Rapid recovery of soil bacterial communities after wildfire in a Chinese boreal forest

Fires affect hundreds of millions of hectares annually. Above-ground community composition and diversity after fire have been studied extensively, but effects of fire on soil bacterial communities remain largely unexamined despite the central role of bacteria in ecosystem recovery and functioning. We investigated responses of bacterial community to forest fire in the Greater Khingan Mountains, China, using tagged pyrosequencing. Fire altered soil bacterial community composition substantially and high-intensity fire significantly decreased bacterial diversity 1-year-after-burn site. Bacterial community composition and diversity returned to similar levels as observed in controls (no fire) after 11 years. The understory vegetation community typically takes 20–100 years to reach pre-fire states in boreal forest, so our results suggest that soil bacteria could recover much faster than plant communities. Finally, soil bacterial community composition significantly co-varied with soil pH, moisture content, NH4+ content and carbon/nitrogen ratio (P < 0.05 in all cases) in wildfire-perturbed soils, suggesting that fire could indirectly affect bacterial communities by altering soil edaphic properties.

izing the main factors driving soil bacterial community composition and diversity after fire, which may improve understanding of postfire forest ecosystem recovery process.

Results
Effects of fire on soil biogeochemical properties. Fire significantly altered soil biogeochemical properties (Table S1). Burning reduced microbial biomass carbon and nitrogen, with 82% less microbial biomass carbon and 71% less microbial biomass nitrogen in 1year-post-fire burned soils than unburned samples, and 63% less microbial biomass carbon and 72% less microbial biomass nitrogen in 11-year-post-fire burned soils. Wildfire increased soil pH, available nitrogen and phosphorus, and decreased soil moisture and carbon/nitrogen ratio 1-year-post-fire, but after 11 years, those properties were not significantly different from the levels observed in the unburned control site.
Bacterial community composition. Across all soil samples, we obtained a total of 319,618 quality sequences with 4,123-9,130 sequences per sample (mean 5,327), and were able to classify 84.3% of those sequences. The dominant phyla (or subphyla in the case of Proteobacteria) across the Greater Khingan Mountains soils were Alphaproteobacteria, Actinobacteria, Acidobacteria, Betaproteobacteria and Bacteroidetes, accounting for more than 76% of the bacterial sequences from each of the soils (Fig. S1). In addition, Gammaproteobacteria, Planctomycetes, Chloroflexi, Deltaproteobacteria, Gemmatimonadetes and Firmicutes were present in most soils but at relatively low abundances, and 24 other rare phyla were identified (Table S2).
Fire significantly shifted the relative abundance of dominant phyla except Actinobacteria (Fig. 1). Fire greatly increased the relative abundances of Betaproteobacteria and Bacteroidetes and decreased the abundance of Alphaproteobacteria, Acidobacteria, as well as Planctomycetes and Deltaproteobacteria with low abundances 1year-post-fire. Interestingly, the relative abundances of dominant phyla returned to a similar level to the controls after 11 years, with the exception of Alphaproteobacteria (Fig. 1). However, by comparing OH (1 year post high intensity fire) with control at class or lower levels certain taxa demonstrated significant responses. In the phylum Acidobacteria, all OTUs (operational taxonomic unit) had a significantly lower abundance except OTU_786 and OTU_32455 in response to fire, and for Bacteroidetes, in the family Sphingobacteria, all OTUs had a significantly higher abundance (Fig. 2). In the phylum Proteobacteria, Alphaproteobacteria demonstrated significantly lower abundance, except for OTU_62141, OTU_58113 and OTU_72620; Betaproteobacteria had a significantly greater abundance, except for OTU_25883; while Deltaproteobacteria showed lower abundance in response to fire (Fig. 2). In contrast, Actinobacteria taxa showed a highly variable response (Fig. 2). Similar patterns were observed when comparing OL (1 year post low intensity fire) with the unburned controls (Fig. S2). Bacterial community composition in soils across the Greater Khingan Mountains showed that fire resulted in a dramatic shift in soil bacterial communities 1-year-postfire (P 5 0.001), but after 11 years the communities were indistinguishable from unburned forest soil (P 5 0.549, Fig. 3, Table S3). Fire intensity had no significant impact upon recovery (Fig. 3).
The soil bacterial community was related to soil biogeochemical variables in both pre-and post-fire soils. Mantel tests showed that bacterial community composition was significantly correlated with soil pH in control soils, while the community composition was significantly correlated with soil pH, moisture content, NH 4 1 content, and C/N ratio in soils 1 and 11 years post fire (Table 1). Among all the measured soil variables, soil pH showed the highest correlation with bacterial community composition in both control and fire-impacted soils ( Table 1). Canonical correspondence analysis (CCA) indicated that soil pH had the strongest effect on bacterial community com-position, while soil moisture content, NH 4 1 content, C/N ratio and TN content also had less, but significant effect on the community composition (Fig. S3). In addition, soil pH showed a significant correlation with the relative abundance of Alphaproteobacteria, Actinobacteria, Acidobacteria, Betaproteobacteria, as well as three less abundant phyla (Fig. S4). Soil NH 4 1 content, C/N ratio, moisture content, TC content and TN content were also significantly correlated with the relative abundance of different dominant phyla (Table  S4).
Bacterial diversity. In terms of both phylotype richness (i.e. number of OTUs) and phylogenetic diversity (Fig. 4), which were surveyed at a depth of 4,000 randomly selected sequences per sample, the diversity of bacterial communities exhibited significant differences (P , 0.001 in both cases). High intensity fire quickly decreased bacterial phylotype richness and phylogenetic diversity, but after 11 years, these parameters returned to match the unburned controls. Interestingly, the highest bacterial diversity was found 11 years after low intensity fire. Soil bacterial phylotype richness was positively correlated with soil pH (P 5 0.013, Fig. 5), dissolved organic carbon (P 5 0.039, Table S5), and negatively correlated with elevation (P 5 0.042, Table S5), while phylogenetic diversity was positively correlated with soil pH (P 5 0.022, Fig. 5) and dissolved organic carbon (P 5 0.049, Table S5).

Discussion
Records of wildfire occurrences provided an opportunity to examine the effects of fire on soil bacterial community composition, diversity and succession. In this study, we found Proteobacteria, Actinobacteria, Acidobacteria and Bacteroidetes were the main phyla in boreal forest soil (Fig. S1), similar to observations from other soils collected from Arctic and subalpine soil environments 19,20 , showing that dominant bacterial phyla in soils are similar. We found that burning had a dramatic impact on the soil bacterial community composition and diversity 1 year following a fire ( Fig. 1, 3, 4). Bacterial communities from 1 and 11 years post-burn were significantly different not only in the OTUs present, but also in the proportional abundances of phyla ( Fig. 1, 4, S1; Table S2). An earlier study 21 found that fire dramatically altered soil bacterial community composition and diversity 4 and 16 weeks after fire and our results showed that the effect of fire on bacterial community could last for more than 1 year, suggesting that fire had a strong impact on bacterial community. Our results go beyond these findings by showing that the composition of bacterial community 11 years after fire returned to a similar state compared to the unburned control site (Fig. 3, Table S3). In addition, bacterial diversity decreased 1-year-post-fire but the diversity recovered 11year-post-fire with the highest diversity at low intensity fire ( Fig. 4), which might be due to the successful colonization and survival of many rare phyla into soil during the process of bacterial succession post fire. Other studies have shown that the post-fire understory vegetation community reaches its pre-fire level in boreal forest 20-100 years after a fire, depending on pre-fire stand age and site conditions 22,23 . Our results therefore suggest that bacterial communities may recover much faster than understory vegetation. In the present study, we have only two time points after fire occurrence (1 and 11 years post fire), and in future study more time points post fire might be needed to clarify when the bacterial communities recover to the unburned level and how bacterial communities succeed after forest fire. Human influence and global warming are rapidly increasing the frequency of forest fire, therefore understanding different recovery rate between soil microbial community and vegetation community in response to wildfire may be important for understanding the recovery of forest ecosystems as a whole.
The results in this study showed that bacterial community composition and diversity were mainly correlated with soil pH. The www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 3829 | DOI: 10.1038/srep03829 relative abundance of dominant phyla was also correlated with soil pH. For example, the relative abundance of Acidobacteria has been shown to increase with decreased pH (Table S4), which is consistent with most of previous studies 20,24 . However, the relative abundance of Alphaproteobacteria was shown to decrease toward higher pH in our study (Table S4), which is contrary to other studies in different systems 19,25 . These results indicated that although bacterial community composition was clearly influenced by pH, there were some dif-ferences in the responses of specific phylum to changes in soil pH. The overriding importance of soil pH has been demonstrated as a key factor in driving soil bacterial distribution across a variety of spatial scales, including continents 19,26 , national 24 , land-use types at a given location 27 , small and sub-meter scales 28 , and even along elevational gradient 20 . In this study, we observed that the bacterial community composition and diversity were primarily correlated with soil pH in both control and post fire soils (Fig S3; Table 1), suggesting that pH might have predictive power for bacterial distribution in not only undisturbed but also recently fire-perturbed ecosystems.
As noted above, disturbance may trigger both direct and indirect effects on soil microbial community structure. We found that wild-fire altered soil pH, moisture content, NH 4 1 content and C/N ratio which were significantly co-varied with bacterial community composition in soils both 1 and 11 years after fire (Table 1). These results might suggest that fire could indirectly affect bacterial communities  by altering soil properties. In addition, the significance of fire as a shaper of vegetation composition and structure is well known. Major links between plant species and soil microorganisms include the quantity of resources produced, competition for nutrients, quality of resources and mutualism 6 . How the successional growth of vegetation after a wildfire will influence and possibly be influenced by soil microbial community structure is a topic warranting future investigation in this and other study systems.

Methods
Site selection and soil sampling. Our study area was located in the Greater Khingan Mountains in northeast China (51u179N 122u429E to 51u569N 123u189E), and encompassed approximately 167,213 ha. The area has a cold, continental climate, with average annual temperature declining from 1uC at its southern extremes to 26uC at its northern extremes, and precipitation declining from 442 mm in the south to 240 mm in the north. More than 60% of the annual precipitation falls in the summer season from June to August 18 . The vegetation of this area is representative of cool temperate coniferous forests, forming the southern extension of the eastern Siberian boreal forests. Historically, fires were caused primarily by lightning 18 . Dendrochronological studies have indicated that the historical fire regime was characterized by frequent surface fires, mixed with infrequent stand-replacing fires, with the interval between fires ranging from 30 to 120 years 18,29 . However, forest harvesting and fire suppression have altered fire regimes in this region 30 .
Soil samples were collected on July 24 th to August 19 th of 2011 in the Huzhong National Natural Reserve of the Greater Khingan Mountains. The study area is primarily covered by mature larch (Larix gmelinii) forest with little human disturbances since the establishment of the Reserve 29 . The parent material is granite bedrock and the soil is a dark brown forest soil 31 . We used a stratified sampling design to select sample plots based on fire history and fire severity. Fire history (surface fire) has three levels: 1-year-after-fire, 11-year-after-fire and unburned control. Fire severity was defined into two levels: low and high severities. Fire severity levels were defined based on differenced normalized burn ratio (dNBR) of remote sensing Landsat TM images, which have been proved applicable to our study area 29 . The dNBR 32,33 was calculated using the equation NBR pre-fire 2 NBR post-fire , while NBR was calculated using the equation (TM4 2 TM7)/(TM4 1 TM7). TM4 and TM7 refer to Thematic Mapper bands 4 (the near-infrared wave) and 7 (the medium-infrared wave), respectively, which were calculated according to pre-and post-fire images. In the stratified random sampling design, the dNBR value was classified into two levels according to its histogram: high severity ($743) and low severity (,743) 29 . In addition, we selected 12 unburned locations (plots), which were classified as mature forest without fire disturbance scattered among fire occurrence region as control. In summary, those samples included no fire (control), 1 year after low intensity fire (OL), 1 year after high intensity fire (OH), 11 years after low intensity fire (EL) and 11 years after high intensity fire (EH). A total of 59 selected samples, 12 from unburned soils and 47 (10 for EL, 13 for EH, 12 for OL and 12 for OH) from burned soils, were analyzed in this study. In each plot (40 m 3 40 m), soil was collected from five points (four vertices and the center) at a depth of 0-5 cm and then mixed as one sample. After sampling, the soils were kept in a cooler and shipped refrigerated to the lab. The samples were thoroughly mixed and sieved to remove grassroots and stone, and divided into two parts: one part was stored at 4uC for biogeochemical analysis; the other was stored at 240uC for DNA analysis.
Soil nutrients and microbial biomass analyses. Soil pH was measured using a pH Meter after shaking a soil water suspension (155 wt/vol) for 30 minutes. Soil moisture was measured gravimetrically. Total carbon (TC) and total nitrogen (TN) were determined by dichromate oxidation and titration with ferrous ammonium sulfate 34 . Soil dissolved organic C (DOC) and dissolved total N (DTN) and mineral nitrogen were extracted by adding 50 ml of 0.5 M K 2 SO 4 to 10 g fresh soil, shaking for 1 h and then vacuum filtering through glass fiber filters (Fisher G4, 1.2 mm pore space). ) and nitrate (NO 3 2 ) contents in the extracts were determined colourimetrically by automated segmented flow analysis (Bran 1 Luebbe AAIII, Germany) using the salicylate/dichloroisocyanuric acid and cadmium column/ sulphanilamide reduction methods, respectively. DOC and DTN were determined using a TOC-TN analyzer (Shimadzu, Kyoto, Japan). Dissolved organic N (DON) was calculated as follows: DON 5 DTN 2 (NH 4 1 2 N) 2 (NO 3 2 2 N). Microbial biomass C (MBC) and biomass N (MBN) were analyzed by the chloroform fumigation and extraction method 35 , and the final values were calculated using 0.35 (k C ) and 0.4 (k N ) correction factors 36 .
Soil DNA extraction. Soil DNA was extracted from the 0.5 g soil after sieving using a FastDNAH SPIN Kit for soil (MP Biomedicals, Santa Ana, CA) according to the manufacturer's instructions. The extracted soil DNA was dissolved in 60 ml TE buffer, quantified by NanoDrop and stored at 220uC.
Bacterial 16S rRNA genes amplification and 454 Sequencing. An aliquot (50 ng) of purified DNA from each sample were used as template for amplification. The V4-V5 hypervariable regions of the bacterial 16S rRNA genes (Escherichia coli positions 515-907) were amplified using the primer set: F515: GTGCCAGCMGCCGCGG with the Roche 454 'A' pyrosequencing adapter, and a unique 7 bp barcode sequence, while primer R907: CCGTCAATTCMTTTRAGTTT contained the Roche 454 'B' sequencing adapter at the 59-end of each primer, respectively. The targeted gene region has been shown to be the most appropriate for the accurate phylogenetic reconstruction of bacteria 37 . Each sample was amplified in triplicate with 50 ml reaction under following: 35 cycles of denaturation at 94uC for 45 s, annealing at 55uC for 45 s, and extension at 72uC for 45 s; with a final extension at 72uC for 10 min.
PCR products were pooled together and purified by Agarose Gel DNA purification kit (TaKaRa). An equal amount of PCR product for each sample qualitative determination by bioanalyzer (Agilent 2100) and quantitative analysis by NanoDrop was combined in a single tube, and run on a Roche FLX 454 pyrosequencing machine (Roche Diagnostics Corporation, Branford, CT, USA), producing reads from the forward direction F515.
Processing of pyrosequencing data. Data were processed by the Quantitative Insights Into Microbial Ecology (QIIME) pipeline 38 . Specifically, bacterial sequences with the same barcode were assigned to the same sample after denoising by denoiser v. 0.91 39 . The barcode and primer sequences were removed, and only the first 350 bp after the proximal PCR primer was included for further analysis. Bacterial phylotypes were identified using uclust 40 and assigned to operational taxonomic units (OTUs, 97% similarity). Representative sequences from each phylotype were aligned using PyNAST 41,42 . The taxonomic identity of each phylotype was determined using the ribosomal database project (RDP) Classifier 43 . To correct for survey effort, we used a randomly selected subset of 4,000 sequences per sample to compare relative difference between samples.
Statistical analysis. Phylogenetic diversities (PD) were estimated by Faith's index 44 , which provides an integrated index of the phylogenetic breadth across taxonomic levels. The relationships between the taxonomic diversity for the group with geochemical features were tested with linear regression analyses using SPSS 17.0 for Windows. The response ratio (RR), calculated using the SAS program (SAS version 9.1. SAS Institute, Cary, North Carolina, USA), was used to analyze the effects of fire on phylogenetic composition and structure of bacterial communities 45 . NMDS (Non- Figure 4 | Changes in bacterial OTUs phylotype richness and phylogenetic diversity across the different groups. Diversity indices were calculated using random selections of 4,000 sequences per soil sample. Error bars denote standard deviation; Different letters represent significant differences from Tukey's HSD comparisons (P , 0.05). OL: one year after low intensity fire; OH: one year after high intensity fire; EL: 11 years after low intensity fire; EH: 11 years after high intensity fire.
www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 3829 | DOI: 10.1038/srep03829 metric multidimensional scaling) using Bray-Curtis dissimilarity and ANOSIM (Analysis of Similarity) based on the OTU table were completed in the vegan package (Version 2.0-2) of R v.2.8.1 project (R Development Core Team. Vienna, Austria) to compare community composition in burned and unburned samples. Mantel tests 46 were performed in the vegan package (Version 2.0-2) of R v.2.8.1 project (R Development Core Team. Vienna, Austria) were used to identify environmental factors that significantly correlated with community composition (abundance of OTUs), and the factors that significantly correlated with the bacterial community composition were tested by variance inflation factor (VIF), which is used to judge the colinearity. The VIF value of factors less than 20 were selected to perform canonical correspondence analysis (CCA) in the vegan package (Version 2.0-2) of R v.2.8.1 project (R Development Core Team. Vienna, Austria). Figure 5 | The relationship between soil pH and bacterial OTUs phylotype richness and phylogenetic diversity by linear regression analyses. The communities were randomly sampled at the 4,000 sequences level. Individual points represent different samples across all the treatments. OL: one year after low intensity fire; OH: one year after high intensity fire; EL: 11 years after low intensity fire; EH: 11 years after high intensity fire. P , 0.05, significant convention. www.nature.com/scientificreports