Tillage practices and straw-returning methods affect topsoil bacterial community and organic C under a rice-wheat cropping system in central China

The objective of this study was to investigate how the relationships between bacterial communities and organic C (SOC) in topsoil (0–5 cm) are affected by tillage practices [conventional intensive tillage (CT) or no-tillage (NT)] and straw-returning methods [crop straw returning (S) or removal (NS)] under a rice-wheat rotation in central China. Soil bacterial communities were determined by high-throughput sequencing technology. After two cycles of annual rice-wheat rotation, compared with CT treatments, NT treatments generally had significantly more bacterial genera and monounsaturated fatty acids/saturated fatty acids (MUFA/STFA), but a decreased gram-positive bacteria/gram-negative bacteria ratio (G+/G−). S treatments had significantly more bacterial genera and MUFA/STFA, but had decreased G+/G− compared with NS treatments. Multivariate analysis revealed that Gemmatimonas, Rudaea, Spingomonas, Pseudomonas, Dyella, Burkholderia, Clostridium, Pseudolabrys, Arcicella and Bacillus were correlated with SOC, and cellulolytic bacteria (Burkholderia, Pseudomonas, Clostridium, Rudaea and Bacillus) and Gemmationas explained 55.3% and 12.4% of the variance in SOC, respectively. Structural equation modeling further indicated that tillage and residue managements affected SOC directly and indirectly through these cellulolytic bacteria and Gemmationas. Our results suggest that Burkholderia, Pseudomonas, Clostridium, Rudaea, Bacillus and Gemmationas help to regulate SOC sequestration in topsoil under tillage and residue systems.

Soil organic C (SOC) is the main source of energy for soil microorganisms 1,2 and SOC content profoundly affects soil properties, including aggregate stability, soil moisture and nutrient cycling 3 . Thus, SOC plays an important role in maintaining long-term sustainability of agro-ecosystems and global biogeochemical cycles 4 . SOC is regulated by many factors, such as tillage practices 1 , residue management 5 , soil aggregate sizes 6 , and microbial functional diversity 2 . Optimizing agricultural management can reduce SOC loss, or even increase the content of SOC 7 . Intensive and continuous soil tilling has been practiced for thousands of years in China 5 . Frequent soil disturbance by intensive conventional tillage (CT) reduces soil aggregate sizes, thereby accelerating SOC oxidation 8 and decreasing SOC content. Moreover, crop residue burning or removing, a common farming practice, reduces the amount of organic substances retained in the soil and the water storage capacity of the entire soil 9 , and decreases soil microbial biomass and functional diversity 2 . In contrast, no-tillage (NT) and straw returning (S) may enhance SOC content in agricultural ecosystems and facilitate sustainable agricultural production 5,7 .
The abundance, diversity and composition of soil microbial communities and their interactions with environment factors have great impacts on SOC dynamics 6,10-12 . In agricultural ecosystems, bacteria and fungi are the main drivers of soil processes including nutrient cycling and the decomposition of soil organic matter, such Links between soil bacterial communities and SOC. SEM revealed that the predictors explained 75.0-84.0% of the variances in SOC (Fig. 5). In Fig. 5a1,a2, tillage and straw systems had different levels of MUFA/ STFA and G + /G − , and thus likely affected SOC directly and indirectly through the presence of 5 kinds of cellulolytic bacteria (Pseudomonas, Rudaea, Bacillus, Burkholderia and Clostridium) and Gemmatimonas.

Discussion
This study investigated the effects of tillage practices and straw-returning methods on soil bacterial communities and their relation to SOC after a 2-year rice-wheat cropping cycle in central China. The results supported our hypotheses that NT and S practices increase the abundance of bacterial genera in the topsoil bacterial communities, and that the composition of the bacterial community is correlated with SOC.
Effects of tillage practices and straw returning methods on soil bacterial community. NT and S practices generally increased the abundance, activity and diversity of soil microbial communities in the topsoil layer 2,28,29 , probably because NT minimizes soil disturbance and S contributes to greater accumulation of crop residues on the soil surface 8 , thus improving soil nutrition condition for soil microbial communities 7 . Soil nutrition condition can be indicated by G + /G − ratio 30 . Lower G + /G − ratio under NT treatments compared with under CT treatments (Table 1) suggests that nutrients are rich in the topsoil layer under NT treatments, which demonstrates the improvement of soil environment for microorganisms under NT 7 . Guo et al. 7 and Wang et al. 31 also       Table 3. Relationships between SOC and bacterial genera based on stepwise regression analysis. * P < 0.05; ** P < 0.01. reported that NT could significantly increase soil N, organic C and SOC fractions compared with CT, whereas CT could negatively affect soil microbial biomass and SOC. Moreover, a greater MUFA/STFA ratio in NT treatments compared with in CT treatments (Table 1) suggests that NT treatments may improve soil gas permeability as suggested by Bossio et al. 32 , because of the accumulation of crop residues on the soil surface under NT 8 . Straw returning, as an input of organic residues to improve soil nutrition condition, can increase soil surface residue C and SOC 7 and provide energy sources for soil microbes, thus enhancing soil microbial biomass 2,7 . Residue amendment improves soil moisture and temperature and promotes soil aggregation, thus boosting microbial growth 20 , which is supported by the result of lower G + /G − under S treatments than under NS treatments (Table 1). However, in the study of the impacts of residue management on soil properties and soil microbial community structure, Wang et al. 33 did not find significant differences in bacterial abundance between S and NS treatments. In addition, S treatments had a higher MUFA/STFA ratio compared with NS treatments (Table 1). This result indicates that soils under S treatments may have greater gas permeability 32 , possibly because straw returning decreases the sensitivity to surface sealing 34 and increases the porosity of the top soil layer 35 . Good soil gas permeability and enrichment of organic matter in soil surface under NT and S practices 7,8 also promote the decomposition of exogenous crop straw, thus improving soil nutrition condition. Therefore, NT and S practices improve soil nutrition, leading to the increase of soil bacterial abundance in the topsoil layer.

Relationships between soil bacterial communities and SOC.
Our results showed that there were seven predominant bacterial genera (Gemmatimonas, Rudaea, Caulobacter, Sphingomonas, Dokdonella, Rhodanobacter and Mycobacterium) in the 0-5 cm soil layer, which accounted for 67.7% of total bacterial abundance (Fig. 2). Multiple analysis results showed that soil bacterial communities were closely related to SOC, and Pseudomonas, Rudaea, Bacillus, Gemmatimonas, Burkholderia and Clostridium greatly contributed to SOC, together explaining 75.6% of the variances in SOC. Pseudomonas, Rudaea, Bacillus, Clostridium, Burkholderia and Dyella belong to cellulolytic bacteria 32 , and together explained 66.3% of the variances in SOC, suggesting that SOC is mainly regulated by these six cellulolytic bacteria. Cellulose is unavailable to most soil microorganisms because the crystallinity of cellulose is extremely recalcitrant for enzymatic degradation 36 . Some studies have suggested that cellulolytic bacteria help to regulate the C cycle 37 because they play an important role in the decomposition of plant residues in the soil ecosystem 36 .
In this study, most of the cellulolytic bacteria screened by stepwise regression analysis are aerobic microorganisms (Table 3), which can be attributed to the high permeability in the 0-5 cm soil layer 35 . Generally, cellulose is mainly degraded in aerobic environments, while up to 5-10% of cellulose is degraded by physiologically diverse bacteria under anaerobic conditions 37 . Many studies have indicated that Clostridium, one of important cellulolytic anaerobic bacterial genera 38 , is highly efficient in degrading cellulose 36,39 because it excretes several kinds of enzymes including cellulase and hemicellulase 40 . In the present study, NT and S treatments had significantly greater Clostridium abundance (Fig. 2). Multiple analyses suggested that Clostridium may play important roles in SOC (Figs 1, 2, 3 and 4). Clostridium is negatively affected by greater oxygen availability in the soil and soil Arrow thickness represents the magnitude of the path coefficient. Values associated with solid arrows represent the path coefficients. Solid and dashed arrows indicate significance (P < 0.05) and non-significance (P > 0.05), respectively. Straw, straw systems; DOC, dissolved organic C; MBC, microbial biomass C; SOC, soil organic C. disturbance 8 . Hence, its high abundance under NT and S treatments was not unexpected as it is likely that the higher residue mulching under S practice and/or less soil disturbance under NT 1,7 create anaerobic zones in the surface soil.
Gemmatimonas (22.6%, 245 OTUs) is the most abundant bacterial genera in this study (Fig. 2), and explained 12.4% of the variances in SOC (Fig. 4). The SEM also showed the key function of Gemmatimonas to SOC sequestration under NT and S practices in this study (Fig. 5). The reason may be that Gemmatimonas can use the metabolic products as sole C sources 41 , such as acetate and propionate [42][43][44] , but most of other bacterial genera, such as Bellilinea 45 and Sphingomonas 46 (Fig. 2), cannot or can only weakly use the metabolic products of cellulose. Thus, it is likely that Gemmatimonas has greater ability to use available C sources compared with other soil microorganisms. The SEM further showed that Gemmatimonas plays an important role in SOC dynamics (Fig. 5), which can be attributed to the fact that Gemmatimonas can reduce the metabolic products of cellulose 41 and thus indirectly promotes the degrading process of cellulose.
Tillage practices and straw returning methods affect the activity and structure of soil microorganisms by changing the habitat characteristics for soil microorganisms such as soil gas permeability and the substrates for soil microorganisms 16,33 , thus affecting SOC. Both NT and S practices promote the accumulation of straw on the soil surface, in which the major component is cellulose 47 , thus improving soil physical conditions 21 and also providing C sources (specifically cellulose) for cellulolytic bacteria 36 . Therefore, NT and S practices promoted the growth of cellulolytic bacteria (Fig. 2), thus increasing the decomposition of cellulose and subsequently the SOC (Figs 2, 3, 4 and 5 and Table 3). Decomposition of exogenous crop straw provides C sources for other soil microorganisms, and therefore increases soil microbial biomass 48 , which contributes to the developing and increasing of soil organic matter 2,6 . However, exogenous organic matter from broken down cellulose promotes C sequestration in soil aggregates, especially in > 250 μ m aggregates because the broken down exogenous organic matter could be bound to the walls of the mineral particles that surround them 2,21 . Yin et al. 49 also reported that bacteria play critical roles in the production of soil aggregates and the conversion of plant residue to soil organic matter. The results of this study suggest that tillage changes the habitats for Pseudomonas, Rudaea, Bacillus, Burkholderia, Della and Clostridium, and then changes the decomposition process of residue, thus affecting SOC in the 0-5 cm soil layer.
This study indicates that after two cycles of rice-wheat rotation, NT and S practices promote SOC in the 0-5 cm soil layer presumably by increasing the abundance of bacterial genera. Redundancy analysis showed a close relationship between SOC levels and the abundance of specific bacterial genera in the soil community.
Stepwise regression analysis and relative influence analysis indicated that Gemmatimonas, Rudaea, Spingomonas, Pseudomonas, Dyella, Burkholderia, Clostridium, Pseudolabrys, Arcicella and Bacillus are positively correlated with SOC. SEM results further suggested that NT and S practices specifically increase the abundance of 5 kinds of cellulolytic bacteria (Burkholderia, Pseudomonas, Clostridium, Rudaea, and Bacillus) and Gemmatimonas in the upper soil layer, likely promoting SOC levels. However, the mediation of bacterial communities on SOC under long-term NT and S practices in the rice-wheat cropping system should be further discussed. Long-term (5+ years) NT and S practices may change SOC in the whole plough layer (0-20 cm); however, the ability of bacterial communities to regulate these effects remains unclear. Therefore, further studies should be conducted to reveal the mechanism of the effects of long-term NT and S practices on soil bacterial communities and their contributions to SOC in the whole plow layer.

Methods
Experimental site. The study site was located at an experimental farm of Huazhong Agricultural University Research (29°51′ N, 115°33′ E) in the town of Huaqiao Town, Wuxue City, Hubei Province, China, which has been described by Guo et al. 2 . The soil is a silty clay loam classified by the Food and Agriculture Organization (FAO) as a Gleysol 2 . The experimental soil (0-20 cm depth) has a pH of 4.79, an organic C content of 16.89 g kg −1 , a total nitrogen (N) content of 2.20 g kg −1 , a total phosphorus (P) content of 0.45 g kg −1 , and a bulk density of 1.21 g cm −3 . The cropping regime was dominated by two crops: summer rice (HHZ, Oryza sativa L.) and winter wheat (ZM9023, Triticum aestivum L.). Experimental design. The detailed experimental design was described by Guo et al. 2 . In brief, field treatments followed a split-plot design of a randomized complete block with tillage practices (conventional intensive tillage, CT; no tillage, NT) as the main plots and straw returning methods [crop straw removal (NS) and crop straw return (S)] as the subplots. The experiment involved four treatments: CTNS, CTS, NTNS and NTS, with each replicated for three times. For CTNS and NTNS treatments, crop residues were removed and not returned to the field. For CTS and NTS treatments, residues were chopped into pieces 5-7 cm in length and returned to the field. The chopped straw was mulched in NT soil and tilled into CT soil. For CT treatments, the soil was moldboard ploughed twice to a 20 cm depth before throwing of rice seedlings and once before sowing of wheat. The soil was not disturbed for NT treatments. Commercial compound fertilizer (15% N, 15% P 2 O 5 , and 15% K 2 O), urea (46% N), single superphosphate (12% P 2 O 5 ), and potassium chloride (60% K 2 O) were used to provide 180 kg N ha −1 , 90 kg P 2 O 5 ha −1 , and 180 kg K 2 O ha −1 during the rice-growing seasons, and 144 kg N ha −1 , 72 kg P 2 O 5 ha −1 , and 144 kg K 2 O ha −1 during the wheat-growing season. P and K fertilizers were only applied as basal fertilizers, and N fertilizers were used with 50%, 20%, 12%, and 18% at the seedling, tillering, jointing, and earring stages of rice-growing seasons, and with 50%, 30%, and 20% at the seedling, tillering, and boosting stages of wheat-growing seasons, respectively. The plots were irrigated to a depth of 8 cm whenever the water depth above soil surface decreased for 1-2 cm during the rice growing season, and were drained in the tillering and maturing stages. We did not irrigate during the wheat-growing season.
Scientific RepoRts | 6:33155 | DOI: 10.1038/srep33155 Soil sampling. Soil samples were collected from the topsoil (0-5 cm depth) using a soil sampler (7 cm diameter) immediately after wheat harvest in June 2013 at eight random points in each plot. After sampling, visible plant residues and stones were removed, and large soil clods were gently broken by hand. Soils were sieved through a 5 mm screen for uniformity, and stored at − 20 °C, and all determinations were finished within two weeks.
The SOC and its fractions (microbial biomass C (MBC) and dissolved organic C (DOC)) in the 0-5 cm soil layer were reported previously by Guo et al. 2 . Phospholipid fatty acid (PLFA) analysis. PLFA analysis was conducted to measure the composition of soil microbial communities according to the methods of Blair et al. 50 and Bossio et al. 32 and detailed measurements were performed as described by Guo et al. 7 . Briefly, lipids were extracted in a single-phase chloroform-methanol-citrate (1:2:0.8) buffer system. Polar lipids were separated from neutral lipids and glycolipids on solid phase extraction columns (Supelco Inc, Bellefonte, PA, USA) by eluting with CHCl 3 , acetone, and methanol. The phosholipid fractions were saponified and methylated to fatty acid methyl esters (FAME). Nonadecanoic acid methyl ester was used as internal standard and was added to calculate the absolute amounts of FAMEs before measurements. PLFAs were analyzed as FAMEs on a gas chromatograph/mass spectrometry system (6890-5973N series GC/MS Agilent Technologies, Palo Alto, CA, USA).

DNA extraction, PCR amplification, 16S rDNA gene amplification and 454 pyrosequencing.
High-throughput sequencing technology, a common method for identifying bacterial communities in various habitats and environmental samples, was used to identify bacteria in the soil samples 51  An aliquot (10 ng) of purified DNA from each sample (one biological replicate) was used as template for amplification. The primer 357F (5′ -CCTACGGGAGGCAGCAG-3′ ), which was modified with the addition of the 454 FLX-titanium adaptor "B" sequence (5′ -CCTATCCCCTGTGTGCCTTGGCAGTCTCAG-3′), was used to amplify the V3, V4 and V5 hypervariable regions of the bacterial 16S rDNA 52 , and 926R: 5′ -CCGTCAATTCMTTTRAGT-3′ was modified with the addition of unique 6-8 nucleotide barcode sequences and the 454 FLX-titanium adaptor "A" sequence (5′ -CCATCTCATCCCTGCGTGTCTCCGACTCAG-3′ ) 52 . Each sample was amplified in triplicate in a 25 μ l reaction and the program was as follows: initial denaturation at 95 °C for 4 min, followed by 25 cycles of denaturation (94 °C for 30 s), annealing (55 °C for 45 s), extension (72 °C for 1 min), and a final elongation step for 8 min at 72 °C. PCR Purification Kit (Axygen, Union City, CA, USA) was used to purify the PCR products. The amplicons of each sample were then pooled in equimolar concentrations into a single tube prior to 454 pyrosequencing. Pyrosequencing was performed on a 454 GS-FLX Titanium System (Roche, Basel, Switzerland) by Shanghai Personal Biotechnology Co., Ltd. Quality filtering of data was conducted following Fierer et al. 53 using the Quantitative Insights into Microbial Ecology (QIIME) pipeline (http://qiime.sourceforge.net) 54 . In brief, sequences with an average quality score of less than 25, sequences with lengths less than 200 nt or greater than 1,000 nt, with ambiguous bases greater than 1, with homopolymer lengths greater than 6, or with maximum primer mismatches greater than 0 were removed from the dataset. And chimeric sequences were removed using the uchime algorithm in mothur (Version 1.21.2, http://www/mothur. org/) 55 . Sequences were clustered into operational taxonomic units (OTUs) using the QIIME implementation of cd-hit with a threshold of 97% pairwise identity. The longest sequences were extracted and taken as representatives for taxonomic identification by BLAST searches against the non-redundant GenBank sequence database. In order to study the function of soil bacterial communities in the SOC dynamics, the relative abundance of soil bacterial genera (OTUs) was used for multiple analysis in this study.

Statistical analyses.
General linear model analysis of variance with SAS 9.0 designed for split plot with tillage practice and straw returning methods as fixed factors and replicates as random factors was conducted to test the main effects and interactions of tillage and straw returning. The least significant difference (LSD) test was used determine the significance of the effects of tillage, straw returning or their interactions. Only the means statistically different at P ≤ 0.05 were considered. Detrended correspondence analysis performed by CANOCO software showed that the data of characteristics of soil microbial communities and bacterial abundance were fitted with the linear model. Thus, redundancy analysis was performed using CANOCO software to explain the relationships between SOC fractions and bacterial communities. Monte Carlo permutation test performed by Cannoco 4.5 was used to assess the statistical significance of explanatory variables. Stepwise regression analysis (SAS 9.0) was performed to determine the relationships between SOC and bacterial genera. The contributions of bacterial genera to SOC were estimated by the relative importance analysis, using the "relaimpo" package in R 56 . Structural equation modeling (SEM), a multivariate statistical method that enables hypothesis testing of complex path-relation networks 57 , was used to evaluate whether bacterial communities mediate the change of SOC in response to the conversion of CT to NT, or NS to S. We constructed an a priori model according to a literature review and our knowledge of how these predicators are related. The initial model comprised eight predictors: tillage systems (Tillage), straw systems (Straw), monounsaturated fatty acids/saturated fatty acids (MUFA/STFA), gram-positive bacteria/gram-negative bacteria (G + /G − ), dissolved organic carbon (DOC), microbial biomass carbon (MBC), soil organic carbon (SOC), and key genera of the soil bacterial communities (Pseudomonas, Rudaea, Bacillus, Burkholderia, Clostridium, and Gemmatimonas picked by stepwise regression analyses), which greatly contributed to the SOC. A 'robust' maximum likelihood estimation procedure of AMOS 7.0 software was conducted for the analysis. χ 2 -test, comparative fit index (CFI), goodness-of-fit (GFI) and root square mean error of approximation (RMSEA) were performed to evaluate model fit.