Aridity gradient overrides degradation in shaping the topsoil microbiome of the Tianshan wild fruit forest

Highlights

• •Aridity index, rather than degradation level, is the paramount driver shaping soil bacterial and fungal communities (explaining ∼43% and ∼30% of the variation, respectively).
• •The ecosystem harbors an exceptionally stable core microbiome, with 18,510 bacterial ASVs shared across aridity gradients.
• •Dry sub-humid (alternating wet-dry) conditions promote the most complex and stable microbial co-occurrence networks.
• •Community assembly of both bacteria and fungi is predominantly governed by stochastic processes (ecological drift).
• •Microbial functional profiling reveals nitrogen cycling as the dominant process, with distinct adaptations along the aridity gradient.

Abstract

The Tianshan wild fruit forest, a critical relict ecosystem and biodiversity hotspot, is experiencing severe degradation, a process widely assumed to be the primary driver of negative shifts in soil microbial communities. However, in arid regions, the overarching influence of climatic aridity may supersede localized degradation effects, creating a pivotal scientific question: which factor—aridity or degradation—dominantly regulates the soil microbiome in this unique habitat? To address this, we analyzed 360 topsoil samples across degradation and aridity gradients using high-throughput sequencing. The results demonstrate that the aridity index, not degradation level, is the paramount factor shaping microbial community structure, explaining the largest proportion of variation in both bacterial (∼43%) and fungal (∼30%) communities. The ecosystem harbors an exceptionally stable core microbiome, with community assembly predominantly governed by stochastic processes. Notably, alternating wet-dry (dry sub-humid) conditions significantly enhanced the complexity and stability of microbial co-occurrence networks compared to semi-arid or humid regions. Additionally, soil microbes mediated multiple core ecological processes, with nitrogen cycling as the most abundant—dominated by Rhodoplanes, Alcaligenes, and rhizobial taxa (Bradyrhizobium, Allorhizob), among others—wherein nitrate reduction was exceptionally active in the Tianshan wild fruit forest, particularly in humid and semi-arid habitats. These findings challenge the prevailing degradation-driven paradigm, highlighting aridity as the master regulator of microbial communities. This insight is crucial for guiding conservation strategies, emphasizing that managing water availability and protecting dry sub-humid habitats are essential for maintaining the microbial stability and functional resilience of this invaluable ecosystem under climate change.

Graphical abstract

1. Introduction

The Tianshan wild fruit forests, located in the Xinjiang region of China, represents a precious relict of Tertiary temperate broad-leaved forests and a globally rare, unique ecosystem (Tao et al., 2021). This ecosystem is not only recognized as a center of origin for modern cultivated fruit trees such as apples but also serves as a critical biodiversity hotspot and an invaluable gene pool for wild fruit tree germplasm (Duan et al., 2017). Beyond its biogeographical significance, the wild fruit forest plays an irreplaceable role in sustaining regional ecological balance by providing essential services, including soil conservation, water retention, and carbon sequestration (Wu et al., 2024). However, this invaluable ecosystem has been experiencing severe degradation over recent decades, driven by a combination of climate change impacts and escalating human activities. This degradation manifests as large-scale mortality of keystone species, notably Malus sieversii (Ledeb.) M. Roem., leading to a significant decline in ecosystem functionality and a substantial threat to local and regional biodiversity (Sun et al., 2020).

Soil microorganisms are fundamental components of forest ecosystems, acting as key drivers of ecosystem stability (Guo and Gong, 2024). They mediate critical biogeochemical processes, including soil organic matter decomposition, nutrient cycling (e.g., carbon, nitrogen, and phosphorus), and the establishment of synergistic relationships with plant roots that influence plant health and productivity (Delgado-Baquerizo et al., 2016). Consequently, significant shifts in the structure and function of soil microbial communities are both a cause and a consequence of ecosystem degradation. In the context of the Tianshan wild fruit forest, previous research has begun to illuminate the microbial dimension of degradation. Studies have indicated that the degradation of M. sieversii populations is associated with reduced β-diversity of symbiotic bacterial and fungal taxa in the topsoil, an increased relative abundance of pathogenic fungal groups, and a disruption in the complexity of microbial co-occurrence networks (Rong et al., 2024). These changes suggest a breakdown in the stable, functional microbial core that underpins ecosystem health.

The conventional paradigm for understanding microbial shifts in degraded ecosystems like the Tianshan wild fruit forest has predominantly focused on the direct and unidirectional impacts of vegetation degradation. Biotic and abiotic stressors, such as the invasion of the apple buprestid beetle (Agrilus mali Matsumura), unsustainable human land use, and extreme climate events, are rightly identified as primary drivers of forest degradation (Tao et al., 2021). The logical extension of this view is that the visually apparent process of ecosystem degradation—characterized by dead branches, reduced canopy cover, and soil nutrient loss—is the principal force restructuring the soil microbiome. This perspective naturally leads to restoration strategies focused on mitigating these degradation drivers, for instance, through pest control, reforestation, and soil amendment.

However, this degradation-centric view may represent an oversimplification of the complex interplay of environmental filters shaping microbial communities in large-scale, heterogeneous landscapes. A critical, yet potentially overlooked, factor is aridity. The Tianshan Mountains are situated in a geographically arid region, and global climate change is predicted to increase the frequency and intensity of drought conditions in this area. Water availability is recognized as a primary abiotic factor regulating soil microbial biomass, diversity, and community composition in arid and semi-arid ecosystems (Nielsen and Ball, 2014). An emerging hypothesis from dryland ecology suggests that in systems where water is a major limiting factor, the large-scale filtering effect of climatic aridity may override the influences stemming from more localized processes like land-use change or degradation (Maestre et al., 2015; Zhang et al., 2024). This poses a pivotal scientific question for the Tianshan wild fruit forest: is the structure and assembly of the topsoil microbiome primarily driven by the overarching climatic gradient of aridity, or is it predominantly shaped by the localized severity of ecosystem degradation?

Resolving this question has profound implications for accurately predicting ecosystem responses to global change and for formulating effective, evidence-based ecological restoration strategies. If microbial community changes are primarily a consequence of degradation, then restoration measures should rightly focus on improving vegetation structure and soil quality. Conversely, if aridity is the dominant regulator, then management and restoration efforts may need to prioritize water resource management and the selection of drought-tolerant plant-microbe combinations to enhance ecosystem resilience. To date, no clear consensus has been reached regarding the relative importance of these two factors in this unique ecosystem. Most studies have focused on the impacts of degradation alone, while systematically overlooking the potential dominant role of aridity—a core climatic characteristic of the Tianshan Mountains (Guo and Gong, 2024). This constitutes a significant cognitive gap in our understanding of the ecosystem's microbial ecology.

Therefore, this study was designed to systematically investigate the drivers of topsoil microbial community structure, assembly mechanisms, and network characteristics across the environmental gradients of the Tianshan wild fruit forest. We hypothesized that: 1) the aridity gradient, rather than the forest degradation level, is the primary environmental factor regulating the structure, assembly, and co-occurrence patterns of topsoil bacterial and fungal communities; and 2) the microbiome of this ecosystem exhibits inherent stability, governed largely by stochastic assembly processes, and that alternating wet-dry environments promote more complex and stable microbial networks than consistently semi-arid or humid conditions.

To test these hypotheses, we collected 360 topsoil samples from wild fruit forest habitats spanning northern and southern Tianshan, encompassing a range of degradation states and natural aridity gradients. We employed high-throughput sequencing of the 16S rRNA and ITS gene regions to characterize the bacterial and fungal communities. This study aimed to: 1) compare the effects of degradation level and aridity index on microbial alpha and beta diversity; 2) identify the core microbiome and its distribution across environmental gradients; 3) quantify the relative contributions of deterministic versus stochastic processes in community assembly using null model analysis; and 4) characterize the complexity and stability of microbial interaction networks under different aridity regions.

2. Materials and methods

2.1. Sampling and physicochemical characterization

A total of 360 samples, from 60 sample sites, were collected from wild fruit forest ecosystem on July 2023. The topsoil samples from different area including degraded rates, reflected by the mortality rate of branches, were collected at depth from 0–10 cm (Fig. 1). Six trees were selected randomly from each sample site for getting six duplicates, in order to be more scientific and representative, five samples around one tree from five directions were mixed into one sample at each tree. All the samples were separated into two parts, one for molecular samples and another for physicochemical samples. The samples for DNA extraction were stored in liquid nitrogen during transportation, and stored at −80 °C upon arrival in the laboratory until further analysis (Liu et al., 2023). In addition, the samples for physicochemical analysis of environmental parameters such as SOC (soil organic carbon), EC (electrical conductivity), pH, TN (total nitrogen), TP (total phosphorus), TK (total kalium), AN (available nitrogen), AP (available phosphorus), and AK (available kalium) were air dried at room temperature. They were then ground into a homogeneous powder and sieved through the 0.25 mm and 2 mm sieve, respectively. Considering that the aridity index at each sampling site is the same, we mixed the six duplicate samples from each sampling site before measuring the physicochemical properties of the soils. Measurement of the nutrients was done by FOSS-fully automatic azotometer, Cary 60-spectrophotometer and flame spectrophotometer at the Research Center for Ecology and Environment of Central Asia (Liu et al., 2019).

Fig. 1.

The sampling sites of the Tianshan wild fruit forest. I, II, III, IV, and V represent the degrees of degradation.

2.2. DNA extraction and Illumina sequencing

Community DNA was extracted from 0.5 g sample using the PowerSoil Kit (MP Biomedical, OH, USA) following the manufacturer’s direction. The genomic DNA extracted from soil samples was first quantified using a NanoDrop™ 2000 spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, USA) to assess purity (A260/A280 ratio: 1.8–2.0) and concentration. For microbial community profiling, the V3-V4 hypervariable regions of the bacterial 16S rRNA gene were amplified using the barcoded primer pair 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) (Caporaso et al., 2011), while the ITS1 hypervariable region of the fungal ITS rRNA gene was amplified with the barcoded primer pair ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) (Gardes and Bruns, 1993) and ITS2 (5′-GCTGCGTTCTTCATCGATGC-3′). PCR amplification was performed in a 25 μL volume reaction mixture containing 25 ng of template DNA, 12.5 μL PCR Premix, 2.5 μL of each primer, and PCR grade water to adjust the volume (Li et al., 2021). The PCR products from each sample were pooled with equimolar concentrations and then sequenced on a Novaseq 6000 platform (Illumina Inc., USA) (250 bp paired-end reads). Paired-end reads were assigned to samples based on their unique barcode and truncated by cutting off the barcode and primer sequence, then merged using FLASH (version 1.2.11). For obtaining high-quality tags, the sequences data were filter by using Trimmomatic (version 0.39), denoising and chimera checking were accomplished by using the DADA2 (Callahan et al., 2016) tool in QIIME2 software (Bolyen et al., 2019) (version 2020.6).

2.3. The amplicon sequencing data processing and bioinformatic analysis

Paired-end reads were assigned to samples based on their unique barcode and truncated by cutting off the barcode and primer sequence, then merged using FLASH (v1.2.8) (Magoc and Salzberg, 2011). The preprocessed reads were pooled and clustered into amplicon sequence variants (ASVs) with a frequency threshold of < 0.01% of the total sequences. Only ASVs present in at least two samples were retained for subsequent analysis (Callahan et al., 2016; Bolyen et al., 2019). Alpha diversity indices, including Shannon and Simpson index were calculated using QIIME2 (https://qiime2.org/) and BiodiversityR (v2.12–1) package in R (v3.6.0). The beta diversity of the samples was evaluated using principal coordinate analysis (PCoA) based on unifrac metrics. The results were then visualized using R software (version. 4.2.2). For group comparisons of Alpha diversity (presented in violin plots), the Wilcoxon rank-sum test was used, and the Benjamini-Hochberg false discovery rate (FDR) correction was applied to control for type I errors. Beta diversity analysis (principal coordinate analysis, PCoA) was addressed using permutational multivariate analysis of variance (PERMANOVA) based on Bray-Curtis distance with 999 permutations, using the Adonis function in the R vegan package, and the significance in boxplots/violin plots was indicated by * for Wilcoxon test results. Venn and UpSet plot diagrams were performed in the R project VennDiagram (v.1.6.20) and UpSetR (v1.4.0) package to show the unique and common taxonomic units across the samples. Linear discriminant analysis (LDA) effect size (LEfSe) analysis was performed to identify microbial biomarkers that discriminated among different aridity gradient groups by using the lefser package (version 1.19.1) in R (version ≥ 4.5.0) via Bioconductor (version 3.22), which is an R implementation of the original LEfSe method. Additionally, the significance of compositional differences between microbiomes from soil samples Semi-arid regions, Dry sub-humid regions, and Humid regions was assessed using the PERMANOVA test, with a significance threshold set at P < 0.05.

2.4. Microbial community analysis

The taxonomic identity was determined using the SILVA 138 database (Quast et al., 2013). The annotations of each ASV were assigned to different taxonomic levels including phylum, class, order, family, and genus. This comprehensive annotation allowed for further analysis and interpretation of the taxonomic composition of the samples. The distance-based redundancy analysis (dbRDA) was conducted to identify the dominant environmental factors shaping the microbial communities in the wild fruit forests of Xinjiang. The community assembly mechanism of stochastic and deterministic processes was disentangled using a statistical framework based on the β nearest-taxon index (βNTI) by using the R packages "picante" (version 1.7) and "vegan" (version 2.5–2) (Zhou and Ning, 2017). To explore the co-occurrence patterns and interaction networks of bacterial and fungal communities across different aridity gradients, microbial co-occurrence network analysis was performed following a standardized workflow. All figures were generated using the R package microeco (version 1.15.0; Tutorial for R microeco package) (Liu et al., 2021) and its dependent packages.

2.5. Co-occurrence network analysis and functional annotation by FAPROTAX

Microbial co-occurrence network analysis was performed based on the genus-level abundance data using R software (v4.3.0). First, the Spearman correlation coefficient matrix was calculated with the Hmisc package to reduce the impact of compositional bias and data sparsity, and the correlation threshold was set to |r| ≥ 0.6 with P < 0.05 to filter significant associations. The adjacency matrix was then constructed and converted into an undirected network using the igraph package, with nodes representing genera and edges indicating significant correlations. The co-occurrence networks were visualized and aesthetically optimized using Gephi (v0.9.7) (Bastian et al., 2009). For functional prediction of the microbial community, the Functional Annotation of Prokaryotic Taxa (FAPROTAX, v1.1) tool was employed, which infers ecological functions based on manually curated taxon-functional associations from published literature. The genus-level taxonomic abundance table (annotated against the Silva database) was input into the FAPROTAX pipeline via the collapse_table.py script, with default parameters to generate functional group abundance profiles, focusing on biogeochemical cycling processes (e.g., nitrogen, carbon, and sulfur cycling).

3. Results

3.1. Amplicon sequencing data and microbial diversity based on degradation

According to the results by 16S rRNA gene sequencing, a total of 26,669,108 clean reads were obtained from the 360 samples with an average of 74,081 clean reads per sample; by ITS1 gene sequencing, a total of 24,580,990 clean reads were obtained from the 360 samples with an average of 68,281 clean reads per sample after removing low-quality sequences. The sequence length in this study was mostly distributed in the range of 400∼440 bp of bacteria and 200–280 bp of fungi. The high coverage (>99% for all samples) confirms the sufficiency of the sequencing depth for reliable analysis. Based on these sequencing data, we first conducted an alpha diversity analysis of the microbial communities in these 360 samples. According to the preliminary experimental design, the 360 samples were divided into five groups based on the degradation status of the wild fruit forest, ranging from Level I to Level V, where Level I represents healthy apple trees and Level V represents dead apple trees. As shown in (Fig. 2), no significant differences in alpha diversity were observed among these five groups, whether measured by the Shannon index, Simpson index, Chao1, or ACE index. Therefore, we further analyzed the alpha diversity of fungi, and the results also demonstrated the same results.

Fig. 2.

Microbial diversity of surface soil in wild fruit forests based on five degrees of degradation. I, II, III, IV, and V represent the degrees of degradation.

3.2. Aridity is the main factor affecting the microbial community of wild fruit forests in Xinjiang

Given that the degradation level of the wild fruit forest exerted no significant impact on the topsoil microbiome, we hypothesized that other environmental factors must govern the microbial community structure. To identify these factors, we measured the physicochemical properties of the soil (Supplementary Table S1), and conducted a comprehensive analysis using distance-based redundancy analysis (dbRDA) and Random Forest models on key soil physicochemical properties. The results demonstrated that collectively, all measured soil physicochemical factors significantly shaped both bacterial and fungal community composition (Fig. 3). The dbRDA1 and dbRDA2 explained 63.0% (36.8% and 26.2%, respectively) of the total variation in bacterial communities (Fig. 3a) and 43.4% (25.8% and 17.6%, respectively) for fungal communities (Fig. 3b). Random Forest analysis further quantified the relative importance of individual factors. For bacterial communities, the aridity index (AI) was the main effect factor, accounting for approximately 43% of the explained variation (P < 0.001), followed by electrical conductivity (EC), available phosphorus (AP), and pH (Fig. 3c). For fungal communities, AI remained the paramount factor (P < 0.001, ∼30% contribution), with total nitrogen (TN), available potassium (AK), soil organic carbon (SOC) also exhibiting extremely significant and important influences (P < 0.001) (Fig. 3d). In conclusion, these findings indicate that aridity (AI) is the core environmental factor regulating the microbial community structure in the wild fruit forest, while soil nutrient content (e.g., N, P, K), salinity, and pH collectively contribute to shaping the distinct assembly patterns of both bacterial and fungal communities.

Fig. 3.

The dbRDA analysis of the community composition of bacteria (a) and fungi (b) in wild fruit forests, soil physicochemical factors, and the contribution of random forest from each factor (c, bacteria; d, fungi).

For further, analysis of soil physicochemical properties under different aridity gradients (BGH: semi-arid region; GS: dry sub-humid region; SR: humid region) in the wild fruit forest revealed significant inter-group variations (Fig. S1). Soil organic carbon (SOC) and total nitrogen (TN) exhibited congruent trends, with both parameters peaking in the GS group, where values were significantly elevated compared to the BGH and SR groups (P < 0.05). Total phosphorus (TP) was likewise significantly enriched in the GS group. Although the BGH and SR groups did not differ significantly from one another, both registered significantly lower TP than the GS group. In the case of total potassium (TK), while no significant difference was detected between BGH and SR, the GS group differed significantly from both. Available nitrogen (AN) was present at significantly higher levels in the GS group relative to BGH and SR, indicating a greater potential for nitrogen availability in this group. Available phosphorus (AP) was generally significantly lower in the SR group than in GS, despite an isolated high-value outlier observed in the BGH group. Available potassium (AK) demonstrated a distinct gradient across the aridity spectrum: GS > BGH > SR, with statistically significant differences (P < 0.05) separating all groups. Soil pH was significantly more acidic in the BGH group than in either GS or SR (P < 0.05), with no significant difference observed between the latter two groups. Electrical conductivity (EC) increased progressively along the gradient, being significantly higher in SR than in GS, which in turn was significantly higher than in BGH (P < 0.05), suggesting an accumulation of salts in the SR soils. The aridity index (AI) varied markedly across groups in the order SR > GS > BGH (P < 0.05), consistent with the predefined gradient. Finally, the dead branch ratio (DBR) was significantly lower in BGH than in both GS and SR (P < 0.05), with no significant difference between these latter two groups.

3.3. The impact of aridity on microbial diversity in wild fruit forests in Xinjiang

A comprehensive analysis of the physicochemical properties of topsoil in the wild fruit forest indicated that aridity is a predominant factor influencing the topsoil microbiota. Accordingly, following established aridity classification criteria, we re-categorized the 360 samples into three distinct groups based on the aridity index (AI): BGH (semi-arid: 0.2 < AI < 0.5), GS (dry sub-humid: 0.5 < AI < 0.65), and SR (humid: AI > 0.65) (Weber et al., 2022). Following this reclassification, significant differences in both bacterial and fungal communities were observed across the aridity groups (Fig. 4). Distinct alpha-diversity patterns confirmed the substantial influence of aridity on the wild fruit forest ecosystem. Further analysis revealed that bacterial diversity was highest in the GS group, with BGH and SR groups showing lower diversity (Fig. 4a, 4b). Specifically, bacterial species richness in the GS group was significantly greater than in both BGH and SR groups (P < 0.05 and P < 0.0001, respectively), with a significant difference also existing between BGH and SR (P < 0.05). The bacterial Shannon index similarly showed significantly higher diversity in GS compared to both BGH and SR (P < 0.0001), while no significant difference (ns) was detected between BGH and SR.

Fig. 4.

The alpha diversity and beta diversity of bacteria and fungi in wild fruit forest under different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid. a, bacteria species richness; b, bacteria shannon index; c, fungi species richness; d, fungi shannon index; e, PCoA of bacteria based on Bray-Curtis distance; f, PCoA of fungi).

In contrast, fungal distribution patterns differed notably from those of bacteria (Fig. 4c, 4d). Fungal diversity was lowest in BGH, higher in GS, and highest in SR. For fungal richness, the GS group was significantly higher than BGH (P < 0.0001). Although SR group did not differ significantly from GS group, it showed significantly higher richness than BGH group, indicating a fungal preference for more humid environments where increased moisture availability generally supports greater fungal diversity. In summary, aridity significantly influenced soil microbial alpha-diversity in the wild fruit forest. The GS group (dry sub-humid) exhibited optimal conditions for both bacterial and fungal diversity, reflecting differential response patterns of bacteria and fungi along the aridity gradient.

Besides alpha-diversity, we further assessed the influence of aridity on the microbiota of the wild fruit forest ecosystem through beta-diversity analysis. Principal coordinate analysis (PCoA) based on Bray-Curtis distances, combined with PERMANOVA, revealed significant differences in community composition between the three groups (BGH, GS, SR) defined by the aridity index (AI) for both bacteria (Fig. 4e; Adonis R² = 0.0509, P = 0.001) and fungi (Fig. 4f; Adonis R² = 0.0313, P = 0.001). For bacterial communities, PCoA1 and PCoA2 explained 10.22% and 7.41% of the variance, respectively, with samples from the BGH, GS, and SR groups showing clear separation and clustering trends. In fungal communities, PCoA1 and PCoA2 accounted for 6.22% and 4.08% of the variance, respectively, while the communities still exhibited significant clustering patterns corresponding to the aridity gradient. Furthermore, this result demonstrated that the microbial community has no apparent difference according to degradation level (Fig. 4), reinforcing that degradation is not a primary driver of topsoil microbial community structure in wild fruit forest. These results collectively substantiate that aridity exerts a fundamental influence on the wild fruit forest ecosystem.

3.4. Composition and distribution characteristics of topsoil microorganisms

To further elucidate the total number of microbial taxa, their distribution patterns, and the specific taxa significantly influenced by aridity in the Xinjiang wild fruit forest ecosystem, we quantified the bacterial and fungal communities. The results revealed that the bacterial community encompassed 29 phyla, 74 classes, 205 orders, 401 families, and 751 genera, while the fungal community consisted of 19 phyla, 70 classes, 186 orders, 468 families, and 1466 genera.

Bar charts and bubble plots of community distribution further demonstrated that Acidobacteriota, Proteobacteria, and Actinobacteriota were the dominant bacterial phyla in the Xinjiang wild fruit forest ecosystem (Fig. 5a), whereas Ascomycota, Basidiomycota, and Mortierellomycota represented the major fungal phyla (Fig. 5b). However, along the aridity gradient (BGH: semi-arid region; GS: dry sub-humid; SR: humid region), the relative abundances of these dominant taxa exhibited distinct gradient-dependent variations. For instance, the relative abundance of Ascomycota—acting as the absolutely dominant phylum—exhibited significant differences in relative abundance across aridity groups, while other major phyla such as Basidiomycota displayed regular shifts in their relative proportions as aridity increased from BGH to SR. Further analysis revealed that the majority of bacterial and fungal taxa were distributed in dry sub-humid regions, with relatively fewer species present in semi-arid and humid regions (Figures S2a, S2b). Collectively, these findings indicate that aridity intensity serves as a key environmental driver shaping the divergence of soil bacterial and fungal community structures at the phylum level in the wild fruit forest ecosystem. Additionally, the relative abundances of the dominant taxa display distinct response patterns to the aridity gradient.

Fig. 5.

The relative abundance of bacterial (a) and fungal (b) communities at the phylum level under different aridity gradients (BGH, semi-arid; GS, dry sub-humid aridity; SR, humid).

3.5. The microbiome is extremely stable in Tianshan wild fruit forest ecosystem

To assess the stability of microbial communities in the Xinjiang wild fruit forest ecosystem, the shared and unique microbial taxa across different habitats was conducted. The results revealed that aridity significantly regulates the species distribution patterns of both bacterial and fungal communities in the wild fruit forest. Notably, the majority of bacterial taxa were shared across habitats with varying aridity levels, and these taxa were distributed almost universally throughout the Xinjiang wild fruit forest ecosystem. The number of shared ASVs reached 18,510 (Fig. 6a). Further taxonomic analysis identified that these shared bacterial taxa were predominantly affiliated to Vicinamibacteria, Gammaproteobacteria, Gemmatimonadetes, and Thermoleophilia (Fig. 6b). This finding demonstrates that these taxa possess strong ecological adaptability, being capable of thriving in semi-arid, humid, and alternating wet-dry environments, thereby functioning as the core bacterial founders of the Xinjiang wild fruit forest habitat.

Fig. 6.

The common/unique groups of bacterial (a) and fungal (c) communities analyzed by Upset and the relative abundances of the corresponding groups (b, d) (BGH, semi-arid; GS, dry sub-humid; SR, humid).

In addition, we analyzed the representative taxa specific to each of the three habitats. As illustrated in Fig S3a and S3b, the representative bacterial taxa in the BGH (semi-arid) habitat were Actinobacteria and Actinobacteriota, while Gammaproteobacteria and Methylomirabilota dominated the GS (dry sub-humid) habitat, and Vicinamibacteria were the key representatives in the SR (humid) habitat. In contrast, fungal communities exhibited substantial divergence across the three habitats, with most taxa being habitat-specific. For instance, the BGH habitat harbored the highest number of unique fungal ASVs (approximately 20,841; Fig. 6c), primarily belonging to the Saccharomycetes (Fig S3a). The SR and GS habitats followed, with 19,080 and 13,864 unique ASVs, respectively (Fig. 6c). The representative unique fungal taxa in the SR habitat were affiliated with Ascomycota, whereas those in the GS habitat included Sordariomycetes, Mortierella, and Mortierellaceae (Fig S3a and S3b). Nevertheless, a subset of fungal taxa with strong habitat adaptability was found to persist across all three habitats, mainly consisting of Sordariomycetes and Dothideomycetes (Fig S3b).

In summary, aridity largely shapes the distribution patterns and habitat adaptation capabilities of microorganisms in the Xinjiang wild fruit forest ecosystem, and significantly drives the compositional patterns of shared and unique taxa. Both bacteria and fungi possess highly stable core microbiomes, with bacterial core microbiomes showing greater stability compared to their fungal counterparts.

3.6. Stochastic processes are the dominant assembly mechanism of microbial communities

This study quantified the assembly processes of topsoil microbial communities by using βNTI. The results demonstrate that aridity intensity (AI) significantly influences the assembly mechanisms of both soil bacterial and fungal communities, although their response patterns diverged markedly. As shown in Fig. 7a and 7b, stochastic processes (|βNTI| < 2) overwhelmingly dominated the assembly of both bacterial and fungal communities. Nevertheless, significant differences (P < 0.001) were detected in these processes across the three aridity groups: semi-arid (BGH), dry sub-humid (GS), and humid (SR). Quantification of the assembly processes revealed distinct trajectories for each domain. For bacterial communities (Fig. 7c), assembly in the BGH group was governed by a combination of stochastic processes and deterministic processes (primarily homogeneous and heterogeneous selection). In contrast, the contribution of stochastic processes dominated by drift increased substantially and became paramount in both the GS and SR groups. Regarding fungal communities (Fig. 7d), deterministic processes (mainly heterogeneous selection) still constituted a considerable proportion in the BGH group. As aridity decreased from GS to SR, stochastic processes centred on ecological drift strengthened progressively, eventually becoming absolutely dominant, while the relative contribution of deterministic processes declined significantly.

Fig. 7.

The βNTI values of bacterial (a) and fungal (b) communities in wild fruit forests under different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid), and quantification results of community assembly processes (c.bacteria; d, fungi).

3.7. The dry sub-humid condition promotes the interaction among microbial network

The co-occurrence network structures of topsoil bacterial and fungal communities in the wild fruit forest exhibited significant differences and taxon-specific responses across the aridity gradient (BGH: semi-arid region; GS: dry sub-humid region; SR: humid region) (Fig. S4). Bacterial networks displayed a progressive structural shift across the gradient. The semi-arid (BGH) group exhibited compact modularity, suggesting cooperative adaptation to environmental stress. The dry sub-humid (GS) group showed the highest network complexity, with significant increases in node and edge numbers and key topological metrics (e.g., average degree, centralities; P < 0.05) (Fig. S5a), indicating intensified interactions and enhanced ecological influence of core taxa. In contrast, the humid (SR) group displayed a more simplified and sparsely connected network, implying reduced selective pressure for strong species interactions (Fig. S4a). Fungal networks, however, demonstrated a distinct response pattern. In the BGH group, networks were dispersed with low modularity, potentially reflecting limited interactions under moisture stress. Complexity peaked in the GS group, which formed a large, well-connected core module and showed significantly higher edge numbers and centrality measures (P < 0.0001) (Fig. S5b), pointing to fostered synergistic interactions. Notably, in the SR group, the fungal network simplified again, suggesting a strategic optimization for functional stability under consistently high moisture (Fig. S4b).

In summary, the dry sub-humid conditions (GS) consistently promoted the most complex and interconnected networks for both bacteria and fungi. However, their structural trajectories differed: bacterial networks transitioned from tight modularity to peak complexity and then to simplification, whereas fungal networks shifted from dispersion to integrated modularity before also simplifying. These findings underscore that aridity gradient critically shapes microbial interaction architectures, with the GS environment most strongly enhancing network complexity, interspecific relationships, and the functional role of core taxa.

3.8. Nitrogen cycling dominated soil microbial functional in the Tianshan wild fruit forest

Based on the FAPROTAX functional annotation results, the soil microbial functions in the Tianshan wild fruit forest cover multiple core ecological processes, including nitrogen cycling, carbon cycling, transformation of elements such as sulfur/arsenic/manganese/iron, fermentation, and pollutant degradation (Fig. 8). Among these, nitrogen cycling-related functions were the most microbial taxa and the highest abundance, encompassing key processes (Fig. 8b) such as denitrification (primarily mediated by Rhodoplanes and Alcaligenes), nitrogen fixation (dominated by rhizobial taxa such as Bradyrhizobium and Allorhizob), nitrate reduction (involving 14 taxa, with Rubrobacter and Rhodoplanes as core taxa) (Fig. 8a). Additional nitrogen-related processes, including nitrite ammonification, nitrate respiration, and nitrogen respiration, were detected with low abundance, primarily mediated by Escherichia and Ochrobactrum, forming a complete "nitrogen input-transformation-loss" functional network. Carbon cycling functions were centered on organic matter decomposition (e.g., chitinolysis by Lysobacter and cellulolysis by Blastocatella, and the phototrophy/photoheterotrophy driven by Rhodoplanes and Cyanobacterial (Fig. 8a).

Fig. 8.

Biogeochemical cycles and the corresponding microbial community distribution based on FAPROTAX functional annotation. a, taxa distribution in different biogeochemical cycle processes; b, biogeochemical cycle under different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid).

Other element transformation (S/As/Mn/Fe), pollutant degradation, fermentation, and ureolysis were also identified in the study. For sulfur cycling, sulfur compound oxidation (mediated by Alcaligenes) and dark oxidation of sulfur compounds (mediated by Bosea and Alcaligenes) indicated microbial involvement in sulfur transformation and detoxification of reduced sulfur species. Regarding arsenic and heavy metal dynamics, Alcaligenes drove arsenite oxidation detoxification and dissimilatory arsenite oxidation, while manganese oxidation was mediated by Leptothrix and Acinetobacter, and iron respiration/oxidation by Aciditerrimonas and Leptothrix—collectively reflecting microbial adaptation to the geochemical characteristics of Tianshan mountain soil. Pollutant degradation potential was evident through aromatic compound degradation (dominated by Nocardioides), hydrocarbon degradation (mediated by Methylosinus and Blastochloris), and plastic degradation (mediated by Alcaligenes and Acinetobacter), highlighting the microbial community’s environmental remediation capacity. Fermentation was distributed across 7 taxa (e.g., Opitutus, Skermanella), contributing to anaerobic carbon metabolism in soil microsites. Additionally, ureolysis (urea hydrolysis) was dominated by Massilia and Rhizobacter, which regulate soil ammonium availability via urea decomposition (Fig. 8a).

The heatmap further compared the biogeochemical cycling processes mediated by microorganisms in the wild fruit forest in across the aridity gradient (BGH: semi-arid region; GS: dry sub-humid region; SR: humid region). As shown in the Fig. 8b, the nitrate reduction process was exceptionally active in the humid environment (SR) and semi-arid habitat (BGH), while it was the weakest in dry sub-humid (GS) condition. Interestingly, however, in the semi-arid habitat (BGH), the entire biogeochemical cycling—especially the nitrogen fixation process within the nitrogen cycle—was clearly the dominant ecological process in the Tianshan wild fruit forest. Similar to nitrogen fixation, the ureolysis activity was relatively high in semi-arid habitat (BGH) but low in the humid environment and dry sub-humid (GS) condition. In addition, phototrophic microbes (particularly photoheterotrophic taxa), nitrogen respiration, nitrate respiration, and denitrification were the predominant groups in the humid environment (SR).

4. Discussion

4.1. The overriding influence of aridity gradient on microbial community structure

The aridity index (AI) is the paramount factor shaping both bacterial and fungal communities, which fundamentally recalibrates our understanding of the forces structuring soil microbiomes in the Tianshan wild fruit forest. The initial analysis revealed a striking absence of significant differences in microbial alpha diversity across five distinct levels of forest degradation (from healthy to dead trees), prompting a re-evaluation of the driving environmental factors. The distance-based redundancy analysis (dbRDA) and Random Forest models unequivocally identified AI as the dominant contributor, explaining approximately 43% of the variation in bacterial communities and 30% in fungal communities, far surpassing the influence of degradation-related metrics or soil nutrients like pH, EC, and TN. This finding aligns with a growing body of evidence from dryland ecology suggesting that broad-scale climatic filters can override the effects of localized disturbance. Our research corroborates studies in other arid mountain systems where aridity has been shown to be a stronger predictor of microbial community composition than land-use change or degradation (Maestre et al., 2015; Dong et al., 2024; Zhang et al., 2024). Aridity exerts its influence primarily by modulating soil water availability, which is the ultimate limiting factor for microbial metabolic activity, diffusion of substrates, and survival in arid and semi-arid ecosystems (Schimel et al., 2007; Nielsen and Ball, 2014). The distinct separation of microbial communities along the aridity gradient (BGH: semi-arid region; GS: dry sub-humid; SR: humid region) in the PCoA plots (Fig. 4e, 4f) visually underscores this strong environmental filtering. The systematic shifts in the relative abundance of dominant phyla, such as the metabolic versatility of Proteobacteria, along the aridity gradient likely reflect adaptations to changing water potential and oxygen availability (Fierer and Jackson, 2006). Consequently, while degradation-related factors (e.g., plant mortality, nutrient loss) undoubtedly impact microbial communities at a local scale, their effects appear to be secondary and are often mediated by the overarching changes in soil microclimate and resource availability induced by the aridity gradient (Gao et al., 2021; Witzgall et al., 2024).

In addition, bacteria and fungi exhibit different response strategies when adapting to environments with different aridity gradients, mainly due to fungi possess a rigid chitinous cell wall and a filamentous hyphal structure, which form dense mycelial networks to resist water loss and expand the range of water and nutrient acquisition in dry habitats. In contrast, bacterial cells lack such structural protection, and their survival under arid conditions relies more on biofilm formation or spore dormancy. Besides, fungi dominate the decomposition of recalcitrant organic matter (such as cellulose and lignin), and their functional traits are less affected by aridity because the hyphal decomposition process is independent of immediate water availability (Manzoni et al., 2012). On the other hand, bacteria are the main drivers of labile organic matter decomposition and nutrient cycling (including nitrogen fixation and nitrate reduction), and their functional activity is closely coupled with soil water availability. This tight coupling leads to more significant shifts in bacterial functional groups along aridity gradients. This distinction is critical for guiding conservation strategies, suggesting that managing regional water availability and protecting mountainous hydrological processes may be more effective for maintaining microbial functional stability than interventions focused solely on degradation remediation.

4.2. Exceptional microbiome stability and enhanced network complexity in dry sub-humid conditions

It is notable that microbiome exhibits exceptional stability within the Tianshan wild fruit forest. This stability is characterized by the persistence of a core suite of bacterial (Acidobacteriota, Proteobacteria, Actinobacteriota) and fungal (Ascomycota, Basidiomycota, Mortierellomycota) taxa across both degradation levels and aridity gradients. This indicates a high degree of functional redundancy and ecological resilience, likely honed by long-term adaptive evolution to the region's historical climate fluctuations (Duan et al., 2022). This core microbiome composition is consistent with that found in other temperate forest and arid-region soils, where these phyla are known for their critical roles in ecosystem functioning (Rousk et al., 2010; Jansson and Hofmockel, 2020). Acidobacteriota and Actinobacteriota are renowned for their ability to decompose complex organic matter (McReynolds et al., 2025), while Proteobacteria are key players in nitrogen cycling. Similarly, Ascomycota and Basidiomycota include vital saprotrophic and symbiotic taxa, and Mortierellomycota are important for phosphorus solubilization (Hibbett et al., 2025). Their dominance underscores their indispensable role in maintaining soil fertility and plant health in this ecosystem.

Network complexity is a key indicator of ecosystem stability and functional robustness (Banerjee et al., 2016). The topological characteristics of these networks, including the number of nodes, edges, and average degree, were significantly higher in the GS group compared to the semi-arid (BGH) and humid (SR) groups (Fig. S4). It suggests that periodic wet-dry cycles promote niche differentiation and facilitate positive, synergistic interactions between taxa. For example, drought-tolerant bacteria may produce extracellular polymeric substances that improve soil water retention, benefiting less tolerant fungi (Costa et al., 2018), while fungi may decompose complex organic matter into labile carbon sources that support bacterial growth (de Boer et al., 2005). This divergence may stem from the inherent biological traits of bacteria and fungi. Bacteria, typically possessing smaller cell sizes and higher surface area-to-volume ratios, may exhibit faster resuscitation and growth capabilities under rapidly fluctuating dry-wet conditions. In contrast, fungi, particularly through their mycelial networks, hold an advantage in accessing spatially dispersed water and nutrients; however, prolonged drought can disrupt hyphal continuity, potentially leading to the lowest diversity under extreme aridity (BGH). Furthermore, as primary decomposers of recalcitrant organic matter such as lignin, fungi likely occupy a more significant ecological niche in humid habitats (SR) where litter accumulation may be greater.

Alternating dry-wet environments create a soil microhabitat mosaic (alternating oxic/anoxic patches, water-filled/air-filled pores), which breaks the competitive exclusion of single dominant taxa in stable moisture habitats (semi-arid/humid) and promotes the coexistence of microbial taxa with divergent moisture adaptabilities. These mutualistic interactions increase network connectivity and modularity, reducing the vulnerability of the network to species loss (Montoya et al., 2006). In contrast, the semi-arid (BGH) environment likely imposes strong water stress that selects for a limited set of drought-tolerant taxa, reducing opportunities for diverse interactions. The humid (SR) environment, while resource-rich, may intensify competitive exclusion, simplifying the network structure. Our findings align with emerging concepts that moderate environmental fluctuations can enhance microbial network complexity by promoting functional complementarity, as observed in agricultural and grassland systems (Jiao et al., 2022; Yang et al., 2024; Yakubu et al., 2025). This positions the dry sub-humid conditions as a critical "buffer" that maintains the architectural integrity and functional resilience of the soil microbiome.

4.3. Stochastic processes dominate microbial community assembly

Stochastic processes overwhelmingly govern the assembly of both bacterial and fungal communities in the Tianshan wild fruit forest. This finding presents a significant contrast to many temperate forest ecosystems, where deterministic processes like environmental filtering often dominate (Zhao et al., 2019). The prevalence of stochasticity, primarily ecological drift, can be attributed to the specific environmental context of this mountainous arid ecosystem. First, the high spatial heterogeneity of soil properties (Fig. S1) and moisture availability creates a mosaic of micro-niches. This heterogeneity reduces the strength of consistent environmental filtering across the landscape, allowing historical contingencies and dispersal limitations to play a larger role (Zhou and Ning, 2017). Second, the fragmented nature of the wild fruit forest itself may enhance dispersal limitation, preventing microbial taxa from reliably reaching all suitable habitats, thereby increasing the influence of drift on local community composition (Hanson et al., 2012). It is noteworthy that the relative influence of stochastic and deterministic processes varied with aridity. Deterministic processes (particularly heterogeneous selection) maintained a considerable proportion in the semi-arid (BGH) group for fungi, likely due to the strong environmental stress of water limitation imposing a strict filter. As this stress relaxed along the gradient to humid (SR) conditions, stochastic processes became absolutely dominant (Fig. 7c, 7d). This trajectory underscores that while stochasticity is the overarching rule, the aridity gradient modulates the specific balance of assembly mechanisms.

4.4. Nitrogen cycling dominance and microbial functional adaptations

The functional annotation results underscore nitrogen cycling as the dominant microbial-driven process in the Tianshan wild fruit forest, a finding consistent with the critical role of nitrogen in maintaining soil fertility within relict mountain ecosystems. The detection of a complete "nitrogen input-transformation-loss" network, mediated by key taxa such as Rhodoplanes (denitrification, nitrate reduction) and rhizobial groups like Bradyrhizobium (nitrogen fixation), highlights a highly adapted microbial consortium. This functional dominance aligns with studies indicating that nitrogen-cycling microorganisms are primary regulators of soil multifunctionality in arid and semi-arid regions (Zhang et al., 2024). Notably, the predominance of nitrogen fixation in the semi-arid (BGH) habitat reflects a key microbial adaptation to water-limited conditions, where biological nitrogen input compensates for limited atmospheric deposition, thereby supporting plant productivity under nutrient stress (Xu et al., 2025).

The spatial variation in nitrogen-cycling processes along the aridity gradient reveals nuanced microbial adaptations to moisture regimes. The heightened nitrate reduction activity in both semi-arid (BGH) and humid (SR) habitats, but its relative weakness in the dry sub-humid (GS) conditions, suggests that frequent moisture fluctuations may disrupt the stability of nitrate-reducing microbial populations. This observation resonates with findings that dry-wet cycles can decouple carbon and nitrogen mineralization dynamics, potentially limiting the niche optimization for specialist taxa (Yang et al., 2024). Conversely, the enrichment of denitrification and nitrogen respiration processes in the humid (SR) habitat points to enhanced nitrogen loss pathways under consistently high moisture, likely due to increased anaerobic microsites favoring facultative anaerobes. The parallel increase in photoheterotrophic microbial groups in SR further suggests tighter coupling between phototroph-mediated carbon cycling and nitrogen transformations in more humid soils.

4.5. Limitations, future perspectives, and practical implications

While this study provides compelling evidence for the dominance of aridity in shaping the Tianshan wild fruit forest microbiome, several limitations should be considered when interpreting the findings. Firstly, our analysis was confined to the topsoil layer (0–10 cm). Although this horizon is highly dynamic and critical for microbial activity, microbial communities and their drivers can differ significantly in deeper soil layers and also rhizosphere, where moisture and organic matter dynamics are distinct. Future studies employing stratified sampling designs are needed to capture the vertical dimension of microbial responses to aridity and degradation. Furthermore, soil samples in this study were collected from the 0–10 cm surface layer around fruit trees without specifically distinguishing between rhizosphere and bulk soil. The plant rhizosphere exerts a strong shaping force on microbial communities. Future studies that incorporate rhizosphere sampling would enable a more precise dissection of the direct effects of plant health status (degradation) on microbes via rhizosphere processes, thereby allowing for a finer-resolution analysis of the relative contributions of aridity and degradation factors.

Secondly, the reliance on amplicon sequencing (16S and ITS) provides robust data on taxonomic composition but offers limited direct insight into functional genes and metabolic pathways. The inferred functional capabilities based on taxonomy, and the ecological implications of the observed network structures, require validation through metagenomic, metatranscriptomic, and metabolomic approaches to elucidate the mechanistic links between community structure, microbial processes, and ecosystem functioning. Thirdly, our study captured a spatial snapshot of the ecosystem. The dynamic nature of climate change and degradation processes necessitates long-term temporal monitoring to understand the trajectory of microbial community shifts, their resilience to extreme aridity events, and the potential for tipping points in ecosystem stability. Finally, while we measured key soil physicochemical properties, other unmeasured factors, such as specific plant root exudates, soil microfauna, and viral interactions, could contribute to the observed patterns. Integrating these components will provide a more holistic understanding of the complex web of interactions governing this relict ecosystem.

Our study confirms that aridity gradient overrides degradation in shaping the topsoil microbiome of the Tianshan wild fruit forest, which provides a pivotal evidence-based direction for the conservation and ecological restoration of this relict biodiversity hotspot under global climate change. For local forestry management and protection practices, targeted and feasible strategies should be formulated with water availability as the core regulation factor and dry sub-humid habitats as the key protection object. Firstly, prioritize the establishment of protected zones in dry sub-humid habitats with high microbial network complexity and stability, prohibit excessive human disturbance such as random grazing and cutting, and preserve the natural moisture fluctuation pattern of the soil to sustain the synergistic interactions of soil microbes. Secondly, carry out long-term dynamic monitoring of soil moisture and microbial communities in the wild fruit forest along the aridity gradient, build a micro-ecological early warning system, and adjust restoration measures in a timely manner according to the monitoring results. Thirdly, combine local afforestation and vegetation restoration with water resource management, select drought-tolerant native tree species and associated microbial inoculants to improve the adaptability of the forest ecosystem to aridity stress. These measures will effectively safeguard the microbial functional resilience of the Tianshan wild fruit forest, and further maintain the structural and functional stability of this unique relict ecosystem in the context of increasing climatic aridity.

5. Conclusion

This study demonstrates that climatic aridity gradient, rather than vegetation degradation, dominates the structure, assembly, and interaction patterns of topsoil microbial communities in the degraded ecosystem of the Tianshan wild fruit forest. Aridity, not degradation level, predominantly explains variations in bacterial and fungal communities, while the relict microbiome maintains high stability via resilient core taxa. Community assembly is primarily shaped by stochastic processes, and moderate dry sub-humid fluctuations promote the most complex and stable microbial networks. Community assembly is primarily shaped by stochastic processes, and moderate dry sub-humid fluctuations promote the most complex and stable microbial networks.These findings revise the conventional degradation-centric perspective and establish aridity as a key regulator in arid-region ecosystems. For conservation, water resources management and dry sub-humid habitats protection are essential to sustain microbial functional stability, providing a scientific basis for safeguarding this precious gene pool and improving predictions of ecosystem responses to global change.

Authors and contributors

Ye Tao, Chen-Quan Gu, and Cheng-Cheng Wang obtained the soil samples. Ye Tao, Cheng-Cheng Wang, and Chen-Quan Gu carried out the experiments. Jian-Rong Huang, Yong-Hong Liu, and Cheng-Cheng Wang did the data analysis and diagram figures. Yong-Hong Liu and Hong-Chen Jiang, conceived the paper and prepared the manuscript. All authors contributed to the article and approved the submitted version. Tao Ye designed the project outline and provided financial support.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this manuscript. No authors have any conflicts of interest to disclose, including but not limited to employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties related to the subject matter of this study. All authors confirm that the above statement is accurate and complete, and they take full responsibility for the authenticity of the information provided.

Appendix. Supplementary materials

Fig. S1 Comparison of soil physicochemical factor in wild fruit forests under different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid) (different letters indicate significant differences at P < 0.05)

Fig S2 The distribution of microbial community under different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid)

Fig. S3 The analysis of LEfSe in the bacterial (a) and fungal (c) communities of in wild fruit forests under different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid) and relative abundances of the top 20 significantly different groups (b, d)

Fig. S4 The co-occurrence network of bacterial (a) and fungal (b) communities in wild fruit forests under different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid).

Fig S5. Comparison of topological characteristics (number of nodes, number of edges, degree, and centrality) of bacterial (a) and fungal (b) co-occurrence networks in the wild fruit forest across different aridity gradients (BGH, semi-arid; GS, dry sub-humid; SR, humid).