu River permafrost soils (Table 1, Figs 2 and 3). However, the weighted UniFrac analyses detected seasonal shifts in bacterial OTU levels, which suggests that the micro-organisms present here may have specialized physiologies adapted to the environmental conditions associated with particular seasons (Fig. 5).
When the soil physicochemical parameters were considered individually, soil moisture was shown to be the most important factor in determining microbial diversity in Beilu River permafrost soils (Figs 4 and 8). Previous studies have shown that the moisture content of Tibetan permafrost soils was positively correlated with SOC stocks and vegetation cover (Wang et al. 2008; Wu et al. 2012). In addition, the labile light fraction of the organic carbon content stored in the M soil was higher than that in the S soil, and there was an exponentially decreasing trend as the vegetative cover decreased (Wang et al. 2008). This may explain why the microbial communities in M soils with higher soil moisture were different from those in S and DS soils (Fig. 4). This finding is in agreement with surveys of Antarctic soils (Yergeau et al. 2007; Chong et al. 2010) and the Canadian low Arctic tundra (Chu et al. 2011) where soil moisture had a strong effect on the composition of soil bacterial communities.
The significant covariation between moisture and the bacterial community composition may be related to Proteobacteria, a dominant group of soil bacteria that was positively related to soil moisture (Table 2, Fig. 2). We also detected significant correlations between H′ and soil moisture (r = 0·506, P < 0·05) and the C/N ratio (r = 0·527, P < 0·05) (Table 2, Fig. 4). The highest average bacterial H′ values were found in M soils, followed by S and DS soils. This was probably because the adequate organic carbon derived from the denser vegetation litter and/or root exudates in higher moisture M soils could support a more diverse bacterial community. Soil C/N ratio was the second most important factor that affected bacterial communities (Fig. 8), in agreement with the findings from both Arctic and temperate ecosystems (Chu et al. 2010, 2011; Yuan et al. 2010). These results suggest that differences in the soil moisture and C/N ratio are probably the main drivers of bacterial community structure in Tibetan permafrost tundra.
The soil pH is known to have a strong effect on the structure and diversity of soil bacterial communities (Fierer and Jackson 2006; Lauber et al. 2009; Chong et al. 2010; Chu et al. 2010; Zinger et al. 2011). This effect was evident in the present study, although the soil pH varied only slightly (8·21–9·58) (Fig. 8). A unique cluster of bacterial and fungal community profiles was associated with the DS soils, which had the highest pH (Fig. 4a,b). However, the soil pH was not the most important soil parameter, in contrast to other studies (Fierer and Jackson 2006; Chong et al. 2010; Chu et al. 2010). The narrow pH range of the seasonal samples at our four sites may account for this difference. For example, the pH of our tundra soils ranged from 8·21 to 9·58, whereas the pH ranges in other investigations were 4·0–7·64 (Chu et al. 2010), 3·30–7·24 (Nacke et al. 2011), 3·30–7·37 (Zinger et al. 2011) and 3·50–8·50 (Fierer and Jackson 2006). Thus, there was less pH range in our samples with which to determine correlations. Indeed, the correlation between the bacterial community and the soil pH might actually be related to the soil C or moisture availability, which co-varies with soil pH (Rousk et al. 2010a).
Most bacterial groups were relatively stable among the M, S and DS vegetation types, but several groups exhibited changes in their relative abundance. For example, the copiotrophic Proteobacteria were consistently more abundant in M (Fig. 2a), which might correspond to the a higher labile carbon content, abundant water and lower pH in M soils (Smit et al. 2001; Fierer et al. 2007; Wang et al. 2008). By contrast, Bacteroidetes and Actinobacteria abundance increased as the vegetation cover decreased from M, to S and DS soils. This may have been due to increases in recalcitrant organic carbon fraction as the vegetation cover decreased (Wang et al. 2008), given the physiology of these bacterial groups to produce enzymes targeting complex and recalcitrant biopolymers (Kirchman 2002; Pankratov et al. 2006; Edwards et al. 2010). The Biolog analysis strongly supported these results, that is, the bacterial community in M significantly degraded amines and amino acids, whereas the S and DS communities degraded polymers more rapidly (Fig. 7).
The UniFrac analysis indicated that the DS bacterial communities were distinct from those of the M and DS soils, which agreed with the RDA patterns, the water content, pH and Proteobacteria abundance in these samples (Table 1, Fig. 5). This might further suggest that the soil moisture and pH were controlling factors for the bacterial species structure in the Beilu River permafrost soils. In particular, the UniFrac analysis detected evident seasonal changes where the November communities were clearly different from those in the spring and summer, even in the DS samples (Fig. 5). This seasonal pattern may have responded to changes in the soil moisture, temperature and substrate availability. The November community correlated with the low soil temperature (~4·2°C at 10 cm depth) (Fig. 1). A previous study indicated that the soil contained 5–6% unfrozen water by weight at −1·5°C (Rivkina et al. 2000). Thus, the liquid water film from the higher water content in November soils (Table 1) may have retained sufficient liquid water within the soil particles to support a cold-adapted microbial community (Fig. 8). The spring community (May and June) clustered together, which will have coincided with the release of thaw water from the soil at melting temperature (1·8°C at 10 cm on 10 May) and the substrate available from microbial turnover (Lipson et al. 2002). The August communities were supported by higher temperatures (Fig. 1), relatively higher moisture-derived precipitation, and nutrition of the root exudates, thus differing from others (Lipson and Schmidt 2004). This suggested that seasonal environmental changes caused community shifts among taxa with similar functions but different environmental tolerances (Wallenstein et al. 2007).
The RDA patterns of the fungal communities were similar to those of the bacterial communities among the M, S and DS vegetation types (Fig. 4b). The dominance of Ascomycota in M soils may have been related to their ability to degrade cellulose and hemicellulose (Liers et al. 2006; Shary et al. 2007), while the higher abundance of Basidiomycetes in DS soils may be responsible for the decomposition of more complex lignocellulose components (Taina et al. 2010). Differences in fungal communities were also correlated with vegetation types in an Alaskan tundra site (Wallenstein et al. 2007), Canadian low Arctic tundra (Chu et al. 2011) and French alpine soils (Zinger et al. 2011). These results suggested that the alpine vegetation type may have a substantial effect on the fungal community, which mainly depends on the plant litter.
In contrast to the bacterial H′, there was a positive correlation between the fungal H′f and the soil pH (Table 1). The H′f showed the opposite trend to the H′, where the highest average H′f was detected in DS, followed by S and M soils. Studies have discerned opposing pH relationships of fungal and bacterial growth due to competitive interactions between fungi and bacteria (Rousk et al. 2010b). Different effects of soil pH on bacterial and fungal communities were also detected in Arctic tundra and alpine grassland soils (Chu et al. 2011; Zinger et al. 2011). The arbuscular mycorrhizal fungi biomass was also reported to co-vary with the soil pH (Djukic et al. 2010). This was consistent with our RDA patterns of significant co-variation between fungal communities and the soil pH (Figs 4b and 8).
In contrast to the seasonal variation of bacterial UniFrac patterns, we observed marked variation in the species of the fungal communities among replicate sampling sites of the M and S vegetation (Fig. 6), whereas no distinct seasonal heterogeneity was detected. This difference in the UniFrac patterns of the bacterial and fungal communities may be attributed to their distinct physiologies. Fungi, especially Ascomycota and Basidiomycota, are capable of degrading woody plant litter such as lignin or cellulose, and many fungi have symbiotic relationships with plant roots (Marschner et al. 2001; Taina et al. 2010). In contrast, bacteria are dependent on root exudates or other more labile small molecule substrates (Rasche et al. 2011). Therefore, fungal communities tend to have a higher responsiveness to the vegetation type compared with bacterial communities (Chu et al. 2011; Zinger et al. 2011). However, a study in an Alaskan tundra site found that seasonal differences were more evident at the species level, where Zygomycota and Ascomycota dominated (Wallenstein et al. 2007). This difference may be attributable to the physiology of the dominant Zygomycota phyla, which cannot degrade wood and tend to depend on sugary substrates associated with seasonal variations (Taina et al. 2010).
To the best of our knowledge, this is the first study to estimate the relative contribution of different drivers on microbial community distribution patterns in Tibetan Plateau permafrost soils. Further studies with more environmental variables at a larger scale will provide further insights into the factors that drive microbial communities in these unique environments.
This study was funded jointly by grants from the National Science Foundation of China (nos. 40871041 and 40671042) and the National Basic Research Project of China ‘Investigation on Tibet Plateau Permafrost Baseline’ (no. 2008FY110200). We thank the Editor and anonymous reviewers for helpful comments that improved this manuscript.