Continuous‐cropping tobacco caused variance of chemical properties and structure of bacterial network in soils

Continuous cropping of the same crop leads to land degradation. This is also called the continuous‐cropping obstacle. Here, we investigated how long‐term continuous cropping of tobacco influences soil biochemical properties and bacterial networks in the mountain lands of China. Two different fields were sampled: one with 25 years of continuous cropping tobacco and one with noncontinuous cropping tobacco. Soil chemical and biological properties were measured including available phosphorus and potassium, soil organic matter, pH, alkali‐hydrolysable nitrogen, micronutrients contents, and activity of urease, catalase, invertase, and phosphatase as well as tobacco agronomic characteristics. Bacterial communities of the two different soils were sequenced by metabarcoding of the 16S ribosomal RNA, and, with these data, network analysis was done. Soil chemical properties and tobacco agronomical properties were negatively affected by the continuous‐cropping obstacle, and this treatment has a less complex network (less modules, nodes, and connectivity) than the soil with noncontinuous cropping treatment. For continuous cropping, there were less generalists, which were key species that connect network, than noncontinuous cropping. Moreover, the taxonomic composition of bacterial network was different in the two different treatments. In the continuous‐cropping network, 40% nodes had negative interactions, suggesting that more competition or antagonism existed among bacterial species. It concluded that continuous cropping has a detrimental effect on soil chemical and that the bacterial network properties under continuous cropping are more sensitive to soil variables (so more unstable and inefficient) because there are less bacterial species that interact each other and this is due to limited nutrients or excessive toxic nutrient.

Tobacco has been used as a model organism in many studies, laying the groundwork for agricultural biotechnology (Sierro et al., 2014). Tobacco is also a major crop for millions of Chinese farmers and widely planted in Hubei Province, China. Due to limited cultivated lands and lacking of reasonable planting concept, the majority of tobacco is usually planted by a continuous cropping (CC) pattern that often leads to continuous-cropping obstacle with poor seedling growth and yield reduction . Tobacco selectively extracts some nutrients (e.g., phosphorus, nitrogen, and potassium) from soil, hence resulting in poor soil fertility and imbalanced soil nutrients. The fundamental difference between tobacco and other crops is the specific biological active substances in tobacco such as secondary metabolites aroma components. Almost all of allelopathic substances are secondary metabolic products, and autotoxic allelopathy caused by allelochemicals is the main reason for continuouscropping obstacle (Ren et al., 2015). Generally, tobacco more easily generates autotoxic allelopathy than other plants (Deng et al., 2017;Farooq, Hussain, Wakeel, & Cheema, 2014). For example, the extracts of tobacco rhizosphere soil significantly inhibit tobacco seed germination and negatively influence plant growth and development (Ren et al., 2015). Thereby, continuous-cropping obstacle may be caused by deterioration of soil physicochemical properties, nutrient imbalance, and allelopathy . For this reason, tobacco was used as a model to study land degradation caused by continuouscropping obstacle in this study.
Soil microorganisms are involved in many key processes of soil ecosystem including nutrition cycling, organic matter turnover, soil structure maintenance, and toxin degradation (Brussaard, Ruiter, & Brown, 2007). Due to their quick response to environmental changes and agricultural practices, soil microorganisms are considered as efficient biological indicators for soil fertility and land management (Avidano, Gamalero, Cossa, & Carraro, 2005;Wu et al., 2017). Many studies found that long-term monocropping can result in a decrease of microbial activity and diversity and change the composition and structure of soil microbial community (Nayyar et al., 2009;van Elsas, Garbeva, & Salles, 2002). In the CC soil, microbial community is continuously exposed to root exudates of the same crop year after year that potentially enriches the crop-specific microorganisms in soil. Therefore, CC can lead to the reduction in abundances of beneficial bacteria (e.g. Arthrobacter and Lysobacter) and the multiplication of detrimental bacteria (e.g., soil-borne pathogens) in soil (She et al., 2017;Wu et al., 2017). In our previous studies, we found that the abundances of beneficial bacteria and fungi were significantly decreased in the soils with monoculture of tobacco (Wang et al., 2017). Similarly, monoculture of potato also causes the shift of overall bacterial communities and the decrease of abundances of Acidobacteria and Nitrospirae in soils (Liu et al., 2014). However, until now, the focus of many studies has been put on the diversity measures and community composition through analysis of metabarcoding data without taking into account the interrelation between bacterial taxa (network analysis), and few studies have investigated the changes of microbial network in the CC land. The complex interactions among different microbial species and the key microbes changed by continuous-cropping obstacle are still unknown.
In soils, various microorganisms form a complex network of interactions. Different soil microbial species interact with each other through various types of interactions including predation, parasitism, competition, amensalism, commensalism, or mutualism (Olesen, Bascompte, Dupont, & Jordano, 2007). All of these interactions among different microbial species together shape the overall structure of a microbial community and determine the ecosystem stability (Hibbing, Fuqua, Parsek, & Peterson, 2010;Zhou, Deng, Luo, He, & Yang, 2011). In the complex ecosystems such as soil, the microbial interactions are considered more important to ecosystem function than the species diversity and abundance (Bascompte, 2007).
Microbe-microbe interactions are a vital part of the soil microbiome . By interacting with each other, soil microbes significantly drive global biogeochemical circulation (Gans, Wolinsky, & Dunbar, 2005;Montoya, Pimm, & Solé, 2006). Knowledge of the microbial interactions can help us to identify microbial environmental niches, reveal microbial ecological roles, and expand our understandings for millions of uncultured microbes in soil. Co-occurrence networks of microbial species can ascertain which microbial species share the same niches and play the key roles in the complex networks. The key microorganisms that maintain the stability and operation of microbial community can be discerned by analysing their roles in the networks (Barberán, Bates, Casamayor, & Fierer, 2012).
Recently, phylogenetic molecular ecological networks (pMENs) have been used to investigate the ecosystem, which can provide us with new understanding of microbial ecology (Barberán et al., 2012;Ding et al., 2015;Lu et al., 2013;Shi et al., 2016;Zhou et al., 2011). By pMENs, Zhou et al. (2011) found that the network compositions and structures under ambient CO 2 are substantially different from the network under elevated CO 2 level. The network structure of microbial communities subjected to long-term warming is affected by environmental traits including temperature and soil pH (Deng et al., 2012). Lu et al. (2013) found that the healthy soils, in which the crops grew well with a high yield, had more interrelated operational taxonomic units (OTUs) than the microbial network of diseased soils, in which the crops grew poorly with a low yield.
Despite the importance of microbial interactions for soil ecosystem, the relationship between land degradation and ecological clusters of soil microbial taxa (a group of microbial species that strongly interact each other) has not been previously investigated at a level of network analysis.
As we suspected, land degradation such as continuous-cropping obstacle may result in significant changes in the correlation network of soil microorganisms; however, empirical evidence for such assumptions is currently lacking. By analysis of soil microbial network, we can forecast the variation tendency of soil ecosystem and their responses to land degradation. To investigate the bacterial networks in response to land degradation caused by long-term CC, we analysed the bacterial 16S ribosomal RNA gene sequences by high-throughput sequencing technologies from the soil samples of CC fields and noncontinuous cropping (NCC) fields, and then the ecological networks of soil microbial communities under NCC and CC were constructed by the random matrix theory-based approach, respectively (Luo et al., 2007). These networks could provide us more information about the potential effects of land degradation on microbial interactions and microbial communities than the previous studies that mainly focused on the diversity and composition of microbial community.
General topography is mostly high mountains, with elevations ranging from 959 to 1,140 m in this area.
The soil samples included two groups: soils with CC tobacco and soils with NCC tobacco. For CC soils, farmers' fields were annually mono-cultivated with tobacco for more than 25 years without rotation with other crops. The intensive agricultural management practices accompanying with increased chemical fertilizer applications (97.5 kg/ha chemical fertilizers including N:P 2 O 5 :K 2 O = 1:1.5:2.5 was applied annually) and non-application of organic amendments had caused severe land degradation in these fields. For NCC soils, farmers' fields were cultivated with tobacco-corn rotation for more than 25 years (1 year for tobacco and another year for corn). In the NCC fields, 1,950 kg/ha organic fertilizer (35% organic matter, 2% N, 0.4% P 2 O 5 , 1.5% K 2 O) and 58.5 kg/ha chemical fertilizers (N:P 2 O 5 :K 2 O = 1:1.5:2.5) were applied annually. Other soil management measures (such as irrigation, planting, and pesticide application) were same for the NCC and CC fields. Tobacco grew dwarf and stunted in the CC soils but grew vigourous and robust in the NCC soils (Table 1).
Soil samples were collected from 18 plots before transplanting tobacco. One composite soil sample was collected for each plot. In each plot (667 m 2 ), soils were randomly sampled at five different sites from tillage layer soils (0-to 15-cm depth and 10-cm diameter) and then well mixed together to form a composite soil sample. Totally, 10 CC soil samples and eight NCC soil samples were obtained. Samples were homogenized through a 2-mm sieve following divided into two subsamples: one for DNA extraction and another for soil chemical properties and enzymatic analysis.

| Analysis of soil chemical properties
Soil chemical properties were measured as described previously (Bao, 2013). Briefly, contents of available P and available K (AK) were determined photometrically by a flame spectrophotometer. Soil organic matter (SOM) content was assayed with acidified potassium dichromate (K 2 Cr 2 O 7 -H 2 SO 4 ). Soil pH was determined using a pH metre. Alkali-hydrolysable nitrogen content was determined by alkaline hydrolysis diffusion method. Contents of micronutrients (Ca, Mg, Na, Fe, Mn, Zn, and B) were measured as previously described in the protocol of Zheng (1996).

| Determination of agronomic characters of tobacco
Ninety days after transplanting, a total of 200 tobacco seedlings were randomly selected from each field to investigate the height, stem girth, and leaf number of tobacco. The maximum leaf area and top leaf area in each tobacco seedling were also detected (Wang et al., 2017).

| High-throughput sequencing of soil microbial communities
For each sample, DNA was extracted from 0.4 g of soils using the FastDNA Spin Kit (MP Biomedicals, USA) according to the manufacturer's protocol, quantified by NanoDrop spectrophotometer (Thermo Fisher Scientific, USA), and then used as polymerase chain reaction templates. Primer pair 338F (5′-ACTCCTACGGGAGGCA GCA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) were used for amplifying V3 and V4 regions of bacterial 16S ribosomal RNA genes, which were routinely used to characterize soil bacterial communities and bacterial species (Kim, Morrison, & Yu, 2011), and then the amplified products were purified by AMPure XP beads.
Amplicons were sequenced on Illumina-MiSeq platform (Illumina Inc., USA) at SHBIO Technology Co. Ltd (Shanghai, China). Raw sequence data were preliminarily filtrated using FASTX Toolkit 0.0.13 software package to remove low mass base at the tail of sequence (Q value <20) and the sequences less than 35 BP, and then the valid reads were chosen with an approximate length about 250 BP (Wang et al., 2017). OTUs were assigned at 97% similarity level, and then OTU abundances were normalized to the standardized relative abundances (SRA).

| Construction and analysis of microbial network
We used the online tool named Molecular Ecological Network Analysis Pipeline (http://ieg2.ou.edu/MENA) to construct microbial networks. Briefly, a matrix of OTUs SRA, a matrix of soil variables, and an OTU annotation file were prepared respectively, and then the SRA matrix was submitted to Molecular Ecological Network Analysis Pipeline to construct networks. Thereafter, modules were detected by the greedy modularity optimization, and three files were generated for network graphs visualization by Cytoscape 3.4.0 software (Su, Morris, Demchak, & Bader, 2014). Network graph was represented using different OTUs (nodes) with positive or negative interactions (edges). Positive interactions indicate that the abundance of OTUs changes along the same trend across different soil samples, whereas negative interactions indicate that the abundance of OTUs changes along the opposite trend in different soil samples (Lu et al., 2013). pMENs were constructed for the CC and NCC soils, respectively. Each network was separated in different modules, which are different interrelated units in the microbial community. A module is a group of microbial species (nodes) that strongly interact each other among themselves but rarely interact with species in other modules (Bascompte & Stouffer, 2009). Modularity (M) measures the degree on how the network is organized into clearly delimited modules. Various network indexes, including average connectivity (avgK), geodesic distance (avgGD) and clustering coefficient (avgCC), and modularity, were used to describe the topology properties and structure of networks. Connectivity is the number of links of a node with other nodes. Geodesic distance is the shortest path between two nodes. Clustering coefficient describes how well a node is connected with the neighbour nodes. By calculating correlations among module eigengenes, eigengene network analysis can reveal the high-order organization of networks structure and identify the key populations of networks. Module membership (MM), the square of Pearson correlation between the abundance profile of a given OTU and a given module eigengene, shows the key OTUs within a module. The groups of eigengenes with considerable correlations are referred as a meta-module (Langfelder & Horvath, 2007).

| Statistical analysis
One-way analysis of variance was performed using SPSS 20.0 software to test all parameters both at 0.05 and 0.01 significance level.
Statistics of networks was calculated including network properties, module eigengene, and randomization of network structure. For assessing statistical significance of networks indexes, a total of 100 randomly rewired networks were generated, and then the differences of indexes between pMEN and random networks were detected by Z test for the NCC and CC networks, respectively. Correlation coefficients for networks were calculated on the basis of global network properties, individual nodes' centrality, module separation, and modularity (Clauset, Newman, & Moore, 2004). Mantel test was conducted to discern the relationships between module eigengenes and soil properties using R vegan package. Singular value decomposition was used for data reduction in eigengene network analysis (Langfelder & Horvath, 2007 Tobacco height, stem girth, leaf numbers, maximum leaf area, top leaf area, and yield were all significantly greater in the NCC soils than those in the CC soils (Table 1c).

| Comparing topological properties and structure of networks between CC and NCC
Two networks were constructed for the CC and NCC soils, respectively ( Figure 1) The avgCC of empirical networks was higher than the corresponding random networks (0.004-0.011), suggesting that these two networks had both the typical small-world characteristics (Table 2). Modularity (M) values of the NCC network and CC network were both higher than the threshold value for modular structure (0.4), suggesting that these two networks were both modular (Newman, 2006). These results suggested that the topological structures of two networks were quite different. Overall, the soil microbial network of NCC was composed of more highly connected OTUs to form a more clustered topology structure than the CC network.

| Comparing composition of networks between CC and NCC
Most of nodes in these two networks belonged to 13 bacterial phyla ( Figure 1). Among them, Acidobacteria, Actinobacteria, and   The high-order structure of two networks was further analysed.
Eigengenes from different modules were significantly correlated to form meta-modules. Three meta-modules were clustered in the NCC

| Key microorganisms of networks
Topological roles of nodes in two networks were shown in Figure 5. Firmicutes was closely related to Bacillus, which is popularly used to control plant diseases and decompose organic matters (Meng, Jiang, & Hao, 2016;Shah et al., 2013). OTU 869 (module hub) was assigned to aerobic denitrifying Hyphomicrobiaceae. Four module hubs (OTU 9, OTU 13, OTU 44, and OTU 164) and one connector (OTU 59) were all belonged to Acidobacteria with a leading position in the network.

Iron (Fe) is an essential micronutrient both for plants and soil
microbes (Raymond, Dertz, & Kim, 2003). Many bacteria belonging to Firmicutes and Proteobacteria (e.g., Bacillus, Burkholderia, Enterobacter, Erwinia, Paenibacillus, Pantoea, Pseudomonas, Serratia, and Stenotrophomonas) can synthesize siderophores to absorb Fe from the surrounding environment , hence occupying favourable niches via depriving Fe from the neighbouring competitors (McNally & Brown, 2015). The siderophore producers also play an important role in promoting plant uptake of Fe .
Here, the MMs belonging to Firmicutes and Proteobacteria were enriched in the NCC network (Figure 2b), possibly promoting the growth of tobacco and soil bacteria by Fe absorption.
Boron (B) is an essential plant micronutrient, but excessive B is toxic to both of plants (Aitken & McCallum, 1988) and of bacteria  2007). Here, B content was significantly higher in the CC soils than the NCC soils (Table 1b). High content of B in the CC soils was possibly detrimental for the bacterial community and tobacco growth and thus might be one of the constraints to limit soil quality that is defined as the capacity of soil to sustain biological productivity, maintain environmental quality, and promote plant and human health (Andrews, Karlen, & Mitchell, 2002). For this reason, reduced application of B fertilizer might be favourable for alleviating continuous-cropping obstacle in this area. These results indicated that long-term CC tobacco had led to land degradation (limited nutrients or excessive toxic nutrient), hence hampering plant growth in the CC fields (Table 1c).
Interactions among different microbial species are very important for ecosystem community's function in the complex soil ecosystem (Barberán et al., 2012;Lupatini et al., 2014). Thus, network analysis has been increasingly used to explore the potential microbial interactions in different ecosystems (de Menezes et al., 2015;Eiler, Heinrich, & Bertilsson, 2012). However, few attempts have been done to study the change of co-occurrence network patterns at degraded lands. In this study, we employed network analysis to visualize and quantify Two pMENs were constructed based on random matrix theory approach, which can automatically identify the thresholds for network construction; hence, the constructed networks are considered accurate and reliable (Deng et al., 2012;Luo et al., 2007;Zhou et al., 2010). This study provides a new insight into the crucial roles of microbial interactions that can enhance our understanding of the response of microbial communities to land degradation caused by continuous-cropping obstacle.
Tobacco continuous-cropping obstacle had a significantly negative impact on the network interactions of microbial communities and had changed the structures and compositions of microbial network. In the CC soils, the microbial network had fewer interactions among different microbial populations when compared with the NCC soils ( Figure 1). Microbe-microbe interactions within communities contribute to cause the complexity in community structure and maintain the stability of ecosystem (e.g., soil microbiome; Barberán et al., 2012;Hibbing et al., 2010;Schmitt et al., 2012;Zhang et al., 2014;Zhou et al., 2011). In the CC soils, the microbe-microbe interactions within communities seemed to be weak for maintaining the complexity of community structure and the stability of ecosystem.  (Freilich et al., 2010). These results suggested that CC dramatically changed the network interactions of various microbial populations in soils.
Generalists are key organisms that promote exchanges of energy, information, and materials among different species in network, hence playing important roles in maintaining the balance of microbial community. In ecology, generalists can utilize a wide range of food sources, thus growing well in many habitats and distributing broadly across the soil ecosystem, whereas specialists have special nutritional requirements, thus surviving only in some restricted habitats (Deng et al., 2012;Olesen et al., 2007;Tao et al., 2018). In this study, only four generalists were found in the CC network, whereas 13 generalists were found in the NCC network ( Figure 5). More generalists in the NCC network were favourable for maintaining the microbial community more ordered, stable, and efficient by promoting exchange of energy, nutrients, and metabolites among different microbes (Olesen et al., 2007;Székely & Langenheder, 2014 formed a more stable community than the CC network (Figure 3 a,b). Complicated interactions among module eigengenes also indicated that the NCC network was more stable and ordered than the CC network (Lu et al., 2013;Tao et al., 2018).
This could have been a beneficial factor for maintaining the stability of microbial network under NCC soil.
Microbial community is usually affected by soil chemical properties. We found that the change of network structure and role shift of key microorganisms was significantly correlated with soil chemical variables. Most of module eigengenes (56%) in the NCC network were positively correlated with soil variables, whereas majorities (66%) of module eigengenes in the CC network were negatively correlated with soil variables (Table 3). The results suggested that continuouscropping obstacle has a detrimental effect on the bacterial network.
Possibly, soil variables were more favourable for the exchanges of nutrients, energy, and information among different bacterial populations in the NCC network when compared with the CC network. This study linked land degradation with microbial associations, and the changes of networks can act as indicators or predictors for land degradation in the future study. However, our results only considered the bacterial community, whereas other microorganisms, for example, fungi and archaea, might play important roles in the microbial interactions as well (Kallenbach, Frey, & Grandy, 2016; Pereira e Silva, Dias,