Changnienia amoena typically flowers in April, which is a key phenological stage for plant-microbe interactions (Chaparro et al., 2014). Therefore, in April 2023, a total of 32 individuals of C. amoena were collected from 6 major native habitats across China: 6 individuals of C. amoena from Kang County, Longnan City, Gansu Province (LongNan, 105°43′E; 33°18’N); 6 from Jigongshan National Nature Reserve, Xinyang City, Henan Province (XinYang, 114°2′E; 31°47’N); 6 from Xuan’en County, Enshi Tujia and Miao Autonomous Prefecture, Hubei Province (EnShi, 109°49′E; 29°57’N); 5 from Shennongjia Forestry District, Hubei Province (ShenNongJia, 110°22′E; 31°26’N); 3 from Gaowangjie National Nature Reserve, Guzhang County, Xiangxi Tujia and Miao Autonomous Prefecture, Hunan Province (XiangXi, 110°4′E; 28°37’N); and 6 from Tongjiang County, Bazhong City, Sichuan Province (BaZhong, 107°11′E; 31°56’N). During the field sampling, no visible disease symptom was observed on C. amoena individuals, and all sampled plants were healthy. At the same time, bulk soil samples were collected from each habitat, including 6 from LongNan, 6 from XinYang, 6 from EnShi, 5 from ShenNongJia, 3 from XiangXi and 6 from BaZhong. Sampling sites within each habitat were spaced at least five meters apart. Soil samples were air-dried, ground, and passed through a 2-mm sieve prior to analysis.

Rhizosphere soil was sampled according to Attia et al. (2022) with some modifications. In detail, loose soil attached to the roots were gently brushed off using a clean brush, leaving an approximately 1 mm thick layer of soil tightly attached to roots. The roots were then immersed in 25 mL of Phosphate Buffered Saline (PBS) buffer in a sterile tube and then vortexed vigorously for 15 s to obtain microbes in the thin layer of soil, representing the rhizosphere fraction. The same roots from previous step were transferred in a new tube containing 25 mL of PBS buffer and subjected to an ultrasonic water bath to further detached microbes on the root surface, representing the rhizoplane fraction. Sonication was performed for 30 s at 70 Hz followed by a pause for 30 s, repeated for 10 cycles. The entire process was carried out at 4 °C to prevent overheating. The suspensions obtained from both steps were intentionally combined and centrifuged at 12,000 g for 10 min at 4 °C. The resulting pellet was collected as the rhizosphere soil sample, while the remaining roots were collected as the root sample. Both root endosphere and rhizosphere samples were flash-frozen in liquid nitrogen and stored at −80 °C until DNA extraction.

Soil total nitrogen (TN) was determined using the Kjeldahl digestion method, total phosphorus (TP) was measured by the molybdenum antimony colorimetric method after fusion with NaOH, while total potassium (TK) was quantified by flame photometry after fusion with NaOH (Jiang et al., 2024). Soil available nitrogen (AN) was determined using the alkaline diffusion method, available phosphorus (AP) was measured by the molybdenum-antimony colorimetric assay after extraction with NaHCO₃, available potassium (AK) was quantified by flame photometry after extraction with NH₄OAc, while soil organic carbon (SOC) was measured using the K₂Cr₂O₇ oxidation–reduction titration method (Wang et al., 2024). Differences in the TN, AN, TP, AP, TK, AK, and SOC of soil collected from different locations were statistically evaluated using One-Way Analysis of Variance (ANOVA).

The total DNA was extracted from root and rhizosphere samples using the E. Z. N. A.^® Plant DNA Mini Kit (OMEGA Bio-tek Inc., Norcross, GA, United States) and the FastDNA™ SPIN Kit for Soil (MP Biomedicals, Solon, OH, United States) according to the manufacturer’s instructions, respectively. The quality of DNA was controlled by agarose gel electrophoresis, and the concentration of DNA was quantified using a NanoDrop™ 1,000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, United States). DNA was stored at −80 °C until subsequent sequencing.

The concentration of DNA in all samples was normalized to 10 ng/μL. The primers 799F (5’-AACMGGATTAGATACCCKG-3′) and 1223R (5′-CCATTGTAGTACGTGTGTA-3′) were used to amplify the V5-V7 hypervariable region of the bacterial 16S rRNA gene (Biget et al., 2024). The primers ITS86F (5’-GTGAATCATCGAATCTTTGAA-3′) and ITS4 (5’-TCCTCCGCTTATTGATATGC-3′) were used to amplify the fungal internal transcribed spacer (ITS) 2 region (Jacquemyn et al., 2014). Each Polymerase Chain Reaction (PCR) was performed in 10 μL mixture containing 0.2 μL TransStart^® TopTaq DNA Polymerase (TransGen Biotech Co., Ltd., Beijing, China), 0.2 μM of each primer (10 μM), and 1 μL template DNA. The thermal cycling program consisted of an initial denaturation at 94 °C for 2 min, followed by 27 cycles of denaturation at 94 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 1 min, with a final extension at 72 °C for 10 min. PCR products showing a distinct band of approximately 400 bp for bacteria and 450 bp for fungi were visualized by agarose gel electrophoresis, excised from the gel, and purified using AxyPrep™ DNA Gel Extraction Kit (Corning Inc., Glendale, AZ, United States). Products from three independent PCR replicates per sample were pooled in equal volumes to minimize amplification bias. Amplicons of the expected size were sequenced on the Illumina NovaSeq 6,000 platform (Illumina, Inc., San Diego, CA, United States) to generate paired-end reads of approximately 250 bp. A standard bacterial/fungal genomic DNA mix was used as the positive control, and sterile water was as the negative control.

Raw sequencing reads were processed using QIIME 2 (Bolyen et al., 2019). Adapter and primer sequences were trimmed using the cutadapt plugin. Quality control and amplicon sequence variant (ASV) identification were performed using the DADA2 plugin (Callahan et al., 2016). Rarefaction curves of both bacterial and fungal communities approached saturation with increasing sequencing depth (Supplementary Figure S1), indicating that the sequencing effort was sufficient to capture the majority of microbial diversity. Taxonomic assignments of bacterial ASV representative sequences were conducted using a pre-trained Naive Bayes classifier based on the SILVA database (version 138.1) with a confidence threshold of 0.8, while taxonomic assignments of fungal ASV representative sequences were conducted using a pre-trained Naive Bayes classifier based on the UNITE database (version 9.0) with a confidence threshold of 0.6. The raw 16S rRNA and ITS amplicon sequencing data have been deposited in the NCBI Sequence Read Archive (SRA) under accession numbers PRJNA1322054 and PRJNA1321960, respectively.

Downstream data analyses were conducted in R (version 4.5.0). To minimize biases arising from unequal sequencing depths, bacterial and fungal ASV tables were rarefied to the lowest sequencing depth among all samples, which was 59,289 reads for bacteria (Supplementary Table S1) and 51,102 reads for fungi (Supplementary Table S2). Alpha diversity indices were calculated using the vegan package (Oksanen et al., 2007). Differences in Shannon-Wiener diversity index across sampling locations and root-associated compartments were statistically evaluated using Two-Way ANOVA, respectively. Bray–Curtis dissimilarity matrices were computed using the vegdist function, and principal coordinates analysis (PCoA) was performed using the cmdscale function, both from the vegan package. Permutational multivariate analysis of variance (PERMANOVA) was employed to assess the individual effect of location and compartment on microbial communities (permutation number = 999). The top 20 most abundant bacterial and fungal genera were identified based on their average relative abundance across all samples. Stacked bar plots were created using the tax_stackplot function in the ggplot2 package to visualize taxonomic composition (Liu et al., 2021). Core Amplicon Sequence Variants (ASVs) were identified following the framework proposed by Qiao et al. (2024). Briefly, Venn diagrams were drawn to determined bacterial and fungal ASVs across locations with each compartment. Ecological niche breadth was then calculated for individual ASVs within the same location and compartment, and the top 10% bacterial or fungal ASVs were classified as generalists. ASVs that were both shared across locations and classified as generalists were defined as core microbiome members.

To explore the influences of soil nutrients on bacterial and fungal communities, several statistical analyses were performed (Zhou et al., 2024). Firstly, the ASV count was subjected to Hellinger transformation to reduce the impact of rare taxa, while the soil variables, including TN, AN, TP, AP, TK, AK, and SOC, were normalized by log1p-transformation. Canonical correspondence analysis (CCA) was performed to test the relationship between bacterial/fungal ASV matrix and soil nutrients. Generalized additive model fitting was incorporated in the PCoA ordination (Bray–Curtis distance) to assess the relationship between the root endosphere/rhizosphere bacterial/fungal community composition and each individual soil nutrient using the ordisurf function in the vegan package. The species richness of bacterial and fungal communities in each location was calculated using the specnumber function in the vegan package. Next, the linear models were fitted using the lm function in the stats package to show the relationship between bacterial / fungal richness and each individual soil nutrient.

Microbial community assembly processes were evaluated using a neutral model, implemented via the neutral.fit function (Sloan et al., 2006). Partial least squares path modeling (PLS-PM) was conducted using the plspm package to disentangle the relationships among geographic location, soil nutrient, and root endosphere/rhizosphere bacterial/fungal community (Tenenhaus et al., 2005). The first principal coordinate axis (PCoA1), derived from community dissimilarity matrices, was used to represent bacterial and fungal community structure in each compartment. Path coefficients (β) represent standardized direct effects among latent variables, while R² values indicate the proportion of variance explained in endogenous variables. The functions of bacterial community were predicted based on Functional Annotation of Prokaryotic Taxa (FAPROTAX) database (Louca et al., 2016), while the ecological guilds of fungal community were assigned based on FUNGuild database (Nguyen et al., 2016). Co-occurrence networks were constructed by jointly analyzing bacterial and fungal ASVs based on root endosphere and rhizosphere samples collected from the same native habitat. Prior to network construction, only ASVs present in at least 20% of samples were retained. Among these, ASVs with a total relative abundance greater than 0.01% across all samples were selected for Spearman correlation analysis. Only strong and significant correlations (Spearman’s ρ ≥ 0.7, p < 0.0001) were retained as robust associations. To reduce false positives, p values were adjusted for multiple testing using the Benjamini–Hochberg procedure prior to network construction (Li et al., 2023).

Plant sampling was conducted in six native habitats of C. amoena (Figure 1A). For each location, the root and rhizosphere soil samples were collected (Figure 1B). Then, the amplicon sequencing was performed to reveal the bacterial and fungal community composition and diversity. The composition of root-associated bacterial communities varied significantly across geographic locations (R² = 0.3254, p = 0.0001) and root-associated compartments (R² = 0.0576, p = 0.0001), with location having a stronger influence. Bacterial communities from XiangXi were clearly distinct from those at other locations, while those from EnShi, ShenNongJia, and BaZhong were highly similar to each other (Figure 1C). A similar pattern was observed for fungal communities (Figure 1D), with location having a stronger influence (R² = 0.2953, p = 0.0001) compared to compartment (R² = 0.00252, p = 0.0003). Fungal communities from XiangXi were clearly distinct, while those from EnShi, ShenNongJia, BaZhong, and LongNan were highly similar. Further analysis revealed that the rhizosphere of C. amoena harbored more diverse bacterial (Figure 1E) and fungal communities compared to the root endosphere (Figure 1F). The Shannon diversity indices of bacterial communities did not show significant difference among locations in either the root endosphere or rhizosphere. Similarly, the Shannon indices of fungal communities in the root endosphere did not differ significantly across locations; however, that in the rhizosphere from EnShi was higher than those observed at other locations.

The dominant bacterial genera associated with the roots of C. amoena were relatively consistent across locations (Figure 2A). The top 10 most abundant ones included Bradyrhizobium, Pseudomonas, Flavobacterium, Dyella, Mucilaginibacter, Sphingomonas, Rhizobium, Bacillus, Mesorhizobium, and Streptomyces. In detail, the relative abundance of Bradyrhizobium was higher in the endosphere than in the rhizosphere. Pseudomonas was more abundant in the endosphere of C. amoena from ShenNongJia and LongNan, while Flavobacterium was more abundant in both compartments from EnShi. Notably, Bradyrhizobium, Rhizobium, and Mesorhizobium were consistently present in all the samples, which may be associated with mycorrhizal symbiosis in C. amoena. Similarly, the dominant fungal genera associated with the roots of C. amoena were also relatively consistent across locations (Figure 2C). The top 10 most abundant ones included Mortierella, Fusarium, Ilyonectria, Tetracladium, Hygrocybe, Dactylonectria, Trichoderma, Solicoccozyma, Exophiala, and Astraeus. The relative abundances of most genera were higher in the endosphere than in the rhizosphere. Notably, Hygrocybe and Russula were most abundant in both compartments of C. amoena from ShenNongJia, while Astraeus and Amanita were most abundant in those from LongNan. More importantly, Astraeus, Amanita and Russula are frequently reported as mycorrhizal fungi and are likely to play essential roles in the growth and development of C. amoena.

Furthermore, a total of 28 bacterial functional groups and 27 fungal guilds were predicted using FAPROTAX and FUNGuild, respectively. The top 20 most abundant bacterial functional potential were similar in both root endosphere and rhizosphere across habitats (Figure 2B). Notably, several predicted bacterial functional potentials were involved in nitrogen cycling processes, including nitrogen fixation, nitrate reduction, nitrogen respiration, nitrate respiration, nitrite respiration, nitrate denitrification, nitrite denitrification, nitrous oxide denitrification, denitrification, and nitrification. These functional potentials indicated that bacterial communities might play key roles in both nitrogen input and transformation in the root-associated compartments of C. amoena. Among these, nitrogen fixation was the most dominant predicted nitrogen-related function, suggesting a potential enhancement of nitrogen acquisition by host plants, which was consistent with the enrichment of Bradyrhizobium, Rhizobium, and Mesorhizobium (Figure 2A). On the other hand, the ecological guild prediction of fungal communities showed a higher relative abundance of ectomycorrhizal and orchid mycorrhizal trophic modes, particularly in samples from ShenNongJia and LongNan (Figure 2D). This pattern was consistent with the higher relative abundances of ectomycorrhizal genera such as Astraeus, Amanita, and Russula in these locations (Figure 2C). Additionally, saprotrophic fungi were also abundant in the root-associated compartments across habitats, likely contributing to organic matter decomposition and nutrient cycling. These findings suggested that the root-associated fungal communities might facilitate nutrient acquisition and improve the growth and adaptation of host plants.

In this study, bacterial and fungal ASVs that were both shared among habitats and exhibited generalist ecological characteristics in the root endosphere and rhizosphere were defined as core microbiome members. Venn diagrams revealed that 345 bacterial ASVs (Supplementary Figure S2A) and 38 fungal ASVs were consistently shared across all habitats in the root endosphere (Supplementary Figure S2C), while 403 bacterial ASVs (Supplementary Figure S2B) and 62 fungal ASVs were shared across habitats in the rhizosphere (Supplementary Figure S2D). Ecological niche breadth analysis further identified a large proportion of ASVs as generalists, including 4,061 bacterial (Supplementary Figure S2A) and 2,102 fungal ASVs as generalists in the root endosphere (Supplementary Figure S2C), and 4,054 bacterial (Supplementary Figure S2B) and 2,102 fungal ASVs as generalists in the rhizosphere (Supplementary Figure S2D). Based on their occurrence across habitats and generalist ecological characteristics, 345 bacterial (Supplementary Figure S2A) and 37 fungal ASVs in the root endosphere (Supplementary Figure S2C), as well as 403 bacterial (Supplementary Figure S2B) and 62 fungal ASVs in the rhizosphere (Supplementary Figure S2D), were defined as core microbiome members.

Taxonomic classification showed that bacterial core ASVs in root-associated compartments were dominated by several genera commonly associated with plant-microbe interactions. In the root endosphere, bacterial core members were mainly affiliated with Bradyrhizobium, Pseudomonas, Dyella, Rhizobium, and Sphingomonas (Figure 3A), whereas the rhizosphere core bacterial community exhibited a similar taxonomic composition, with Bradyrhizobium, Pseudomonas, Dyella, Flavobacterium, and Pseudolabrys as dominant genera (Figure 3B). Fungal core ASVs displayed more compartment-specific taxonomic patterns. In the root endosphere, Ilyonectria, Exophiala, Keithomyces, Mortierella, and Fusarium were dominant (Figure 3C), whereas Mortierella, Ilyonectria, Saitozyma, Cladosporium, and Solicoccozyma were dominant genera in the rhizosphere (Figure 3D). Overall, both the number and taxonomic composition of core microbial members differed between the root endosphere and rhizosphere, indicating compartment-specific patterns of core microbiome assembly.

Network analysis revealed distinct microbial co-occurrence patterns across six native habitats of C. amoena (Figure 4). The root-associated microbial communities in XiangXi formed a highly connected network with a larger size (nodes = 964), greater connectivity (edges = 3,298), and higher complexity (average degree = 6.842), whereas the network in XinYang exhibited the weakest connectivity, characterized by fewer nodes (nodes = 95) and edges (edges = 69), and lower complexity (average degree = 1.453). However, all the microbial networks exhibited high modularity values (ranging from 0.831 to 0.966), with the highest observed in LongNan, indicating a high degree of community compartmentalization. Obviously, positive correlations dominated all the networks, reflecting prevalent cooperative interactions among root-associated microbes of C. amoena. However, some negative correlations existed in Bazhong and XiangXi, suggesting the potential microbial competition or antagonism. It should be noted that the co-occurrence networks were inferred from Spearman correlations, which reflect statistical associations rather than direct ecological interactions; thus, potential indirect or compositional effects could not be completely excluded.

Bacterial ASVs belonging to Proteobacteria, Bacteroidetes, and Actinobacteria consistently forming large and well-connected hubs with fungal ASVs, which were particularly observed in Bazhong and XiangXi. These fungal ASVs were dominated by Basidiomycota and Ascomycota. In contrast, less abundant fungal phyla such as Mortierellomycota and Chytridiomycota appeared more scattered and less interconnected in the networks. Notably, ASVs affiliated with mycorrhizal fungi frequently occupied central positions in the co-occurrence networks across all habitats except XinYang. These ASVs were assigned to Ceratobasidiaceae, Cortinariaceae, Hydnaceae, Inocybaceae, Leptodontidiaceae, Russulaceae, Sebacinaceae, Serendipitaceae, Thelephoraceae, and Tulasnellaceae. Interestingly, these mycorrhizal fungal ASVs exhibited significant positive relationships with numerous bacterial ASVs, many of which were affiliated with Bradyrhizobium, Rhizobium, Mesorhizobium, and Nitrospira. These findings suggested that interactions between mycorrhizal fungi and bacteria in the root-associated compartments of C. amoena might play essential roles in facilitating nutrient acquisition by host plants.

The assembly processes of root-associated bacterial and fungal communities of C. amoena were evaluated using the neutral community model. Generally, the model provided a good fit for both microbial communities, explaining 82.5% of the variation in bacterial communities (Figure 5A) and 79.8% in fungal communities (Figure 5B). However, stochastic processes exert a stronger influence on the assembly of bacterial communities than on fungal communities. Additionally, the estimated migration rate (Nm) was substantially higher for bacterial community assembly (Nm = 362.658) than for fungal community assembly (Nm = 41.083). All these results indicated that the root-associated fungal communities exhibited more limited dispersal than bacterial communities, implying that the fungal community assembly was more strongly governed by spatial constraints and environmental factors.

To investigate the biogeographic patterns of root-associated microbial communities of C. amoena, we first performed univariate linear regression analysis to assess the correlation between microbial community similarity and geographic distance among habitats. A significant negative correlation was observed for bacterial communities (R² = 0.2039, p < 0.001), indicating that the similarity of bacterial communities significantly decreased with increasing geographic distances (Figure 6A). Although the correlation for fungal communities was slightly lower (R² = 0.1865, p < 0.001), a significant decline in similarity was also detected as geographic distance increased (Figure 6B). Additionally, the correlations between microbial Shannon-Wiener diversity and the latitude of sampling locations were analyzed. The diversity of bacterial communities exhibited weak correlations with latitude (Figure 6C). In contrast, fungal diversity significantly decreased with increasing latitude (Figure 6D). These results indicated that geographic factors influence both bacterial and fungal community similarity, whereas fungal community diversity shows a stronger association with geographic gradients than bacterial diversity.

Since geographic locations had pronounced influences on root-associated microbial communities of C. amoena (Figure 6), we speculated that soil nutrient availability could shape microbial communities across habitats. Therefore, TN, AN, TP, AP, TK, AK, and SOC in soil collected from different native habitats were measured. The trends of TN and AN were consistent across habitats, with the highest levels observed in EnShi and the lowest in XiangXi and BaZhong (Supplementary Figures S3A,B). However, TP and AP exhibited divergent patterns. TP was highest in EnShi and lowest in XiangXi and LongNan, while AP peaked in BaZhong (Supplementary Figures S3C,D). TK and AK showed similar trends, with the highest level observed in ShenNongJia. However, TK was lowest in EnShi and LongNan, while AK was lowest in BaZhong (Supplementary Figures S3E,F). Moreover, SOC was highest in EnShi and lowest in XiangXi (Supplementary Figure S3G). Correlation analysis revealed strong associations among TN, AN, and SOC, with coefficients greater than 0.95 (Supplementary Figure S3H). These strong correlations might be attributed to the role of SOC as an energy source that stimulated microbial activities involved in N cycling, thereby enhancing the decomposition and mineralization of organic nitrogen (Shu et al., 2025).

Next, canonical correspondence analysis (CCA) and PERMANOVA were performed to evaluate the influences of soil nutrients on root endophyte and rhizosphere microbial communities of C. amoena, respectively. The results showed that TN, AN, TP, AP, and AK significantly contributed to the distribution of bacterial communities (Figure 7A), and TN, AN, TP, AP, TK, and AK to fungal communities in both compartments (Figure 7B). PERMANOVA showed that TN and TP significantly explained variation in bacterial and fungal community composition in both the root endosphere and rhizosphere, whereas SOC did not show significant effect in any compartment (Table 1). Additionally, most nutrients consistently explained a larger proportion of variation in rhizosphere microbial communities than in endophytic ones, indicating that rhizosphere communities exhibit stronger overall responses to soil nutrient gradients.

To gain deeper insights into the influences of soil nutrients on the root-associated microbial communities of C. amoena, a generalized additive model was used to characterize the relationships between individual soil nutrient and microbial taxonomic composition in the root endosphere and rhizosphere, respectively. All the tested nutrients, except TK and AK, were significantly associated with bacterial community composition in both root endosphere and rhizosphere (Figure 8A). Thus, the effects of soil nutrients on bacterial communities were consistent in both root-associated compartments. However, the effects of soil nutrients on fungal community composition differed between the root endosphere and rhizosphere. Only AP and TK were significantly associated with fungal communities in the root endosphere, whereas all tested nutrients except TP and AK showed significant effects in the rhizosphere (Figure 8B). Overall, fungal community composition in the rhizosphere was more sensitive to soil nutrient variations than that in the root endosphere, which is consistent with the PERMANOVA results (Table 1). Furthermore, linear models were used to assess the relationships between soil nutrients and microbial richness. None of the measured nutrient exhibited a significant influence on bacterial richness in either the root endosphere or the rhizosphere (Supplementary Figure S4A), whereas fungal richness in the root endosphere was significantly and positively correlated with TN, and fungal richness in the rhizosphere was significantly and positively correlated with TN, AN, and SOC (Supplementary Figure S4B). Together, soil nutrients consistently shaped bacterial community composition in both compartments, while their effects on fungal communities and richness differed between the root endosphere and rhizosphere.

Finally, a partial least squares path model (PLS-PM) was employed to elucidate the multivariate relationships among geographical location, soil nutrients and bacterial and fungal communities in the root endosphere and rhizosphere (Figure 9). Geographical location exhibited a strong positive effect on soil nutrients (β = 0.7224), explaining 52.18% of its variance (R² = 0.5218). Soil nutrients, in turn, showed significant positive effects on both root endosphere bacterial (β = 0.5390) and fungal communities (β = 0.3923). Additionally, root endosphere bacterial and fungal communities were strongly linked to their corresponding rhizosphere communities, with significant positive path coefficients from endophytic to rhizosphere bacteria (β = 0.8122) and from endophytic to rhizosphere fungi (β = 0.7800). However, the direct effects of geographical location on bacterial and fungal communities in both root endosphere and rhizosphere were weak, indicating that geographical influences on microbial communities were largely mediated by soil nutrients rather than acting directly. Notably, the standardized path coefficient from rhizosphere bacterial communities to root rhizosphere fungal communities was negative (β = −0.6864), whereas rhizosphere bacterial communities showed a strong positive effect on root endosphere fungal communities (β = 0.8548). Thus, these pathways suggest that the effect of bacterial communities on fungal communities in the root endosphere of C. amoena is primarily indirect, mediated through rhizosphere bacterial communities. Overall, the PLS-PM indicates that soil nutrients mediate variation in root-associated bacterial and fungal communities across different native habitats.

The root-associated bacterial and fungal communities of C. amoena varied across both locations and compartments, with locations exerting a more pronounced influence (Figures 1C,D). Moreover, similarity of both bacterial and fungal communities significantly decreased with increasing geographic distances (Figures 6A,B), demonstrating clear biogeographic patterns in the root-associated microbiomes of C. amoena. Similar patterns have also been reported in other plants previously. For example, bacterial communities in the root endosphere and rhizosphere of Pinus muricata were shaped more strongly by locations than compartments (Berrios et al., 2023), while root-associated fungal communities across several plant families were found to differ significantly across locations (Hogan et al., 2023). Notably, the PLS-PM model showed that geographical location explained a large proportion of soil nutrient variation (Figure 9). Then the soil nutrients showed significant positive effects on bacterial and fungal communities in the root endosphere, indicating that the observed microbial biogeographic patterns may result from habitat-driven differences in soil nutrient availability (Supplementary Figure S3). Previously, TN, AN, TP, TK, and AK have been reported as major drivers of rhizosphere bacterial and fungal communities of Chromolaena odorata (Zhang et al., 2024). Likewise, TP and AP were shown to explain major variations in root-associated fungal communities (Hogan et al., 2023). In this study, TN and TP had the strongest influences on both bacterial and fungal communities in the root endosphere and rhizosphere of C. amoena (Table 1). Additionally, rhizosphere microbial assemblages are more strongly structured by soil nutrients compared to endophytic communities.

In the root-associated compartments of C. amoena, bacterial diversity showed no significant correlation with latitude (Figure 6C), whereas fungal diversity exhibited significant negative correlations with latitude (Figure 6D). Soil is regarded as a seed bank of plant root-associated microbiota (Bai et al., 2022), and thus changes in soil microbial communities, which can be driven by nutrients, pH, temperature, moisture, or other factors (Frindte et al., 2019), may lead to changes in root-associated microbiomes. Some studies show declining soil bacterial diversity with increasing latitude, while fungal diversity follows a unimodal distribution pattern (Shi et al., 2014; Fu et al., 2023); others show a hump-shaped pattern of bacterial diversity with latitude, with fungal diversity either decreasing or showing weak latitudinal trends (Zhang et al., 2019; Liu et al., 2020). Notably, these patterns of soil microbiomes are somewhat distinct from those observed in root-associated microbiomes observed in this study, highlighting that the distribution of root-associated microbial communities is shaped not only by soil factors but also by plant–soil feedbacks (Kaur and Sharma, 2021). For example, plant root architecture and exudates can alter nutrient conditions in surrounding soils, thereby influencing rhizosphere microbial communities, which can in turn affect host plant performance (Zhao et al., 2021). Generally, the root-associated microbiomes of C. amoena show pronounced biogeographic patterns particularly shaped by soil nutrients with TN and TP as key factors. It should be acknowledged that, in addition to soil nutrients, other environmental factors, such as soil pH, moisture, and light availability, are also known to influence plant-microbe interactions and root-associated microbiome assembly, and thus should be considered in future studies.

Among the top 20 most abundant bacterial genera detected in the root endosphere and rhizosphere of C. amoena, Pseudomonas, Flavobacterium, Sphingomonas, Rhizobium, Bacillus, and Streptomyces are commonly observed as orchid root-associated bacteria and show plant growth-promoting abilities (Kaur and Sharma, 2021). Notably, nitrogen-fixing genera such as Bradyrhizobium, Rhizobium, and Mesorhizobium were enriched, particularly in the roots of C. amoena (Figure 2A), which might be related to the enriched bacterial predicted functions related to nitrogen cycles (Figure 2B). Although these bacteria typically form root nodules in legumes, they can also promote the growth of non-legume plants by producing phytohormones, solubilizing precipitated phosphorus, mineralizing organic phosphorus, secreting siderophores, degrading ethylene, and inhibiting pathogens (Mehboob et al., 2009; Ormeño-Orrillo and Martínez-Romero, 2019). Moreover, Bradyrhizobium possesses the capability for photosynthesis (Avontuur et al., 2023), which is consistent with the enriched bacterial functional potential of phototrophy, photoheterotrophy, and photoautotrophy observed in root-associated compartments (Figure 2B). This trait may also contribute to the dominant distribution of Bradyrhizobium in forest soil (VanInsberghe et al., 2015). Since soil is regarded as a seed bank of plant root-associated microbiota (Bai et al., 2022), this ecological advantage may help to explain why Bradyrhizobium is the most abundant genus in the root-associated compartments of C. amoena (Figure 2A). However, the direct influences of Bradyrhizobium, Rhizobium, and Mesorhizobium on seed germination, protocorm development, seedling establishment, and then distribution of C. amoena deserve further exploration, which can support the function prediction in this study.

Among the top 20 most abundant fungal genera in the root endosphere and rhizosphere of C. amoena, Astraeus, Amanita, and Russula were identified as typical OMF (Figure 2C). Notably, Astraeus and Amanita represented dominant genera in the roots of C. amoena collected from LongNan, whereas Russula was most abundant in the rhizosphere of C. amoena from ShenNongJia. Although OMF are indispensable for orchids, the Waiting Room Hypothesis proposes that mycorrhizal partners originate from saprobic fungi through an intermediate endophytic stage (Selosse et al., 2022). In line with this hypothesis, ecological guild prediction of the fungal communities revealed an enrichment of saprobic, endophytic, and mycorrhizal trophic modes in both root endosphere and rhizosphere of C. amoena (Figure 2D). Among non-OMF taxa, Mortierella was consistently abundant across all habitats in both root-associated compartments (Figure 2C), which corroborates a previous report of its isolation from C. amoena roots (Jiang et al., 2011). Mortierella has been reported as a saprophytic fungus with multiple agricultural benefits, including enhancing gibberellic acid production in planta and increasing soil phosphorus availability (Li et al., 2020). Given that AP significantly influenced both bacterial and fungal communities (Figures 7, 8), Mortierella may affect the root-associated microbiomes of C. amoena by enhancing phosphorus availability. Hygrocybe, which has been reported neither mycorrhizal nor saprotrophic (Tello et al., 2014), was most abundant in the root endosphere of C. amoena from ShenNongJia. Other genera enriched in root-associated compartments were also reported to promote plant health and development, such as Trichoderma and Exophiala (Xu et al., 2020; Contreras-Cornejo et al., 2024). Fusarium was enriched in root-associated compartments from XiangXi, XinYang, and BaZhong, with a higher relative abundance in the root endosphere than in the rhizosphere. However, this genus is reported as both root endophytes (Shamsudin et al., 2025) and pathogens in orchids (Srivastava et al., 2018). Fusarium may act as conditional or opportunistic pathogens, coexisting with host plants without causing disease symptoms under non-stressful environmental conditions and in the presence of a stable root-associated microbiome (Hardoim et al., 2015). Collectively, these findings highlight the diversity of fungi in the root endosphere and rhizosphere of C. amoena, including several groups with plant growth-promoting potentials, which may contribute to the health, adaptation, and distribution of this endangered orchid.

In the microbial co-occurrence networks across habitats, several ASVs were identified as mycorrhizal fungi (Figure 4). Among these, Ceratobasidiaceae, Tulasnellaceae, and Serendipitaceae are collectively referred to as rhizoctonias and are known to establish mycorrhizal associations exclusively with orchids (Selosse et al., 2022). Additionally, ectomycorrhizal fungi, mainly from Sebacinaceae and Thelephoraceae, are frequently detected in orchid roots as well. These mycorrhizal fungal ASVs exhibited significant positive relationships with many bacterial ASVs (Figure 4), indicating intimate interactions among C. amoena, mycorrhizal fungi, and bacteria. For example, MHBs can enhance mycorrhizal colonization by stimulating spore germination and mycelial expansion, as well as increasing root surface area to facilitate fungal colonization. Consequently, MHBs can double the colonization rate of mycorrhizal fungi and promote plant growth (Berrios et al., 2023). In the root-associated microbial networks of C. amoena, Bradyrhizobium, Rhizobium, Mesorhizobium, and Nitrospira showed significant positive correlations with mycorrhizal fungal ASVs (Figure 4). Bradyrhizobium has been reported to positively associate with mycorrhizal fungal abundance and identified as a strong predictor of mycorrhizal colonization in pine forests (Berrios et al., 2023). A similar trend was observed in the root-associated compartments of C. amoena. In detail, Bradyrhizobium was more abundant in BaZhong and XiangXi compared to other habitats (Figure 2A), which exhibited more complex microbial co-occurrence networks with higher numbers of mycorrhizal fungal ASVs (Figure 4). Additionally, mycorrhization can also be stimulated by Pseudomonas, Streptomyces, Rhizobium, and Bacillus (Frey-Klett et al., 2007), which were also abundant in the root-associated compartments of C. amoena (Figure 2A). Importantly, the multipartite interactions among plants, mycorrhizal fungi and bacteria can support plant growth and productivity (Kaur and Sharma, 2021) by enhancing nutrient mobilization from soil minerals, fixing atmospheric nitrogen, and protecting plants against root pathogens (Frey-Klett et al., 2007). Notably, although the soil nutrient levels in XiangXi and BaZhong are relatively lower than those in other habitats (Supplementary Figure S3), these two locations exhibit the most complex microbial networks (Figure 4). Therefore, the C. amoena individuals in XiangXi and BaZhong may benefit from their root-associated microbiomes. Overall, our results highlight close associations among C. amoena, mycorrhizal fungi, and root-associated bacteria, which may be relevant to plant performance and persistence across its native habitats. Therefore, the simulation of microbial interactions of native habitats may provide an efficient strategy for the conservation of endangered plants.

Our results demonstrate that the persistence of C. amoena in its native habitats is closely associated with the composition, functional potential, and interaction structure of its root-associated microbiomes, highlighting the importance of integrating microbial ecology into conservation strategies for endangered orchids. The widespread occurrence of OMF across root-associated compartments and habitats (Figures 2C,D) indicates that these fungal taxa are essential for the long-term survival of C. amoena, particularly given the obligate dependence of orchids on mycorrhizal symbiosis for seed germination and early development (McCormick et al., 2018). Additionally, the root-associated fungal communities can be affected by soil nutrients (Figures 7B, 8B), suggesting their vulnerability to habitat disturbance. Therefore, effective in situ conservation should prioritize the protection of undisturbed native habitats and intact mycorrhizal networks that support plant population persistence and regeneration. In parallel, ex situ conservation of C. amoena may benefit from inoculation with locally adapted OMF to enhance seed germination, seedling establishment, as previously demonstrated in other orchids (Yang et al., 2024).

In addition to fungi, the enrichment nitrogen-cycling bacterial genera such as Bradyrhizobium, Rhizobium, and Mesorhizobium (Figures 2A,B) suggests that root-associated bacterial partners may contribute to nitrogen acquisition of C. amoena and partially compensate for nitrogen limitation in native habitats (Trivedi et al., 2020). Notably, these nitrogen-cycling bacterial genera were identified as core microbiome members associated with C. amoena across native habitats (Figure 3A). Therefore, soil disturbance and excessive fertilization should be minimized in native populations of C. amoena, as maintaining microbiome-mediated nutrient cycling processes is critical for sustaining ecosystem resilience.

Further, the strong positive associations between OMF and beneficial bacterial taxa revealed by microbial co-occurrence network analyses (Figure 4) indicate that C. amoena relies on multipartite plant-fungus-bacterium interactions rather than single microbial partners. More importantly, geographically distinct populations of C. amoena are supported by habitat-specific microbiomes rather than a uniform microbial assemblage (Figures 1C,D). Collectively, the conservation of this endangered orchid should not focus solely on the plant itself but also consider the preservation of its associated microbiome.