Fig. 3 (a)-(b): Compositions and structures of soil microbial communities between different stand ages in the L. gmeliniiplantations. The NMDS, the dynamics of microbial community during the increase of L. gmelinii plantations age and then assessed through a PERMANOVA.
The plot of NMDS (Figs. 3c and 3d) suggests variations in microbial beta diversity among the soil samples, with the response of soil bacterial beta diversity to afforestation time being larger than that of fungal beta diversity (bacteria, PERMANOVA, r = 0.48,p  < 0.001; fungal, PERMANOVA, r = 0.31,p  < 0.001), despite the significant variation in both species with afforestation time, and the differences in diversity between the restorative stages were much larger than the differences in diversity among soil depths.
Six bacterial phyla (Acidobacteria, Actinobacteria, Proteobacteria, Rokubacteria, Gemmatimonadetes, and Chloroflex) dominated forest soils, accounting for 95% of bacterial abundance. The relative abundances of these taxa varied by forest ages stages and soil depth. Notably, Rokubacteria significantly increased with increasing forest ages. Proteobacteria showed the opposite trend. Other phyla showed different responses with increasing stand age but these responses were insignificant. In the case of the fungal community, fungi could be classified into two different phyla, predominantly Ascomycota and Basidiomycota. The relative abundance of Basidiomycota increased but Ascomycota decreased with increasing forest age. The relative abundance of Basidiomycota gradually decreased with increasing soil depth in the younger stand (10 years old).
Using random-forest machine learning to discriminate samples with different samples based upon the microbiota composition (Zhang et al., 2018). The importance of indicator prokaryotic classes was assessed by 10-fold cross-validation (Xue et al., 2021). When the cross-validation error curve was at its lowest, a total of 24 OTUs were detected as biomarker fungi taxa and 42 OTUs as biomarker bacteria taxa. Among these classes, taxa exhibited different relative abundances in soils of different stand ages, and these results suggest that these taxa may be useful as biomarkers of soil microorganisms in planted forests (Fig. S2).
Using FUNGuild, 11 of all OTUs were assigned to putative fungal functional groups, which accounted for 54.5% of total biomarker fungi taxa. Pathotroph-saprotroph-symbiotroph (45.8%) were the abundant functional groups across all samples (Fig. S2). For trophic modes, the relative abundances of some dominant taxonomic groups changed apparently with forest ages. The relative abundances of saprotroph and symbiotroph in 30 and 50 years forests were greater than those of 10 years forest group, while the relative abundance of pathotroph aerobic chemoheterotrophy (AC) and aromatic compound degradation (ACD) was significantly lower than that of 10 years forest. Relative abundance of aerobic nitrite oxidation (ANO) significantly increased with stand age. These changes are closely related to soil nutrient limitation.