Seasonal Variations Affect The Ecosystem Functioning And Microbial Assembly Processes In Plantation Forests


 Background: While afforestation mitigates climate concerns, the impact of afforestation on soil microbial compositions, ecological assembly processes, and multiple soil functions (multifunctionality) in afforested areas remains unclear. The Xiong'an New Area plantation forests (Pinus and Sophora forests) were selected to examine the effects of plantation types in four contrasting seasons on soil microbiomes.Results: We evaluated three functional categories (nutrient stocks, organic matter decomposition, and microbial functional genes) of multifunctionality, and the average (net) multifunctionality was quantified. The results showed that net soil multifunctionality as a broad function did not change seasonally, unlike other narrow functional categories. Bacterial communities were deterministically (variable selection and homogenous selection) structured, whereas the stochastic process of dispersal limitation was mainly responsible for the assembly and turnover of fungal and protist communities. Additionally, we showed that winter triggered an abrupt transition in the bacterial community assembly from deterministic to stochastic processes in Pinus forests that was closely associated with a reduction in the bacterial Shannon diversity, with functional patterns of a high level of nutrient cycling (nutrient stocks and organic matter decomposition). Conclusions: Overall, the present study contributes local-ecosystem prospects to model the behavior of soil biota seasonally and their implied effects on soil functioning and microbial assembly processes in plantation forests.

abundance was signi cantly different between these two plant species plots (Supplementary Table 6). The overall bacterial community was dominated by sequences assigned to the class that belonged to Actinobacteria (28.48%) (Supplementary Figures 4A, B; Supplementary Table 6). The difference in bacterial abundance at the genus level increased between spring and summer in the CP (62.72%) and CS (52.71%) plots, whereas differences in abundance decreased with seasonal changes from summer to autumn and autumn to winter for both plant species (Supplementary Figure S5A). The seasons induced signi cant changes in the bacterial Shannon diversity (F 7, 23 = 6.8806, P = 0.0007) and richness (F 7, 23 = 3.4191, P = 0.026) of the bacterial community for both plant species plots. The results showed that bacterial diversity and richness were signi cantly different in spring and in winter between the two different plant species (Figs. 1A, B; Supplementary Table 7

Fungal Communities
After quality ltering, the remaining 1,618,594 fungal sequences were obtained from 24 soil samples, and an average of 67,441 (±3,636; ±SE) sequences were shown per sample. The normalized reads were grouped into 3,902 OTUs with a 97% similarity threshold. The fungal Venn diagram revealed that the core fungal community consisted of 266 OTUs among all samples (Supplementary Figure 3B). The abundance of the dominant Ascomycota reached its highest level in winter for both plant species, and its abundance did not signi cantly differ between spring and winter in the CS plot (Supplementary Figure 4C; Supplementary Table 6). The highest abundance of fungal classes belonged to Sordariomycetes (42.20%) (Supplementary Figure 4D). Furthermore, the relative abundance of fungal genera did not signi cantly increase with changing seasons from summer to autumn and from autumn to winter for  Table 9). Our LDA results (total = 119 biomarkers) showed that the highest number of biomarkers for fungal taxa was observed in SSP samples (n = 24) (P < 0.05, LDA > 2) (Supplementary Figure 8B). These biomarkers accounted for 3.04% of all taxa retrieved. The hierarchical clustering results showed that the segregation of the fungal compositions mostly occurred in autumn.
For instance, the structure of the fungal communities of the CAT samples and CS samples formed two segregated clusters, and the remaining fungal samples formed one cluster (Supplementary Figure 9B).

Protist Communities
After quality ltering and removal of the Metazoa, Streptophyta, and Fungi [11], 546,303 protist sequences remained among all soil samples, and an average of 22,763 (±1,415; ±SE) sequences were calculated for each sample. The normalized sequences were clustered into 2,162 protist OTUs with a 97% similarity threshold. The core protist community consisted of 370 OTUs among all samples (Supplementary Figure 3C). The protist communities were dominated by the phyla Cercozoa (26.87%), Chlorophyta (14.51%), and Apicomplexa (12.77%), followed by unclassi ed Eukaryota (10.98%) and Lobosa (8.75%) (Supplementary Figure 4E). The abundances of the most dominant protist phyla and classes, Cercozoa and Filosa-Sarcomonadea, respectively, were the highest in both plant species plots in spring (Supplementary Figure 4F; Supplementary Table 6). Another example of the impact of season on protist abundance was Chlorophyta, which had a high frequency in winter (Supplementary Table 6). The abundances of protist genera increased from spring to summer for both plant species. In contrast, their abundances showed different trends within seasonal changes from summer to autumn and autumn to winter for both plant types (Supplementary Figure 5C). The protist Shannon diversity (F 7, 23 = 1.9374, P = 0.0486) and richness (F 7, 23 = 2.7219, P = 0.0461) were signi cantly different among seasons. The results showed that the protist richness was high in winter but was not signi cantly different from that in other seasons, whereas the protist Shannon diversity did not follow this trend (Figs. 1E, F). The seasonal trends of other α-diversity indices are summarized in Supplementary Figure 6C. The homogeneity in the protist community composition was corroborated by the concentric ordination space covered in the NMDS (ANOSIM, R = 0.1667, P = 0.032) analysis in comparison with bacterial and fungal compositions (Supplementary Figures 7E, F; Supplementary Table 10). Our LDA results showed that the highest number of biomarkers for protist taxa was observed in CSP samples (n = 12) (P < 0.05, LDA > 2) (Supplementary Figure 8C). These biomarkers (in total = 42) accounted for 1.94% of all taxa retrieved. The hierarchical clustering results showed that the protist compositions of the Pinus and Sophora forests formed segregate clusters in spring and winter, respectively, and other seasons grouped in the same cluster (Supplementary Figure 9C).

Relationships Among Diversity, Abundance, And Eeas
Although the microbial Shannon diversity was linked with several edaphic factors, such as soil pH, fewer soil factors showed a signi cant association with microbial richness (Chao1) (Fig. 2). Four (PD, Chao1, Sobs, and ACE) and one protist alpha diversity (Simpson) index were negatively and positively correlated with PPO and POD activities, respectively (Supplementary Figure 10).

Quanti cation Of Microbial Abundance Using Real-time Quantitative Pcr
Quantitative PCR showed that the fungal gene copy number was lower than that of other soil biota and followed the trend bacteria > protists > fungi (Supplementary Figure 11). The 16S rRNA gene copy number of bacteria ranged on average from 8.65 × 10 8 to 34.84 × 10 8 copies/g dry soil, the fungal gene copy numbers ranged on average from 0.68 × 10 8 to 6.84 × 10 8 copies/g dry soil, and the protist gene copy numbers ranged on average from 1.17 × 10 8 to 7.22 × 10 8 copies/g dry soil.

Drivers Of Soil Microbiome Community Variations
The PERMANOVA showed that the impact of season on bacterial and fungal compositions was signi cant, explaining 26.83% (P = 0.029) and 43.30% (P = 0.002) of the bacterial and fungal variations, respectively (Figs. 3A, B; Supplementary Tables 11-13). In contrast, the in uence of season on protist compositions was not signi cant, explaining 17.91% (P = 0.163) of the variation in protists (Fig. 3C, Supplementary Table 13). The second most important factors for predicting the β-diversity of bacterial, fungal and protist taxa were AGC, TP, and POD, which explained 17.02% (P = 0.011), 35.57% (P = 0.003), and 12.18% (P = 0.029), respectively, of the variance (Fig. 3C). Bacterial and fungal RDA analyses of soil enzymes did not separate hydrolytic and oxidative activities onto separate axes (Figs. 4A, C). However, interestingly, RDA of soil enzymes condensed hydrolytic and oxidative activities onto separate axes (Fig. 4E). In addition, fungal RDA of edaphic parameters was much concentric compared with bacterial and protist RDA analyses of soil properties (Figs. 4B, D, and F).

Effect of seasonal variations and plantations on microbial trophic groups
The results showed that protist consumer taxa were the dominant group across our experimental plots (42.24%, on average). The second most dominant protist trophic group was phototrophs (eukaryotic algae, 31.68%), with the highest abundance in SWI samples and the lowest abundance in CSU samples. The protist communities were composed of taxa putatively assigned as 11.64% parasites, with the highest abundance in SSU samples and the lowest abundance in CSP samples. The photophagotroph group was found across sampling plots with a lower abundance compared with other trophic groups (0.35%, on average); the lowest abundance of this group was observed in CAT samples (0.15%), and the highest abundance was observed in SAT samples (0.53%). Across sampling plots, 14.07% of the overall protist communities were composed of the unclassi ed group (Supplementary Figure 12). The results of PICRUSt2, fungal guilds and modes as well as the bacterial TAX4FUN analysis are presented in Supplementary Figures. S13-16.

Co-occurrence Patterns In Microbial Networks
To determine how plantation types and seasonality affected microbial network complexity, topological network parameters such as the 'average degree (AD)', 'average clustering coe cient (ACC)', and 'average path length (APL)' were considered. The higher the AD and ACC, the more complicated the networks. The more moderate the APL, the closer the association among the members. Applying this rationale and considering plantation type, our results indicated that the soil-biome relationship was more complex and intense in the Pinus forest; considering seasonality, the microbial co-occurrence network was more complex in spring, whereas this network was more intense in winter (Fig. 5, Supplementary Table 14). Interestingly, the highest number of microbial links (bacteria to protists) was recorded in spring (Supplementary Table 14). Based on the key nodes, our results suggested that spring, which was characterized by those nodes, was a liated with bacterial and protist networks and abiotic factors such as soil enzymes that belonged to the N and P cycles (Supplementary Table 15). More details of microbial interactions are presented in Supplementary Table 16. All network parameters were signi cantly associated with the three different classes of air temperature. Remarkably, oxidative enzymes such as PPO and N-related enzymes such as LAP were also associated with most network parameters (Supplementary Figure 17).

Abundance And Composition Of Microbial Functional Genes
Absolute functional gene abundance differed signi cantly (P < 0.05) among soil samples (Supplementary Figures. S18-S20). More speci cally, the results of QMEC based on the HT-qPCR approach showed that the abundance of those genes involved in C degradation (ANOVA, P = 0.0139), N cycling (ANOVA, P = 0.0026), and S cycling (ANOVA, P = 0.009) was signi cantly affected by plantation type seasonally (Supplementary Figure 21). The PCA revealed that the CNPS cycling gene compositions in different plantation types with seasonal variations formed distinct but concentric clusters in which the CNPS cycling genes of the Pinus forest in autumn and winter separated from other clusters (Supplementary Figure 22). Additionally, a signi cant impact of different plantations with seasonal variations on the overall pro le of CNPS cycling genes was observed (Adonis, P = 0.004, Anosim, P = 0.006, and MRPP, P = 0.008). The abundance of keystone C-hydrolysis genes signi cantly differed seasonally among plantation types, and the abundance of those genes involved in hemicellulose degradation was higher than that of other C-hydrolysis genes (Supplementary Figure 23). Our results also showed that those genes involved in C degradation and N cycling signi cantly differed between the two plantations (Supplementary Figure 24).
Contribution of CNPS cycling genes to microbial diversities and compositions Spearman analysis showed that there was no association between microbial richness and diversity and the microbial CNPS cycling genes, whereas fungal and protist β-diversity was signi cantly associated with microbial C degradation and C xation. In contrast, bacterial β-diversity was signi cantly associated with microbial P cycling genes (Supplementary Figure 25). Interestingly, the protist RDA results of the microbial CNPS cycling genes indicated signi cantly condensed C degradation and C xation on separate axes (P < 0.05) (Supplementary Figure 26).
Variation in ecosystem functioning is impacted by biotic and abiotic parameters Among the four functional categories (Supplementary Table 17), narrow functional categories (nutrient stocks, organic matter decomposition, and microbial functional genes) signi cantly differed between plantation types with seasonal variations (ANOVA, P < 0.05), whereas 'net soil multifunctionality' did not differ signi cantly (P = 0.4306) (Supplementary Figure 27). Different levels of statistical associations were observed between multifunctionality indices and microbial compositions (Supplementary Figure 28

Assembly Processes Of Microbial Communities
We observed that the βNTI distributions differed signi cantly across plantation type, with seasonal variations for the bacterial (F 7, 575 = 3.5549, P = 0.0009), fungal (F 7, 575 = 6.9355, P = < 0.0001), and protist communities (F 7, 575 = 3.6975, P = 0.0006) (Figs. 6A, B, and C). Null model analysis showed that the relative contributions of deterministic (|βNTI|≥2) and stochastic (|βNTI|< 2) processes in various microbial compartments differed greatly. The deterministic processes of variable selection and homogenous selection were primarily responsible for the assembly and turnover of the soil bacterial communities (average: 62.49%) (Fig. 6D), whereas the stochastic process of dispersal limitation was mainly responsible for the assembly and turnover of the fungal and protist communities (average: fungi: 42.36%, protists: 76.39%) (Figs. 6E, F). Interestingly, among microbial compartments, the relative contribution of undominated processes to the fungal community (average 42.36%) was greater than that to the bacterial and protist communities (bacteria: 25.34%, protists: 14.93%, on average). The results showed that the bacterial deterministic process of homogenous selection shifted toward the dominance of stochastic processes (dispersal limitation) in the Pinus forest in winter (Fig. 6D).
In uence of biotic and abiotic parameters on microbial diversity, composition, assembly processes and multifunctionality We built theoretical frameworks using PLS-PM to disentangle direct and indirect relationships by focusing on the community diversities and compositions of soil biota with climatic and environmental variables, multifunctionality and microbial community assembly. In both models based on α-diversity ( Fig. 7A) and β-diversity ( Fig. 7B) indices, annual air temperature substantially and negatively structured the microbial meta-community co-occurrence network parameters and soil enzyme activities. In the model based on β-diversity indices, annual air temperature negatively in uenced multifunctionality. In both models, soil enzyme activities and soil properties were greatly and positively associated with multifunctionality. In the model based on α-diversity, soil enzyme activities signi cantly in uenced protist diversity. In both models, a signi cant link between soil enzymes and soil properties was observed.

Microbial diversities under seasonal variations and in different plantations
The seasonal dynamics of the observed richness and Shannon diversity of each soil microbiome showed different patterns (Fig. 2, Supplementary Table 7). These ndings suggest that seasonal variations had an additional in uence on soil microorganisms concerning different plant species, which is consistent with other studies [17,19,20,[23][24][25][26][27][28]. As an example, unlike other studies in which potential biotic and abiotic environmental variables such as soil moisture [29,30], plant types [31], pH [12,13,29,32], litter chemistry [33], and fertilization/season-induced changes [34] contributed to shaping the composition of the protist structures, our PERMANOVA pro ling revealed that the key factor driving the protist community's structure was peroxidase (POD) (Fig. 3). PERMANOVA showed that the protist community, compared with the bacterial and fungal communities, was not in uenced by either seasonal variation or plantation type, suggesting that protist taxa have a broad tolerance to the wide uctuation of seasonal variabilities. Thus, we suggest that the protist composition is strongly but indirectly in uenced by soil bacterial and fungal metabolic activities, particularly C-related functions. If this pattern holds true across a range of ecosystem types, then it implies that protists appear to be impacted to a unique degree compared with other soil biotas by alterations in soil oxidative activities and nitrogen uctuations. Moreover, the GAM results indicated the importance of N cycles and oxidative enzymes to predict the diversity and composition of the soil protist taxa (Supplementary Figure 10). This nding is partly consistent with previous nding which highlighted that protists were the most susceptible soil biota to the application of nitrogen fertilizers [34]. PERMANOVA also showed that seasonal variations were the best predictor of bacterial and fungal β-diversity community variations. In accordance with the present results, previous studies have demonstrated that the soil bacterial and fungal communities showed more seasonal than spatial variation in alpine tundra soils [18,20,35]. Interestingly, the number of soil variables that signi cantly explained fungal β-diversity and variation was higher than that of soil parameters, which proportionally explained the bacterial β-diversity variation. A possible explanation might be that fungal lineages are fostered to better access more soil volume due to their hyphal growth and thereby obtain access to more substrate and nutrients than other biotas [36]. Thus, the potential alteration in edaphic parameters might have a substantial in uence on the fungal structure rather than bacterial taxa. Additionally, our results showed that unlike the β-diversity of the fungal and protist taxa, soil moisture strongly and signi cantly explained bacterial β-diversity variation, which is consistent with a previous study that highlighted the importance of the soil water content for soil bacterial communities [25]. However, PERMANOVA showed that none of the microbial β-diversity variations were signi cantly explained by the impact of different plantations. This nding is consistent with previous report that the effect of soil type on shaping the bacterial rhizosphere was stronger than that of plant species [37].
The relative abundance of bacterial taxa was disproportionately higher than that of other soil biotas Our results showed that the high proportion of assigned reads and quantitative PCR belonged to bacterial taxa in our experimental plots in the plantation forest over the course of one year. This observation might be described by some plausible reasons. First, the annual accumulation of aboveground litter in the young plantation forest oor in the XNA (forest stand age ~ 4 years) is much lower than that in hyperdiverse and layered forests (mature temperate forests). Second, other studies highlighted that alkaline pH was not optimal for fungal growth [38]. Unlike fungal diversity indices without an association with soil pH, bacterial Shannon diversity was signi cantly associated with soil pH (Fig. 3), suggesting that the bacterial diversity was more signi cantly affected by pH than the fungal diversity, which might be due to comparatively narrow optimal pH ranges for bacterial growth but wide pH ranges for fungal growth [39]. Third, the fungal growth network is more sensitive to anthropogenic activities [11,40] and forest management [41] than the bacterial community by creating plantation forests. Our results suggest that the development of the new forest may have resulted in a loss of niches for other soil biotas. Moreover, we infer that the microbial community structure can be altered by the potential effect of alkalinity by favoring high-pH adapted or alkaliphilic microorganisms. We concluded that several reasons, such as the pH level, the paucity of plant litterfall, anthropogenic activities and land management practices, created an opportunity for bacterial taxa to take over other soil biota, such as fungal and protist lineages, in young plantation, seasonal, open-canopy forests in the XNA. Further studies that take these variables into account will need to be undertaken.
Our results showed that the spring and summer microbial communities were signi cantly bacterialdominated genera, whereas the autumn and winter communities were fungal-dominated genera, but these fungal differences were not signi cant (Fig. S6). Previous studies that evaluated the seasonal uctuations of the microbial community in diverse ecosystems [18,20,42] observed consistent results on whether fungi that dominated under-snow biomass and bacterial taxa were more active in summer. The increase in bacterial abundance from spring to summer was associated with an increased abundance of protist genera, some of which may be bacterial feeders (Supplementary Figure 5). Several reasons can explain the domination of bacterial taxa in summer. First, fungi commonly target recalcitrant substrates such as plant litter (higher C to N ratio), whereas bacteria generally target labile substrates such as root exudates. In summer, warmer temperatures can increase root exudates compared with winter, and fresh plant litter input produces a relatively more moderate C to N ratio [40]. Therefore, the domination of bacterial taxa was expected in summer. Another reason that can explain the trend mentioned above might be related to temperature and soil moisture as the two main factors that drive soil biota abundance [43,44]. We realized that the soil moisture was relatively high in winter, particularly in the Pinus forest (Table S5). Additionally, the GAM analysis showed that both fungal and bacterial Shannon diversity indices decreased with increasing soil moisture. The decrease in the bacterial Shannon diversity was much greater than that of the fungal Shannon diversity (Supplementary Figure 10), suggesting a threshold for soil moisture and that fungi will exhibit less of a response to changes in moisture compared with bacteria [40]. However, it has been highlighted that the bacterial community was more sensitive to lower soil moisture than the fungal community [41]. A consistent pattern for the effect of soil moisture on soil biota abundance remains to be elucidated. Another possible explanation for the increased abundance of fungi in cold seasons could be partially related to their natural resistance to freeze-thaw perturbations [45,46] and may not be related to substrate preference. More discussion about these reasons can be found in the additional le 1.

Effect Of Seasonality On Soil Enzymes
Consistent with previous literatures [47][48][49], the PLS-PM results predicted that the activity of soil enzymes was highly temperature-dependent (Fig. 7). In analyzing the projected future climate, the XNA would become warmer and wetter [50], in which the rising temperature had a more signi cant in uence on respiration than on assimilation, suggesting the suppression of vegetation growth at any increase in temperature in the future in this area [50]. Consequently, we inferred that less plant growth would be followed by more miniature plant rhizodeposition and subsequently less microbial enzyme activities. Thus, this outcome likely can explain the negative association of the temperature and enzyme activities in our study. It has been highlighted that some environmental constraints, such as acidic pH and high phenolic compounds, can greatly decrease enzyme activities [49], consequently leading to weak or no temperature sensitivity of soil enzymes. Thus, it appears that the alkaline pH of the XNA, in part, drove the soil enzymes to be substantially affected by the temperature. There are similarities between our results and those of other studies [21,51], highlighting the association of temperature with many EEAs. However, it has been suggested that soil peroxidase activity can decline with warming and that other enzymes may be less sensitive to warming, indicating that EEAs are generally more sensitive to nutrient addition than atmospheric and climate change [52]. The results of the ecoenzyme activities indicated that seasonal variations and plantation types signi cantly in uenced the EEA (Supplementary Table 1), indicating different isoenzyme pools and consequently, the domination of different microbial taxa in each season, plantation, gene pool, and expression pattern. Other researchers documented comparable seasonal alterations in the EEA of soils [53][54][55][56]. As an example, the production of BGA increased in winter in the Pinus forest, whereas its activity was high in autumn in the Sophora forest. A possible explanation is that different plantation types have individual ecological and physiological features [14], resulting in different effects on the edaphic conditions and microenvironment [57], conclusively impacting the secretion of microbial nutrient-acquiring exoenzymes and plants [58]. High bacterial and fungal gene copy numbers generally showed higher enzyme activity in the Pinus forest than in the Sophora forest, indicating that key microbial-mediating biogeochemical processes vary seasonally in different plantations (Supplementary Figure 11). Additionally, these higher gene copy numbers likely explain the high potential enzyme activities in winter, but our oxidative enzyme activities were higher in summer. A similar nding showed that oxidative activities were more variable than hydrolytic activities and increased with soil pH [59]. In addition, as mentioned previously, microbes typically invest more in enzymes, especially lignin-degrading enzymes, in summer [60]. In contrast, potential enzyme activities were potentially expected to be high in winter. The most recent study, highlighted that the same EEA dataset could be interpreted in contrasting ways [61]. Higher enzyme activities can be interpreted as more nutrient availability [59] or reduced nutrient availability [62]. In accordance with the resource allocation strategy, soil biota may secrete more soil ecoenzymes under low nutrient conditions [7,[63][64][65][66][67][68][69]. In our study, we inferred that the higher activity of some enzymes in cold seasons might be related to the lack of plant litter and plant photosynthesis products. However, other studies showed that higher enzyme activities in hyperdiverse forests were limited to autumn, which provided fresh and readily available substrates for more secretion of soil ecoenzymes [56,70,71], suggesting the synchronous intensi cation of belowground biomass and plant litter driving a considerable increase in the metabolism of microorganisms to produce more soil ecoenzymes. Our study showed the positive relationships of BGC: ALP, LAP + NAG: ALP with TC, LAP + NAG with TP and SOM (Supplementary Figure 1), as well as signi cant associations between BGC: LAP + NAG and BGC: ALP with TC/TN ratio (Supplementary Figure 1); these associations might not be the conclusion to match enzymatic acquisition ratios with nutrient stoichiometry [64] because nutrients such as organic matter elements can be stored in various forms, and only a tiny portion can be used by microbes [72].
The results of the investigated C:N:P stoichiometry ratios indicated a deviation from the global [59,66] and regional scale of 1:1:1 for China's forests based on a nationwide dataset [64,69], suggesting that ecoenzyme activity stoichiometry in the plantation forests of the XNA was mostly based on nutrient resource availability and demand for microbial nutrients and was not homeostatic. This result is consistent with that from several studies conducted in China [57,73,74], indicating that this ratio can potentially be different based on the type of ecosystem and soil. Our results also showed that the Sophora forest faced more substantial P limitation (with an average lower TP) than the Pinus forest with comparatively lower C:P and N:P ecoenzyme activity ratios, particularly in warmer seasons (Supplementary Tables 2-5). This is similar to the previous nding, which highlighted that P de ciency is a common problem in forest ecosystems and often intensi es in the summer season [75]. However, the abundance of P cycling genes in both forest types did not show a signi cant difference (Supplementary Figure 21). A detailed discussion is presented in the additional le 1.

Microbial trophic regulation by seasonal variations
Unlike the most recent study, in which biotic factors were identi ed as the major biogeographical predictor for protist consumers and mean annual temperature was the best predictor of the diversity and composition of phototrophs on a large scale in the Southern Hemisphere [76], our ndings showed that seasonality and plantation type were important factors shaping protist trophic interactions at the local scale. The highest level of protist consumers was observed in spring and summer in the Pinus forest. The possible explanation may be related to the temporal variation in the supply of carbon and other nutrients in warm seasons from Chinese pine roots to soil, as well as other compounds such as amino acids, organic acids, and sugars lost via this route, which provide high-quality nutrient sources for microbes, stimulating their growth and thereby increasing prey availability for consumers such as protists, which in turn can enhance nutrient uptake by plants [42,77,78]. The high abundance of consumers indicates the importance of protists' role in controlling the frequency of other soil biotas [13]. In support of this, PLS-PM analysis predicted that the protist α-diversity indices strongly in uenced the microbial cooccurrence network parameters (Fig. 7A), indicating the robust impact of predator community compositions on microbial networks. This strong in uence of predator diversity on the microbial co-occurrence network indicates that the predation pressure of protists signi cantly in uenced the microbial co-occurrence network, and in turn, bacteria and fungi, as the key food source of protists, may shape the diversity of protist taxa, suggesting the bottom-up regulation of soil biota [79]. The high presence of protist parasites was observed in summer in the Sophora forest, which might be related to the preference of this group for a more arid atmosphere and the existence of favorable hosts. Interestingly, the relative abundance of 'phototroph-dominated' protist taxa was high in winter in the Sophora forest, which might be related to the fact that plant photosynthesis and consequently plant exudates are less abundant in the cold season, and this group of protists try to compensate for the paucity of root exudates to x carbon and soil nutrients [80]. The results of predictive tools such as Tax4Fun, FUNGuild, and PICRUSt2 showed that the abundance of gene families differed seasonally for bacteria, fungi, and protists in each plantation plot, suggesting that a given microbial lifestyle might play unique roles within different plantations at different seasonal stages.

Soil microbial co-occurrence network complexity under seasonal successions and in different plantations
The segregated soil biota co-occurrence network patterns and their topological characteristics in each plantation plot showed obvious seasonality (Fig. 5, Supplementary Table 14). Our results showed that the spring microbial network was complex in comparison with that of other seasons. The upward shift in network complexity from one season to another may have happened because of, or was induced by, several biotic or abiotic factors that generated an adaptive response in soil microbiomes seasonally.
More speci cally, many factors might contribute to this intricacy, such as controlling the temporary partitioning of nutrients between soil biota and plants due to the rapid alterations in microclimate from winter to spring, which consequently lead to transitions in some groups of soil microbiomes, abiotic stresses (wet-dry and freeze-thaw cycles), and the consumption of labile C composites, which drives the turnover of microbial community attendants to release labile N for plant uptake [42]. The increased network complexity in spring can indirectly be attributed to an increase in soil thermal variability and increased resource availability in spring that foster microbial diversity and network complexity. It has been highlighted that network complexity is closely related to stable ecosystem functioning [81].
Furthermore, spring consisted of nodes with high BC values (Supplementary Table 15), indicating the importance of the control potential that an individual node exerts over the interactions of other nodes [82]. These complex interactions in spring might consequently enhance system durability and resistance to adverse environmental conditions such as wet-dry and freeze-thaw cycles with changing seasons [83].
Additionally, the soil microbiome networks could be used to visualize the scenarios in which the highest percentage of links among soil biota was observed between bacteria and protists (spring), indicating that bacteria and protist taxa were tightly linked within the microbiomes (Supplementary Table 14). It has also been shown that protist communities create a dynamic hub in soil biota [34,84]. As bacteria and fungi are principal prey for phagotrophic and/or consumer protists as the dominant soil protist functional group in soils [12,13,77,85,86], biotic interactions within the soil microbiome can in uence protist diversity. The taxon-speci c manner of protist taxa was observed for their potential microbial prey by detecting different and/or speci c links between bacteria/fungi and protist lineages. These results are consistent with previous studies showing that protist taxa selectively graze on fungal or bacterial lineages [34,77,87]. Interestingly, however, we observed changeable connections of some protist groups with different microbiome taxa in different seasons and the different plantations, suggesting that the substitution of preferable feeding impacts the predator composition structure [86]. The network results showed that the young plantation forests induced more positive interactions (mutualism or commensalism) than negative interactions (competitive) in each plantation type and even in different seasons, suggesting that cooperative interactions might play a key role in shaping microbial interactions and structures in different plantations seasonally. This might correspond to stable ecosystem functioning [11]. However, competitive interactions do not necessarily match unstable and/or poor ecosystem functioning [88]. Whether these positive and negative interactions mutually impact the microbial network assembly seasonally needs to be further assessed. The association of environmental variables with cooccurrence network parameters ( Fig. 5; Supplementary Figure 17) showed that the assembly of the ecological network was shaped by several vital parameters, among which temperature appeared to be the strongest. This striking observation is consistent with the pervious study, which highlighted that microbial co-occurrence networks are mainly modulated by temperature, followed by precipitation, soil nitrogen, latitude, and plant diversity [89].

Diversity-function-assembly relationship with seasonality
Our results suggest that different afforested plantations changed ecosystem functioning seasonally, consistent with previous nding [60]. We found a positive relationship between microbial diversity and composition and multifunctionality in plantation forests seasonally, corroborating the positive biodiversity-ecosystem function relationships. These results are in line with recent studies on the northeastern and central Chinese Tibetan Plateau at the local scale [90] and the global scale [91]. First, these outputs suggest that microbial diversity and composition have a leading role in maintaining ecosystem functioning [36]. Second, this association indicates high functional redundancy in soil biotas [92]. Previous study highlighted that functional redundancy is a part of microbial communities [93], and the functional redundancy of the microbial communities may explain why microbial taxonomic and functional gene diversity were correlated with different environmental variables in our study. Therefore, any changes in microbial diversity resulting from both soil biotic and abiotic factors might in uence the soil's multifunctionality, suggesting that the estimation of the causal association between microbial diversity and ecosystem functioning would be complex [94]. Furthermore, our results suggested that broad functions, such as net soil multifunctionality, may be more functionally redundant and thus better buffered against microbial shifts that are caused by seasonality in different plantation types or under other biotic and abiotic disturbances.
The null model supports the notion that distinct assembly processes drive the structure of different soil microbiomes, supporting our main hypothesis. Deterministic processes, particularly variable selection, tended to be more critical in shaping the assembly of the soil bacterial communities. In contrast, stochastic processes dominated the soil fungal and protist community assemblies, with dispersal limitation playing a more critical role in both plantation types. A similar nding was recently reported for the assembly of the soil bacterial community [4,95]. In accordance with the present results, previous study demonstrated that stochastic processes are more important than deterministic processes for microbial community assembly at small scales [96]. Moreover, our results showed that seasonality played a decisive role in mediating the balance between stochastic and deterministic processes and showed a signi cant association with the diversity of soil microbiomes seasonally. The soil bacterial community assembly was governed by both deterministic and stochastic processes, with deterministic processes exerting a more substantial in uence than stochastic processes. However, the relative importance of deterministic processes versus stochastic processes in bacterial community assembly varied between the different plantations, with seasonal variations in the Pinus forest. More speci cally, seasonal transition in particular plantations markedly diminished the relative importance of homogenous selection and increased dispersal limitation of the bacterial community assembly in winter in the Pinus forest, corresponding to signi cantly lower bacterial Shannon diversity (Fig. 2), higher bacterial gene copy numbers (Supplementary Figure 11), the association of the bacterial Shannon diversity with soil pH (Fig. 3) and the bacterial βNTI value (Supplementary Figure 34), suggesting seasonality signi cantly increased the importance of stochastic processes, speci cally in the bacterial community with stronger deterministic assembly. This result is likely related to the seasonal transition leading to the selection of a particular group of soil microbiomes and in uencing several soil edaphic conditions. Thus, selected microbes may have eminent potential to increase functions connected to nutrient stocks and organic matter decomposition and to decrease functional genes involved in CNPS cycling in the Pinus forest in winter (Supplementary Figure 27), which was associated with a shift in the deterministic process to the stochastic process for the bacterial community with the transition of season from autumn to winter for this plot (Fig. 6). In accordance with the present results, previous studies have demonstrated that determinism-dominated assembly processes generally selected limited taxa, which led to limited stress and perturbance tolerance [97]. Therefore, we propose that microbial communities such as bacteria, which were characterized by the highest gene copy numbers and the lowest Shannon diversity in winter, are potentially more susceptible to the assembly transition from deterministic to stochastic with seasonal variations. In contrast, microbial communities such as fungi and protists with stochastic assembly are potentially more resistant to the assembly transition in alkaline soils seasonally. Thus, the uctuation of the microbial assembly is ultimately bene cial for ecosystem stability.
Studies have determined that environmental variables (such as air temperature, soil pH, and moisture) and habitat heterogeneity are vital determinants of community assembly [4,[98][99][100][101]. We observed that soil pH was positively correlated with the soil bacterial Shannon diversity. Therefore, the decrease in the soil bacterial Shannon diversity in winter (Fig. 2) may be associated with the decrease in soil pH (Supplementary Table 1), suggesting that the uctuation of soil pH in alkaline soil exerted substantial effects on the bacterial community assembly compared with the fungal and protist communities. Soil pH is regarded as a crucial environmental factor that shapes bacterial community assembly processes in agricultural soils [98]; speci cally, acidic soil led to the stochastic assembly of the bacterial community. In contrast, it has been reported that the dominant process for the bacterial community was homogenous selection in more acidic and alkaline soils, whereas stochastic assembly processes dominated at closeto-neutral pH in nonagricultural soils [99]. We speculate that in alkaline soil with low habitat heterogeneity (monoculture plantation), the bacterial community assembly may be driven by deterministic processes with a high possibility of seasonal in uence. By contrast, fungal and protist communities are more likely to be driven by stochastic processes, suggesting that stochastic processes may be more vital for soil microbial communities [96].

Conclusions
The results predicted the signi cant association of temperature with co-occurrence network parameters and soil enzymes, suggesting rst the impact of temperature as the main aspect of seasonal variations in soil biota occurrence, and second, that in any future global change scenario such as climate warming, the microbial interactions and functioning ecosystem will likely be greatly affected, at least in alkaline soils. The protist community composition was uniquely structured with C-related functional activities (lignindegrading enzymes, C-degradation and C-xation) relative to bacterial and fungal β-diversity variations, which were mostly explained by seasonal variations (Supplementary Figure 36). Our study highlighted the importance of the protist phagocytosis process on soil microbial interactions through the predicted impact of protist α-diversity on microbial cooccurrence network parameters. This association might be driven by the high abundance of protist consumers as the main predator of bacterial and fungal lineages in our sampling plots. Some functional categories, such as nutrient stocks and functional groups involved in the CNPS cycling genes, were signi cantly associated with the microbial cooccurrence network parameters, suggesting that the abundance of this functional group can be partly driven by microbial interactions. Bacterial communities were deterministically (variable selection and homogenous selection) structured, whereas the stochastic process of dispersal limitation was mainly responsible for the assembly and turnover of the fungal and protist communities. Additionally, we showed that winter triggered an abrupt transition in bacterial community assembly from a deterministic to a stochastic process in the Pinus forest that was closely associated with a reduction in the bacterial Shannon diversity, with the pattern of a high level of nutrient cycling (nutrient stocks and organic matter decomposition functional categories), suggesting that the bacterial community with deterministic assembly is potentially more susceptible to the assembly transition with seasonal uctuations in diversity and soil pH. This study contributes local-ecosystem prospects to model the behavior of soil biota seasonally and their implied effects on soil functioning and microbial assembly processes, which will bene t global-scale afforestation programs by promoting novel, precise and rational plantation forests for future environmental sustainability and self-su ciency.

Materials And Methods
Study area and soil sampling The Xiong'an New Area (38°43 -39°10 N, 115°38 -116°20 E), which includes three counties (Xiong, Anxin, and Rongcheng), is located in the north of China, which was established in 2017, and it is another new national city after the Shenzhen Special Economic Zone and Shanghai Pudong New District [9] (Supplementary Figure 35). Between 2017 and 2020, Xiong'an added more than 27,000 hectares of trees, increasing its forest coverage to 30 percent (http://www.xinhuanet.com/english/2021-04/01/c_139852880.htm). This area is classi ed as having a warm temperate continental monsoon climate with four distinct seasons. The average annual temperature and precipitation are approximately 12.1°C and 560 mm, respectively [9,11]. The soil samples were collected from a P. tabulaeformis plot (Chinese pine tree, hereinafter referred to as 'CP', with an approximate size of 3,706.01 m 2 ) and an S. japonica plot (Chinese scholar tree, hereinafter referred to as 'CS', with an approximate size of 4,114.78 m 2 ) in plantation forests with a history of planting either Zea mays L. or Triticum aestivum L., followed by the closing of the land for afforestation. These species were sampled one year post afforestation age, and the approximate age of the trees at the time of sampling was four years. The time frame for soil sampling was arranged considering four different seasons. Seasonal samplings were performed as follows: in July 2019 (summer: plant growth period and rhizodeposition; hereinafter referred to as 'SU'), in October 2019 (autumn; during the late phase of litterfall; hereinafter referred to as 'AU'), in January 2020 (winter; snow-covered time and carbon polymers/phenolics; hereinafter referred to as 'WI'), and in May 2020 (spring: three weeks after the emergence of leaves; hereinafter referred to as 'SP'). The time-scale sampling was selected according to the weather climate of the northern part of China. Sampling details were followed according to our previous studies [9,102]. Brie y, three random lines were randomly selected with the distance between each line at approximately 100 m, and by walking along each line, a soil core was collected every 5 m. The resulting soil cores were mixed to yield a composite sample from each line; three samples were collected per plot for each season. Thus, we collected 120 soil subsamples at depths of 15-30 cm in the root-zone soil (24 composite soil samples) from two de ned experimental plots over one year. After each set of sample collections, visible grass roots and pebbles were removed.
All soil samples were divided into two parts; the rst part (~ 10 g) was immediately frozen at -20°C using a portable refrigerator (Foshan Aikai Electric Appliance Co., Ltd, Guangdong, China) for DNA extraction and was stored at -80°C after the samples were transferred to the laboratory. The second part (~ 500 g) was transferred at an approximate temperature of 22 ± 2°C for samples collected in spring, 29 ± 2°C in summer, 12 ± 2°C in autumn, and 4 ± 2°C in winter and was stored at 4°C for geochemical measurements at the molecular ecology laboratory at Hebei Normal University.

Dna Extraction And Amplicon Sequencing
Soil genomic DNA extraction and quality checking were performed according to our previous publications [9,11]. PCR ampli cation and sequencing were individually performed for each replicate. The abundance of soil microbial communities was estimated using PCR ampli cation techniques. Target genes were bacterial 16S rRNA, fungal ITS, and protist 18S rRNA. The primers 338F/806R, 1737F/2043R, and TAReuk454FWD1F/TAReukREV3R were used to amplify the V3-V4 region of the bacterial 16S rRNA gene, fungal ITS1, and 18S rRNA, respectively, by an ABI GeneAmp® 9700 PCR thermocycler (ABI, CA, USA).
PCRs, ampli cations, and sequencing were performed in the framework of our previous studies [9,11,102].
Quantitative Real-time Pcr (Qpcr) And High-throughput Quantitative Pcr (Ht-qpcr) According to our previous protocol [11], the 7300 Real-Time PCR System (Applied Biosystems, California, USA) was used to determine the gene copy numbers of soil biotas for each soil sample. The details can be found in Supplementary Table 20. The abundance and diversity of the functional genes involved in C, N, P, and S cycling in different plantation plots and seasons were estimated using QMEC based on HT-qPCR [22], which enabled the simultaneous qualitative and quantitative determination of 72 genes (Supplementary Table 21) on the WaferGen Smart-Chip Real-time PCR system (Bio-Rad, Herculers, CA, USA).

Bioinformatics And Statistical Analysis
The details of the bioinformatics and statistical analysis are provided in the additional le1.
Page 33/36    The microbial community assembly processes across two forest types seasonally. The values of the beta nearest taxon index (βNTI) for soil bacterial (A), fungal (B), and protistan (C) communities present. The upper and lower signi cance thresholds for the βNTI index was + 2 and -2, respectively. The relative turnover in soil bacterial (D), fungal (E), and protistan (F) community assemblies, governed principally by deterministic processes (homogeneous and variable selections), stochastic processes (dispersal limitation and homogenizing dispersal), or undominated process. Bars with different lowercase letters indicate signi cant differences within the microbial taxa phyla (P < 0.05) across two forest types with seasonal variations, as revealed by one-way ANOVA with Turkey's post hoc test. Figure 7