Statistical analyses
To assess how different environmental factors shaped the network parameters of IDEN and MEN, we used cluster analysis to assess the collinearity or redundancy of environmental variables by the “varclus” function in the “Hmisc” R package before further analyses. Only one variable was selected if pairs factors were highly correlated (Spearman’s R2> 0.6) as the representative variable. Using these criteria, MAP, MAT, carbon: nitrogen (C: N) ratio, SOC, STP, pH, SM, fungal richness, bacteria richness, plant richness and plant productivity were reserved in our analysis (Fig S1). Those environmental factors of 60 samples in desert steppe, alpine steppe and alpine meadow were showed by box plots, based on Kruskal-Wallis tests (Fig S2).
A random forest analysis was applied to identify the major environmental factors contributing to the variation in network topological features. The analysis was performed using the “randomForest” function in the “randomForest” package in R (Svetnik et al. 2003). Using the “a3” function to examine the significance values of the cross-validated R2 in the “A3” package; the significance of each predictor on the response variables was assessed with 2000 response variable permutations using the “rfPermute” function in the “rfPermute” package in R.
The correlation coefficients between topological features and environmental factors were calculated. The importance of environmental factors for topological features was estimated with multiple regression on distance matrices (MRM) in “ecodist” packages. The Euclidean distance matrices for environmental factors and topological features standardized with “decostand” of “vegan” package were used in MRM models. We furthermore quantified the relationship between these factors and topological features by linear fitting.
To assess the variation trend in function of microbial nodes along the plant richness gradient, we divided sites into three groups of low (0-8 species per sites), medium (9-13 species per sites) and high (14-28 species per sites) plant richness. The function of microbiota linked to plants were predicted, and assigned to three groups through sites they belong to. Ecological guilds of fungal OTUs linked to plants were assigned using the “funguild_assign” function in the “FUNGuildR” package in R. Only sequence taxonomy identity above 97% and the guild confidence ranking assigned to ‘Highly probable’ and ‘Probable’ was accepted (Nuskeet al. 2018). The PICRUSt 2 was used to predict the function (referring to KEGG pathway database) for each sequences of bacterial OTU linked to plants (Douglas et al. 2019).