Tillage practices and residue management manipulate soil bacterial and fungal assembly and co-occurrence network patterns in maize agroecosystem

Background: Tillage practices and residue management are highly important agricultural practices. However, very few studies have examined the influence of tillage practices and residue management on both bacterial and fungal communities and network patterns in consecutive years. Results: We examined the effects of different tillage practices, including no tillage, rotary tillage, and deep tillage, on the soil bacterial and fungal communities and co-occurrence networks following residue removal and residue retention in 2017 and 2018. This study showed that both bacterial and fungal communities were unaffected by tillage practices in 2017, but they were significantly influenced in 2018. In addition, soil fungal operational taxonomic unit (OTU) richness was significantly enhanced by deep tillage compared with no tillage in 2018, while bacterial OTU richness was unaffected in either year. Tillage practices had differing effects on the soil microbial network patterns, with rotary and deep tillage increasing the complexity of bacterial networks but simplifying fungal networks. However, residue retention only induced a shift in the fungal community in 2018 without an obvious effect in the bacterial community in both years. In addition, residue retention simplified soil bacterial and fungal networks in 2018. Conclusions: This study highlighted the dissimilar responses of bacterial and fungal networks to tillage practices and emphasized that tillage practice is more important than residue management in shaping soil microbial communities.

account for more biomass than bacteria in agroecosystems [28], the responses of fungal communities to tillage practices and residue management have been studied less frequently, and even fewer studies have examined the populations of both bacteria and fungi. Although both bacteria and fungi are critical components of soil microorganisms, they differ greatly in their growth rates, stress tolerance and substrate utilization [29]. Owing to their many differences, a comparison between bacterial and fungal communities would improve our understanding of the manner in which different components of the soil microorganisms respond to tillage practice and residue management.
Moreover, simultaneously considering the fungal and bacterial communities would be more comprehensive compared to considering these two microbial kingdoms separately [30].
Studies have reported the soil microbial alpha-or beta-diversity in response to tillage practices and residue management, while missing the correlations among soil microorganisms. It should be noted that soil bacteria and fungi coexist in complicated environments and form complex interactions with each other; these interactions maintain the soil microbial community assemblage and ecosystem functions [31]. Co-occurrence network analysis has proven to be a particularly powerful tool to understand how microbe-microbe associations change in response to agricultural management, providing considerable information beyond community analyses [31]. Increasing numbers of studies have shown that agricultural management alters the microbial network structure and keystone taxa [3]. For instance, Ling et al. [32] and Yang et al. [33,34] reported that organic inputs increased the complexity of bacterial and fungal networks in arable soils. Similarly, the application of rice straw generated more substantial bacterial network connectivity in residue retained soils [35]. In a recent study, Banerjee et al. [36] compared no tillage, conventional and organic agricultural systems and observed a significantly higher complexity of fungal networks in organic systems, whereas they did not find a difference between no-till and conventional tillage systems. However, information on the manner in which microbial co-occurrence networks respond to tillage and residue management is still limited.
In this study, we explored the impact of tillage practices and residue management on the bacterial and fungal community structure in two consecutive years using Miseq Sequencing and co-occurrence network analyses. Our objectives were to address the following questions: (1) Do tillage practices and residue management influence the soil bacterial and fungal community? (2) Do network structures and the topology of soil bacteria and fungi differ among treatments? (3) Do bacterial and fungal community and co-occurrence networks respond differently to tillage practices and residue management?
The fungal OTU richness was significantly affected by tillage practice and marginally affected by the interactive effect between tillage practices and residue management in 2018 (Supplementary Table   S1). Under residue removal, the DT treatment harbored significantly higher fungal OTU richness than that of the NT treatment (Fig. 1A). In contrast to what has been found for fungi, the bacterial OTU richness did not differ significantly among treatments in both years (Supplementary Table S1, Fig.   1A).

Bacterial and fungal community composition
The differences between microbial communities were then visualized in each year using PERMANOVA and NMDS analyses. A PERMANOVA analysis indicated that both the bacterial and fungal communities were unaffected by tillage practices and residue management in 2017 (all P > 0.05). In 2018, the bacterial community was significantly affected by tillage practices (r 2 = 0.18, P = 0.02), and the fungal community composition was significantly affected by tillage practices (r 2 = 0.18, P = 0.006) and residue management (r 2 = 0.11, P= 0.008). Based on NMDS ordination, distinct clustering with respect to tillage practices was observed for fungal communities in 2018, whereas bacterial communities overlapped slightly among tillage systems (Fig. 1B). Pairwise PERMANOVA comparisons detected significant differences between the NT and DT treatments and marginal differences between the NT and RT treatments in both bacterial and fungal communities in 2018 (Supplementary Table   S2).
Mantel tests revealed that the soil bacterial community composition significantly correlated with soil pH in 2017 and correlated with SC in 2018 (Table 1). The soil fungal community composition significantly correlated with soil pH and SOM in 2018, while none of these variables correlated in 2017 (Table 1). In 2018, our SEM explained 28% of the variation in bacterial communities and 25% of the variation in fungal communities (Fig. 2). The final SEM showed that tillage practices and residue management exhibited indirect effects on the soil fungal community, primarily mediated through soil nutrient availability (Fig. 2). In contrast, the effect of tillage practices on bacterial community was primarily mediated through soil pH (Fig. 2).
Next, we investigated changes in the relative patterns of abundance of bacterial and fungal phyla.  Table S3). The classification information of the other enriched and depleted OTUs is provided in Supplementary Table S3.

Co-occurrence networks
Bacterial, fungal and combined bacterial-fungal co-occurrence networks were constructed separately in each tillage system in both years (Fig. 5, Supplementary Fig. S2). The complexity and connectivity of soil bacterial networks were higher in the RT and ST treatments than in NT treatment, particularly in 2018 (Fig. 5A). This pattern was demonstrated by the network topological properties, i.e., the connectedness and average degree was significantly higher in the RT and DT treatments than in NT treatment ( Fig. 5A, Supplementary Table S4). The modularity of the NT network was considerably higher than those of RT and DT, indicating a more modular bacterial network in the NT treatment (Fig.   5A). Fungal co-occurrence networks were smaller and less connected than the fungal networks ( Fig.   5B). In particular, the fungal co-occurrence network responded differently to tillage practices compared with bacterial networks. In 2017, we observed a significantly higher complexity of fungal networks in the NT treatment than in RT and DT treatments (Fig. 5B). In the same manner, the topological properties, including connectedness and average degree, were markedly higher in the NT treatment than in RT and DT treatments, while the modularity trended in the opposite direction ( Fig.   5B). However, the fungal networks did not display marked differences in their structure and topology in 2018 (Fig. 5B). The positive/negative link (P/N) ratio was also calculated for each co-occurrence network. In both years, the P/N ratio of bacterial co-occurrence network was lowest in the DT treatment and significantly lower than that in NT and RT treatments (Fig. 5A). Similarly, the P/N ratio of fungal network gradually decreased along with the tillage intensity in 2017, but this pattern was not apparent in 2018 (Fig. 5B). When considering the combined bacterial-fungal networks, we We also constructed soil bacterial and fungal co-occurrence networks in the -R and +R treatments,  Table S5).

Predicted bacterial and fungal functions
A predicted metagenome was constructed for each sample from the 16S rDNA sequencing data by analysis of the raw OTU copy number using PICRUSt. The most abundant Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways related to the 'metabolism' category. Furthermore, a PERMANOVA analysis revealed that the tillage practice significantly influenced the bacterial community composition upon "carbon metabolism" and marginally affected the "nitrogen metabolism"-related bacterial community composition in 2018 but not in 2017 (Fig. 7A, Supplementary Table S6). Using BugBase, we predicted nine potential bacterial phenotypes, including aerobic, anaerobic, those containing mobile elements, facultatively anaerobic, biofilm forming, Gram negative (G-), Gram positive (G+), potentially pathogenic and stress tolerant. Among these phenotypes, the relative abundance of facultative anaerobic bacteria significantly increased following residue retention, and the G+/G-ratio significantly decreased along with the tillage intensity (Fig. 7B, Supplementary Table   S7). The soil fungal community was assessed in terms of fungal guilds, and 45.5% of the fungal OTUs were assigned to a fungal guild. In 2018, the relative abundance of pathotrophs decreased significantly in ST compared with NT ( Fig. 7C, Table S7). However, symbiotrophs and saprotrophs were unaffected by tillage practices and residue retention in both years (Fig. 7C, Supplementary   Table S7).

Effects of tillage practices on the soil bacterial and fungal communities
We observed that tillage practices significantly impacted the bacterial community composition in 2018, which was supported by previously published research [17,37,38]. Several mechanisms could be responsible for the phenomenon. First, the vertical distribution of bacteria could be a factor.
Bacteria s are generally small and easily influenced by the soil microenvironment [9], and therefore, they are distributed heterogeneously along the soil depth [9]. Both the RT and DT treatments would change the vertical distribution of soil bacteria by mixing topsoil and subsoil layers [9], thus influencing bacterial communities as compared with no-tilled soils. Second, change in the soil physiochemical variables is another factor. Soil pH has long been recognized as the main driver of the soil bacterial community [39]. Previous research confirmed that no tillage generally decreased the soil pH, owing to the accumulation of organic matter in the top soil [40,41]. Notably, the soil pH decreased in NT treatment (Supplementary Table S8) in this study. Therefore, the tillage practices may influence the bacterial community through soil pH, which was confirmed by the SEM analysis. In addition to pH, the soil oxygen level is another important factor that shapes soil bacterial community [9,24]. It has been revealed that NT systems generally reduce soil aeration porosity and oxygen content [42] by compacting the soil, while RT and DT systems alleviate soil compaction and increase the oxygen level (Supplementary Table S8). Consequently, tillage practices could possibly influence the distribution of aerobic, anaerobic and facultative anaerobic bacteria, leading to different degrees of soil C sequestration among the tillage systems [43]. Therefore, we hypothesized that the "carbon metabolism"-related bacteria would be more sensitive to tillage practices. Interestingly, our NMDS ordination revealed that bacterial functional community upon "carbon metabolism" in the NT treatment was clearly separated with DT treatment by conducting "PICRUSt" function predicting.
The shift of bacterial community was also reflected at the phylum level. Tillage practice was reported to favor fast-growing copiotrophs, while depressing oligotrophs [44]. Bacteroidetes, which typically have copiotrophic attributes and thrive in environments with high C availability, was favored by the RT and DT treatments in this study. At the OTU level, OTUs belonging to Bradyrhizobium and Candidatus Solibacter were observed to be enriched in the DT treatment in 2018, and they have been  Table S8). In addition, the homogenization of soil layers and improved soil aeration would be other reasons for the fungal community shift [38].
One concern for some farmers in adopting no tillage practice is a potential increase in plant diseases.
In a study conducted in a black soil region, Wang et al. [22] reported that long-term NT practices increased the abundances of two pathogenic fungi (Fusarium graminearum and F. moniliforme). In the same manner, Govaerts et al.
[51] conducted a 5-year field study in Mexico and concluded that the incidence of root rot under NT was higher than that under conventional tillage practice. Our results indicate that the NT treatment harbored many more pathotrophs than other treatments, which is consistent with the results of previous studies. Taken as a whole, these examples combined in our results suggest that a no-till system may favor potentially pathogenic fungi, which may threaten plant growth and crop yields. The higher abundance of pathotrophs in a no-till system might be explained, in part, by protecting them from high temperature, limited water availability and the repeated disruption of mycelia [52,53].

Effects of residue management on soil bacterial and fungal communities
Residue retention is an effective management technique to enhance SOM content and improve soil nutrient availability [12,54,55] However, the climate in Northeast China is characterized by an extremely long and cold winter [33]. Therefore, the residue buried in soil after harvest in mid October had difficulty decomposing during the winter, owing to the deficiency of water and heat [8]. Together with the relatively short duration of the experiment, the SOM and nutrient contents (e.g., AP, AK and NH 4 + -N) were not obviously improved by residue management in this study (Supplementary Table   S8). As reported previously, the beneficial effect of residue retention on soil variables was usually not apparent in short-term studies [56]. Therefore, the soil bacterial richness and community composition was unaffected by residue retention in both years. These results were not surprising considering the lack of change in most soil physiochemical variables (Supplementary Table S8 and Fernandez et al.
[58], who reported no effect or minor effects of residue application on the would result in greater storage of carbon in no-tilled soils [65], while bacteria-dominated communities would drive decomposition and nutrient cycles in the rotary or deep tillage systems [66]. Thus, the relatively lower soil nutrient availability in DT systems could be explained by the network analysis in this study.
Our results indicated that soil bacterial and fungal network patterns respond differently to tillage  Table S8). The environment that is richer in nutrients may alleviate the physical disturbance and tighten the fungal interactions. Additionally, we found that the P/N ratio gradually decreased with tillage intensity for both bacterial and fungal networks, suggesting that many bacterial and fungal species in the RT and DT soils competed for resources or spaces and repelled each other.
Since organic inputs provide a substantial supply of substrates and nutrients for soil microorganisms, previous studies indicated that organic inputs generally increased the complexity of soil microbial networks [33,35,36]. In contrast, we observed that residue retention simplified both soil bacterial and fungal co-occurrence networks in 2018. However, the studies described above used animal manure or compost as organic inputs, which were easily decomposed in soils compared with maize straw, making it difficult to compare them with this study. One likely explanation for this pattern is that the reduction of available N in residue retained soils may simplify the microbial networks [13].
Alternatively, the maize residues incorporated in soils may disrupt the microbial habitats and prevent their interaction.

Conclusions
Our results indicated that tillage practices not only affected soil microbial community compositions but also the co-occurrence networks, with potential implications for agriculturally relevant functions suggesting that residue retention may be a secondary factor that affects soil bacteria and fungi.
Another substantial contribution of this study is the analysis of the bacterial and fungal community through consecutive years. Our results showed that the amplitudes of the changes in soil alpha-and beta-diversities and bacterial networks increased with time after the adoption of tillage and residue management. However, the need remains to reveal the long-term effects of tillage practices and residue management on soil bacteria and fungi.

Field experiment design
The

Bioinformatics analysis
The bioinformatics analysis in this study was previously described [33,34]. Raw sequences of bacteria and fungi of low quality were discarded using QIIME Version 2017.6.0 [74]. Chimeric sequences were detected and eliminated using the 'chimera.uchime' command [75] against the Greengene 13.8 database [76] and UNITE 7.0 [77] database in Mothur, respectively. After removing potential chimeras, the clean sequences were clustered into different operational taxonomic units (OTUs) at a 97% identity threshold using the USEARCH v10.0 pipeline [78] after dereplication and the removal of all singletons. The taxonomic assignment of bacterial and fungal OTUs was conducted using local BLAST against the Greengene and UNITE databases, respectively. OTUs that were not assigned as bacteria and fungi were removed. To ensure equal sampling depth among the samples, data were subsampled to the lowest sequence count using the 'sub.sample' command in Mothur [75].

Data analysis
Bacterial and fungal alpha-diversity indices, such as OTU richness, Shannon diversity index and Pielou evenness index, were calculated in the "vegan" package [79]. Two-way analyses of variance (ANOVAs) were then performed to examine the effects of tillage practice, residue management and their interaction on bacterial and fungal alpha-diversity indices. All the data were tested for normality and homogeneity of variance before two-way ANOVAs. Paired comparisons among the treatments were determined by a Tukey's HSD post-hoc test at a 5% significance level.
We applied a structural equation model (SEM) to detect the causal relationships among tillage practices, residue management, soil pH, soil nutrient and soil bacterial and fungal community composition using AMOS 22.0 [80]. We assumed a conceptual model based on our knowledge of soil ecological causal relationships. "Soil nutrient" is a synthetic variable derived from the SOM, AP, AK, NH 4 + -N and NO 3 --N. The adequacy of model was determined using χ 2 tests (P > 0.05), a goodness-offit index (GFI > 0.9), Akaike Information Criteria (AIC), and root square mean errors of approximation (RSMEA < 0.05) [81].
Data on the soil bacterial and fungal community compositions were tested using a permutational multivariate analysis of variance (PERMAVONA) to evaluate the effects of tillage practice, residue management and their interaction in the vegan package [80] with 999 permutations. The bacterial and fungal community compositions were subsequently ordinated using non-metric multidimensional scaling (NMDS) based on the Bray-Curtis dissimilarity matrices in the "vegan" package [80]. The correlations between soil physiochemical variables and soil microbial community compositions were calculated using Mantel tests with the "ecodist" package [82]. A ternary plot was used to demonstrate the distribution of bacterial and fungal OTUs across the NT, RT and DT treatments. The enriched and depleted OTUs by residue retention were calculated in the "Deseq2" package [83] and visualized in a volcano plot.
Bacterial, fungal and combined bacterial-fungal co-occurrence networks from the NT, RT and DT treatments were built for each year, respectively. Therefore, each network was based upon six communities. In addition, soil bacterial and fungal networks in -R and +R treatments (each containing nine communities) were also built. Bacterial and fungal OTUs with relative abundances > 0.5% and occurred in > 50% communities were included in the networks to allow us to focus only on the abundant OTUs. Spearman's correlation coefficients were calculated between OTUs using the "Psych" package [84]. P-values for multiple tests were calculated using the false discovery rate (FDR) as described by Benjamini and Hochberg [85]. The correlations with a Spearman's coefficient < 0.6 and a P value > 0.01 were eliminated [86]. Subsequently, each network was used to sub-set network matrices for each sample by selecting OTUs that were detected as present within the sample using the "induced_subgraph" function in "igraph" package [87]. The number of positive and negative correlations, average degree, connectedness and modularity were calculated in each network.
Stepwise multiple-regression models were performed to detect the best predictor of network connectedness.
The functional profiles of bacteria were predicted using the "phylogenetic investigation of communities by reconstruction of unobserved states" (PICRUSt) [88]. The bacterial phenotypic information was obtained using BugBase (http://bugbase.cs.umn.edu). The fungal functional guilds (pathotrophs, saprotrophs and symbiotrophs) were characterized using FUNGuild v1.0 [89]. We excluded OTUs that were assigned as "possible", and OTUs with multiple assignments were further assigned as described by Tedersoo

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.