4 DISCUSSION
4.1 The resistance of multifunctional and bacterial groups under different land use intensity
Our study showed that land-use change affected the resistance of ecosystem functions related to C, N and P cycling, resulting in changes in ecosystem multifunctional resistance as proposed in our first hypothesis. It is interesting that the main bacterial species (>3%) are the same different land uses. This indicates that the bacterial community has self-recovering ability during the succession process (Da C Jesus et al., 2009). However, the resistance of bacterial groups differed among different land use (Fig. 3). This is consistent with previous studies that different groups showed particular responses to land use change (Shange et al., 2012; Griffiths, 2012; Gomez-Acata et al., 2016). The responses are generally classified as copiotroph and oligotroph (Fierer et al., 2007). Proteobacteria and Actinobacteria can be categorized as copiotrophic, being well adapted to nutrient-rich environment (Zhang et al., 2016; Tian et al., 2017;Zheng et al., 2019). Studies have shown that the higher the organic matter content and pH in forest soil, the more the system tends to have larger relative populations of Actinomycetes and Proteobacteria (De Vries et al., 2012; Ramirez et al., 2014). In this study, the resistance of Actinomycetes and Proteobacteria in secondary forests were the highest phylums (Fig. 3a), which is consistent with the higher TOC content in secondary forests than that in soils that have been cleared and used for agriculture (AL and CL). Alphaproteobacteria, Betaproteobacteria and Deltaproteobacteria at class level were in accord with the Proteobacteria at phylum level (Fig. 3b). In contrast to copiotrophic groups, Acidobacteria grow rapidly in low nutrient environments (Navarrete et al., 2015a; Zheng et al., 2019), and is generally classified as oligotroph. Therefore, the resistance of Acidobacteria in abandoned land is also higher than in SF in this study (Fig. 3a). And at the class level, the resistance of Acidobacteria is consistent with their position at the phylum level (Fig. 3b). Verrucomicrobia has higher relative abundance in cultivated and abandoned land (Fig. 3a), being categorized as oligotroph and having strong resilience in low nutrient conditions (Battistuzzi & Hedges, 2009; Pan et al., 2014). Chloroflexi has strong adaptability in low-carbon soils (ability to survive desiccation and low nutrient availability conditions; Battistuzzi, 2009). Bacterial resistance corresponds to its correlation with its abundance (Fig. 2), that is, the stronger the resistance of the bacteria, the higher the abundance in the soil, the resistance and abundance of the bacteria have a close relationship with land use. This indicates that microbial communities can regulate the stability of ecosystems based on the resistance of the bacteria.
4.2 The linkage of the composition of microbial communities and ecosystem multifunctional resistance
The SEM analysis showed important effects of bacterial community structure on the multifunctional resistance of ecosystems, confirming our second hypothesis. This is consistent with previous studies emphasizing that multifunctionality resistance was regulated by changes in microbial composition (Delgado-Baquerizo et al.,2017b). In the response to global climate change, microbial functional capacity displayed a positive linear relationship with multifunctionality resistance, and buffered negative impacts on ecosystem functioning caused by climate change via regulating rich microbial communities (Luo et al., 2019). Our research indicates that in addition to protecting ecosystems against impacts of climate change, microbial communities also have a regulatory effect on the negative effects of land use change on ecosystem functions.
Of 98 bacterial classes selected through random forest models, 20 bacterial species were identified as the main predictors of multifunctional resistance (Fig.5a). This indicates that multifunctionality resistance is driven by multiple key microbial taxa, consistent with De Vries & Shade (2013). Consistent with our second hypothesis, oligotrophic bacteria (Verrucomicrobia and Chloroflexi) contributed more to predicting the multifunctional resistance of ecosystems than copiotrophic bacteria (Fig. 5a). Among bacterial species, OPB35_soil_group (phylum Verrucomicrobia) contributed the most (Fig. 5a), and were also selected as main drivers of soil single function (including DOC, TN, NO3- and TP; Fig. 6). Research shows Verrucomicrobial abundance is extremely sensitive to changes in chemical factors linked to soil fertility, such as soil total carbon, nitrogen, and phosphorus (Navarrete et al., 2015b), which supports our research. At the same time, Verrucomicrobia may exert a great impact with regard to nitrogen availability in certain ecosystem, including oligotrophic environments (Wertz et al., 2011). Chloroflex is another main predictor of multifunctional resistance, together with 8 other taxa at class level (Fig.5a). Chloroflexia and Anaerolineae were significantly positively correlated with β-1,4-glucosidase and phosphatase, promoting starch degradation and mineralization of organic phosphorus (Table S2). Ardenticatenia was significantly positively related to N-acetylglucosaminidase, which promotes the degradation of chitin. These findings support previous studies that oligotrophic bacteria promote the resistance of functions related to C cycle (De Vries & Shade 2013; Trivedi et al., 2013), but our research further shows that oligotrophic bacteria also contribute to the nitrogen and phosphorus cycles of the ecosystem. Actinobacteria and Proteobacteria, as copiotrophic bacteria (Navarrete et al., 2015c; Zheng et al., 2019), also play an important role in predicting multifunctional resistance (Fig. 5a). Thermoleophilia affected ecosystem functions by affecting starch degradation, leucine decomposition and organic phosphorus mineralization (Table S2). Our research showed Thermoleophilia not only significantly predicted multifunctional resistance (Fig. 5a), but also TOC, TN, TP (Fig. 6). Furthermore, Actinobacteria are significantly related to the total nutrient content in the soil and play an important role in soil carbon and nitrogen storage. Alphaproteobacteria contribute more for NH4+ concentrations (Fig. 6), catalyzing the rate of N cycle and mediating multifunctionality resistance by regulating NH4+. At the same time, cyanobacteria are not only predictors of multifunctional resistance (Fig. 5a), but also predictors of TOC and DOC (Fig. 6), which indicates that they also participate in the soil carbon cycle. Chamizo et al. (2018) considered that they play multiple roles in soil including synthesis of extracellular polysaccharides and contributing to accumulation of total organic carbon and total nitrogen. In addition, previous studies have shown that functional genes associated with nitrification and denitrification are the major predictors of multifunctional resistance (Luo et al., 2019); consistent with this, our research also shows that the nitrifier Nitrospira (phylum Nitrospirae) can predict multifunctionality (Fig. 5a).
4.3 Soil pH and elements driving multifunctional resistance
Besides soil bacterial communities, the SEM further indicated a significant effect of soil pH and certain elemental concentrations on multifunctional resistance (Fig. 4). pH is an important driving factor affecting many soil properties and microbial populations and acttivities (Fierer et al., 2007) and could affect multifunctional resistance via direct and indirect effects (Li et al., 2019). In this study, pH had a significant negative effect on N-related function resistance (Fig. 4e). This may be because pH significantly influenced substrate availability (Kemmitt et al., 2006).Pearson correlation analysis shows that pH has a negative effect on NO3- and NAG (Fig. S2), indicating that in neutral or slightly alkaline soils (average pH of the soil in the study area is 7.09), pH can reduce NO3- content and nitrogen functional enzyme activity to reduce N-functional resistance.
Soil vegetation cover and land management directly influence the cycling of chemical elements, and are key factors for soil biogeochemistry and also Al behavior in soil (Tejnecký et al., 2020). We found significant differences in soil Al, Mg and Ca concentration between land uses and these had significant effects on functional resistance (Fig. 4). Don et al. (2011) considered that the accumulation of some mineral elements has a negative impact on multifunctional resistance. Our study also showed that Al had a negative effect on multifunctional resistance (Fig. 4), which may be because Al reduced the activity of soil functional enzymes responsible for carbon, nitrogen and phosphorus transformations (Fig. S2). At the same time, the differences in soil organic matter content affect the differences in cation exchange capacity, especially Al3+ (Jiang et al., 2018). In our study, aluminum was positively correlated with TOC, DOC (Fig. S2), this explained the promoting effect of aluminum on C-functional resistance (Fig. 4). Similar to Al, Mg also promotes resistance to carbon function. This is because Mg can increase the content of DOC in the soil (Zhu et al., 2019), significantly correlated with DOC content (Fig. S2).We also found that soil Ca was significantly related to N-functional resistance (Fig. 4) and positively related to soil nitrogen content (Fig. S2).Previous studies have shown a positive correlation between soil Ca and nitrogen metabolism (Tang et al., 2019), soil Ca2+ was beneficial to nitrogen saturation and nitrate leaching in long-term soil nitrogen enrichment (Perakis et al., 2013), and Ca is related to the process of soil microbial nitrogen-fixation (Xie et al., 2016).