Alteration of soil bacterial interaction networks driven by different long-term fertilization management practices in the red soil of South China

Applied Soil Ecology 120 (2017) 128–134

Contents lists available at ScienceDirect

Applied Soil Ecology journal homepage: www.elsevier.com/locate/apsoil

Research Paper

Alteration of soil bacterial interaction networks driven by different longterm fertilization management practices in the red soil of South China

MARK

Weibing Xuna,b, Ting Huangc, Wei Lia, Yi Rena, Wu Xionga, Wei Rana, Dongchu Lid, Qirong Shena, ⁎ Ruifu Zhanga,b, a Jiangsu Provincial Key Lab for Solid Organic Waste Utilization, National Engineering Research Center for Organic-Based Fertilizers, Jiangsu Collaborative Innovation Center for Solid Organic Waste Resource Utilization, Nanjing Agricultural University, Nanjing, 210095, PR China b Key Laboratory of Microbial Resources Collection and Preservation, Ministry of Agriculture, Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, PR China c Hanlin College, Nanjing University of Chinese Medicine, Taizhou, 225300, PR China d Qiyang Red Soil Experimental Station, Chinese Academy of Agricultural Sciences, Qiyang 426182, PR China

A R T I C L E I N F O

A B S T R A C T

Keywords: Soil bacterial community Fertilization Ecological network Indicator

Interactions among soil bacteria occur widely and play important roles in the maintenance of soil functions. Long-term fertilization management practices have distinct effects on soil fertility and the soil microbial activity and community, which are closely associated with soil microbial interactions. Red soil is typically low-productivity soil in South China. Chemical nitrogen fertilization caused serious soil acidification and low-productivity (defined as the acidified soil group, Ac), whereas the application of lime to the acidified soil increased the soil pH (defined as the quicklime improvement soil group, Qlime). Long-term manure or fallow treatment maintained the soil pH and increased the soil fertility (defined as the high-productivity potential soil group, HPP). A molecular ecological network analysis method was used to analyze 454 pyrosequencing data of bacterial communities from the HPP, Ac and Qlime soils. Several major differences were observed among the three constructed networks. First, the HPP network contained the largest ratio of positive to negative correlations, whereas the Ac network contained the smallest. Second, the HPP and Ac networks shared only 8.67% of their operational taxonomic units (OTUs), whereas Ac and Qlime shared 27.04%. Third, the HPP network contained the most “module hubs” (A set of OTUs that have strong interactions or common functions are grouped as a “module” in network analysis. These OTUs are called “nodes”. And the nodes with high connectivity to many other nodes within the same module are “module hubs”.), whereas Ac contained the fewest. These results demonstrated that the bacterial community of HPP was a better organized or a better operated community than Ac and that quicklime application helped to order the bacterial community. By comparing the topological roles of nodes in different networks, we proposed that there should be more module hubs in the networks of higherproductivity soils and hypothesized that these OTUs could be indicators of high-productivity.

1. Introduction Soil biodiversity is usually associated with ecosystem functions (Hooper et al., 2005; Naeem and Wright, 2003; Petchey and Gaston, 2006). Fuhrman (2009) attempted to link marine microbial community structures to their functional implications. However, the interactions among different microorganisms were ignored. Soil is a complex ecological environment in which microorganisms do not exist in isolation (Faust and Raes, 2012). Tens of thousands of bacterial species (Torsvik et al., 1990) in soils may form complex ecological interaction networks through various interaction types (Lidicker, 1979) (e.g., mutualism,

⁎

commensalism, parasitism, predation, and competition). The ecological interaction pattern of a network can reflect the general situation of a whole community structure (Freilich et al., 2010) or even the ecosystem functions (Zhou et al., 2010). Biological networks were previously constructed at the intracellular levels for protein-protein, protein-DNA, and protein-metabolite interactions (Barabási and Oltvai, 2004; Han et al., 2004; Maslov, 2002; Zhang and Horvath, 2005) or in complete systems like food-web structures (Cattin et al., 2004; Dunne et al., 2002). Global networks can be constructed from different data sets independently (e.g., 16S rRNA datasets) (Chaffron et al., 2010).

Corresponding author at: College of Resources & Environmental Sciences, Nanjing Agricultural University, 6 Tongwei Road, Nanjing, Jiangsu Province 210095, PR China. E-mail address: [email protected] (R. Zhang).

http://dx.doi.org/10.1016/j.apsoil.2017.08.013 Received 5 March 2016; Received in revised form 6 June 2017; Accepted 22 August 2017 0929-1393/ © 2017 Elsevier B.V. All rights reserved.

Applied Soil Ecology 120 (2017) 128–134

W. Xun et al.

organized in a randomized design with two replications and began in 1990 with the same rotation system. The experiment began with ten original treatments: (i) control (CK), no fertilizer application; (ii) chemical nitrogen (N); (iii) combined chemical nitrogen-potassium fertilizer (NK); (iv) combined chemical nitrogen-phosphorus fertilizer (NP); (v) combined chemical nitrogen-phosphorous-potassium fertilizer (NPK); (vi) combined chemical phosphorus-potassium fertilizer (PK); (vii) combined of swine manure and chemical nitrogen-phosphoruspotassium fertilizer (NPKM); (viii) swine manure (M); (ix) combined of straw and chemical nitrogen-phosphorus-potassium fertilizer (NPKS); and (x) fallow (Fallow). Five treatments (N, NK, NP, NPK and NPKS) resulted in strong acidification after 20 years (2010) of fertilization. These plots were split into two parts. Half-plot was fertilized as usual, and quicklime (CaO) was added into the other half-plot. More detailed description of the treatments and fertilizations can be found in my previous publication (Xun et al., 2016).

Network analysis (Zhou et al., 2010) is performed to represent the ecological interactions of different species in a complex bacterial community. Moreover, random matrix theory (Wigner, 1967) is a powerful approach for studying the interactions of complex systems. Therefore, the combination of these two techniques will be a very good analytical tool for microbial ecology. Briefly, after estimating the pairwise correlations between any two OTUs according to their relative abundances, their similarity is measured using the absolute value of the pairwise correlation coefficient. Then, based on an RMT approach, a similarity threshold is applied to transform the similarity matrix into adjacency matrix, which provides the strengths of the connections between nodes. Module analysis and network characterization are then performed according to the adjacency matrix. The topological network parameter, for example, connectivity, is estimated to represent how strongly that node is connected to all of the other nodes in the network. Additionally, the topological roles of nodes are evaluated to understand their importance in driving community functions. Previous works (Deng et al., 2012; Zhou et al., 2011) studied the responses of soil microbial communities to elevated CO2 levels and provided good examples of phylogenetic molecular ecological networks based on the random matrix theory conceptual framework. Recently, microbial association networks have been used to study various complex microbial ecological systems inferred for a range of communities from the ocean (Aylward et al., 2015; Fuhrman et al., 2015; Steele et al., 2011) to natural soils (Barberán et al., 2012; Eldridge et al., 2015) to the human body (Kelder et al., 2014; Lagkouvardos et al., 2015). Microbial association networks have also been widely applied in agricultural soils. Nielsen et al. (2014) compared microbial communities from soils with enhanced biochars and traditional fertilizers and found that enhanced biochar application provided similar sweet corn yields although the community composition and networks were quite different compared with standard fertilizers. Menezes et al. (2014) demonstrated that fungi and bacteria were co-correlated and formed distinct associations in soils. In the agricultural soil of Southern China, different fertilizations practices have significantly altered crop yields and soil characteristics (Xun et al., 2016), resulting in the formation of the major low-productivity and high-productivity soils. Soil bacteria are indispensable maintainers of soil productivity (Smith and Paul, 1990). Thus, elucidating the ecological network of the bacterial communities from soils with different productivities is meaningful. However, differences between the networks of the microbial communities in the major lowproductivity and high-productivity arable soils are poorly understood. To investigate the alteration of soil bacterial interaction networks driven by different long-term fertilization management practices, we collected samples, from soils that could be identified as high-productive potential soils (HPP), low-productivity soils (Ac) and artificially remediated soils (Qlime), and analyzed the 16S rRNA genes using 454 high-throughput pyrosequencing. We hypothesized that several specific OTUs might play mainstay roles in a holonomic biological network and could serve as indicators of soil productivity.

2.2. Soil sample characteristics Soil samples were collected in May, 2012. Each plot was divided into two parts, and fresh samples were obtained from the upper 20 cm. Each replicate was a mix of 12 soil cores that were 5 cm in diameter. Four soil samples of each treatment were obtained; three of the samples were randomly selected for analysis. All samples were sieved through a 2 mm sieve. The subsamples used to measure physico-chemical properties were air-dried, and the subsamples used for the molecular analyses were stored at −80 °C prior to DNA extraction. Several physico-chemical and biological properties were measured to assess soil fertility: (i) The soil pH was determined using a PHS 3C mv/pH detector (Shanghai, China) at a soil-to-water ratio of 1:5; (ii) Available N (AN) was measured using the NaOH pervasion method (Bao, 2000); (iii) Available K (AK) in the soil was extracted with ammonium acetate and determined by flame photometry, and available P (AP) in the soil was extracted with sodium bicarbonate and then determined using the molybdenum blue method (Olsen et al., 1954); (iv) Soil organic matter (SOM) (Schollenberger, 1931) was determined by the potassium dichromate volumetric method; and (v) The microbial biomass C (MBC) (Vance et al., 1987) concentration was measured using the chloroform fumigation-extraction method and these values were then transformed to microbial biomass using a kEc factor of 2.64 (Zhong and Cai, 2007). 2.3. 454 pyrosequencing analysis Soil DNA was extracted from 0.25 g subsamples using the PowerSoil DNA Isolation Kit (Mo Bio Laboratories Inc., Carlsbad, CA, USA), and three successive DNA extractions of each sample were pooled before PCR. The DNA quality was assessed with a NanoDrop ND-2000 Spectrophotometer (NanoDrop ND2000, Thermo Scientific, Wilmington, DE, USA). 454 pyrosequencing analysis of the V1-V3 hypervariable regions of the bacterial 16S rRNA gene (Zhao et al., 2014) was performed on a 454 GS-FLX Titanium System (Roche, Switzerland) by Majorbio Bio-pharm Technology Co., Ltd (Shanghai, China). The pyrosequencing data were processed using Mothur (version 1.27.1) (Schloss et al., 2009) following the Schloss standard operating procedure (SOP) (http://www.mothur. org/wiki/454_SOP). More detailed description of the raw sequences analysis can be found in my previous publication (Xun et al., 2016).

2. Materials and methods 2.1. Experimental site description This study was established in the Red Soil Experiential Station (RSES) of the Chinese Academy of Agricultural Sciences, Qiyang (111°53′E, 26°45′N), Hunan Province, southern China. Red soil, which developed from Quaternary red clay, is one kind of Ferralsols according to the World Reference Base for Soil Resources (WRB) (IUSS Working Group, 2014). After three years of homogenization in one field by performing annual rotations of winter wheat (Triticum aestivum L.) and summer maize (Zea mays L.), the experimental field was divided into successions of plots (20 m × 10 m × 0.4 m). All fertilization treatments were

2.4. Soil grouping and network analysis The crop yields, soil properties (Table S1) and bacterial communities were different among these fifteen fertilization managements (Xun et al., 2016). With pH values higher than 5.0 and greater yields, the CK, Fallow, NPKM, M and PK treatments were defined as the highproductivity potential group (HPP); with low pH values and low129

Applied Soil Ecology 120 (2017) 128–134

W. Xun et al.

productivity, the chemical nitrogen fertilization treatments (N, NP, NK, NPK and NPKS) were defined as the acidified group (Ac); the remaining five short-term lime improvement counterparts (NCa, NPCa, NKCa, NPKCa and NPKSCa) were defined as the quicklime improvement group (Qlime). Therefore, all treatments were divided into 3 groups for network constructing. A network graph can be constructed to represent the ecological interactions of different species instead of only considering the actions of individual species in a microbial community. In a network, the interactions between OTUs (nodes) are represented by pairwise positive or negative correlations. Since a module may contain pivotal nodes that are connected to other nodes in the network, the topological role of an individual node showing its potential importance can be calculated according to how well it is connected to other nodes in the same module or in other modules. The normalized relative abundances of OTUs across different samples were used to define the adjacency matrix by calculating the pairwise Pearson correlations. Thus, each pair of linked OTUs changes in the same/opposite tendency according to the positive/negative correlations. Module based high-order organizations in one ecosystem could probably be recognized as stronger alliances in which cooperation (positive correlation) or competition (negative correlation) might appear. However, in fact, these links only indicated the co-occurrence but not necessarily physical interactions between OTUs. The potential interactions could be simply understood as every pair of OTUs respond to the same soil property rather than interacting directly. Network analysis was performed using the Molecular Ecological Network Analyses Pipeline (MENAP) (http://ieg2.ou.edu/MENA/main. cgi) (Zhou et al., 2011). The network analysis could be divided into two steps based on the pipeline: (i) network construction, which included high throughput sequencing data updates, data standardization, pairwise similarity of abundance across different samples and the determination of the adjacency matrix by an RMT-based approach and (ii) network analysis, which included network module detection, eigengene network analyses, network overall topological structure, topological role of node, association of network properties to environmental characteristics, identification of module topological roles for individual nodes and network comparisons between conditions (Zhou et al., 2010).

Fig. 1. (A) Log10 transformation of the abundances of normalized OTU counts for one independent replicate and a test in which both replicates were from the same sample. The intersection of the split lines shows where an OTU with 20 reads would lie in both replicates. (B) Progressive drop-out analysis displaying the R2 correlation of the data in A if OTUs with low read numbers were discarded. Only OTUs with more than 20 reads are considered (red line). The R2 is acceptable at 0.874 (P < 0.05).

For ecological network construction, soil bacterial taxonomic information was collected from the HPP, Ac and Qlime groups. Then, a 20 × 3 threshold (Fig. 1) was used to filter low-quality OTUs, resulting in 256 OTUs, 231 OTUs and 306 OTUs for HPP, Ac and Qlime, respectively. The RMT-based network method was performed using the chosen OTUs from each soil group. Finally, three ecological networks were constructed (Fig. 2), in which different OTUs (nodes) were connected by pairwise positive or negative interactions (links) to investigate the soil bacterial interaction patterns under different soil productivities. The bacterial community compositions were different among three networks. A total of 8.67% (32) of the nodes (OTUs) were shared between Qlime and HPP. Besides, although the bacterial communities from Qlime were the same as these from Ac two years before, some differences were observed in the network composition (Fig. 2). Only 27.04% (83) of the nodes were shared between Ac and Qlime. Such a small proportion of OTUs shared between networks indicated that the soil bacterial interaction patterns were strongly affected by soil management.

2.5. Statistical analysis The relative abundances of all OTUs were used for the network analysis. A progressive drop-out analysis was used to calculate the R2 correlations under different thresholds. Duncan’s multiple range test was used to calculate the significance among samples. Turkey’s HSD test was used to calculate the significance between two samples. Correlations were calculated using Spearman’s rank correlation. All statistical analyses were performed with the Vegan package (v.2.0-8) (Oksanen et al., 2013) in R software version 3.0.1. Cytoscape (v3.0.2) (Shannon et al., 2003) was used to construct models of the ecological networks. 3. Results and discussion 3.1. Ecological networks construction The soil properties, bacterial diversity and community composition were significantly altered by soil fertilization management. The ordination analysis result revealed that the bacterial communities could be separated into 3 groups (as we mentioned previously in Section 2.4 of Materials and Methods), mainly associated with soil pH and SOM. The detailed report on these results have been published elsewhere (Xun et al., 2016). Accordingly, we suggested that soil management would have strong effects on bacterial communities and interaction patterns, which could be well represented by the ecological network.

3.2. Overall structure of ecological networks Following the RMT-based network analysis, very close similarity thresholds (St) were obtained for HPP (0.88), Ac (0.87) and Qlime (0.86), and three networks were constructed with 218 nodes, 183 nodes and 207 nodes, respectively (Table 1). The network fitted the regular 130

Applied Soil Ecology 120 (2017) 128–134

W. Xun et al.

Fig. 2. An overview of the bacterial networks in the HPP (A), Ac (B) and Qlime (C) soil samples. Modules with more than 5 nodes are marked with “HPP*", “Ac*" and “Qlime*" (* indicates a number). Node colors indicate different major taxonomic units. Red lines between nodes (links) indicate negative interactions, whereas blue lines indicate positive interactions.

power-law with R2 values of 0.892, 0.884 and 0.858 for HPP, Ac and Qlime, respectively. These degrees of distribution fit well with regular power-law model, indicating the scale-free property of the three networks (Goh et al., 2001). To describe the differences between networks, we used four network indices: (i) Average connectivity (avgK), which indicates the complexity of a network (higher avgK indicates a more complex network); (ii) Average clustering coefficient (avgCC), which is used to measure the extent of the module structure present in a network; (iii) Average path distance (GD), which measures the efficiency of information or mass transport in one network, with a smaller GD indicating that all of the nodes in one network are closer; and (iv) Modularity, which measures the degree of clarity into which the delimited modules of the network are organized. We observed that the avgK value of Qlime was larger than the values of Ac and HPP, indicating that the Qlime network was a more complex network. Besides, the GD values and avgCC values of the empirical networks were significantly higher than the values of their

corresponding random networks, indicating the small-world behavior of the constructed networks (Brown et al., 2004; Deng et al., 2012; Watts and Strogatz, 1998). Moreover, the modularity values of the empirical networks were higher than their corresponding random networks, indicating that the networks we constructed in the present study were also modular. In addition, we observed that HPP had the highest R2 value and Qlime had the highest avgK and avgCC, whereas the GD and modularity showed the opposite trend. All of these pivotal topological properties of the constructed networks suggested that our networks appeared to be scale-free, small world and modular. These constructed networks may be useful for further research because many complex networks shared the same general features (e.g., scale-free, small world, and module) (Alon, 2003; Barabási and Oltvai, 2004; Clauset et al., 2008). The attributes of edges (positive and negative interactions) could reflect the ways the nodes interacted among modules or within a network. So, the numbers of positive and negative links were counted and the P/N ratios were calculated for the whole network and major

Table 1 Topological properties of the empirical networks of microbial communities under HPP, Ac and Qlime conditions and their associated random networks. Condition

HPP Ac Qlime

Random networksa

Empirical networks No. of original OTUsb

Similarity threshold (St)

Network sizec

R square of powerlaw

Average connectivity (avgK)

Average clustering coefficient (avgCC)

Average path distance (GD)

Modularity (no. of modules)

Average path distance (GD)

Average clustering coefficient (avgCC)

Average modularity

256 231 306

0.88 0.87 0.86

218 183 207

0.892 0.884 0.858

4.202 4.601 13.246

0.113 0.174 0.221

3.74 3.718 3.137

0.629 (19) 0.602 (21) 0.337(6)

3.340 ± 0.154 3.121 ± 0.158 2.485 ± 0.025

0.046 ± 0.009 0.058 ± 0.010 0.110 ± 0.012

0.455 ± 0.007 0.416 ± 0.009 0.187 ± 0.005

HPP, high productive potential soil group; Ac, acidified soil group; Qlime, quicklime improvement soil group. a The random networks were generated by rewiring all of the links of a network with the identical numbers of nodes and links to the corresponding empirical network. b Number of originally used OTUs for network construction. c Number of nodes (OTUs) in a network.

131

Applied Soil Ecology 120 (2017) 128–134

W. Xun et al.

Table 2 Numbers of positive and negative links between nodes from each module and the P/N ratios. Edge Attribute

Positivea Negativeb P/N ratioc

Module-HPP

Module-Ac

Module-Qlime

HPP1

HPP2

HPP3

HPP4

HPP5

HPP6

Total

Ac1

Ac2

Ac3

Ac5

Ac6

Ac7

Total

Qlime1

Qlime2

Qlime3

Total

155 4 38.8

128 13 9.8

8 20 0.4

23 8 2.9

7 21 0.3

11 4 2.8

332 70 4.7

38 77 0.5

70 70 1

26 23 1.1

11 3 3.7

9 8 1.1

7 3 2.3

161 184 0.9

307 199 1.54

134 241 0.56

122 31 3.94

563 471 1.2

HPP, high productive potential soil group; Ac, acidified soil group; Qlime, quicklime improvement soil group. a Positive links between two nodes in one module. b Negative links between two nodes in one module. c Ratio of the number of positive links and negative links.

links within modules (Pi = 0). The other 3.78% of the nodes were generalists (module hubs and connectors), including 1.81% that belonged to connectors and 1.97% that belonged to module hubs. Module hubs or connectors are requisites in networks, although sometimes they do not exist simultaneously (Guimera et al., 2005). More module hubs could keep the community structure in order. In the three networks constructed in this study, seven module hubs were from HPP (OTU09960, Acidobacteria_Gp3; OTU09965, Acidobacteria_Gp6; OTU07660, Acidobacteria_Gp6; OTU08465, β-Proteobacteria; OTU08414, γ-Proteobacteria; OTU07737, γ-Proteobacteria and OTU00858, unclassified Bacteria), four OTUs were from Qlime (OTU09390, Actinobacteria; OTU07859, Firmicutes; OTU07446, Gemmatimonadetes and OTU07638, α-Proteobacteria) and only one OTU was from Ac (OTU09939 Actinobacteria). Thus, the OTUs belonging to module hubs from HPP occupied a large fraction (58.33%), followed by Qlime (33.34%) and Ac (8.33%). Therefore, the HPP network is more orderly than Ac and Qlime. On the other hand, more connectors could organize a series of modules into a complete community, resulting in more efficient energy metabolism, nutrient cycling and substance transformation in the soil. Here, we observed eleven connectors, including three OTUs from HPP (OTU09192, Acidobacteria_Gp1; OTU08581, Actinobacteria and OTU02740, unclassified Bacteria), four OTUs from Ac (OTU08996, Acidobacteria_Gp1; OTU05928, Actinobacteria; OTU08271, Actinobacteria and OTU08853, unclassified Bacteria) and four OTUs from Qlime (OTU01740, Gemmatimonadetes; OTU05516, α-Proteobacteria; OTU01368, β-Proteobacteria and OTU05482, Acidobacteria_Gp1). Thus, the OTUs belonging to the connectors from the three sample groups were nearly the same. Oldham et al. (2008) performed a comprehensive analysis of gene coexpression network organization in the human brain through the oftinvoked principle of guilt by association and demonstrated that coexpressed genes with strong module memberships in transcriptomes are robustly driven by the same underlying cellular composition. Thus, the bacterial OTUs with strong module memberships (module hubs) might play roles as functional or only physical associators in a bacterial community, but their topological roles would be changed by underlying factors. Here, we observed that only one OTU (OTU09960) of the seven module hubs from HPP could be found in the specialists from Ac, whereas two of the four module hubs from Qlime (OTU09390 and OTU07638) could be found in the specialists from Ac and three (OTU07446, OTU07859 and OTU09390) could be found in the specialists from HPP. Additionally, we calculated the correlations between the relative abundances of all OTUs and the soil properties (Table S2) and observed that all of the module hubs had significant relationships with the soil properties, whereas the connectors did not. Thus, the topological roles of individual nodes could be altered by different soil managements. Taken together, higher number of generalists in HPP indicated a more orderly community. In this case, the HPP network could appear to be a better organized network than Ac. Quicklime remediation increased the number of module hubs, indicating that quicklime provided

modules (with more than 10 nodes in one module) (Table 2). Firstly, we observed more positive interactions (links) in HPP than Ac, indicating that HPP might contain more cooperators (bacteria having cooperative interactions) while Ac might contain more defectors (bacteria having competitive interactions). Therefore, the HPP network appeared to be a better operated network than Ac. Secondly, for the separate modules, four of the six modules in HPP had high P/N ratios (> 2.8), whereas the other two modules had low P/N ratios (< 0.3). The P/N ratios in Ac were closer to 1 (0.5 ≤ P/N ratio ≤ 2.3) with the exception of “Ac5”. Moreover, the P/N ratio deviated from the value of 1 in Qlime (Table 2) and the P/N ratio of Qlime increased to 1.2 from the 0.9 observed in Ac, indicating that more cooperators appeared after the quicklime improvement. Generally, cooperators make contributions whereas the defectors do not; the contributions are gathered, transformed, and then redistributed to everybody (including to the crops). The cooperators benefit more from their own contributions than the defectors (Archetti et al., 2011). Since quicklime amendment brought positive bacterial links, we suggested that quicklime has a positive effect on the bacterial community of acidified soil. However, the higher avgK and avgCC values and lower GD and modularity values in Qlime compared to HPP and Ac demonstrated a more complex network in Qlime, probably due to the shortterm improvement. 3.3. Different topological roles of individual nodes Each ecological network was constructed with a certain number of nodes. The topological roles of individual nodes reflect the potential importance of OTUs in the bacterial community and were usually described with Zi (“How well-connected”; used to describe how well a node is connected to other nodes in the same module) and Pi (“How well-distributed”; used to describe how well a node is connected to the nodes in other modules) (Guimerà and Amaral, 2005). Therefore, different roles of individual nodes from each network in the Z-P plot indicated that a variety of soil properties shifted the ecological network structure and the potential ecological function. According to the thresholds of Zi and Pi simplified in the pollination network (Olesen et al., 2007), the whole network was divided into five categories (Zhou et al., 2011): (i) fringing nodes with Zi < 2.5 and Pi = 0, indicating species only interacting with others in the same module; (ii) peripherals with Zi < 2.5 and 0 < Pi < 0.62, indicating species not interacting much with others within and among module; (iii) module hubs with Zi ≥ 2.5 and Pi < 0.62, indicating species interacting much with others within module; (iv) connectors with Zi < 2.5 and Pi ≥ 0.62, indicating species interacting much with others among modules; and (v) network hubs with Zi ≥ 2.5 and Pi ≥ 0.62, indicating species that play roles as both module hubs and connectors. In this study, no network hubs were observed, and all of the nodes were divided into four categories (Fig. 3). The majority (96.22%) of the nodes were specialists (fringing nodes and peripherals) with most of their links inside their own module, including 66.67% that only had 132

Applied Soil Ecology 120 (2017) 128–134

W. Xun et al.

Fig. 3. Zi-Pi plot showing the distribution of OTUs based on their topological roles. Each symbol represents an OTU in Ac, HPP or Qlime. OTUs pointed out belonging to generalists indicate the OTU numbers and taxa. OTUs pointed out belonging to specialists indicate the OTUs are specialists in Ac or HPP while generalists in Qlime.

Appendix A. Supplementary data

remissions on acidified soils and make the bacterial community more organized. Consequently, it is reasonable to provide our speculations on agricultural upland red soil bacterial communities. First, connectors always appeared either in high-productivity soils or in low-productivity soils. Second, module hubs would be the most important members for the identification of high-productivity soils and low-productivity soils. Third, module hubs are usually significantly restricted by soil properties whereas the connectors are not, suggesting that the global roles of the module hubs in the network might be described as indicators (Guimerà and Amaral, 2005).

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.apsoil.2017.08.013. References Alon, U., 2003. Biological networks: the tinkerer as an engineer. Science 301, 1866–1867. Archetti, M., Scheuring, I., Hoffman, M., Frederickson, M.E., Pierce, N.E., Yu, D.W., 2011. Economic game theory for mutualism and cooperation. Ecol. Lett. 14, 1300–1312. Aylward, F.O., Eppley, J.M., Smith, J.M., Chavez, F.P., Scholin, C.A., DeLong, E.F., 2015. Microbial community transcriptional networks are conserved in three domains at ocean basin scales. Proc. Natl. Acad. Sci. 112, 5443–5448. Bao, S.D., 2000. Soil and Agricultural Chemistry Analysis. China Agriculture Press, Beijing, pp. 355–356. Barabási, A.L., Oltvai, Z.N., 2004. Network biology: understanding the cell’s functional organization. Nat. Rev. Genet. 5, 101–113. Barberán, A., Bates, S.T., Casamayor, E.O., Fierer, N., 2012. Using network analysis to explore co-occurrence patterns in soil microbial communities. ISME J. 6, 343–351. Brown, K.S., Hill, C.C., Calero, G.A., Myers, C.R., Lee, K.H., Sethna, J.P., Cerione, R.A., 2004. The statistical mechanics of complex signaling networks: nerve growth factor signaling. Phys. Biol. 1, 184–195. Cattin, M.F., Bersier, L.F., Banašek-Richter, C., Baltensperger, R., Gabriel, J.P., 2004. Phylogenetic constraints and adaptation explain food-web structure. Nature 427, 835–839. Chaffron, S., Rehrauer, H., Pernthaler, J., von Mering, C., 2010. A global network of coexisting microbes from environmental and whole-genome sequence data. Genome Res. 20, 947–959. Clauset, A., Moore, C., Newman, M.E.J., 2008. Hierarchical structure and the prediction of missing links in networks. Nature 453, 98–101. Deng, Y., Jiang, Y.H., Yang, Y., He, Z., Luo, F., Zhou, J., 2012. Molecular ecological network analyses. BMC Bioinf. 13, 113. Dunne, J.A., Williams, R.J., Martinez, N.D., 2002. Food-web structure and network theory: the role of connectance and size. Proc. Natl. Acad. Sci. 99, 12917–12922. Eldridge, D.J., Woodhouse, J.N., Curlevski, N.J., Hayward, M., Brown, M.V., Neilan, B.A., 2015. Soil-foraging animals alter the composition and co-occurrence of microbial communities in a desert shrubland. ISME J. 9, 2671–2681. Faust, K., Raes, J., 2012. Microbial interactions: from networks to models. Nat. Rev. Microbiol. 10, 538–550. Freilich, S., Kreimer, A., Meilijson, I., Gophna, U., Sharan, R., Ruppin, E., 2010. The largescale organization of the bacterial network of ecological co-occurrence interactions. Nucleic Acids Res. 38, 3857–3868. Fuhrman, J.A., Cram, J.A., Needham, D.M., 2015. Marine microbial community dynamics and their ecological interpretation. Nat. Rev. Microbiol. 13, 133–146. Fuhrman, J.A., 2009. Microbial community structure and its functional implications.

4. Conclusions We have shown the interactions and network organizations of bacterial communities in high-productivity potential soils, low-productivity soils and artificially remediated soils using network analysis. Our results demonstrated that the bacterial community of HPP was a better organized or a better operated community than Ac and that quicklime application helped to order the bacterial community. The soil management altered the bacterial community and shifted the potential ecological roles of OTUs. We also suggested that the module hubs could be important indicators for the identification of high-productivity soils and low-productivity soils.

Acknowledgements The authors thank the staff at the Qiyang Red Soil Experimental Station for managing the field experiments and helping with the collection of soil samples. This research was financially supported by the National Natural Science Foundation for Young Scientists of China (41601252), General Financial Grant from the China Postdoctoral Science Foundation (2016M601833), the National Key Basic Research Program of China (973 program, 2015CB150500), the Fundamental Research Funds for the Central Universities (KJQN201748). 133

Applied Soil Ecology 120 (2017) 128–134

W. Xun et al.

Nat. Neurosci. 11, 1271–1282. Olesen, J.M., Bascompte, J., Dupont, Y.L., Jordano, P., 2007. The modularity of pollination networks. Proc. Natl. Acad. Sci. 104, 19891–19896. Olsen, S.R., Cole, C.V., Watanabe, F.S., Dean, L.A., 1954. Estimation of Available Phosphorus in Soils by Extraction with Sodium Bicarbonate. US Department of Agriculture, Washington, DC. Petchey, O.L., Gaston, K.J., 2006. Functional diversity: back to basics and looking forward. Ecol. Lett. 9, 741–758. Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., Lesniewski, R.A., Oakley, B.B., Parks, D.H., Robinson, C.J., Sahl, J.W., Stres, B., Thallinger, G.G., Horn, D.J.V., Weber, C.F., 2009. Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities. Appl. Environ. Microbiol. 75, 7537–7541. Schollenberger, C.J., 1931. Determination of soil organic matter. Soil Sci. 31, 483–486. Shannon, P., Markiel, A., Ozier, O., Baliga, N.S., Wang, J.T., Ramage, D., Amin, N., Schwikowski, B., Ideker, T., 2003. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 13, 2498–2504. Smith, J.L., Paul, E.A., 1990. The significance of soil microbial biomass estimations. Soil Biochem. 6, 357–396. Steele, J.A., Countway, P.D., Xia, L., Vigil, P.D., Beman, J.M., Kim, D.Y., Chow, C.-E.T., Sachdeva, R., Jones, A.C., Schwalbach, M.S., 2011. Marine bacterial, archaeal and protistan association networks reveal ecological linkages. ISME J. 5, 1414–1425. Torsvik, V., Goksøyr, J., Daae, F.L., 1990. High diversity in DNA of soil bacteria. Appl. Environ. Microbiol. 56, 782–787. Vance, E.D., Brookes, P.C., Jenkinson, D.S., 1987. An extraction method for measuring soil microbial biomass C. Soil Biol. Biochem. 19, 703–707. Watts, D., Strogatz, S., 1998. Collective dynamics of small-world networks. Nature 393, 440–442. Wigner, E.P., 1967. Random matrices in physics. SIAM Rev. 9, 1–23. Xun, W., Zhao, J., Xue, C., Zhang, G., Ran, W., Wang, B., Shen, Q., Zhang, R., 2016. Significant alteration of soil bacterial communities and organic carbon decomposition by different long-term fertilization management conditions of extremely lowproductivity arable soil in South China. Environ. Microbiol. 18, 1907–1917. Zhang, B., Horvath, S., 2005. A general framework for weighted gene co-expression network analysis. Stat. Appl. Genet. Mol. Biol. 4, 1–45. Zhao, J., Zhang, R., Xue, C., Xun, W., Sun, L., Xu, Y., Shen, Q., 2014. Pyrosequencing reveals contrasting soil bacterial diversity and community structure of two main winter wheat cropping systems in China. Microb. Ecol. 67, 443–453. Zhong, W.H., Cai, Z.C., 2007. Long-term effects of inorganic fertilizers on microbial biomass and community functional diversity in a paddy soil derived from quaternary red clay. Appl. Soil Ecol. 36, 84–91. Zhou, J., Deng, Y., Luo, F., He, Z., Tu, Q., Zhi, X., 2010. Functional molecular ecological networks. mBio 1 e00169–10–e00169–19. Zhou, J., Deng, Y., Luo, F., He, Z., Yang, Y., 2011. Phylogenetic molecular ecological network of soil microbial communities in response to elevated CO2. mBio 2 e00122–11–e00122–11.

Nature 459, 193–199. Goh, K.I., Kahng, B., Kim, D., 2001. Universal behavior of load distribution in scale-free networks. Phys. Rev. Lett. 87, 278701. Guimerà, R., Amaral, L.A.N., 2005. Functional cartography of complex metabolic networks. Nature 433, 895–900. Guimera, R., Mossa, S., Turtschi, A., Amaral, L.N., 2005. The worldwide air transportation network: anomalous centrality, community structure, and cities’ global roles. Proc. Natl. Acad. Sci. 102, 7794–7799. Han, J.D.J., Bertin, N., Hao, T., Goldberg, D.S., Berriz, G.F., Zhang, L.V., Dupuy, D., Walhout, A.J.M., Cusick, M.E., Roth, F.P., Vidal, M., 2004. Evidence for dynamically organized modularity in the yeast protein?protein interaction network. Nature 430, 88–93. Hooper, D.U., Chapin, F.S., Ewel, J.J., Hector, A., Inchausti, P., Lavorel, S., Lawton, J.H., Lodge, D.M., Loreau, M., Naeem, S., Schmid, B., Setälä, H., Symstad, A.J., Vandermeer, J., Wardle, D.A., 2005. Effects of biodiversity on ecosystem functioning: a consensus of current knowledge. Ecol. Monogr. 75, 3–35. IUSS Working Group, 2014. World Reference Base for Soil Resources 2014 International Soil Classification System for Naming Soils and Creating Legends for Soil Maps. FAO, Rome. Kelder, T., Stroeve, J.H., Bijlsma, S., Radonjic, M., Roeselers, G., 2014. Correlation network analysis reveals relationships between diet-induced changes in human gut microbiota and metabolic health. Nutr. Diabetes 4, e122. Lagkouvardos, I., Kläring, K., Heinzmann, S.S., Platz, S., Scholz, B., Engel, K.H., SchmittKopplin, P., Haller, D., Rohn, S., Skurk, T., 2015. Gut metabolites and bacterial community networks during a pilot intervention study with flaxseeds in healthy adult men. Mol. Nutr. Food Res. 59, 1614–1628. Lidicker, W.Z., 1979. A clarification of interactions in ecological systems. Bioscience 29, 475–477. Maslov, S., 2002. Specificity and stability in topology of protein networks. Science 296, 910–913. Menezes, A.B., Prendergast-Miller, M.T., Richardson, A.E., Toscas, P., Farrell, M., Macdonald, L.M., Baker, G., Wark, T., Thrall, P.H., 2014. Network analysis reveals that bacteria and fungi form modules that correlate independently with soil parameters. Environ. Microbiol. 17, 2677–2689. Naeem, S., Wright, J.P., 2003. Disentangling biodiversity effects on ecosystem functioning: deriving solutions to a seemingly insurmountable problem. Ecol. Lett. 6, 567–579. Nielsen, S., Minchin, T., Kimber, S., van Zwieten, L., Gilbert, J., Munroe, P., Joseph, S., Thomas, T., 2014. Comparative analysis of the microbial communities in agricultural soil amended with enhanced biochars or traditional fertilisers. Agric. Ecosyst. Environ. 191, 73–82. Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., Minchin, P.R., O’Hara, B., Simpson, G.L., Leslie, G., Solymos, P., Stevens, H., Wagner, H.H., 2013. The Vegan R Package Version 2.0-8: Community Ecology. Available at: http://CRAN.R-project.org/ package=vegan. Oldham, M.C., Konopka, G., Iwamoto, K., Langfelder, P., Kato, T., Horvath, S., Geschwind, D.H., 2008. Functional organization of the transcriptome in human brain.

134