3.3 Factors driving rhizosphere microbial community structure and assembly
We analyzed the correlations between rhizosphere microbial community structure and soil environmental properties using the Mantel test and Spearman correlation analysis. Bacterial community structure was significantly correlated with MBC, AP, TP, NH4+-N, ON3--N, and SOC contents (P< 0.01), whereas fungal community structure was significantly correlated with NH4+-N, TN, TP, and SOC contents (P < 0.01; Supplemental information Fig. S4). Furthermore, we used random forest analysis to predict the contribution of soil environmental properties to β-diversity in rhizosphere bacterial and fungal communities. TP and NH4+-N were the major factors driving bacterial β-diversity (Fig. 2E), whereas NH4+-N and pH were the major factors driving fungal β-diversity (Fig. 2F).
3.4 Co-occurrence patterns of rhizosphere microbial communities across different forest ages
We constructed co-occurrence networks for rhizosphere bacterial and fungal communities separately, and determined network topographical properties to explore the effects of different forest ages. Consistent with the results of β-diversity analysis, RS15 and RS25 samples clustered in a module, while RS35 and RS45 samples clustered in another module for both bacteria and fungi (Fig. 3). We also mapped the fsOTUs (Fig. S5) into the rhizosphere microbial networks, and observed their clustering according to forest age (Fig. 3A–B).
We then explored the distribution patterns of fsOTUs in the co-occurrence networks of rhizosphere bacterial and fungal communities (Fig. 3, Table1). The abundances of inter-kingdom microbial associations also varied based on forest age, and there are difference among different modules (Fig. 3C–D). The module members were sensitive to forest age and their distribution in the network partially reflected the drivers of community dissimilarity illustrated in the PCoA plot (Fig. 2). Module 1 mainly contained fsOTUs specific to RS15 and RS25, which were not fully separated from other modules (Module 3; Fig. 3A–B). Numerous fsOTUs assigned to RS35 and RS45 were predominantly located in Module 2 (Fig. 3A–B). Our results suggested that the species with similar community phylogeny had more similar community structure, closer symbiotic relationship and easier aggregation in the network.
All the forest age-responsive modules comprised a taxonomically broad set of bacteria (Supplemental information Fig. S6A), revealing that the different forest ages did not target specific bacterial lineages. However, in the case of fungi, Olpidiomycota andKickxellomycota only existed in Module 1, andAphelidiomycota only existed in Module 2, whileZoopagomycota mainly existed in both Module 1 and Module 2 (Supplemental information Fig. S6B).
To reveal the effects of soil multinutrient cycling on the structures of sensitive bacteria and fungi, we used rhizosphere soil multinutrient index for regression analysis with normalized abundances of fsOTUs in different modules. The abundances of fsOTUs exhibited positive correlations with rhizosphere soil multinutrient index in Module 1 (bacteria: R 2 = 31, fungi:R 2 = 30); however, in Module 2 there were negative correlations (bacteria: R 2 = 24, fungi: R 2 = 38, Fig. 3E−F).
3.5 Identification of core microbiota and their potential roles in soil nutrient cycling
We identified 37 bacterial and 16 fungal classes as core microbiota (Fig. 4). Nitrososphaeria , Nitrospiria , Verrucom- icrobiae , Alphaproteobacteria, Gemmatimonadetes ,Bacteriodia, Bacillus , and Acidimicrobiia were considered core bacterial microbiota. Agaricomycetes ,Tremellomycetes , Motrierellomycetes ,Glomeromycetes , and Sordariomycetes were considered core fungal microbiota. Subsequently, we explored the role of core microbiota in rhizosphere soil nutrient cycling through co-occurrence network and random forest analyses. The co-occurrence networks of core bacteria and fungi were established based on correlation analysis (Fig. 4A). Core bacterial microbiota comprised 272 nodes, and other bacterial microbiota had 2,967 nodes. Core fungal microbiota contained 64 nodes, and other fungal microbiota had 1532 nodes. The degrees of different sub-communities were significantly higher (P < 0.001; Wilcoxon rank sum tests) in core microbiota than in other microbiota in both fungi and bacteria (Fig. 4B-C).
To disentangle the linkages between core microbiota and soil nutrient cycling, we quantified the contributions of core and other sub-communities to the rhizosphere soil multinutrient cycling index (Fig. 4C). In the course of forest restoration, the bacterial and fungal diversities of the core microbiota were more important than those of non-core microbiota. Furthermore, multiple regression modeling was used to evaluate the associations of core microbiota and rhizosphere soil nutrient contents. Different bacterial and fungal classes contributed to the variation in rhizosphere soil nutrient contents (Fig. 5). Particularly, TN, SOM, SOC, and NH4+-N contents were major factors associated with fungal core microbiota. NH4+-N content was a major factor associated with most bacterial core microbiota, and was positively correlated with Acidobacteriae, Alphaproteobacteria , andVerrucomicrobiae abundance, and negatively correlated with the abundance of most core bacteria.
Discussion