3.2 Biodiversity-functioning relationships across land use types
The positive correlations between microbial richness and ecosystem respiration were found across land use types (Fig. 3a). The AOA OTUs richness was positively correlated with the PNR in the cropland. Non-significant relationship was observed in the shrubland and woodland (Fig. 3c). There was not significant relationship between the AOB richness and PNR across land use types (Fig. 3e). More importantly, the keystone tax was observed to have a contribution to ecosystem functioning. The keystone PLFAs amounts were negatively associated with the mSR (Fig. 3b). Only statistically significant correlation was observed between the keystone taxon abundance and the PNR in the cropland, but the relationship showed the opposite trend (Fig. 3d). Furthermore, the slope of the relationship between microbial richness and the mSR in the cropland was statistically different with the woodland (P < 0.001) and shrubland (P< 0.001). By contrast, the scaling slopes of the relationship between keystone taxon and mSR were close (P = 0.475 across different land use types).
Meanwhile, for total microbes, the distance-corrected dissimilarities of keystone species or total community composition with those of ecosystem functioning and environmental factors were identified (Fig. 4). Overall, for total microbes, the total PLFAs nor the keystone species composition was the strongest correlated with the mSR. The total AOA composition and the keystone AOB species were significant correlated with specialized ecosystem functioning (i.e. PNR) (Fig. 4b, c).
3.3 Microbial andammoxidation genes community’s co-occurrence patterns across land use types
In all land use types, the networks exhibited values of Modularity network, Clustering coefficient and Centralization degree were higher compared to their respective Erdös–Rényi random networks (Fig. 5; Table S2), suggesting all microbiome networks had small-world properties and modular structures. The overall topology indices of PLFAs communities did not significantly change across land use types possibly due to its low resolution. Networks had a greater number of negative correlations between nodes in the woodland of AOA (4.93) and the cropland of AOB (3.42). Meanwhile, the high network modality and average path length were found in the woodland of AOA and cropland of AOB, which suggested a complex and stability of amoA gene community in response to land use types (Fig. 5; Table S3).
3.4 Keystone species and environmental factors influencing the ecosystem functioning
The SEM revealed that the PLFAs richness, PLFAs composition and substrate supplying factors (Bulk density, TC, TN, and C/N) were significantly directly associated with the mSR, and other environmental factors (pH, soil temperature and moisture) indirectly affected the mSR across land use types (Fig. 6a). With regards to specialized function (i.e. NR and PNR), the keystone AOA composition, and substrate supplying factors indirectly regulated the PNR, the environmental factors also dominated the NR (Fig. 6b). Whereas in the AOB community, we found the AOB keystone composition was closely related with the PNR, and the substrate supplying factor directly impacted the AOB community richness nor the AOB community composition, and indirectly impacted the NR (Fig. 6c).