Fungal Communities Are More Sensitive to the Simulated Environmental Changes than Bacterial Communities in a Subtropical Forest: the Single and Interactive Effects of Nitrogen Addition and Precipitation Seasonality Change

Increased nitrogen deposition (N factor) and changes in precipitation patterns (W factor) can greatly impact soil microbial communities in tropical/subtropical forests. Although knowledge about the effects of a single factor on soil microbial communities is growing rapidly, little is understood about the interactive effects of these two environmental change factors. In this study, we investigated the responses of soil bacterial and fungal communities to the short-term simulated environmental changes (nitrogen addition, precipitation seasonality change, and their combination) in a subtropical forest in South China. The interaction between N and W factors was detected significant for affecting some soil physicochemical properties (such as pH, soil water, and NO3- contents). Fungi were more susceptible to treatment than bacteria in a variety of community traits (alpha, beta diversity, and network topological features). The N and W factors act antagonistically to affect fungal alpha diversity, and the interaction effect was detected significant for the dry season. The topological features of the meta-community (containing both bacteria and fungi) network overrode the alpha and beta diversity of bacterial or fungal communities in explaining the variation of soil enzyme activities. The associations between Ascomycota fungi and Gammaproteobacteria or Alphaproteobacteria might be important in mediating the inter-kingdom interactions. In summary, our results suggested that fungal communities were more sensitive to N and W factors (and their interaction) than bacterial communities, and the treatments’ effects were more prominent in the dry season, which may have great consequences in soil processes and ecosystem functions in subtropical forests.


Introduction
Water and nitrogen (N) are the basic elements that affect soil microorganisms and biogeochemical processes. Since the last century, intensified anthropogenic activities and industrial production have caused global changes in the distribution and cycling of both water and N [1,2]. Global ecosystems are thus subjected to more variable precipitation patterns and enhanced atmospheric N depositions [3]. Great changes in water potential and/or reactive N levels in soil may be challenging for the growth and functions of microorganisms. Bacteria and fungi, the two main microbial groups in soil, have disparate characteristics in cell structure and physiology, likely posing differential response patterns to environmental changes. Bacterial and fungal communities could be sensitive [4][5][6] or unaffected [7][8][9] to the changes in precipitation patterns and nitrogen levels [10][11][12]. Fungal communities might be more sensitive than bacterial communities to non-extreme soil moisture variations [13], while it showed more resistance to extreme desiccation and re-wetting [14]. In arid shrubland or wetlands, bacterial communities were more sensitive, while in subtropical forests fungal communities were more sensitive to nitrogen depositions [15][16][17]. It was still hard to generalize a common law to decide whether fungi or bacteria are more sensitive to N deposition (N factor) and/or precipitation change (W factor). The inconsistency may be attributed to ecosystem types, soil original fertility levels, and experimental designs [18,19].
Growing evidences have shown that multiple drivers of global change interact in complex ways that are not predicted by the additive effects of individual drivers. The interaction types can be divided to synergistic, antagonistic, and simply additive [20]. As the two environmental factors that often change simultaneously [21], N deposition (addition) and altered precipitation can solely and interactively affect soil microbial communities and ecosystem functions. In the arid/semiarid ecosystems where both water and N are frequently limited, the changes in precipitation and N level might interact strongly (synergistically or antagonistically) to influence microbial diversity, biomass, and fungi-to-bacteria ratio [22,23]. Bacterial communities showed different sensibility to reduced and increased precipitation in a desert steppe, and the effects of water addition were weakened by nitrogen deposition [23]. In tropical/subtropical ecosystems, whether the interaction between the two factors is significant for microbial communities and what the interaction types are, were seldom understood. In addition to the diversity, the co-occurrence network between different members in a community has become a new dimension in exploring microbial community changes. Though the co-occurrence network does not represent real inter-species interaction [24], it has its own merit for inferring the relationships between different community members [25][26][27]. Both bacterial and fungal community networks could be affected by the N and W factors. In temperate ecosystems, soil bacterial network structures were more susceptible to drought than did fungal networks [28,29], which contrasts with one study conducted in a subtropical forest where fungal networks changed more to precipitation seasonality change [30]. The responses of community diversity and network may be different to environmental changes, for example, soil bacterial diversity showed no significant change but the network structure changed significantly in different seasons in a maize cropland [31]. However, the responses of community network structure to the interaction of different environmental factors (e.g., N and W factors) were not understood for either bacterial or fungal communities.
As a good proxy for microbial functions in soil, soil enzymatic activities were often determined along with microbial community studies. The relationships between diversity with soil functions have been frequently reported. Generally, greater diversity ensures high enzymatic potential, though the functional redundancy among different members may obscure their relationships. The compositional changes of community may or may not explain more than the diversity for the enzymatic variations [32]. Network structure also has good correlations with soil enzyme activities. For example, some module with special keystone species correlated significantly with the activities of laccase in soil [33]. Despite a good collection of literature reporting microbial community traits with soil enzyme activities, there lacks studies to compare different community traits of both bacteria and fungi in correlation with soil enzymatic activities. An inspection of different community traits (diversity, composition, and network structure) for both bacteria and fungi, and their relationships with soil enzymatic activities would have the potential to enhance the predictive capacity for soil functions and ecosystem processes in context of global changes.
Tropical and subtropical forests are central spots of global biodiversity and carbon storage, and also hotpots sensitive to changes in precipitation patterns and N deposition [33]. In South China, subtropical forest ecosystems are facing the challenges of both enhanced N deposition and precipitation pattern change. Previous studies have indicated that in the future, the subtropical forests in South China would be subjected to growing N deposition and changes in precipitation seasonality (drier dry season and wetter wet season with the total precipitation unchanged) [34]. How soil bacterial and fungal communities respond to the sole and interactive effects of N and W factors was largely unknown, which hinders our understanding of soil functions and ecological processes in subtropical forests. Here, we conducted a field experiment simulating enhanced N deposition and precipitation seasonality change as future climate scenarios in the subtropical forests in South China. We aimed to determine (1) how soil bacterial and fungal communities respond to N and/or W factors in terms of community diversity, composition, and network topological features; and (2) whether there are interactive effects of N and W factors on soil physicochemical properties and different microbial community traits. We hypothesized, (1) the interaction of N and W factors could be significant in driving some traits of bacterial and fungal communities. (2) Fungal communities may be more sensitive than bacterial communities to the changes in N level and precipitation in the subtropical forest. Our study would track both fungal and bacterial communities in a variety of traits. The results would deepen our understanding of soil biodiversity, functions, and ecosystem processes in the context of global changes.

Site Information and Experimental Design
The experimental site was located in a subtropical forest at the Heshan National Field Research Station of Forest Ecosystem (112° 50′ E, 22° 34′ N). The forest is an evergreen monsoon subtropical forest with two dominant tree species, Schima superba and Michelia macclurei [35]. The climate in the region is characterized by distinct dry and wet seasons, usually expanding from October to March and April to September, respectively. The mean annual air temperature is 21.7 °C, and the mean annual precipitation is 1700 mm [36,37]. The soil type is Ultisol as per the USDA soil taxonomy classification [38].
A total of 16 plots were established in 2018. The total area was nearly 1 ha, and the mean slope of these plots was 15°. The two factors, enhanced N deposition (N) and precipitation seasonality change (PC), and their combination (NPC) and control (C) were randomly assigned to four plots in each of the four blocks (Fig. S1). At the time of establishment, the above-ground plant community compositions and below-ground soil physicochemical properties did not show significant variations in plots from different treatments (data unpublished). The plot size is 12×12 m 2 , with a distance of at least 2 m from each other. For the N and NPC treatments, solutions of NH 4 NO 3 were sprayed into soil at a height of nearly close to the forest floor at the beginning of each month (on one of the first several days with no rain), with an annual dose of 100 kg N hm -2 . For the PC and NPC treatments, a set of steel frames with a height of 1.5 m above ground was set up to support throughfall reduction shelters and water-adding sprinklers (Fig. S1). To minimize the lateral water flow and interference from other plots, the four sides of each plot were trenched using 1-m height polyvinyl chloride boards to a depth of 60-80 cm. We adopted a throughfall exclusion rate of 67% as per the other studies that were used in the tropical forests [39,40]. Ten to twelve polyethylene sheets, covering a total shade area of 67% of the plot area, were spread to reduce the throughfall during the dry season (from September 15 to April 14) (Fig. S1). These sheets are of 95% light transparency to minimize the shading effect. The excluded rain flowed along with the sheets into the polyvinyl chloride troughs at the lower slope and was drained out of the plots (Fig. S1). Automatic rain gauges (Davis Instrument, MD, USA) were installed at the plot (area without shelters) to record the amount of throughfall. During the wet season, water with 67% of the recorded amount of throughfall in the dry season was added to the PC and NPC treatment plots. In each plot, water was added through 25 automated sprinklers at the center of the steel frames, with a spraying diameter of 2 m and showering ca. 50 L of water per hour (Fig. S1). The throughfall was 454.08 mm in the dry season in the first hydraulic year. For each plot from the PC and NPC treatment, 10.90 m 3 of water was added at the end of each month from May to August 2019. The added water was pumped from a nearby pond, which in general had lower organic C and N contents than and similar pH values to the throughfall. Such differences in water chemistry were considered to have a neglectable impact on the treatment effects [35][36][37].

Sampling and Determination of Soil Physicochemical Properties and Enzyme Activities
To understand the quick responses of soil microbes to shortterm environmental changes, samples were taken during both dry and wet seasons in the first hydraulic year (September 15, 2018-September 14, 2019). Five upper 0-15-cm soil samples were randomly taken from each plot on December 13, 2018, and July 12, 2019, by soil augers with a diameter of 5 cm. The soil samples were pooled together in one clean plastic bag for one composite sample and transferred to the laboratory within 8 h in a heat-insulated box with ice. The samplings were performed at least 10 days after water or N addition. In the laboratory, the soils were sieved through a 2-mm sieve. A subset of the soil sample was stored at 4 °C for chemical and enzymatic analyses. All analyses were done in less than 2 weeks. A subset of the soil sample was stored at −80 °C for DNA extraction and sequencing.
Soil water content (SWC) was determined by the weight method after drying the fresh soil at 105 °C for 24 h. Soil pH was measured in an air-dried soil/water suspension (1:2.5, w/w) using a pH meter (Mettler-Toledo GmbH, Greifensee, Switzerland). Total organic carbon (TOC) was determined using the K 2 Cr 2 O 7 titration method. Dissolved organic carbon (DOC) was measured with a TOC analyzer (Shimadzu, Kyoto, Japan) after filtering the extracts of 10 g of fresh soil in a 0.5 M K 2 SO 4 solution. Total N (TN) was measured by the indophenol blue colorimetric method. Total phosphorous (TP) was measured by the molybdenum antimony blue colorimetric method [41]. Available phosphorus (AvaiP) was measured by Olsen's method [42]. Soil microbial biomass was determined using the fumigation extraction method [43]. The conversion coefficient used to calculate the microbial biomass C (MBC) was 0.45 [44].

Soil DNA Extraction and rRNA Gene Sequencing
Total genomic DNA was extracted from nearly 0.5-g soil with the MP soil extraction kit (MP medicals, USA) as per the manufacturer's protocol. The quantity and purity of extracted DNA were checked in a Nanodrop One spectrophotometer. The primers 341F/806R (5′-CCT ACG GGA GGC AGCAG-3′/5′-GGA CTA CHVGGG TWT CTAAT-3′) were used to amplify the V3-V4 region of the bacterial 16S rRNA gene. The primers ITS3NGS/ITS4NGS (5′-CAT CGA TGA AGA ACG CAG -3′/5′TCCTSSGCTTANTDATA TGC -3′) were used to amplify the ITS2 region of the fungal ITS rRNA gene. All amplicon library preparation and sequencing were performed in the MagiGene Company (Guangzhou, China).

Bioinformatic Analysis of Sequences
The paired-end reads from the sequencing company were mainly processed with the software Mothur v1.44 [47]. Briefly, the forward and reverse reads were merged with the command "make.contigs" with the default parameter. For bacteria, the sequences with an average quality score lower than 25 and a length shorter than 400 were discarded with the command "trim.seqs." Then, the sequences were further quality-controlled with the "pre.cluster" command. The "unoise" method was chosen, and sequences were refined with the threshold of 4 bases' difference. Then, all the sequences were compared with Silva v132 to obtain taxonomic information [48]. Those sequences not affiliated with the bacterial domain were discarded. OTUs were then calculated with the command "cluster" and the method "agc" and a cut-off of 0.03. For fungi, the sequences with an average quality score lower than 25 and a length shorter than 300 were discarded. The sequences containing the ITS2 region of rRNA were then extracted with the software "ITSx" [49]. Then, the sequences were processed with the command "pre. cluster" in Mothur. All the sequences were compared to the UNITE database to obtain taxonomy information [50]. Those sequences not affiliated with fungi were discarded. OTUs were then calculated with the command "cluster" and the method "agc" and a cut-off of 0.02 in Mothur. The original OTU table for the bacterial or fungal community was made with the command "make.shared" in Mothur.

Meta-networks of Soil Microbial Communities
To construct the co-occurrence networks, the top 1000 abundant bacterial and fungal OTUs were pooled together to represent a "meta-community." The Spearman correlation coefficients among OTUs and the significance values adjusted with the "Benjamini and Hochberg" method were calculated with the "WGCNA" and "multtest" packages in R software 4.1.1 [51]. The meta-community network was built based on the correlation coefficients and adjusted P values. The cutoff of correlation coefficients (0.68) was detected with the "RMThreshold" package based on the random matrix theory [52]. The cut-off of the adjusted P value was set as 0.001. The deconvolution method was used to detect the edges arising from the direct interactions in the network [53]. The edge tables connecting those significantly correlated OTUs were exported and then loaded into R. The downstream analyses of network features calculation and network graph visualization were performed with the "igraph" package. Sub-networks were extracted from the meta-network by attaining the OTUs that occurred in the specified groups.

Statistical Analyses
The grouping effects of treatment and season on alpha diversity, soil physicochemical properties, soil enzyme activities, and network features were checked for significance with the permutation-based two-way analysis of variance (ANOVA). The significance of treatment and season in affecting community compositional change (beta diversity) was checked with permutation-based multivariate analysis of variance (adonis in R). For pairwise comparisons of alpha diversity, soil physicochemical properties, soil enzymatic properties, and network topological features, permutation-based pairwise t tests were performed with the "FDR" correction methods for multiple comparisons. The Bray-Curtis dissimilarities were calculated on the square-rooted abundance values across all samples and then visualized with the nonmetric multidimensional scaling (NMDS) plot. The scaled values of soil physicochemical properties (except the pH which kept no change) and enzyme activities were put into the correlation analyses with the "Spearman" method. The Euclidean distances of the scaled values of alpha diversity and network features, and the Bray-Curtis distances of community compositions were correlated with each of the soil physicochemical properties and enzyme activities with the "mantel.test" function in R. The results of the above correlation analyses and Mantel tests were visualized with the "quickcor" function in R. We also checked the relative contributions of soil physicochemical properties and different community traits (the alpha and beta diversity of the bacterial and fungal communities respectively, and the network-level topological features of the meta-community networks, between bacteria and fungi networks, within-bacteria networks, and within-fungi networks) to the variations of soil basic functions (represented by the 4 determined soil enzyme activities ), which were done with the analyses of multiple regression on distance matrices (MRM) [54].
To check the interactive effects of N and W factors in affecting soil physicochemical properties, enzyme activities, microbial community alpha and beta diversity, and network features, the two-way analyses of variance were done, for which the samples were given two dummy variables set as "N" and "PC." In specific, C samples were set as 0, 0; N were samples set as 1, 0; PC samples were set as 0, 1; NPC samples were set as 1, 1 for the "N" and "W" variables, respectively. All the aforementioned statistical analyses were performed with the "vegan,", "lmPerm," "rcompanion," "ggcor," and "ecodist" packages in R. These and aforementioned R packages can be found in the R package repository (https:// cran.r-proje ct. org/ web/ packa ges/).

Soil Physicochemical Properties and Enzyme Activities
We examined 10 soil physicochemical properties and 4 soil enzyme activities (  Only in the dry season, the PC and NPC treatments substantially decreased (with a magnitude of over 25%) the SWC compared with the control; and the N treatment significantly reduced soil pH (permutation-based pairwise t test, P < 0.05). No significant changes in the 4 measured enzyme activities were detected between different treatments in either the dry or wet season. Statistically, we found that in the dry season, both soil SWC and pH were significantly affected by the N factor, the W factor, and their interaction. The N factor also significantly influenced soil AvaiP content. The W factor also significantly influenced the contents of soil DOC, NH 4 + , and soil BG activity (permutation-based two-way ANOVA, all cases, P < 0.05). In the wet season, the contents of soil DOC and NH 4 + were significantly influenced by the N factor. Soil pH was significantly affected by the W factor. The contents of soil MBC and NO 3 were significantly affected by the interaction of N and W factors (permutation-based two-way ANOVA, all cases, P < 0.05) (Table S1).

Alpha Diversity of Soil Bacterial and Fungal Communities
To reduce the biases caused by uneven sequencing effort, all bacterial and fungal samples were rarefied to 26136 and 13143 sequences, respectively. There were 6455 (ranging from 961 to 1424) bacterial and 1514 (ranging from 265 to 412) fungal OTUs for all the samples. For bacterial communities, the alpha diversity (both Shannon diversity and Pielou's evenness) showed significantly lower values in the wet season (permutation-based two-way ANOVA, P < 0.05), but no significant differences between different treatments. For fungal communities, the PC and N treatments significantly decreased fungal diversity compared with the control (permutation-based t test, P < 0.05) in the dry season (Fig. 2). Statistically, we found that only in the dry season, the interaction of the N and W factors significantly affected the alpha diversity of fungal communities (Table 1).

Taxonomy Distributions and Compositions of Soil Bacterial and Fungal Communities
A total of 38 bacterial subphyla and 39 fungal classes were detected in our study. Among them, the top 7 bacterial and fungal groups comprised 91.3% and 89.2% of all the bacterial and fungal sequences, respectively. In general, season affected the relative abundances of these top microbial groups more than treatment did ( Fig. 3(b, d)). The N treatment significantly enhanced the relative abundances and OTU richness of Gammaproteobacteria in the wet season. The NPC treatment significantly increased the relative abundances and OTU richness of Planctomycetota in the dry season compared with the control (permutation-based t test, P < 0.05) (Fig. S2(a, c)). For the fungal class, the relative abundances of Dothideomycetes, unclassified Ascomycota, and Sordariomycetes increased significantly from the dry season to the wet season. The relative abundance and OTU richness of Eurotiomycetes decreased significantly from the dry season to the wet season (permutation-based two-way ANOVA, P < 0.05). In the dry season, the PC treatment significantly enhanced the relative abundance of Eurotiomycetes, and the N treatment significantly reduced the OTU richness of Agaricomycetes, unclassified Ascomycota, Eurotiomycetes, and Sordariomycetes (permutation-based t test, all cases, P < 0.05) (Fig. S2(b, d)). Statistically, both the relative abundance and OTU richness of the Planctomycetota were significantly affected by the N factor in the dry season. There were more taxa significantly affected by the N and W factors (and their interaction) in the dry season than in the wet season (Table S2).
The OTU compositions of the soil bacterial and fungal communities all showed significant seasonal changes (twoway Adonis, P < 0.001) ( Fig. 3(a, c)). In general, the treatment caused marginal significant effects on the changes of both bacterial and fungal communities, and the effect was slightly greater for fungi (P = 0.04) (two-way Adonis, P < 0.1, Fig. 3(a, c)). The volcano plots showed that fungal communities changed more than bacterial communities to different treatments in comparisons with the control, as indicated by the higher ratios of OTUs that were significantly enriched or depleted (Fig. S3).

Microbial Community Co-occurrence Networks
We constructed a meta-community network based on the combined community comprised of the top 1000 bacterial and top 1000 fungal OTUs from all samples (Fig. S4). The meta-network was finally comprised of 391 bacterial OTUs and 556 fungal OTUs, developing 1162 within-bacteria links (60.9% positive relationship), 732 within-fungi links (95.6% positive relationship), and 593 between bacteria and fungi links (57.7% positive relationship).
By preserving the OTUs occurring in a specific group, the sub-networks can be generated from the meta-community network. Generally, the basic node-level network features (the descriptions of the node-level and networklevel features were shown in Table S3) did not change greatly between different treatments or different seasons. The within-fungi network showed more changes than the within-bacteria network in different seasons or treatment. Season significantly affected all the 4 node-level features, and treatment significantly affected 3 out of the 4 features

a a a A A A A a b b ab A A A A a b b ab A A A A a a a a A A A A
(a) ( b) (d) (c) Fig. 2 The alpha diversity of bacterial (a, b) and fungal communities (c, d). The box is drawn to represent values from the 1/4 quantile to the 3/4 quantile. The black horizontal bars denote the medians of the diversity values. Whiskers and black solid circles represent the 95% CI values and outliers, respectively. No same letters above the boxes denote significant differences between different treatments. The tests were done respectively, for the dry and wet seasons Table 1 The effects of N addition (N factor), precipitation seasonality change (water (W) factor), and their interaction (N×W) in affecting microbial community alpha and beta diversity. The significance value for alpha diversity was calculated from the permutation-based two-way analysis of variance. The significance value for beta diversity was calculated with the "adnois" function in the "vegan" package in R. Significant values were indicated by the bold P values (P < 0.05) of the within-fungi network (permutation-based two-way ANOVA, P < 0.05) (Fig. 4). The N treatment significantly reduced the betweenness and closeness for the withinfungi network in the dry season. For the network-level topological features, seasonal differences were more obvious than among-treatment differences (Fig. S5). For the N, W factors, and their interaction, we found that these factors had greater effects on the network-level features in the dry season than in the wet season. The interaction of the N and W factors prominently affected the network features in the within-fungi and within-bacteria networks in the dry season, and the within-bacteria networks in the wet season (Table S4).

Links Between Microbial Community Traits, Soil Physicochemical Properties, and Enzyme Activities
For the soil physicochemical properties, mainly the SWC, NH 4 + , and AvaiP correlated positively with each other (Pearson correlation, r > 0.78, P < 0.001). The TN, TP, and MBC correlated positively with each other (Pearson correlation, r > 0.58, P < 0.001). The Mantel test showed that bacterial alpha diversity correlated significantly with the contents of MBC, NH 4 + , TP, and AvaiP, and the activities of BG and ALP. Fungal alpha diversity community correlated significantly with the SWC and ALP activity. Bacterial beta diversity correlated significantly with the contents of SWC, MBC, TN, TP, and AvaiP, and the  a a a a A A A A  a a a a A A A A  a a a a A A A A  a a a a Fig. 4 The basic node-level features of different networks. Degree is the number of neighbors for a specific OTU. Closeness is defined by the inverse of the average length of the shortest paths to/from all the other vertices in the graph. Betweenness is the number of shortest paths between any two nodes in the graph passing through that node.

B A A A c d a b A B C D b b b a B B B A b b a b B C AB A c d a b B D C A a a b a B A A A a a a a A A A A a a a a A A A A a b c b C A B A a a b ab A A A A a a a a A A A A a a a a A A A A a a a a A A A A
PPE is the proportion of positive edges in all of the edges (links) for a specific OTU in the network. Within each season, sharing no letters on the bars denotes significant differences between different treatments showed that, by order, the soil physicochemical properties, the meta-community network features, the bacterial alpha diversity, the bacteria-fungi network features, the within-bacteria network features, and the bacterial beta diversity mainly explained the variations of soil enzyme activities. The fungal alpha diversity, the within-fungi network features, and the fungal beta diversity explained relatively minor parts (Table 2). enzyme activities were used. The alpha diversity and network features were scaled before calculating the Euclidean distances. For beta diversity, the Bray-Curtis distance based on the square-rooted community abundance was used. The meanings of the abbreviations for soil physicochemical properties and enzyme activities can be found in the main text and the legend of Fig. 1

Table 2
The results of the MRM analyses that linked the soil physicochemical properties and microbial community traits (alpha, beta diversity, and network features) to the variations of soil enzyme activities. Before the analyses were done, the soil enzyme activities, soil physicochemical properties (except the pH which kept no change), and the community traits (except the beta diversity) were both scaled to zero mean and unit variance. The Euclidean distances of soil enzyme activities and community traits (except the beta diversity) were then used in the formula in the MRM function. For the beta diversity of bacteria or fungi, the Bray-Curtis distance was used in the formula. Significant values were indicated by the bold P values (P < 0.05)

Discussion
Different from previous studies containing only a single N or W factor in tropical/subtropical forests, our study involved both the nitrogen addition and precipitation seasonality change (Fig. S1), so the single and interactive effects of the two factors could be checked. The 1-year manipulation of the nitrogen deposition and precipitation seasonality did not significantly change the above-ground tree communities and biomass in comparison with the control (data unpublished). Yet, as the sensitive ecological groups, soil microbial communities showed immediate responses to the N and W factors, which showed great fluctuations between the dry and wet seasons. The strength of W factor were obviously stronger than that of N factor in affecting soil physicochemical traits in the dry season (Table S1), reflecting the overwhelming effects of precipitation reduce on soil SWC and other associated traits. However, no great differences were detected between the N and W factors for their effects on microbial community traits (Table 1). In the dry season, the PC treatment (rainfall reduction) caused drastic changes in the SWC (Fig. 1), which served as one primary property that affected other soil physicochemical properties and microbial community traits (Fig. 5). On the contrary, the PC treatment in the wet season (rainfall increase) did not cause significant changes in the SWC; and the TN and NO 3 did not change significantly when both the N and water additions were applied (in the NPC treatment) (Fig. 1), which implicated high NO 3 leaching loss due to the enhanced surface runoff and interflow in the NPC plots (data unpublished). The enhanced hydrologic leaching may downsize the interaction strength of N and W factors in the wet season [55].

Fungal Community Diversity Were More Sensitive to N and W Factors
In the subtropical forests of southern China, the climate is characterized by the divergence of the dry and wet seasons [34], which caused great seasonal changes of soil abiotic properties and microbial communities By introducing the dummy variables for the N and W factors, we did the statistical tests of the two factors and their interaction in affecting soil physicochemical properties and soil microbial communities. The results showed that fungal communities showed more sensitive to the N and W factors and their interaction than bacterial communities, which could be reflected in both alpha and beta diversity. Similar results showing that fungal communities were more sensitive to the N or W factor had been observed in forest ecosystems in other studies [16,30,56,57]. The reasons might be, first, the PC treatment in the dry season (water reduction) caused a low soil water content (with a mean of 18.5%), corresponding to a value of less than −0.4 Mpa of soil water potential and nearly 30% water holding capacity in the subtropical forest soil [58,59], which might represent a water condition causing mild drought stress for soil microbes [60]. The moderate drought stress (less water in large soil pore) may more readily affect fungal community than bacteria which lives in a finer scale and often develop biofilms in soil [61]. Second, in the dry season, the N treatment caused lower pH, SWC, and AvaiP compared with the control. The contents of soil pH, SWC, and AvaiP were positively correlated (marginally significantly, P < 0.1) with fungal Shannon diversity, while having no significant relationships with bacterial Shannon diversity (Figs. S5-6). That fungal communities were more sensitive than bacteria to the N factor had been indicated by a meta-analysis study, which was generally consistent across global terrestrial ecosystems [11]. The N or PC treatment in the dry season could reduce the OTU richness of Agaricomycetes, Eurotiomycetes, and Sordariomycetes, which revealed that some species within these classes might have low N nutrition demand or less drought tolerance (e.g., [62]), while the interaction of N and W factor acts antagonistically to alleviate the decreasing effects (Fig. S2). The mechanistic explanation may lie in that, in a drought environment, the water and nitrogen could be both limited (as the diffusive capacity of N in soil is hindered), while the addition of exogenous N possibly alleviates the N limitation ( Fig. 1) [63]; thus, an antagonistic interaction could be observed for N addition and precipitation reduce in the dry season [64]. This antagonistic interaction between the N and W factors was also observed in one desert ecosystem [65].

Fungal Community Networks Were More Sensitive to N and W Factors
More significant changes between different treatments for both the node-level and network-level topological features in the within-fungi networks were found than in the within-bacteria networks (Fig. 4, Fig. S5 and Table S4), which revealed that fungal community networks were more sensitive to the N or W factors. This result partly disagreed with one study conducted in a grassland, in which soil bacterial network showed less stable than fungal network when subjected to drought treatment [66]. The accordant point between this study with de Vries' study is that the community networks with more negative edges (fungal networks in de Vries' study but bacterial networks in this study) showed more stable to environmental changes. Ecological networks consisted with strong interactions might be less stable than those consisted with weak interactions [67]. Members connected with positive links in one community may respond in tandem to environmental fluctuations, resulting in positive feedback and co-oscillation [68]. In the dry season, the fungal members might have sparser relationships (lower closeness) and less interaction influence (lower betweenness) in the N treatment, while in the wet season, fungal members might develop sparser relationships and have higher interaction influence in the N treatment in comparisons with the control (Fig. 4) [69]. N addition might downregulate the potential cooperation between different fungal species in acquiring N, while in the wet season, the higher N content (Fig. 1) might favor the growth efficiency and biomass of some fungal species [70], which might exert a higher influence capacity on other members in the community.

The Inter-kingdom Network Explained Great Part of Soil Enzyme Variations
Soil functions, which are often represented by the enzyme activities in soils, are closely linked with microbial activities [71]. In this study, four enzymes related to carbon, nitrogen, and phosphorous cycling were included to represent the basic yet sensitive soil functions to environmental change. Similar to soil physicochemical properties, the seasonal dynamics of soil enzyme activities were more apparent than the differences between different treatments (Fig. 1). For the short-term simulated environmental changes, soil physicochemical properties explained a greater part of the variations of enzyme activities than the community traits ( Table 2 and Fig. 5). Soil physicochemical properties, such as soil water content, pH, NH 4 + , and AvaiP, were significantly correlated with enzyme activities (Fig. 5) and may readily affect enzyme activities through the regulation of reaction conditions or substrate concentrations. Due to the widespread functional redundancy among different microbial taxa, the changes in microbial community traits (e.g., compositional change) may have less influential capacity on soil enzyme activities [72] ( Table 2). The network structure of the meta-community explained a higher proportion of variations of enzyme activities than the alpha or beta diversity of bacterial and fungal communities ( Table 2 and Fig. 5). This implicated the importance of microbial connections or interactions in affecting soil enzyme activities and ecosystem functions [73].
Our results also suggested that the inter-kingdom microbial associations possibly had great effects in affecting soil enzyme activities (even larger than the effects of withinkingdom associations) ( Table 2). Bacteria and fungi may interact far more often than previously thought [74]. They can establish close physical associations ranging from seemingly disordered polymicrobial communities to highly specific symbiotic relationships, such as fungal hyphae and bacterial cells. Their interactions were suggested to be important in gut health, rumen ecosystem functions, and also in biogeochemical processes [75]. The cooperations between bacteria and fungi in degrading litters were also well known [76]. Our results revealed that the links between Ascomycota with a variety of bacteria (such as those from Gammaproteobacteria, Alphaproteobacteria, and Verrucomicrobiota) might be important in mediating the interactions between fungi and bacteria (Table S5). Ascomycota fungi could interact with bacteria through the hyphae or inter-kingdom gene transfer, which promoted nutrient transportation and enzyme activities [77][78][79]. Take the most important edge in the bacteria-fungi network (Table S5) for example. The Archaeorhizomyces are global distributed fungi, which live in soil or around hardwoods roots. It may play great roles in nutrient turnover and can establish links with other fungi or bacteria [80,81]. The uncultured KF-JG30-C25 was also found to have many links with other fungi (such as the Ascomycota), and their potential interactions may contribute to the assimilation of Acidobacterial extracellular polymeric substances [82]. Individually, bacterial diversity (alpha and beta) and network features were more important than those of fungi in explaining the variations of enzyme activities ( Table 2, Fig. S6-S12). This may be due in part to the fact that bacteria may be more effective (higher biomass-specific activities) than fungi in regulating enzyme activities [76]. For a specific season, the relative abundances of 4 bacterial taxa (Acidobacteriota, Gammaproteobacteria, Planctomycetota, Verrucomicrobiota), but only 1 fungal taxa (Eurotiomycetes) were significantly correlated with enzyme activities ( Fig. S8-S12). Besides, the 4 determined enzymes were mainly corresponding to the degradation of labile organics, which were preferentially linked with bacteria's functions [83]. It is possible that when including the enzymes specific for the recalcitrant carbon such as lignin, the importance of fungal community traits might arise in explaining the variation of enzyme activities.

Conclusions
The interactions between N and W factors are complex in affecting soil abiotic and biotic properties, which can be affected by the ecosystem type and the environmental settings. In our study, the precipitation setting was to simulate the predicted future precipitation patterns in South China (drier dry season and wetter wet season with the total precipitation amount unchanged). In the dry season, the rainfall reduction and nitrogen addition treatments interact in an antagonistic way to cause minor changes of soil biotic (such as pH, SWC, and DOC) and abiotic properties (fungal alpha diversity) than the sole treatments. This could be partly attributed to that the water reduction in the dry season enhanced N limitation, which was alleviated directly by the N addition. In the wet season, the interaction between N and W factors was weaker than in the dry season. This may be due to the high hydrologic leaching caused by the water addition in the wet season. Fungi, rather than bacteria, showed more sensitive to the N and W factors (and their interaction) at the community level. Fungal communities might be more readily affected by the intermediate water stress and show stronger responses to physicochemical changes caused by the N addition. We also found that the topological features of the meta-community network were important in explaining the variation of enzyme activities across the samples. Though there lie gaps between co-occurrence network with true interaction, our results implicated that the inter-kingdom associations (cooperations) between fungi and bacteria might contribute greatly to soil enzyme activities, which should be considered along with the traditional diversity index when linking microbial community traits with soil processes and ecosystem functionality.