Malus sieversii decline altered soil bacterial community structure
According to our analysis, M. sieversii decline did not result in differences in bacterial richness (Chao 1) (R2 = 4, 528, F = 0.046, P > 0.05, df = 1) or Shannon’s diversity index (R2 = 0.007, F = 0.167, P > 0.05, df = 1), but bacterial community compositions were altered (Fig. 1A, PERMANOVA, P < 0.05). The bacterial community in soils of healthy M. sieversii had a higher beta diversity compared with that surrounding degraded M. sieversii (Fig. 1B, Wilcoxon test, P < 0.05). Beta-diversity decomposition analyses showed that soil bacterial community compositional dissimilarities among total, healthy, and degenerated M. sieversii trees were dominated by species replacement processes (contributions of 94.2%, 92.8%, and 95.5% for total, healthy, and degenerated M. sieversii, respectively), while richness difference processes only contributed 5.8%, 7.2%, and 4.5% on average (Fig. 1C). The abundance of Proteobacteria, the dominant bacterial phylum, increased with M. sieversii decline (P < 0.05, Supplementary Figure S1), with percentages ranging from 15.4% ± 2.9 in healthy soils to 19.1% ± 3.5 in degenerated soils. In addition, Chloroflexi, Gemmatimonadetes, and Nitrospirae had high relative abundances in healthy soils (P < 0.05, Supplementary Figure S1). The relative abundance of indicator species (OTUs) from these phyla varied greatly between soils of healthy and degraded M. sieversii trees (log2 fold change = 1.85–3.34, P < 0.05, Supplementary Table S1; Supplementary Figure S2).
As revealed by changes in the bacterial network topology, the bacterial community became less complex along with the transition from healthy to degenerated soils (Fig. 2A–C; Table 1). The network associated with M. sieversii decline was much looser than that of soil under healthy trees, with each node from healthy soil exhibiting an average of 11.6 degrees of connection compared with 6.1 degrees in degraded soil. The average clustering coefficient also decreased with M. sieversii degeneration, whereas the average path length score and modularity increased.
The hub OTUs in the network model were based on the top 10 most-connected nodes (Supplementary Table S2). A large fraction of hub or most-connected species under healthy M. sieversii soils were related to Gaiellaceae (Actinobacteria), Pseudonocardiaceae (Actinobacteria), and Micromonosporaceae (Actinobacteria), whereas Geodermatophilaceae (Actinobacteria) and Nitrososphaeraceae (Crenarchaeota) were the most connected taxa under soils of degraded M. sieversii. Finally, the potential functions of members of the bacterial community proved insensitive to M. sieversii decline (Supplementary Figure S3; R2 = 2.825, F = 0.841, P > 0.05, df = 1).
Table 1
Network topological properties from bacterial and fungal communities in the wild fruit forest. Networks for bacteria and fungi are based on 16S rRNA and ITS genes (n = 15 for healthy and degraded M. sieversii soils), respectively.
Microsite | Nodes | Edges | Mean degree | Density | Modularity | Average clustering co.a | Mean path length |
Bacteria | | | | | | | |
Healthy M. sieversii | 279 | 1626 | 11.61 | 0.042 | 0.53 | 0.51 | 4.08 |
Degraded M. sieversii | 286 | 878 | 6.14 | 0.022 | 0.66 | 0.43 | 5.30 |
Fungi | | | | | | | |
Healthy M. sieversii | 252 | 722 | 5.73 | 0.023 | 0.59 | 0.39 | 4.80 |
Degraded M. sieversii | 261 | 484 | 3.71 | 0.014 | 0.77 | 0.38 | 5.69 |
Degradation of M. sieversii trees was correlated with potential soil fungal functions
Fungal richness and Shannon’s diversity index were insensitive to M. sieversii decline (R2 = 1,893, F = 0.251, P > 0.05, df = 1; R2 = 0.166, F = 0.585, P > 0.05, df = 1; Supplementary Figure S1). Healthy and degenerated M. sieversii harbored significantly different fungal communities (Fig. 1D, PERMANOVA, P < 0.05). Furthermore, beta diversity decomposition analyses revealed that fungal community compositional dissimilarities among trees were dominated by species replacement processes (contributions to beta diversity of 82.5%, 82.8%, and 80.4% for total, healthy, and degraded M. sieversii, respectively), whereas the contribution of richness difference processes ranged from 17.2–19.6% on average. When fungi were separated into functional guilds, negative effects on beta diversity were detected for symbiotic and saprotrophic fungi, whereas positive effects were detected for pathotrophic fungi (Fig. 3), thus suggesting that M. sieversii decline exerted large effects on fungal guild compositions. Moderate levels of M. sieversii decline may thus have cascading consequences for fungi or affect higher-order microbial interactions.
The most abundant fungal species present in the wild fruit forest were from phylum Ascomycota, followed by Basidiomycota (Supplementary Figure S1). Malus sieversii decline was associated with increased dominance of Lasiosphaeriaceae, Clavicipitaceae, and Hypocreaceae (log2 fold change 1.41–4.73, P < 0.05, Supplementary Table S1) and decreased dominance of Mortierellales and Trichocomaceae (log2 fold change 1.31–5.15, P < 0.05, Supplementary Table S1).
A lower average degree and a higher average path length were observed in the co-occurrence network of degraded soil compared with that of healthy M. sieversii, which indicates that the interconnectedness of the fungal network was reduced with M. sieversii decline.
Microbial community assembly processes
Community-level habitat niche breadths were used to explore the relative importance of deterministic and stochastic processes in the microbial community assemblies (Fig. 4). A significantly higher niche breadth was observed for bacteria than for fungi in the wild fruit forest (P < 0.001). Moreover, bacterial niche breadth was significantly higher in soils of healthy M. sieversii compared with degraded soil, whereas fungal niche breadth values were similar in both types of soils. NCM prediction successfully estimated a large percentage of the relationship between the occurrence frequency of OTUs and their relative abundance variations, with a better goodness of fit for bacteria than for fungi (72.4% and 61.4% of variation explained, respectively; Fig. 5). Furthermore, bacteria and fungi associated with M. sieversii decline had increased migration rates, which indicates that the microbes were less limited by dispersal. The NST was above the 50% boundary point for bacteria but below this point for the fungal community, thus suggesting that stochasticity played a relatively major role in the bacterial assembly but that deterministic processes were more important for fungi.
Characterization of edaphic variables of healthy and degraded M. sieversii
Degradation of M. sieversii increased both SOC (Mann-Whitney U, P < 0.05, Supplementary Figure S4) and microbial biomass (microbial C; t = 2.474, df = 28, P < 0.01, Supplementary Figure S4), with 9.1% more C and 26.8% more MBC observed in soils surrounding degraded M. sieversii. Degradation of M. sieversii reduced the soil pH from 7.32 to 7.02 (t = 4.213, df = 28, P < 0.001, Supplementary Figure S4). Malus sieversii decline did not significantly change TN, TP, AN, the Corg:Cmicro ratio, or the C:N ratio but boosted SMF (P < 0.05, Supplementary Figure S4).
Linkages of bacterial and fungal community composition, niche overlap, and assembly with soil environmental factors
Among all analyzed environmental factors, total N and the C:N ratio exhibited the highest correlations with niche overlap of the bacterial community (P < 0.05; Fig. 6), whereas available N was the factor most significantly correlated with fungal community niche overlap (P < 0.05; Fig. 6). In regard to healthy M. sieversii, pH directly affected the composition of bacterial communities, but the C:N ratio was significantly correlated with community niche distribution. Mantel tests further indicated that the changes in fungal niche overlap were positively correlated with total N, the Corg:Cmicro ratio, and SMF, whereas soil P pool affected the fungal community assembly. Finally, no factors affecting the bacterial and fungal communities around degenerated M. sieversii were detected.