Introduction

The use of medicinal plants has significantly enhanced public healthcare1 and played a crucial role in advancing global medicine2. A key question that remains unanswered is whether the stark differences in the metabolites of these plants are attributed to the microorganisms they harbor3,4,5. This uncertainty has propelled the plant microbiome to the forefront as an influential factor in plant health and production, attracting considerable interest from both the scientific community and commercial sectors6. Numerous studies have focused on the impact of plant-associated microorganisms on critical plant traits such as growth, disease resistance, environmental stress tolerance, water retention, and the production of plant growth-promoting hormones6,7.

Despite the expanding research demonstrating the influence of host plant species8, their developmental stages9, root exudates10, and rhizodeposits11 on the composition and functional capabilities of plant microbiomes, our understanding of these interactions remains incomplete in medicinal plants. It is known that root-associated microbiomes can vary depending on soil type and host genotype, though the extent of these factors’ contributions to community structure and metabolic potential remains incompletely understood12. However, research on plant-associated microbiomes, particularly their impact on plant health, with an emphasis on the root microbiome, is still in its nascent stages13,14.

Plants belonging to the Apiaceae15 and Araliaceae16 families are recognized for their abundant secondary metabolites, including terpenoids, ginsenosides, phenylpropanoids, polyacetylenes, and alkaloids17,18,19. These compounds hold immense potential in pharmaceutical research and industrial applications. Angelica gigas (AG)20, Peucedanum japonicum (PJ)21, Aralia cordata (AC)22,23, and Panax ginseng Meyer (PG)24,25 have garnered significant interest due to their diverse phytochemical profiles and medicinal potential17,26, however, despite this interest, studies exploring their microbial communities are relatively sparse and warrant further investigation to fully understand their microbial dynamics. Thus, investigating the complex microbiome connections between plant species in the Apiaceae15 and Araliaceae16 families within the order Apiales27 presents a valuable opportunity to unravel the complexities of plant-microbiome interactions.

The root microbiome of medicinal plants plays a crucial role in supporting plant health, resilience, and bioactivity, which is vital for their application in traditional and modern medicine. Despite the extensive research on the pharmacological properties of PG, there is limited understanding of the microbial diversity within its roots and how this microbiome compares to those of other phylogenetically related medicinal plants.

In this study, we aim to investigate the root microbiome of PG and compare it to that of other medicinally significant species, namely AG, PJ, and AC. These plants were selected due to their established medicinal uses and their phylogenetic relationships, which allow us to explore both within-family and PG’s inter-family microbial similarities and differences. By conducting a comparative analysis of these root microbiomes, we aim to uncover potential unique or shared microbial taxa that could contribute to the community of these plants. Understanding these relationships sheds light on the microbial community of PG and its relatives and enhances our knowledge of how root microbiomes may influence medicinal efficacy across phylogenetically related and unrelated species.

Results

Diversity of bacterial and fungal communities

To elucidate the determinants influencing the composition of root microbiota, we first examined the microbial diversity. Through alpha diversity assessments using Chao1, Shannon, and Simpson indices, PG’s microbial communities exhibited a notably lower diversity than those of AC, AG, and PJ (Fig. 1a; Supplementary Fig. 1). When comparing diversity values between plants in the identical family, both bacterial and fungal communities from AC-AG exhibited higher diversity than those from PG-PJ (Fig. 1b). With respect to beta diversity, compositional variations between PG and other plants were found (Fig. 1c,d). In the bacterial communities, PG and PJ samples were clustered together compared to the other plants despite the differences in the sampling areas (Fig. 1c; Supplementary Data 1). In the fungal community, on the other hand, AG and PJ, which belong to the same family (Apiaceae), were closely located compared to the remaining plants (Fig. 1d; Supplementary Data 1), suggesting that their community compositions were similar. Although PERMANOVA showed significant effects of plant species (bacteria: R² = 0.09, p-value = 0.0001; fungi: R² = 0.17, p-value = 0.0001) (Supplementary Data 1), family groups (bacteria: R² = 0.12, p-value = 0.0001(Supplementary Data 1); fungi: R² = 0.18, p-value = 0.0001) (Supplementary Data 1), and geographical regions (bacteria: R² = 0.29, p-value = 0.0001; fungi: R² = 0.34, p-value = 0.0001) on the observed community variances (Supplementary Data 1).

Fig. 1
figure 1

The distinctions between medicinal plants are visualized using alpha and beta diversity plots. Alpha diversity measures with the most common indices in (a, b) bacterial and fungal communities are shown. Ordination analysis of beta diversity in (c, d) bacterial and fungal communities.

Full size image

The composition of the endophytic bacterial and fungal communities

We have honed our relative abundance analysis to specifically examine the microbial communities associated with PG and its comparison to other medicinal plants, namely AC, PJ, and AG (Fig. 2; Supplementary Fig. 2, Fig. 3; Supplementary Data 2). This approach enables us to identify microbial signatures that differentiate PG from other medicinal plants, highlighting the importance of pairwise comparisons in understanding microbial diversity within these plants (Supplementary Fig. 4). The raw sequencing of a total of 13,597,330 sequences was obtained on the Illumina MiSeq. The V5-7 regions of the rRNA genes of bacteria and ITS2 regions of fungi were sequenced after quality control, generating a total of 6,940,039 reads (Supplementary Data 3). For bacterial communities, 422 ASVs were identified in AG, 551 in PJ, 1062 in AC, and 515 in PG. In contrast, fungal communities comprised 200 ASVs in AG, 135 in PJ, 203 in AC, and 156 in PG. The rarefaction curves showed that the number of ASV in each sample increased gradually with the quantity of sequence, thus confirming that the amount of sequencing data is adequate (Supplementary Fig. 5). Our findings reveal that PG is characterized by a notable dominance of Proteobacteria, including Xanthomonadales and Rhizobiales (abundance ratios of 30.18% and 30.05%, respectively), suggesting a microbial profile that differs from those of AC, PJ, and AG (Fig. 2a; Supplementary Fig. 2; Supplementary Data 2). For instance, AC, sharing the Araliaceae family with PG, showed a lesser but still abundant presence of Proteobacteria (Burkholderiales, with an abundance ratio of 17.71%), indicating familial similarities yet distinct microbial compositions. In contrast, AG and PJ, both belonging to the Apiaceae family, exhibited different dominant microbial orders, with AG showing a preference for Bacteroidetes (Flavobacteriales, abundance ratio of 35.28%) and PJ for Proteobacteria (Xanthomonadales, abundance ratio of 30.18%). Upon evaluating the bacteria with the highest abundance at the genus level for each medicinal plant, Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium exhibited the greatest abundance in PG, with this difference confirmed to be statistically significant (Fig. 2b; Supplementary Fig. 3; Supplementary Data 2).

Fig. 2
figure 2

Relative abundance (RA) in the (a) phylum level in bacteria, (b) highly abundant genera in bacteria, (c) phylum level in fungi, and (d) highly abundant genera in fungi in the AG, PJ, AC, and PG. Low abundance taxonomic groups with less than 5‰ (permille) of each sample are highlighted in gray. Unidentified and other taxonomic groups are indicated in black.

Full size image
Fig. 3
figure 3

Amplicon sequence variants (ASVs) responsible for the community differences. The ASVs were identified using pairwise random forest classification models. (a) Responsible bacterial ASVs between AG (Angelica gigas) and PG (Panax ginseng). (b) Responsible fungal ASVs between AG and PG. (c) Responsible bacterial ASVs between AG (Angelica gigas) and PJ (Peucedanum japonicum). (d) Responsible fungal ASVs between AG and PJ. (e) Responsible bacterial ASVs between AC (Aralia cordata) and PG (Panax ginseng). (f) Responsible fungal ASVs between AC and PG.

Full size image

When analyzing the fungal communities associated with PG in comparison to other medicinal plants, a similar abundance of fungi from the phyla Ascomycota and Basidiomycota is evident (Fig. 2c). At the genus level, Phlebopus (Basidiomycota) demonstrates the highest abundance in PG (Fig. 2d). Furthermore, PG’s fungal community, heavily populated by Leotiomycetes (Helotiales, with an abundance ratio of 44.12%), contrasts with the dominance of Dothideomycetes (Pleosporales, abundance ratio of 49.47%) in AC, while showing similarities with the fungal profiles of PJ and AG. On the other hand, PG and AC showed a higher abundance of Agaricomycetes compared to other plants. In particular, PG showed the notable dominance of Boletales (Supplementary Fig. 3).

Random forest models linking root-associated microbiota

To deeply examine the compositional differences between PG and other medicinal plants, we investigated differentially distributed ASVs in the microbiota profiles. This analysis, guided by differential abundance analysis and random forest (RF) classification models, has enabled a precise identification of microbial communities that distinguish PG from these plants, illuminating the microbial community that PG harbors.

The microbial community of PG, marked by the significant presence of B1653_o_Enterobacterales and F8_o_Helotiales, distinguished it from other medicinal plants by showing higher F46_Paraphoma in AC, B48_f_Comamonadaceae in AG and B58_f_Streptomycetaceae in PJ (Supplementary Fig. 6; Supplementary Data 4). Additionally, the particular abundance of B69_f_Flavobacteriaceae in AG and the rise of B2019_f_Pseudonocardiaceae in PJ (Supplementary Fig. 6; Supplementary Data 4), compared to PG, further highlight the distinct microbial relationships intrinsic to PG’s root microbiome.

The deployment of RF classification models highlighted the most influential microbial species within PG (Supplementary Data 5). Remarkably, the model revealed a predominance of Actinobacteria and Proteobacteria, shaping PG’s microbial community (Fig. 3a,b). In the comparative analysis of PG and PJ, PG shows a significant prevalence within bacterial communities, particularly within Proteobacteria families like Rhizobiaceae (Fig. 3c; Supplementary Data 4; Supplementary Data 6), whereas PJ is notably enriched in fungal communities, especially within Ascomycota families such as Phaeosphaeriaceae (Fig. 3d; Supplementary Data 4; Supplementary Data 6). These findings highlight distinct microbial preferences for PG and PJ across bacterial and fungal domains. In comparing PG and AG, AG exhibits a clear dominance, particularly within the bacterial communities where families like Comamonadaceae are notably enriched (Supplementary Data 4; Supplementary Data 6). In fungal communities, AG also shows a significant presence, with Tausonia (Supplementary Data 4; Supplementary Data 6). In the AC-PG comparison, Proteobacteria and Actinobacteria were the dominant bacterial ASVs, with a distinctive presence of B502_Lysobacter in PG, setting it apart from the mostly AC-enriched bacterial community (Fig. 3d,e). Similarly, while Ascomycota and Basidiomycota fungi predominantly favored AG in the AG-PJ comparison, PG was uniquely characterized by the presence of F129_Helotiales from Ascomycota, emphasizing its distinct microbial community in contrast to AG and PJ. The bacterial communities further support this distinction, with PG displaying enrichment of Comamonadaceae and Microbacteriaceae, whereas AG and PJ are dominated by Sphingomonadaceae and Rhizobiaceae, respectively. Additionally, AC exhibits a unique microbial signature, particularly with the enrichment of Streptomycetaceae and Micromonosporaceae among bacteria, and Didymellaceae and Chaetomiaceae among fungi, highlighting its diverse microbial community. These findings underscore each plant’s specific role within the medicinal plant microbiome, with distinct bacterial and fungal ASVs enriched or scarce in their environments, pointing to targeted root-microbiome interactions that vary by plant type (Fig. 4a,b; Supplementary Figs. 6 and 7, Supplementary Data 4; Supplementary Data 6).

Fig. 4
figure 4

Heatmap of the core microbiome analysis to identify core taxa. (a) Bacterial core ASVs. (b) Fungal core ASVs. Core ASVs were calculated for AG (Angelica gigas), PJ (Peucedanum japonicum), AC (Aralia cordata), and PG (Panax ginseng), respectively. The prevalence threshold for core ASVs was 90% (bacteria) and 95% (fungi) of AG, PJ, AC, and PG.

Full size image

Common and specific core microbiome in four medicinal plants

In our quest to illuminate the complex microbial structure within medicinal plants, particularly focusing on PG’s stance within these communities, we embarked on an in-depth exploration of the pan and core microbiomes. By applying a rigorous taxon prevalence criterion—90% for bacteria and 95% for fungi—we were able to discern the core taxa that are central to each plant’s microbiome (Supplementary Fig. 8). PG stands out not only in its microbial community but also in the specificity of its core microbial community, particularly when juxtaposed against AC, PJ, and AG (Fig. 5). When the core microbiome of PG was compared with that of three other medicinal plants, it was observed that the core microbiomes of both bacteria and fungi were more abundant in PG than in other medicinal plants (Fig. 5; Supplementary Fig. 8). This distinction is most pronounced within the PG-AC pair, where 1,573 bacterial pan-microbiomes were identified, yet only two ASVs, B70_Cutibacterium and B10_Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium, were identified as core, highlighting the microbial signature of PG (Supplementary Fig. 8a, Supplementary Table 1). This trend of selectivity extends to the fungal domain, where, out of 353 pan-microbiomes, only three—F36_Moesziomyces, F30_Phaeosphaeria, and F22_Alternaria—were deemed core, further accentuating PG’s distinct microbial community (Supplementary Fig. 8b, Supplementary Table 1).

Fig. 5
figure 5

Comparison of network attributes between bacterial and fungal nodes in microbial co-occurrence networks (a) AG (Angelica gigas) and (b) PJ (Peucedanum japonicum) and (c) AC (Aralia cordata) and (d) PG (Panax ginseng).

Full size image

In contrast to PG’s specialized microbial community, AG and PJ present a broader bacterial and fungal community. With 965 bacterial pan-microbiomes, including four core ASVs—B21_Pseudomonadaceae, B114_Pseudomonadaceae, B5_Comamonadaceae, and B4_Rhodanobacteraceae—this comparison underscores a more generalized microbial community spanning these plants (Supplementary Fig. 8c, Supplementary Table 1). This generalization extends into the fungi, where, among 327 pan-microbiomes, only four—F56_Malassezia, F36_Moesziomyces, F30_Phaeosphaeria, and F22_Alternaria—stand out as core, suggesting a less selective fungal community in AG and PJ (Supplementary Fig. 8d, Supplementary Table 1). This broader microbial diversity contrasts sharply with PG’s distinct composition, where the core bacterial and fungal communities are markedly fewer, highlighting PG’s niche. This comparative analysis illuminates the microbial composition of PG, characterized by a highly specialized core microbiome, setting it apart from the more generalized microbial communities observed in AC, AG, and PJ.

Moreover, we identified a stable core microbiome that remains consistent across different geographical locations and plant species, highlighting the fundamental and ubiquitous role these core microorganisms play within the community. Our analysis of pan-microbiomes revealed a total of 2,273 bacteria and 590 fungal ASVs. The microbial communities associated with medicinal plant genera and families exhibit a limited set of core microorganisms. The bacterial core microbiome includes key taxa such as B5_Comamonadaceae, B10_Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium, and B4_Rhodanobacteraceae (Fig. 5a), while the fungal core microbiome is composed of F56_Malassezia, F36_Moesziomyces, F30_Phaeosphaeria, and F22_Alternaria (Fig. 5b). Particularly noteworthy is PG’s core microbiome, which is markedly characterized by B10_Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium. Moreover, the preeminence of fungal core ASVs over bacterial ones in PG is marked by their substantially higher abundance, illustrating a distinctive fungal profile within PG’s microbiome compared to other medicinal plants. The abundance patterns of a core microbiome within PG, characterized by certain significant taxa, serve to illustrate the distinctiveness of PG’s microbial community.

Network analysis of endophytic microbial communities

We constructed microbial networks for each plant to unravel the intricate and dynamic relationships within root endophytic microbial communities of medicinal plants (Fig. 6; Supplementary Fig. 8). The network for AG consisted of 165 nodes and 982 connections, split between 568 positive and 414 negative associations (Supplementary Fig. 9a). PJ’s network had 101 nodes and 666 connections, with a distribution of 351 positive and 315 negative associations (Supplementary Fig. 9b). For AC, the network had 149 nodes and 740 connections (430 positive and 310 negative associations) (Supplementary Fig. 9c), while PG’s network had 88 nodes and 408 connections (257 positive and 151 negative associations), all with a correlation threshold of > 0.3 and < − 0.3 (p-value < 0.05) (Supplementary Fig. 8d).

Fig. 6
figure 6

The microbial network of the medicinal plant’s microbiome identified common hubs. Hubs are network nodes that have elevated network closeness centrality. (a) AG (Angelica gigas) - (b) PJ (Peucedanum japonicum), (c) AC (Aralia cordata), and (d) PG (Panax ginseng).

Full size image

Each plant displayed bacterial and fungal roles in their respective microbial networks, particularly evident in the PG, AC, AG, and PJ networks. Bacterial taxa in AG and PJ’s networks showed more connectivity than fungal taxa (Fig. 6a, b). Notably, there was a significant divergence in the microbial networks of AC and PG, as highlighted in their network centrality indices (Fig. 6c, d), with fungi playing a more central role in AC’s network compared to others.

To further dissect the network roles, hub nodes were identified using degree and closeness centrality measures. Nodes in the top 1% percentile of these measurements were designated as hubs. In AG, hubs included Actinobacteriota (B74_f_Microbacteriaceae, B194_f_Streptomycetaceae, and B125_f_Streptomycetaceae), Bacteroidota (B80_f_Flavobacteriaceae), and Proteobacteria (B155_f_Oxalobacteraceae) (Fig. 4a; Supplementary Table 2). PJ’s network hubs were Proteobacteria (B100_Caulobacter and B7_f_Oxalobacteraceae) (Fig. 4b; Supplementary Table 2), and in AC, prominent hubs were Proteobacteria (B31_f_Comamonadaceae and B8_f_Sphingomonadaceae) and Ascomycota (F42_Setophoma) (Fig. 4c; Supplementary Table 2). Notably, no hub node was defined in PG’s network (Fig. 4d; Supplementary Table 2). Unlike the microbial networks observed in AG, PJ, and AC, PG’s network was noted for its lack of hub nodes. This feature suggests an egalitarian structure within PG’s microbial community, where interactions are broadly distributed rather than concentrated around a few key taxa. Moreover, the identification of microbial associations within PG’s network, such as interactions involving B13_f_Oxalobacteraceae and B61_Burkholderia-Caballeronia-Paraburkholderia (Supplementary Data 7), underscores the specific and distinct microbial interactions that contribute to the of PG’s root microbiome.

Discussion

The dynamics of plant-microbial interactions are intricate. Plants primarily modulate the composition of the root microbiome by releasing root exudates and activating the plant immune system. In turn, the presence of the root microbiome enhances plants’ adaptability and metabolic capabilities, contributing to several physiological processes like plant development, nutrient uptake, and responses to environmental stressors28,29. Previous studies for Panax have not yet proven that climates affect the diversity of endophytic communities in different geographical regions30,31. Generally, to discover the geographical effects on the plant microbiomes, microbiomes associated with the same species were investigated. However, in this study, microbiomes of different species grown in different locations were examined. In spite of this experimental limitation, a compositional commonality between PG and other medicinal plants was found. The main bacterial phyla for all medicinal plants were Proteobacteria, Actinobacteria, and Bacteroidetes. Prior research has demonstrated that Actinobacteria and Proteobacteria are prevalent phyla found in plant tissues32. These microorganisms play a crucial function in contributing to host metabolism and ensuring the stability of the endophytic microflora33. The relative abundance of Actinobacteria in AC, PJ, PG, and AG was quantified as 32.47%, 17%, 15.27%, and 8.03% respectively. Approximately 46.4% of the detected bacterial orders were shared among all plant species (Supplementary Fig. 4). Through a meticulous pairwise comparison, we have illustrated PG’s distinctive microbial interactions, shedding light on the bacterial and fungal constituents that define its ecological niche. This analysis not only advances our understanding of PG’s microbial community but also underscores the complexity and specificity of microbial communities across different medicinal plants, providing insights into their potential microbial functions and adaptations. These findings suggest that there might be common internal niche conditions, which are possibly attributed to the selective enrichment, spanning the investigated plant species irrespective of external environmental conditions.

Previous research on Panax spp. has predominantly highlighted the dominance of Proteobacteria and Actinobacteria within the bacterial endophytes of roots, with a lesser presence of Firmicutes and Bacteroidetes34. Furthermore, studies have identified specific genera like Bacillus, Pseudomonas35,36 Erwinia, Stenotrophomonas, and Sphingobium37 associated with different Panax species. Furthermore, the fungal microbiome is mostly composed of the phylum Ascomycota, which accounts for 76% of endophytic fungi. Basidiomycota, on the other hand, has been reported at a considerably lower abundance of 7% 38. Our research also reveals a predominance of Ascomycota and Basidiomycota in the endophytic fungal community of PG (Figs. S2 and S3), suggesting a longstanding symbiotic relationship fostered by common niche environments.

The intricate microbial relationships uncovered in our study, particularly within the roots of medicinal plants such as PG, highlight the potential roles these microbes play in plant health and development. For instance, the Flavobacteriaceae family, identified as both a significant RF and a network hub in AG, has been reported to contribute to the transformation of ginsenosides and the production of secondary metabolites with antibacterial and antioxidant properties39. Notably, endophytic bacteria from the PG rhizosphere, especially Flavobacterium spp., have been shown to enhance the production of rare ginsenosides Rg3 and Rh2, underscoring the potential benefits of these compounds in cancer treatment, antioxidant protection, and immune support40. This focused exploration into PG’s microbiome paves the way for future research, promising new avenues for leveraging microbial insights to optimize medicinal plant utility and cultivation, particularly for PG. Understanding these microbial interactions can lead to innovative strategies to maximize the medicinal benefits of PG, enhance its cultivation practices, and improve its overall yield.

Furthermore, Rhodanobacteraceae’s increased abundance in the ginseng rhizosphere, correlated with higher ginsenoside concentrations, suggests a beneficial interaction facilitated by the antibacterial properties of ginsenosides40. This could be corroborated by the greater proportionate prevalence in PG samples in comparison to other medicinal plants (Figs. 2c and 3a; Figs. S2 and S7). The pan microbiome, a thorough catalog of all microorganisms associated with plants, contrasts with the core microbiome, which zeroes in on the resilient and indispensable microorganisms vital for the plant’s health and development. Our pan-microbiome analysis identified a total of 2,273 bacterial and 590 fungal ASVs. Microbial communities associated with medicinal plant genera and families exhibit a limited set of host microorganisms. The microbiome of PG is particularly characterized by B10_Allorhizobium-Neorhizobium-Parararhizobium-Rhizobium, and the higher abundance of fungal host ASVs than bacterial ones indicates a distinct fungal profile in the microbiome of PG. The bacterial communities Allorhizobium-Neorhizobium-Neorhizobium-Parhizobium-Rhizobium were consistently present in both rice genotypes throughout all three experimental fields in a study. These bacterial communities are known to be part of the core-endophytic microbiome of rice genotypes41. The abundance patterns of the core microbiome, characterized by specific important taxa in the microbial community of PG, highlight the uniqueness of the microbial community of PG. This suggests that similar microorganisms in different plants, displaying conserved traits, could be indicative of certain plants supporting similar microbial communities in similar environments. This study not only provides a detailed review of the microbial community within PG and medicinal plants belonging to the same order but also lays the groundwork for further research into optimizing these plant-microbial interactions. Understanding these complex relationships is crucial for improving the efficacy of medicinal plants and harnessing their therapeutic potential, and it is an important step forward in the field of plant biology and phytotherapy. It also points to the probable importance of these microorganisms in influencing PG’s growth, health, and medicinal qualities. These results in PG’s position in the microbial world of medicinal plants. Identifying its distinctive core microbiome offers valuable perspectives on the microbial interactions crucial for PG’s development and its potential for therapeutic use.

In our study, microbial networks were constructed to examine the relationships of root endophytic microbial communities in each medicinal plant. The SparCC analysis uncovered intricate relationships between the endophytic microbial communities. We observed both positive and negative associations among the bacterial and fungal communities of the endophytic microbiome. In AG and PJ, bacterial taxa showed more connections, while the microbial networks of AC and PG differed, with fungi playing a more central role in AC. Specific bacterial and fungal hub nodes were identified in AG, PJ, and AC; however, no hub nodes were identified in PG, indicating a more egalitarian structure. The observations from PG’s network analysis suggest a more resilient microbial community capable of maintaining functional stability under external pressures, thereby supporting PG’s health and enhancing its medicinal properties. This insight into the specific microbial interactions within PG’s root microbiome offers a valuable perspective on the microbial underpinnings of medicinal plants, highlighting the potential to leverage these intricate microbial relationships to optimize plant health and medicinal value.

Materials and methods

Sample preparation

In March 2022, we collected samples for a comparative analysis of root microbiome profiles in phylogenetically related medicinal plants. We collected four plant species from two different families, Araliaceae and Apiaceae, within the same order, Apiales. To minimize environmental variability and emphasize genetic influences, each medicinal plant species was obtained from a single representative location. The selection was based on the assumption that their phylogenetic relationship with PG might indicate similar root microbial communities. For PG, A 5-year-old local landrace was collected from Hoengseong (37°19’31.0"N 128°05’18.3"E), the second-largest region in Gangwon Province for cultivating PG (Ginseng Statistics in 2021, https://www.mafra.go.kr/bbs/mafra/71/331285/artclView.do ). For AC, A 2-year-old cultivar named ‘Baek mi hyang’ developed and commercialized by Gangwon State Agricultural Research and Extension Service, was collected from Suwon (37°15’34.5"N 126°58’57.9"E). For PJ samples, a one-year-old local landrace was collected from Suwon (37°15’34.5"N 126°58’57.9"E). For AG, a one-year-old local landrace was collected from NIHHS in Eumseong (36°56’38.8"N 127°45’17.7"E).

In both field (PG, AC, and PJ) and pot (AG) conditions, we randomly selected three different individuals for each species. After being cleaned with running tap water, the collected root samples were surface sterilized with 75% ethanol for 1 min, 5% NaOCl for 3 min, and 75% ethanol for 30 s. The surface-sterilized root sample samples were chopped into 1 cm long segments using the sterile razor blade. The root segments from identical plant individuals were randomly pooled into Lysing Matrix S tubes included in SPINeasy DNA Kit for Soil (MP Biomedicals, USA) to make the weight 0.3 g for DNA extraction. To prevent DNA denaturation, all tubes were stored in phosphate-buffered saline (PBS). To consider the variability of root microbiomes within each individual, eight replicates (two or three repeats for each plant individual were randomly selected) were prepared for PG, AC, AG, and PJ.

DNA extraction and molecular analysis

The samples were ground using a homogenizer (FastPrep-24™ 5G, MP Biomedicals, USA). A DNA extraction step was performed on each sample at 4 °C. The procedure was executed following the manufacturer’s instruction. The purity and measurement of all DNA samples were assessed using a NanoDrop spectrophotometer (Thermo Scientific™, USA). The extracted DNA underwent polymerase chain reaction (PCR) utilizing the primers 799 F-1193R for the V5-V7 area of the bacterial 16 S ribosomal RNA (rRNA) gene42. The fungal ITS1 region of the ribosomal RNA genes was amplified using the ITS1-F and ITS2-R PCR primers43. The amplification of each sample was performed in triplicate using a 25 µl reaction tube. The reaction tube consisted of 12.5 uL of 0.5 μm of a 2× Platinum SuperFi™ II Green PCR Master Mix (Thermo Fisher Scientific), 1 uL of 0.5 μm each of forward and reverse primers and 0.8 µM of diluted DNA template. The PCR was conducted utilizing the subsequent protocol for amplifying bacterial amplicons. The procedure commenced with an initial denaturation step at a temperature of 98 °C for a duration of 30 s. This was followed by 32 cycles of denaturation at 98 °C for 10 s, primer annealing at 55 °C for 10 s, and extension at 72 °C for 40 s. Subsequently, a final extension step was performed at 72 °C for 5 min, and the reaction was then held at a temperature of 12 °C. The process for ITS PCR amplification remained consistent. The MEGAquick-spinTMPlus DNA Purification Kit (Intron Biotechnology, Korea) was employed for the purpose of consolidating and refining the amplicon replicates derived from a single DNA sample. The sequencing procedure was conducted at the National Instrumentation Center for Environmental Management (NICEM) located at Seoul National University.

Sequence processing

The sequencing sequences were processed using QIIME2 (ver. 2022.11)44. Following the process of demultiplexing, the sequences were merged and subjected to quality filtration within the QIIME2 (v. 2022.11) pipeline, utilizing the DADA2 plugin45. The processed readings were subsequently assigned to amplicon sequence variations (ASVs). The taxonomic classification of non-chimeric ASVs was determined using the Naive Bayes approach, which was implemented in the q2-feature-classifier algorithm46. The taxonomic classifications for the V5-V7 region of the 16 S rRNA gene were derived from the Silva database (ver. 138)47. The taxonomic classifications for the ITS regions were conducted utilizing the UNITE database (UNITE_ver8_dynamic as of May 2021)48. The dataset contained bacterial sequences with lengths ranging from 278 to 400 base pairs (bp) and fungal sequences ranging from 100 to 400 bp. The phyloseq 49 program in R has been used to cleanse ASV profiles. Implementing these filtering procedures decreased the total number of ASVs observed in bacteria to 3,061, while in fungi, the count was reduced to 694. The ASVs classified under the taxonomic category “kingdom Fungi” but lacking identification at the phylum level were subjected to a BLASTn search. The bacterial ASV profile was further filtered to exclude chloroplasts and mitochondria.

Data analysis and visualization

All statistical analyses conducted in R version 4.2.2 (R Core Team, 2013) established a statistical significance threshold of p-value = 0.05. The cumNorm function from the metagenomeSeq R package (v. 3.8) has been used to apply a logarithmic transformation to the ASV abundance table and normalize it using the cumulative-sum scaling (CSS) method50. The diversity function in the Vegan (v2.5-3) R package was utilized to compute the Chao, Shannon, and Simpson indices51. The Kruskal-Wallis test and Wilcoxon rank-sum test were conducted using the R programming language. The Bray-Curtis dissimilarity matrix was computed to generate principal coordinate analyses (PCoA), both with and without imposed limits. The constrained principal coordinates (CAP) analysis employed the cap scale and ordinate functions from the phyloseq and vegan packages. The adonis2 function from the Vegan package51, was employed to perform a permutational multivariate variance analysis (PERMANOVA). The Vegan package was utilized to conduct variance partitioning and significance analysis for experimental variables using 99,999 permutations. Taxonomic composition analysis was conducted using the R package ggplot2 52, focusing on taxa that exhibited a relative abundance exceeding 0.5%. Both main ASVs of medicinal plants were identified. A prevalence criterion of 90% was established for core ASVs pertaining to bacteria, whereas a prevalence criterion of 95% was set for fungal core ASVs. A Venn diagram showing the numbers of shared and relative abundances of bacterial and fungal ASVs among AG, PJ, AC, and PG was generated using InteractiVenn53. The FitZig algorithm from the metagenomeSeq package54 was utilized to construct a statistical model based on a mixture of zero-inflated Gaussian distributions. The Contrasts and eBayes functions were implemented using the R package limma (v. 3.34.9)55. The statistical significance of the abundance differences was determined by examining the FDR-adjusted p values. Abundance differences with p-values below 0.01 were considered significant. The ggplot2 package was utilized to generate volcano plots, which visually represented the ASVs exhibiting different levels of bacterial and fungal abundance.

The random forest classification model

The parameters for the classification model, which included the four different medicinal plant species, were set using the abundance of microbiota. The ROCR (v. 1.0.7) and randomForest packages (v. 4.6–14) were used. The machine learning strategy using the random forest in R was used to investigate ROC curves56. The assessment of the effectiveness of the RandomForest model in predicting four medicinal plants involved the utilization of the average reduction in the Gini coefficient as an evaluation of importance for the ASVs. The significance function from the randomForest package was utilized for this purpose. The ASVs from each kingdom’s RF model were categorized as AG_enriched, PJ_enriched, AC_enriched, PG_enriched, AP_enriched, and AR_enriched or non-differential ASVs based on the differential abundance test.

Networks of connections between microorganisms

In the realm of microbial network analysis, the selection method predominantly employed iNAP57. The input data for SparCC analysis consisted of the merged tables containing bacterial and fungal ASVs58. The estimated correlations59 were limited to those that were statistically significant at a significance level of 0.05, using a two-sided test. ASV associations were limited to correlations of > 0.3 and < − 0.3 (p-value < 0.05). The ForceAtlas2 layout algorithm to visualize the data using Gephi (v.0.9.2) was used60. The calculations of degree, betweenness, closeness, and eigenvector centralities were performed using the R programming language and Gephi (v.0.9.2)60, respectively. The hub nodes were selected as the top 1% in terms of degree and closeness centrality within each network.