Transgenerational Dynamics of Endophytic Microbiome in Soybean Seeds
Article information
Abstract
Seeds harbor diverse microbial communities, including endophytes, some of which are vertically transmitted and may contribute to plant health and productivity. However, the temporal dynamics of seed endophytic communities remain poorly understood in many crop species. In this study, we monitored the composition of bacterial and fungal endophytes in soybean (Glycine max) seeds, along with rhizosphere microbiomes, across three plant generations using a culture-independent approach. Our results revealed two key patterns: seed endophytic communities are distinct from those of bulk soil and rhizosphere microbiomes; and the composition of seed endophytes fluctuates over generations, likely influenced by both pre-existing endophytes, environmental factors, and microbial influx from the surrounding soil and rhizosphere, suggesting possible microbial transmission from the rhizosphere into seeds. Interestingly, despite generational variation, the seed fungal endophyte communities consistently maintained higher phylogenetic diversity compared to bacterial endophytes, which showed limited overlap across generations and were composed of fewer, closely related taxa. Analysis of community assembly mechanisms indicated that both seed and rhizosphere microbiomes significantly contributed to the next generation of seed microbiota, primarily through stochastic drift and homogeneous selection processes. Collectively, our findings offer valuable insights into the intergenerational dynamics of seed endophytes in soybean and provide a foundation for future efforts to harness seed-associated microbiomes for improving crop health and productivity.
Plants function as holobionts, hosting diverse microbial communities that contribute to growth, development, and adaptation to environmental stressors (Rosenberg and Zilber-Rosenberg, 2016; Simon et al., 2019; Vandenkoornhuyse et al., 2015). Plant-associated microbes inhabit not only rhizosphere, endosphere, and phyllosphere but also in and outside seeds. Among these microbial partners, seed endophytes comprise a particularly important group owing to their potential for vertical transmission from parent to offspring and their influence on early plant establishment (Nelson, 2018). In this context, seeds offer a unique mechanism through which beneficial microorganisms can be inherited alongside plant genomes (Berg and Raaijmakers, 2018).
Although a growing body of evidence indicates that seed endophytes can be transmitted across generations in diverse crop species, the mechanisms and consistency of transmission appear to vary by plant species and microbial taxa (Rochefort et al., 2021; Truyens et al., 2015; Zhang et al., 2022). Research in rice (Oryza sativa) has provided detailed insights into the dynamics of transgenerational seed microbiomes, showing that core seed-associated microbes contribute to nutrient acquisition and stress resilience (Kim et al., 2022). The concept of vertical transmission of seed-associated microbes has gained increasing attention in recent years, driven by growing recognition of their ecological and agronomic significance. Some studies report vertical transmission rates exceeding 90% for many fungal endophytes, underscoring the efficiency and potential importance of this process (Ngugi and Scherm, 2006). Understanding the transmission dynamics of seed endophytes is, therefore, critical in agricultural systems, where manipulation of plant-associated microbiomes holds promise for sustainable crop production.
Investigating the transgenerational dynamics of seed endophytes is vital for fundamental ecology, evolutionary biology, and agricultural applications. From a developmental perspective, seed-borne endophytes are among the first microbes to colonize emerging seedlings and can shape the initial microbiome that influences early growth (Rochefort et al., 2021). The ecological concept of the priority effect—where early-arriving species influence community assembly—suggests that seed-transmitted microbes play foundational roles in shaping plant-microbe interactions across the plant’s life cycle (Nelson, 2018). Beneficial seed endophytes can enhance disease resistance, nutrient acquisition, stress tolerance, and overall growth (Nelson, 2018; Shahzad et al., 2016). Furthermore, transmission fidelity has implications for the effectiveness and predictability of microbiome-based interventions. If endophytes are reliably inherited across generations, they may serve as sustainable tools for long-term crop improvement. Conversely, if their transmission is inconsistent or environmentally modulated, this knowledge is crucial for tailoring inoculation strategies to maximize their persistence and impact.
Despite increasing interest, major knowledge gaps remain regarding the mechanisms and consistency of seed endophyte transmission, particularly in staple crops like soybean. Limited understanding exists around how these microbes colonize developing seeds, and how environmental factors, plant genotype, and microbial interactions influence their transgenerational dynamics. Addressing these gaps requires longitudinal studies that track both seed and rhizosphere microbiomes across multiple generations and environmental contexts.
In this study, we provide a comprehensive analysis of bacterial and fungal endophyte communities in soybean seeds across three generations. Soybean (Glycine max), a globally important crop for protein and oil, serves as an ideal model for such studies due to its economic relevance and well-characterized genetics (Ayilara et al., 2023; Messina, 1999). Simultaneously, we characterize rhizosphere microbial communities to investigate potential microbial sources contributing to seed colonization. Using culture-independent methods and a focus on community assembly mechanisms, this work aims to elucidate the ecological processes governing the vertical transmission of seed endophytes and support future efforts to harness these microbial partnerships for sustainable soybean cultivation.
Materials and Methods
Collection of soybean seeds, bulk soils, and rhizosphere soils
Daewon soybean seeds (Kyoungshin Seeds Co., Ltd., Uiseong, Korea), which are cultivated commonly in Korean farms, were used in this study. In order to obtain the next generation of seeds, the soybean seeds were planted in the experimental plot of Dong-A University (location: Gimhae-si, Gyeongsangnam-do, Republic of Korea). In 2021, 2022, and 2023, seeds were sown 9 Jun, 10 Jun, and 15 Jun, respectively. Seeds were harvested from plants in 12 Nov, 16 Nov, and 10 Nov of the corresponding years. Bulk soil used in this study were also obtained in the same field and rhizosphere soils were collected from the cultivated soybean roots.
Extraction of DNA for analysis of microbial community
Soybean seeds were washed with 70% ethanol and dipped into 1% sodium hypochlorite solution for 5 min to disinfect surface and rinsed with sterilized water. Surface-sterilized seeds were pulverized using liquid nitrogen with mortar and pestle. The genomic DNA was extracted from the pulverized seeds with a FastDNA Spin Kit (MPBio, Santa Ana, CA, USA) according to the manufacturer’s instructions. DNA of bulk soils and rhizosphere soils was extracted with a FastDNA Spin Kit for Soil (MPBio) according to the manufacturer’s instructions. Extracted DNA was quantified with a Nano-400A micro-spectrophotometer (Allsheng, Hangzhou, China).
Illumina Miseq amplicon library and sequencing
To analyze microbial community, the library was manufactured using paired-end sequence read. For the analysis of fungal community, internal transcribed spacer (ITS) region was amplified using forward primer ITS3 (5′-GCATCGATGAAGAACGCAGC-3′) and reverse primer ITS4 5′-TCCTCCGCTTATTGATATGC-3′). The 30 μL reaction mixture consisted of 2.5 μL of template DNA, 5 μL of each primer, 12.5 μL of KAPA HiFi (Roche, Basel, Swiss), 2.5 μL of plant chloroplast-targeting peptide nucleic acids (7.5 μM), and 2.5 μL of mitochondrial peptide nucleic acids (7.5 μM) (Supplementary Table 1). The PCR was performed in a thermal cycler (Life Technologies Holdings Pte Ltd., Woodlands, Singapore) using cycling conditions that consisted of an initial denaturation at at 98°C for 3 min, 30 cycles at 98°C for 10 s, 78°C for 10 s (annealing temperature of the PNA blocker), 55°C for 30 s, 72°C for 1 min, and a final extension at 72°C for 5 min.
For the analysis of bacterial community, 16S rRNA region was amplified using forward primer V3 (5′-CCTACGGGNGGCWGCAG-3′) and reverse primer V4 (5′-GACTACHVGGGTATCTAATCC-3′). The composition of the PCR reaction mixture was the same as described above. The PCR was performed using cycling conditions that consisted of an initial denaturation at 98°C for 3 min, 35 cycles at 98°C for 10 s, 78°C for 10 s (annealing temperature of the PNA blocker), 55°C for 30 s, 72°C for 1 min, and a final extension at 72°C for 5 min. The PCR products were purified using the AMPureXP Beads (Beckman Coulter, Brea, CA, USA) and a 96-Well Ring Magnet Plate T480 (Permagen Labware, Peabody, MA, USA). The purified PCR products were analyzed by 1% agarose gel electrophoresis, and the samples were sent for sequencing using Miseq platform (Illumina, San Diego, CA, USA) by Macrogen, Inc. (Seoul, Korea).
ITS and 16S rRNA sequencing data processing
The raw data were analyzed using the EzMAP pipeline (Shanmugam et al., 2021). Chimeras were eliminated and sequences were clustered into operational taxonomic units (OTUs) at 97% sequence similarity. Taxonomical identification was carried out using SILVA (16s rRNA) and UNITE (ITS) for bacteria and fungi, respectively. The taxa not seen more than 3 times in at least 20% of the samples was removed. Relative abundance, alpha diversity, and beta diversity were analyzed using the normalized OTUs to understand the microbial community structure. Alpha diversity indices (Shannon index and Observed species index) were calculated to assess the richness and species diversity. Beta diversity of the microbial communities was studied using principal coordinate analysis.
Analysis of community diversities and assembly mechanisms
To find the source of seed microbial communities, we examined sink-source relationships between seed samples and other plant and soil compartments. For this analysis, we used a fast expectation-maximization for microbial source tracking (FEAST, v 0.1.0) algorithm (Shenhav et al., 2019). To assess the relative importance of sources to the sink formation, we analyzed three analytic sets (Set 1: sink – second-generation seeds [2GS] and source – first-generation seeds [1GS], bulk soil [BS], and second-generation rhizosphere [2RS]; Set 2: sink – third-generation seeds [3GS] and source – 2GS and third-generation rhizosphere [3RS]; Set 3: sink – 3GS and source – BS, 2RS, 3RS, 1GS, and 2GS). Two hyperparameters, convergence threshold and the maximum number of iterations, in FEAST were set to 1 × 10−6 and 1,000, respectively.
We also predicted ecological drivers of the assembly of seed bacterial and fungal communities using the R package iCAMP (v 1.5.12) (Ning et al., 2020). The relative importance of five assembly mechanisms (or ecological drivers), homogeneous selection, heterogeneous selection, dispersal limitation, homogenizing dispersal, and drift were estimated at the community level and in phylogenetic groups (bins). Absolute abundance tables of bacterial and fungal communities from seed samples were used for this analysis. Two parameters, bin size and index, were set to default values, 48 and Confidence, respectively.
Results
Experimental design and sequencing results
To understand the transgenerational dynamics of seed endophytic community, we set out to monitor the changes in microbial community structures in soybean seeds across three generations (Fig. 1A). While doing so, we also monitored the BS and rhizosphere microbiomes in order to infer the contribution of these environments to the seed microbiome. Bacterial and fungal communities were evaluated based on amplicon sequencing of 16S rRNA and ITS regions, respectively, in six replicates for each sample. After pre-processing and removal of chimeric and chloroplast genome sequences, a total of 2,849,203 and 4,125,895 reads were obtained for bacteria and fungi, respectively (Supplementary Table 2). The resulting sequences were clustered into 1,033 bacterial and 108 fungal OTUs based on a threshold of 97% sequence similarity after removing OTUs with less than 3 reads.
Experimental scheme and changes in relative abundances of microbes. (A) Seeds and soil samples (bulk soil [BS] and rhizosphere soil) were collected to monitor the transgenerational dynamics of seed microbiome and effects of soil microbiome on seed. Soybean seeds (1GS, first-generation seeds) were planted into the soil to obtain from the grown plants the second-generation seeds (2GS), which were used to get the third-generation seeds (3GS). These seed samples were collected simultaneously with BS and rhizosphere samples (second-generation rhizosphere [2RS] and third-generation rhizosphere [3RS]). (B) Changes in abundances of bacteria at phylum level. (C) Changes in abundances of fungi at phylum level. Size of circles represent relative abundances.
Relative abundance of bacterial and fungal community
At the phylum level, bacterial communities were absolutely dominated by Proteobacteria in 1GS (average 99.45%) (Fig. 1B, Supplementary Fig. 1A). It was also enriched in the 2GS and 3GS (average 48.3% and 62%, respectively). Proteobacteria was also significantly enriched from BS to rhizosphere (average 29.46%, 56.67% and 40.57% at BS, 2RS, and 3RS, respectively). Bacteroidetes was observed in 2GS and 3GS (average 34.76% and 4.69%, respectively). It was also seen in the BS, 2RS, and 3RS (average 14.95%, 21.24% and 32.23%, respectively). At the genus level, Ralstonia was consistently enriched in the soybean seeds across all generations (average 86.9%, 48.2%, and 62% at 1GS, 2GS, and 3GS, respectively), while it was not found in all soil samples. Elizabethkingia, which was not detected in 1GS, was observed in 2GS (average 33.19%), 3GS (average 2.78%), and 2RS (on average, 2.71%). Paraburkholderia was also observed across all generations of soybean seeds (on average, 9.46%, 4.33%, and 0.72% at 1GS, 2GS, and 3GS, respectively). Five genera (Bradyrhizobium, Elizabethkingia, Pseudomonas, Sphingobium, and Variovorax) overlapped between the seed and soil samples.
Relative abundance of fungi showed that the seed samples consisted of two phyla; Ascomycota and Basidiomycota, and that the soil samples consisted of five phyla; Ascomycota and Basidiomycota, Chytridiomycota, Mortierellomycota, and Mucoromycota (except the unidentified) (Fig. 1C, Supplementary Fig. 1B). Among those, Ascomycota was dominant in all the seeds (on average, 95.38%, 98.62%, and 99.93% at 1GS, 2GS, and 3GS, respectively), and it was also enriched in the soil samples at phylum level (on average, 72.82%, 58%, and 87.08% at BS, 2RS, and 3RS, respectively). The next most abundant phylum was Basidiomycota in soil samples (average 13.85%, 32.9%, and 10.52% at BS, 2RS, and 3RS, respectively), while insignificant in the seed samples (average 4.49% and 1.22% at 1GS and 2GS, respectively, not detected in 3GS).
In the family and genus level, Magnaporthaceae (Pyricularia in genus level, average 85%) was dominant, accounting for 85% on average in 1GS. Pyricularia was also observed in 2GS (three out of six replicates). However, the proportion declined relatively (on average, 19.4%). The abundance of Pyricularia diminished dramatically in 3GS (on average, 1.63%), compared to other generation seeds. In 2GS, the dominant genus for individual replicates was different; Diaporthe was the most abundant genus in three out of six replicates (56.02%, 48.74%, and 39.49%) and Fusarium, Eremothecium, and Pyricularia were dominant in each of the other three replicates (81.33%, 97.9%, and 84.12%, respectively). Furthermore, Eremothecium, which was dominant in one replicate of 2GS, was present as the most dominant fungal genus in 3GS.
In the soil samples, 35 genera were observed, of which 6 genera (Actinomucor, Fusarium, Mortierella, Plectosphaerella, Thanatephorus, and Volutella) were found in both BS and rhizosphere soils. There were 16 genera in 2RS and 21 genera in 3RS, 48% of which, 12 genera, was overlapped; Actinomucor, Arthrobotrys, Cladosporium, Fusarium, Gibellulopsis, Hannaella, Lectera, Mortierella, Plectosphaerella, Pyrenochaeta, Thanatephorus, and Volutella. In addition, 9 genera were detected in both seed and soil samples; Blastobotrys, Cladosporium, Diaporthe, Fusarium, Hannaella, Plectosphaerella, Pyricularia, Trechispora, and Trichoderma.
Alpha- and beta-diversities among seeds and rhizosphere communities
Our analysis showed that alpha diversity of bacterial and fungal communities from the seeds are much lower than those of BS and rhizospheres (Fig. 2A and B). Interestingly, 2GS tends to have higher alpha diversity than other seed samples in both bacteria and fungi. Among soil samples, the highest alpha diversity was observed in BS, followed by rhizosphere samples.
Microbial diversity of the endophytic seed and soil microbiomes. The microbial alpha-diversity indices Species richness (left) and Shannon index (right) for bacteria (A) and fungal (B) communities. Beta-diversity of the seed and soil microbiome was visualized by principal coordinate analysis for bacteria (C) and fungi (D). The microbiome composition based on Bray-Curtis community dissimilarities. 1GS, first-generation seeds; 2GS, second-generation seeds; 3GS, third-generation seeds; BS, bulk soil; 2RS, second-generation rhizosphere; 3RS, third-generation rhizosphere.
The observed variations in abundances and alpha diversities among samples were supported by ordination analyses and permutational analysis of variance (PERMANOVA) (Fig. 2C and D, Supplementary Tables 3 and 4). The beta diversity of seeds and soil samples showed distinct clustering of microbial communities corresponding to the sample sites. Clustering of samples according to the sample sites were more pronounced in bacterial communities than in fungal counterparts. Although seed microbiome communities are quite distinct from those of BS and rhizosphere, there still exist considerable differences among seed endophyte communities across generations, suggesting that seed microbiomes are influenced by both pre-existing communities and external sources.
Distribution of bacteria and fungi in seeds and soils
In an attempt to understand transgenerational dynamics of seed microbiomes, we investigated the shared taxa among samples (Fig. 3). For this analysis, we have discarded low abundance taxa and focused only on the taxa with at least 0.01 of relative abundance. Our analysis showed that there are phylogenetically related group of bacteria present across all generations of seeds (Paraburkholderia and Ralstonia). Although there are other group of bacteria that are found in seeds, most of them did not persist across all three generations. Notably, many of the bacterial endophytes found in seeds except 2GS tend to be absent in soil samples. This observation suggests that those in 2GS is likely to originate from soil, transiently inhabiting the interior of soybean seeds, and that bacterial seed endophytes found across generations are well adapted to soybean seed.
Transgenerational dynamics of microbiomes in seeds and soil samples for bacteria (A) and fungi (B). Taxa with at least 0.01 of relative abundance were used for the analysis. Individual blocks represent replicates for the samples. Intensity of colors indicates relative abundances. (Continued).
In case of fungi, seed endophytes in 2GS also tend to be present in soil samples, suggesting the transient nature of these taxa. Unlike bacterial endophytes, however, fungal endophytes that persist in seeds across generations were phylogenetically scattered, and tend to be found in soil samples as well, suggesting the bidirectional influence between soil and seed in shaping the fungal seed microbiome structure.
Source and assembly mechanisms of seed microbiome
To infer the sources of seed microbiome and assembly mechanisms based on our dataset, we performed sink-source relationship analysis using FEAST (v. 0.1.0) and community assembly mechanism analysis using iCAMP (v. 1.5.12). For analysis of sink-source relationship, all replicates were merged into each sample type. We found that in 2GS, both bacteria and fungi were primarily derived from two major sources: 1GS and 2RS microbiomes (Fig. 4A). In contrast, for 3GS, where only the seed microbiota from the previous generation (2GS) and the rhizosphere microbiome of the current generation (3RS) were considered as potential sources, the fungal communities were more strongly influenced by the seed microbiota of the previous generation compared to bacterial communities. When evaluating the contribution of 1GS across both 2GS and 3GS, the microbial communities of both bacteria and fungi in later generations (3GS) were influenced by 1GS and 2GS microbiotas. Notably, bacterial communities showed a slight but detectable contribution from the rhizosphere microbiome, in addition to inheritance from prior seed generations.
Microbial sources and assembly mechanisms governing the formation of seed bacterial and fungal communities. (A) Microbial sources of progeny seed bacterial and fungal communities. (B–D) Assembly mechanisms affecting the formation of seed bacterial and fungal communities. (B) Community-level relative importance of assembly mechanisms in the bacteria and fungi. (C) Community-level relative importance of assembly mechanisms in each seed generation (top, bacteria; bottom, fungi). (D) Phylogenetic bin-level relative contributions of assembly mechanisms in the bacteria and fungi of each seed generation. Bars indicate the relative importance of assembly mechanisms (homogeneous selection [HoS], blue; heterogeneous selection [HeS], light blue; dispersal limitation [DL], yellow; homogenizing dispersal [HD], orange; ecological drift [DR], red).
To investigate the ecological processes underlying seed microbiota assembly, we analyzed community assembly mechanisms from 1GS to 3GS (Fig. 4B). In bacterial communities, ecological drift emerged as the dominant process, accompanied by simultaneous contributions from homogeneous selection—where similar microbial community structures are formed across different seeds—homogenizing dispersal, and dispersal limitation. These results suggest that bacterial seed microbiota are shaped by a complex interplay of stochastic and deterministic processes. In contrast, fungal communities were primarily governed by homogeneous selection, indicating stronger deterministic forces structuring fungal assemblages across generations.
Generation-specific analysis revealed dynamic shifts in assembly mechanisms over time (Fig. 4C). At the whole-community level, bacterial communities showed increasing influence of homogeneous selection and drift across generations. Homogenizing dispersal played a more prominent role during the formation of 1GS microbiota. In contrast, fungal communities consistently exhibited dominance of homogeneous selection across all generations. However, when examined at the bin level, the dominant assembly processes varied among bins and shifted across generations, suggesting that different phylogenetic groups respond differently to ecological forces over time.
When further analyzed at the phylogenetic bin level (Fig. 4D), we found that in bacterial communities, bin 2 (composed largely of unclassified phyla) and bin 12 (comprising Pseudomonadales and Enterobacterales within Gammaproteobacteria) were strongly influenced by homogeneous selection. Bins affiliated with Acidobacteria, including bin 28 and bin 29, were primarily structured by dispersal limitation, while most of the remaining bins appeared to be governed predominantly by ecological drift. In fungal communities, bin 4 (containing largely unclassified taxa) was strongly shaped by homogeneous selection, while bin 1 (containing Saccharomycetales and Pleosporales) and bin 5 (mainly affiliated with Basidiomycota) exhibited weaker but detectable signals of homogenizing dispersal.
Discussion
Our study revealed significant compositional shifts in both bacterial and fungal communities across three generations of soybean seeds, supporting the previous studies for dynamic seed microbiome inheritance (Abdelfattah et al., 2021; Shade et al., 2017). The bacterial community underwent a marked transition from near-complete dominance of Proteobacteria in 1GS (99.45%) to a more diverse structure in the second (48.3%) and third (62.0%) generations. At the genus level, while Ralstonia remained consistently abundant across generations, novel taxa such as Elizabethkingia emerged in the second generation and persisted. This pattern suggests that the seed microbiome undergoes significant restructuring during transgenerational transmission, consistent with previous observations of dynamic microbial inheritance in plant seeds (Abdelfattah et al., 2023). The initial dominance of Proteobacteria, particularly Ralstonia, may represent a founder effect where specific bacterial lineages establish initial colonization of seed environments. The subsequent diversification in later generations indicates that the seed microbiome is not simply a result of static inheritance but rather a dynamic system subject to ecological succession, as suggested by recent reviews on microbial inheritance mechanisms (Berg and Raaijmakers, 2018; Shade et al., 2017).
Fungal communities displayed even greater variability, with Pyricularia declining dramatically from 85.0% relative abundance in 1GS to 1.63% in the third generation. This decline was accompanied by increased representation of genera such as Diaporthe and Eremothecium, indicating potential susceptibility of fungal communities to environmental filtering or plant-mediated selection. Notably, 2GS exhibited considerable variation across replicates, suggesting a transitional phase in fungal community assembly.
The consistently lower alpha diversity observed in seed communities compared to soil environments reveals the selective nature of the seed microenvironment. This finding indicates that seeds act as biological filters, supporting only a subset of the microbial diversity present in surrounding soil environments, a phenomenon well-documented in seed microbiome studies (Nelson, 2018; Shade et al., 2017; War et al., 2023). The observation that 2GS exhibited the highest alpha diversity among all seed samples is particularly intriguing and suggests that the second generation represents a critical transition point in microbiome assembly.
This diversity pattern may reflect the complex interplay between maternal inheritance and environmental acquisition, as described in the framework of microbial inheritance by Abdelfattah et al. (2023). The second generation may represent a “mixing zone” where maternal and soil-derived microbiota compete or interact, resulting in temporary diversification. The subsequent decline in diversity in 3GS suggests the establishment of stronger host-mediated filtering mechanisms or competitive exclusion processes. From a functional perspective, reduced microbial diversity may reflect the selection of taxa that confer specific benefits, such as enhanced germination, stress tolerance, or early growth promotion.
Analysis of transgenerationally persistent taxa revealed phylogenetic patterns in microbial inheritance. Bacterial genera such as Paraburkholderia and Ralstonia were consistently detected across all generations but were absent from soil samples, implying adaptation to the seed environment and possible specialization for vertical transmission. These findings align with previous observations of Paraburkholderia and Ralstonia as specialized endophytes in various plant species (Compant et al., 2010; Sohaib et al., 2024). In contrast, fungal taxa that persisted across generations were phylogenetically diverse and often present in soil, suggesting that fungal colonization may be more opportunistic. This indicates differing ecological strategies: bacterial persistence may rely on co-evolutionary specialization, while fungal persistence may depend more on environmental flexibility and repeated colonization events. These patterns have implications for understanding the evolution of plant-microbe interactions. Bacterial endophytes that persist across generations may form stable, mutualistic relationships with hosts, potentially influencing host fitness and microbial success. The more opportunistic nature of fungal associations may reflect broader ecological roles, such as decomposers or opportunistic colonizers.
Source tracking using FEAST (Shenhav et al., 2019) revealed distinct patterns of microbiota origin. In 2GS, both bacterial and fungal communities were primarily derived from 1GS and the rhizosphere. In contrast, third-generation fungal communities were more strongly influenced by the seed microbiota of the preceding generation, while bacterial communities showed greater contributions from the rhizosphere, indicating differing transmission strategies between microbial groups. These patterns suggest that bacterial seed endophytes rely more heavily on environmental re-acquisition, whereas fungi are more stably inherited via vertical transmission. Consistent with this, community assembly analysis using iCAMP (Ning et al., 2020) indicated that bacterial communities were shaped by multiple processes, including ecological drift, homogeneous selection, homogenizing dispersal, and dispersal limitation. In contrast, fungal communities were predominantly governed by homogeneous selection, implying stronger deterministic structuring. The dominance of homogeneous selection in fungal communities suggests that consistent selective pressures, potentially imposed by the host seed environment, lead to convergent fungal community structures across generations. Bacterial communities, in contrast, appear to be more influenced by stochastic processes and environmental variability.
Temporal analysis of assembly mechanisms revealed generational shifts in community structuring forces. In bacterial communities, the influence of homogeneous selection and ecological drift increased over time, while homogenizing dispersal was more prominent in the first generation. This trend suggests a progression from environmental dependence toward host filtering and selection. Fungal communities were consistently shaped by homogeneous selection across all generations, reflecting more stable assembly patterns and possibly narrower ecological niches. However, analysis at the phylogenetic bin level revealed variation in dominant processes among bins, suggesting that microbial groups differ in their ecological responses and transmission strategies. These findings suggest that seed microbiomes become increasingly structured and predictable over generations, especially for bacterial communities. This may reflect the evolution of host traits that selectively recruit or exclude specific microbes, or the stabilization of microbial networks that resist colonization by transient taxa.
Taken together, this study provides a comprehensive framework for understanding the transgenerational dynamics of seed microbiomes in a major crop species. By demonstrating that both vertical and horizontal processes shape seed-associated microbial communities, our findings highlight the complexity of plant-microbe co-assembly. The contrasting patterns between bacterial and fungal communities point to distinct ecological roles and transmission strategies. Bacteria may confer flexibility and environmental responsiveness, while fungi may provide more stable, functionally specialized associations. These insights have direct implications for microbiome-informed crop management and breeding.
Understanding how and when to intervene in microbiome assembly—particularly during critical transition points such as the second generation—could improve the design of microbial inoculants or breeding programs that select for favorable plant-microbe interactions. The identification of persistent, seed-adapted microbial taxa offers targets for further functional characterization and biotechnological application.
Notes
Conflicts of Interest
Junhyun Jeon, a contributing editor of the Plant Pathology Journal, was not involved in the editorial evaluation or decision to publish this article. All remaining authors have declared no conflicts of interest.
Acknowledgments
This work was supported by grants from the Rural Development and Administration, Republic of Korea (PJ0157682021) and the National Research Foundation of Korea (RS-2024-00407104 to J.J. and RS-2022-NR072199 to H.K.).
Electronic Supplementary Material
Supplementary materials are available at The Plant Pathology Journal website (http://www.ppjonline.org/).