Network construction
To detect warming effects on potential fungal interactions, we used functional gene data to construct functional molecule ecological networks (fMENs) for control and warmed samples separately, via the pipeline of Molecular Ecological Network Analyses (MENA) (http://ieg4.rccc.ou.edu/mena/) (Deng et al., 2012). Only functional genes detected in all six biological replicates were included to ensure network reliability. To compare topological characteristics between control and warmed networks, we chose the same threshold of 0.98 during network construction, which refers to the Spearman’s correlation coefficients between pairwise OTUs. Network modules, which are groups of nodes that highly connected within the group but less connected outside the group, were separated by the fast-greedy modularity optimization. Network topological characteristics were calculated via MENA, including total nodes, total edges, average connectivity (node degree), average clustering coefficient, average path distance, and modularity. Network randomization (a total of 100 random networks) was performed via MENA. To test whether fMENs of control and warmed samples have typical properties of biological networks (e.g., small-world, hierarchy, modularity), network topological characteristics were compared with their corresponding 100 random networks. To test statistical differences of fMENs between control and warmed samples, we performed two-tailed Student’s t -tests by taking their network topological characteristics as averages and standard deviations of their corresponding 100 random networks as standard deviations (Wu et al., 2016). Cytoscape (version 3.5.1) was used for network visualization. Some network scores (e.g., connectivity) were used to identify putative key genes/species that have large influences in maintaining community structure and function, despite that species identities can be inaccurate owing to horizontal gene transfer. To select network key genes, we set thresholds based on within-module connectivity (Zi) and among-module connectivity (Pi) of nodes: Zi ≥ 2.5 and Pi ≥ 0.62 for network hubs, Zi ≥ 2.5 and Pi < 0.62 for module hubs, Zi < 2.5 and Pi ≥ 0.62 for connectors, Zi < 2.5 and Pi < 0.62 for peripherals (Guimera & Amaral, 2005). Previous studies have shown that network key genes/species were inclined to be more responsive to environmental changes than others (Wu et al., 2016). Therefore, we checked whether the connectivity of network key genes was linked to warming-induced changes in environmental factors by performing partial Mantel tests between the matrix of connectivity of network key genes and the matrix of environmental factors.