3. RESULTS

3.1 Variation in multifunctionality

We first explored the vertical variation in multifunctionality in desert ecosystems. Overall, the multifunctionality index did not exhibit a significant difference between whole superficial soils (0-20 cm) and deep soils (20-100 cm) (p > 0.05) and slightly decreased as soil depth increased through the soil profiles (Fig. 1 a, b). We also found that the multifunctionality index significantly decreased along the desertification gradient (Fig. 1 c) but did not show different changes with depth between the superficial and deep soils (Fig. 1 d).
Three individual functional indexes (C, N, and P processes) were evaluated (Figure S1). Overall, the C and N processes were higher at the potential desertification (PD) sites but did not show different changes between the moderate desertification (MD) and severe desertification (SD) sites. However, the P process did not differ between the whole superficial and deep layers along the desertification gradient. The C process did not exhibit a significant difference between superficial soil and deep soil at the MD and SD sites. The N process differed between superficial soil and deep soil at only the SD sites. The C and N processes typically decreased as the soil depth increased throughout the entire profile.

3.2 Variation in soil microbiomes

We acquired a total of 5490761, 7562145, and 5357782 high-quality bacterial, fungal, and archaeal sequences (from the 125 samples), which were classified into 4890, 7723, and 3246
operational taxonomic units (OTUs), respectively. Bacteria were mainly composed of the phyla Actinobacteria (40.06%), Proteobacteria (27.23%), Chloroflexi (8.26%), and Acidobacteria (7.89%). Fungi were mainly composed of the phyla Ascomycota (69.70%), Basidiomycota (9.85%), and Zygomycota (5.39%). Archaea were mainly composed of the phyla Thaumarchaeota (81.34%) and Euryarchaeota (15.27%). Overall, Acidobacteria, Chloroflexi, and Thaumarchaeota typically increased, while Actinobacteria, Proteobacteria, and Euryarchaeota significantly decreased during the process of desertification (Figure S2).
Regarding the soil profiles, distinct variation trends of the microbiomes were observed along the desertification gradient. First, Ascomycota, Basidiomycota, Euryarchaeota and Thaumarchaeota did not differ between the superficial and deep layers at any desertification stage. In contrast to Proteobacteria, Acidobacteria, Actinobacteria and Chloroflexi in the superficial soils were much richer than those in the deep soils (Figure S3). Second, Acidobacteria, Actinobacteria, Chloroflexi, and Thaumarchaeota significantly decreased and Proteobacteria significantly increased as soil depth increased (Figure S4). The most abundant phyla of bacteria, fungi, and archaea showed distinct variation trends as soil depth increased along the desertification gradient (Figure S5). The significance of the primary phyla composing the microbiomes was confirmed by a Kruskal-Wallis H test and least-squares linear regression analysis.
The alpha-diversity levels of the soil microbiomes were all higher at the PD and MD sites than at the SD sites, except for the Shannon index for archaea (Fig. 2). As soil depth increased, bacterial, fungal and archaeal diversity exhibited different fluctuating trends in the desert ecosystems. The Shannon and phylogenetic diversity indexes of bacterial diversity significantly decreased as soil depth increased, and the phylogenetic diversity index of fungi significantly decreased, but the Shannon diversity index of archaea significantly increased as soil depth increased. Vertical variation tendencies were determined by least-squares linear regression models. In addition, distinct variation tendencies for alpha diversity were detected during the desertification process (Figure S6). The Shannon index of bacteria at the SD sites and the phylogenetic diversity index of bacteria did not exhibit different changes between superficial and deep soils. The alpha diversity of fungi in superficial and deep soils did not exhibit different changes, except at the PD sites. The alpha diversity of archaea in superficial and deep soils did not exhibit different changes, except for the Shannon index at the MD sites.
NMDS analysis suggested that soil microbiomes at different desertification sites formed distinct clusters in ordination space (Fig. 3). The significant spatial variations in species composition were further confirmed by ANOSIM and ADONIS tests (Table S3). These variations were the highest for bacteria, followed by fungi and archaea, suggesting that the bacterial communities were more likely to be reshaped by desertification. We observed different changes in microbiomes between superficial and deep soils (Fig. 3 and Table S4). These changes were larger for bacteria and archaea than for fungi, suggesting that the bacterial and archaeal communities had higher feedbacks with soil depths. The significant variations in beta diversity among different microbiomes were evaluated and identified (Fig. 3). Fungi exhibited the highest beta diversity, suggesting a higher community dispersion level. The vertical spatial variation (VDR) of different microbiomes was compared among the desertification sites. We further found that the VDR slopes of the soil microbiomes at the PD sites were steeper than those at the MD and SD sites. The bacterial and archaeal VDR slopes were significantly steeper than the fungal VDR slopes, suggesting that desertification markedly enhanced the vertical variation in bacterial and archaeal communities but had less effect on the variation in fungal communities.

3.3 Links between the microbial community and multifunctionality

The relationships between the microbial community and multifunctionality in desert ecosystems were explored (Figs. 4 and 5). The dominant bacterial and archaeal phyla were significantly (P < 0.01) related to multifunctionality, but the dominant fungal phyla did not exhibit different relationships (P > 0.05) (Fig. 4). The soil microbial diversity of bacteria and fungi was positively related to multifunctionality, while archaeal diversity showed the reverse pattern (Fig. 5). Further analyses provided evidence that dominant species (Acidobacteria and Chloroflexi) and the fungal phylogenetic diversity index had significantly stronger and more positive correlations with multifunctionality than did microbial diversity (Shannon and phylogenetic diversity indexes). Furthermore, we found significantly stronger and more positive correlations between microbial diversity and individual functions (i.e., C, N, and P processes), as well as between this diversity and combinations of functions (i.e., C and N, N and P, C and P, and C, N, and P processes) (Table S5). In particular, Acidobacteria, Chloroflexi and fungal phylogenetic diversity were positively and strongly related to the individual functions and combinations of functions.

3.4 Accounting for soil multifunctionality drivers in desert ecosystems

Random forest modelling was adopted to determine and compare the most important predictors of multifunctionality in desert ecosystems. We found that Acidobacteria and fungal phylogenetic diversity were more important than the other multifunctionality predictors (Fig. 6 a). The predominant predictors of multifunctionality differed between superficial and deep soils (Fig. 6 b, c). While the fungal phylogenetic diversity index best predicted multifunctionality in superficial soils, Acidobacteria and Thaumarchaeota were instead pivotal in deep soils. Further, the microbial species index was the predominant predictor of multifunctionality (Fig. 6 d).
Our structural equation modelling (SEM) explained 65.0% of the variance in multifunctionality at the desertification sites (Fig. 7). The microbial species index appeared to have a distinctly stronger total positive effect than the biodiversity index on multifunctionality (Fig. 7 and Figure S7). Depth and soil electrical conductivity (EC) had the strongest total positive effects on multifunctionality, as indicated by the standardized total effects from SEM.