Steering root-associated microbiomes via direct and soil legacy effects of neighbors Authors: Mireadili Kuerban, Sofia I F Gomes, Ozlem Erol, T Martijn Bezemer Published: 2026 ABSTRACT AbstractBackgroundRoot-associated microbiomes are widely assumed to be primarily shaped by the host plant. However, other co-occurring plants can also influence the composition of this microbiome. This can be either directly, by stimulating or suppressing microbial recruitment of the focal plant, or indirectly, through soil legacies that shape the microbial pool encountered by the focal plant during establishment. The relative importance of these neighbor effects shaping the focal plant microbiome and the consequences for focal plant growth remain poorly known. Here, we challenge the paradigm that the host plant identity is the main driver of root-associated bacterial and fungal communities.ResultsUnder experimental conditions, we show that the influence of the host plant on root-associated microbiomes is dramatically reduced in the presence of neighbors and their soil legacies. Remarkably, root-associated bacterial communities were primarily shaped by co-occurring neighbor plants, while fungal communities were mainly determined by soil legacies from previously grown plants. Furthermore, culturable microbial communities (NatComs) isolated from the roots of focal plants differed greatly depending on neighbor identity and soil legacies. Inoculating new focal plants with these communities caused strong responses in growth that varied greatly depending on the history of these isolated communities, highlighting the importance of neighbor effects. Finally, we show that metabolites produced by these microbial communities are one mechanistic pathway through which neighbors influence plant growth.ConclusionsOur results stress the often-overlooked ecological and functional importance of both direct and indirect effects of neighbor plants in shaping root-associated microbiomes. By integrating neighbor effects with host identity, this work provides a starting point for understanding how neighboring plants contribute to shaping plant communities, with potential implications for sustainable agriculture. Video AbstractSupplementary InformationThe online version contains supplementary material available at 10.1186/s40168-026-02429-2.Keywords: Root microbiome, Neighbor effect, Soil legacy, Community assembly, Plant–microbe interactions Introduction Plants interact with specific microorganisms that live inside or in the close vicinity of their roots [1]. These root-associated microbes can confer specific functions to their hosts and modulate plant growth, health, and productivity [2–5]. Recent studies have highlighted that the assembly of root-associated microbiomes is a multistep process, largely determined by host plant identity, with a distinct contribution from the soil [6, 7]. Under natural conditions, however, plants rarely grow in isolation. Instead, they interact in direct and indirect ways with neighboring plants both above-ground and below-ground, e.g., through the shared soil substrate in which they grow. As these neighboring plants also shape the soil microbiome, this raises the important question of to what extent co-occurring neighbor plants, beyond focal plant identity, determine the microbial community associated with the roots of a focal plant. Furthermore, it is important to understand the implications of these neighbor effects on the growth of a focal plant via their influence on the microbiome. The research field of plant-soil feedback examines how changes in the soil induced by the growth of a plant affect the survival and performance of plants establishing later in the same soil [8, 9]. An important assumption is that after a period of growth a plant has changed abiotic properties of the soil such as nutrient availability, and biotic properties such as the composition of the soil microbiome, and that these legacies subsequently influence the growth of a succeeding plant, either positively or negatively [10, 11]. Over time, the later-growing plant also reshapes the soil microbiome, leading to fluctuations in both soil microbial communities and plant growth over time [12]. Interestingly, recent work has shown that members of the microbial community that were stimulated by the first plant can be detected in the root-associated microbiome of the later-growing plant for extended periods [13]. Hence, plant-soil feedback research predicts that plant species-specific microbial soil legacies may influence the assembly of root-associated microbiomes in plants that subsequently grow in the same soil. In contrast to plant-soil feedback, the field of plant-plant competition research focuses on direct interactions between plants growing simultaneously in the soil. Interestingly, recent work has indicated that neighbor plants can also influence the root-associated microbiomes of co-occurring plants [14–17]. The direct effect of a neighbor plant occurs after the neighbor has established itself and at that time can begin to influence the focal plant. Indirect neighbor effects, through soil legacies, expose the focal plant already from germination to different initial microbial communities. Therefore, indirect neighbor effects via microbial legacies may exert a stronger influence on the root-associated microbiome of focal plants than direct neighbor effects. Despite existing investigations, our understanding of the relative importance of soil legacy effects and the direct effects of neighboring plants on the root-associated microbiome of focal plants remains nascent. An important gap in our understanding of plant-soil microbiome interactions is how the direct (via co-occurrence) and indirect (via soil legacies) effects of neighboring plants shape the bacterial and fungal communities associated with the roots of focal plants. Bacteria have short generation times, they are likely to respond rapidly to environmental changes, while fungi often grow more slowly and persist longer [12, 13]. Hence, soil legacies, which represent longer-term effects, may have a stronger influence on fungal communities, while the more transient effects of neighboring plants may be more evident in bacterial communities. Another critical gap in our understanding concerns the consistency and specificity of neighbor-mediated effects on the composition of the root microbiome of the focal plant. Each host plant species uniquely influences its root-associated microbiome by stimulating a specific subset of microbes [6, 7], and these microbes can spill over to the roots of neighboring plants [18]. This suggests that there are specific microbial species associated with the identity of a neighbor. The scarce information that is published further suggests that these neighbor effects on the focal plant microbiome may vary between different focal plant species [19]. For example, the presence of common yarrow, Achillea millefolium, as a neighbor decreased the relative abundance of Rhizoctonia solani in two focal plant species, but not in two others [20]. It therefore remains unclear whether a given neighbor plant consistently enriches or depletes particular microbial taxa across different focal plant species, and whether this direct neighbor effect resembles the indirect effects mediated through soil legacies. Testing functional consequences of root-associated microbiomes for plant growth remains a significant challenge, because these microbiomes are inherently diverse and complex, comprising a multitude of microbial taxa that interact with one another and with their host [21]. Replicable and simplified communities are often obtained in standardized systems through bottom-up approaches to study how they interact with host plants [22]. Although reductionist approaches using single microbial isolates or highly simplified synthetic communities have provided important mechanistic insights, they often lack ecological realism and fail to capture emergent properties of natural microbial assemblages [23]. Cultured natural communities (NatComs) are increasingly becoming alternative promising tools to manipulate complex host-microbe interactions [23, 24]. While NatComs necessarily represent a culturable subset of the root-associated microbiome under the selected culture conditions, a previous study showed that they do resemble the initial composition of the microbiome [24], and hence can provide a tractable system for assessing potential ecological functional consequences of microbiome variation. Importantly, microbial communities can influence plant performance not only through their taxonomic composition but also through the collective production of metabolites, including hormones, signaling compounds, and other bioactive molecules [25, 26]. These community-level metabolomes may represent an integrative mechanism by which complex microbial assemblages affect plant growth [27, 28]. Approaches such as NatComs provide an opportunity to explore such metabolite-associated effects while retaining key aspects of community complexity. Here, we determine how the root-associated microbiomes of a set of focal plant species that commonly co-occur in diversified grasslands [29] are influenced by direct effects of neighbor plants and by indirect effects of these plants through soil legacies (Fig. 1). We hypothesize that (i) the primary determinant of the assembly of the root-associated microbiome is the host plant itself, but that direct neighbor interactions and soil legacies modify this; (ii) soil legacies, which are formed over longer time periods, will have a stronger influence on the assembly of fungal communities that are generally slow growing, while the direct effects of neighboring plants are more transient, and will be more evident in bacterial communities that are fast growing; (iii) each neighbor plant species will consistently enrich or deplete particular microbial taxa across different focal plant species; (iv) the direct and soil legacy effects of neighbor plants on the assembly of root-associated microbiomes of focal plants are functionally important for plant growth and this is driven by the metabolites produced by the root-associated microbiomes. To test these hypotheses, we combine a two-Phase-PSF experiment with root-associated microbiome profiling and a complementary inoculation assay using cultured NatComs (Fig. 1B, C and Supplementary Fig. S1). Overview of the main factors influencing the recruitment of root-associated microbiomes (A) and experimental setup (B, C). A Three key factors determine the composition of the root-associated microbiome community of a focal plant when plants grow together: (1) focal effect: different host plants recruit distinct microbiomes through differences in the host immune system, root morphology, chemical features and rhizodeposition [20]; (2) soil legacy effect: the soil acts as a “microbiome bank” providing microbes available for plant recruitment; and (3) neighbor effect: neighboring plants can shape the root-associated microbiome of the focal plant since heterospecific neighbors recruit different microbiomes from the available pool and microbes from these microbiomes can be incorporated into the microbiome of the focal plant. B, C In Phase I, four plant species were planted as monoculture to create distinct microbial soil legacies. In Phase II, each pot within each replicate was then planted with two plants from the same species or two plants belonging to different plant species in soil conditioned by each respective plant. After another 8 weeks, bacterial and fungal NatComs were isolated from the roots of both plants in each pot separately. In Phase III, seedlings were grown in sterile soil inoculated with NatComs that were isolated from plants belonging to the same species or added sterilized metabolite solutions collected from the different bacterial NatComs. Here only for one of the four species the experimental setups are shown. Further details are presented in Supplementary Fig. S1. The abbreviations are Cichorium intybus (CI, forb, blue), Phleum pratense (PP, grass, yellow), Trifolium pratense (TP, legume, red), and Trifolium repens (TR, legume, green) Results Neighbors and soil legacies alter root-associated microbiomes At the end of the first Phase of the experiment the monocultures of the four plant species had significantly different soil microbiomes, and hence each plant species created a unique bacterial and fungal legacy in the soil (Supplementary Fig. S2). In the second Phase, we grew the four species in different pairwise combinations in the soils with the four legacies. We first examined how much of the variation in the root-associated microbiome composition of each focal plant was explained by the identity of the focal plant. Focal species identity explained 70% and 43% of all variation in the bacterial and fungal communities, respectively (Fig. 2A and Supplementary Table S1). In this case, the plants were grown with a neighbor that belonged to the same species and in soil with a legacy created by plants belonging to the same species (Fig. 2A and Supplementary Table S1). While each plant species recruited a unique bacterial and fungal root-associated community (‘focal effect’ in Fig. 2B), this clear species-level structuring became progressively less pronounced when soil effects, neighbor effects, and their combination were incorporated into the design, as reflected by increased overlap among species clusters across the panels (Fig. 2B, C and Supplementary Fig. S3). Overall, for each focal plant species, microbial communities became more variable when plants were grown in heterospecific soils or with heterospecific neighbors, resulting in lower within-species similarity compared to conspecific soil and neighbor conditions (Fig. 2C). Hence, adding soil legacies, neighbor effects, or both substantially reduced the influence of the focal plant on its root-associated microbiome (Fig. 2A). Effects of focal plant species, soil legacies and neighbor plants on root-associated microbial communities. A Variation in bacterial and fungal community composition in roots explained by focal plant identity (grey), soil legacies (dark grey), and neighbor identity (black). The arrows indicate the decline in variation that is explained by the focal plant. The statistical results given in panels A and B are based on redundancy analysis (RDA) and variance partitioning. Detailed statistical results are presented in Supplementary Table S1. B Compositional variation of bacterial and fungal communities associated with the roots of each plant species based on principal component analysis (PcoA). Focal effects disappear when soil effects and neighbor effects are added. By adding these effects, the number of samples also increases. (C) Mean similarity (± 1 SE) among microbial communities for each focal plant species when grown with a neighbor of the same species in conspecific soil (Focal), versus in soils with legacies and with different neighbors. Shown are inverse Bray-Curtis dissimilarities. For the focal effect, n = 3, and for soil legacies (soil), neighbors (neigh), and their combined effects (soil + neigh), n = 9. Statistical values shown above each graph are the results of one-way ANOVA. Significant differences between the focal effect and the other effects are shown and were calculated based on planned contrasts within the same one-way ANOVA model framework. D Variation in bacterial and fungal community composition in roots of each focal plant species explained by soil legacies (dark grey), neighbor identity (black), and interactions between soil legacies and neighbor identity (white). The abbreviations are Cichorium intybus (CI, forb, blue), Phleum pratense (PP, grass, yellow), Trifolium pratense (TP, legume, red), and Trifolium repens (TR, legume, green). Detailed statistical results are presented in Supplementary Table S2. *, **, and *** represent significant levels at 0.05, 0.01, and 0.001, respectively This is also visible when examining the specific effects of neighbors and soil legacies. For each of the four plant species a considerable part of the variation in their microbiome was explained by neighbor and legacy effects (Fig. 2D and Supplementary Table S2). Interestingly, neighbor effects explained substantially more of the variation in bacterial communities associated with the roots of a focal plant than soil legacy effects (Fig. 2D). In contrast, for fungal communities, soil legacy effects explained most of the variation (Fig. 2D). This outcome remained the same when the treatments were analyzed as two-level factors, i.e. soil legacy type (focals grown in “own” soil versus in soil conditioned by the other plant species), and neighbor type (focal plants grown together with another plant of the same species versus a different plant species). Neighbor type explained a larger part of the variation in root-associated bacterial communities (4.78 to 20.95%) while soil legacy type explained a larger part of the variation in root-associated fungal communities (0.26 to 2.26%, Supplementary Fig. S4 and Table S3). The effects of neighbor species and soil legacies on the root-associated microbiome composition of the focal plant varied and clearly depended on neighbor identity, soil legacy identity and focal plant identity (Supplementary Figs. S5 and S6). Consistency of neighbor and soil legacy effects Subsequently, we examined whether neighbor plants and soil legacies of other plants consistently enriched or depleted specific microbes in the roots of different focal plant species, and to what extent these neighbor and soil legacy effects overlap (Fig. 3A and Supplementary Fig. S7). When the three other plant species were grown with Cichorium intybus as neighbor, compared to the three plants grown with another conspecific plant in the same soil, 19 bacterial ASVs and one fungal ASV were enriched, while one bacterial ASV was consistently depleted (Fig. 3B). These 19 bacterial ASVs were taxonomically classified into five bacterial phyla Pseudomonadota (15 bacterial ASVs), Acidobacteriota, Bacteroidota, Chloroflexi and Myxococcota, and 16 bacterial genera such as Steroidobacter, Pseudomonas, Massilia, Rugamonas, Aminobacter and Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium. Cichorium intybus as a neighbor consistently enriched one fungal ASV assigned to the genus Arthrobotrys across focal plants (Fig. 3B). Notably, only four of these enriched indicator taxa (two Pseudomonas taxa, Massilia and Rugamonas) overlapped with Cichorium intybus host-specific indicators (Supplementary Table S5). When the other plant species acted as neighbor, far fewer bacterial and fungal ASVs were consistently enriched or depleted in the different focal plants (Fig. 3B and Supplementary Table S4). Specifically, Phleum pratense as neighbor consistently depleted two bacterial ASVs (Massilia and Anaerolineae) in the roots of the other three plant species, while it enriched one bacterial ASV (Rhizobacter) and one fungal ASV (Poaceascoma). These two enriched ASVs overlapped with Phleum pratense host-specific indicators (Supplementary Table S5). When Trifolium pratense was grown as neighbor, four bacterial ASVs classified into Ohtaekwangia, Niastella, Hydrogenophaga, and Steroidobacter were consistently enriched in the roots of the other three plant species. According to FAPROTAX-based functional predictions, all these enriched and depleted microbes appear to be associated with a range of different functions related to nutrient cycling (e.g. nitrate denitrification, aerobic chemoheterotrophy, methanotrophy, ureolysis) and plant growth (Supplementary Table S6). Consistent effects of neighbor plants and soil legacies shaping root-associated bacteria and fungi. A Conceptual illustration of neighbor and soil legacy impacts on the root microbiome of a focal plant species. Consistent neighbor effects are observed when a neighboring plant species (e.g., Phleum pratense), grown alongside different focal species in soils conditioned by the focal species, consistently enriches or depletes specific taxa (indicator microbes) in the root microbiome of these focal plants. Consistent soil legacy effects occur when different plant species, grown alongside a plant of the same species in soils previously conditioned by another species (e.g., Phleum pratense), consistently recruit the same indicator microbes associated with that specific soil legacy, regardless of focal species identity. Detailed comparison groups are shown in Supplementary Fig. S2. B Bacterial and fungal taxa identified as indicators for neighbor and soil legacy effects for Cichorium intybus (CI, blue), Phleum pratense (PP, yellow), Trifolium pratense (TP, red), and Trifolium repens (TR, green). Indicator taxa shown are significantly enriched or depleted (p < 0.05) and are present in more than 50% of the analyzed samples. Circle size represents the log fold change. Detailed statistical results are presented in Supplementary Table S4 Growing in a particular soil legacy also consistently enriched or depleted particular ASVs in the root-associated microbiome of focal plants, as shown by the indicator taxa analysis (Fig. 3B and Supplementary Table S4 and S6). Interestingly, soil conditioned by the plant Phleum pratense exhibited the strongest and most persistent legacy. When other plant species were grown in soil previously conditioned by Phleum pratense, 13 bacterial ASVs (including members of the genera Acidibacter and Lechevalieria, and the families Comamonadaceae, BIrii41 and A4b) and seven fungal ASVs (including members of the genera Cladorrhinum, Ophiospharella, Poaceascoma and the family Lindgomycetaceae) were consistently enriched in all these plant species (Fig. 3B and Supplementary Table S4). In addition, one fungal ASV (Dactylonectria) was consistently depleted in all plant species when they were grown in soil with a legacy of Phleum pratense (Fig. 3B and Supplementary Table S4). Interestingly, 15 of these 21 soil legacy indicator taxa overlapped with Phleum pratense host-specific indicators (Supplementary Table S5). Importantly, there was virtually no overlap in indicator taxa enriched or depleted by neighbors and by soil legacies, except for two taxa for Phleum pratense (Fig. 3B and Supplementary Table S4), highlighting that the direct effects of neighbor plants and their soil legacy effects on the microbiome of a focal plant do not overlap. Neighbors and soil legacies influence plant growth Next, we investigated the relative importance of direct effects of neighbor plants and soil legacy effects on focal plant growth. In conditions that plants grow with neighbors, plant-plant competition typically greatly influences plant growth, as space, light, water, and nutrients are inherently constrained [30, 31]. Hence, we expected that neighboring plants would influence focal plant growth more than soil legacies of other plants. For three of the four plant species, Phleum pratense, Cichorium intybus and Trifolium pratense, the biomass was significantly influenced by neighbor plant identity and soil legacy origin (Fig. 4A and Supplementary Fig. S8 and Table S7). However, most of the variation in focal plant growth was explained by soil legacies rather than by neighbor plants (Fig. 4A). Interestingly, the growth of Trifolium repens was neither affected by soil legacies nor by neighbors (Fig. 4A and Supplementary Fig. S7), highlighting that these responses are species-specific. Assessing correlations with root microbiomes indicated that the biomass of Phleum pratense and Trifolium pratense was significantly correlated with the composition of their root-associated bacterial communities (Fig. 4B and Supplementary Table S8), highlighting a potential link between bacterial community structure and plant performance. In contrast, there was no relationship between biomass and fungal root-associated communities for any of the species tested (Fig. 4B and Supplementary Table S8), indicating that under the growth conditions that we tested, bacterial, rather than fungal root-associated communities may be more influential in mediating the observed plant growth responses. Analysis of root-associated microbial communities and bacterial communities in NatComs, and their effects on plant growth. A Variation in shoot biomass of focal plants in Phase II explained by soil legacies (dark grey), neighbors (black), and the interaction between soil legacies and neighbors (white) estimated with a two-way ANOVA model. Detailed statistical results are presented in Supplementary Table S5. Generally, soil legacies explain more variation than neighbor plant identity. B Variance in bacterial and fungal communities associated to the roots of each focal plant explained by the shoot biomass of focal plants in Phase II, estimated with an RDA model. Detailed statistical results are presented in Supplementary Table S6. Variation in bacterial, but not fungal communities, is explained by the shoot biomass of two species Phleum pratense and Trifolium pratense. C Relative abundance of the top 30 ASVs in bacterial NatComs. Most of these top abundant ASVs originate from roots and/or soil. D Principal coordinates analysis of the bacterial communities cultivated in the NatComs showing a focal species effect. E Variation in shoot biomass of the inoculated plants explained by the identity of the NatComs estimated with a one-way ANOVA model. Detailed statistical results are presented in Supplementary Table S8. F Shoot biomass of inoculated plants in Phase III. The abbreviations are Cichorium intybus (CI, forb, blue), Phleum pratense (PP, grass, yellow), Trifolium pratense (TP, legume, red) and Trifolium repens (TR, legume, green). *, **, and *** represent significant levels at 0.05, 0.01 and 0.001, respectively Effects of isolated NatComs on plant growth So far, using various analytical approaches, we provide several lines of evidence that other plants, as neighbors and via soil legacies that they create, greatly influence root-associated microbiomes of focal plants, and that these changes in the composition of the microbiome, in particular the bacterial community composition, are related to plant growth. However, this evidence stems from plants grown with neighbors and in soils with different legacies and these confounding factors also influence plant growth. To test the functional importance for plant growth of changes in the composition of bacterial communities associated with a focal plant, resulting from direct effects of other plants and from soil legacies of other plants without confounding factors, we isolated culturable bacterial NatComs from roots of the four focal plant species exposed to the different neighbors and soil legacies. In the bacterial NatComs, 195 bacterial strains were detected using 16S amplicon sequencing, taxonomically classified into 63 bacterial genera. Most of these bacterial ASVs were also detected in soil or root samples (Fig. 4C). The NatComs isolated from the four focal plant species differed significantly (Fig. 4D). We then initiated a community inoculation experiment with seedlings growing in sterilized soil to examine the consequences of inoculating different bacterial communities for plant growth. For each of the four tested plant species, the bacterial communities used for inoculation were isolated from roots of conspecific plants that had been grown with either a conspecific neighbor in different legacy soils, with a heterospecific neighbor in soil conditioned by the focal plant itself, or with a heterospecific neighbor in soil conditioned by the neighbor plant. So, all test plants received bacterial NatComs isolated from plants of the same species, but these communities had been shaped by different combinations of soil legacies and neighbor plants. Remarkably, the shoot biomass of the test seedlings varied greatly and depended strongly on the bacterial NatCom that was inoculated (Fig. 4F and Supplementary Fig. S9). Between 30 and 60% of the variation in biomass in Phase III was explained by the identity of the bacterial NatComs (Fig. 4E and Supplementary Table S9). For Cichorium intybus, in all inoculated communities its biomass was lower than that of the control plants that were grown in sterilized soil. A similar pattern was observed for Trifolium pratense. In contrast, for Phleum pratense, inoculation with most bacterial communities increased plant growth, while for Trifolium repens, increases and decreases in biomass were observed after inoculation with bacterial communities, when compared to the control (Fig. 4F and Supplementary Fig. S9). We also repeated the experiment with fungal NatComs that were extracted from plant roots using the same approach and set-up as described for bacterial NatComs. The shoot biomass of the test seedlings also varied depending on the identity of the fungal NatCom that was added (Supplementary Fig. S10). Metabolic profiles of isolated bacterial NatComs and their effect on plant growth Subsequently, we examined a potential mechanism explaining why the bacterial NatComs influenced the biomass of the host plant so greatly. We focused on the effects of metabolites produced by the bacterial NatComs and did this for three of the examined plant species. We analysed the metabolite composition of the bacterial NatComs using Ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS). The bacterial NatComs produced more than 1,200 metabolites, and these metabolites differed significantly depending on which host plant the bacterial communities were isolated from (Fig. 5A and Supplementary Fig. S11). To investigate whether the isolated bacterial NatComs explained metabolic compositional differences in NatComs, we performed a Mantel test and Procrustes analysis between metabolic data and the composition of the bacterial NatComs based on sequencing data. Both analyses showed that isolated bacterial NatComs and metabolic composition were significantly correlated, suggesting that each bacterial NatCom produces a specific metabolic profile (Supplementary Table S10). Metabolic profiles of isolated bacterial NatComs and their effect on plant growth. A Partial least-squares discriminant analysis (PLS-DA) of the metabolomics of different bacterial NatComs. R2, fitness of the model, and Q2, predicted variance, are presented. B Correlation between shoot biomass after inoculating with bacterial NatComs and shoot biomass when plants were exposed to metabolites of bacterial NatComs for Phleum pratense. The abbreviations are Cichorium intybus (CI, forb, blue), Phleum pratense (PP, grass, yellow), and Trifolium pratense (TP, legume, red) To examine whether the effects of the inoculated bacterial NatComs on host plant growth were driven by the metabolites produced by these communities, seedlings were grown in sterilized soil with either metabolites collected from sterilized bacterial NatComs or live bacterial NatComs. In the metabolite addition experiment, shoot biomass of the test seedlings varied greatly and depended on the treatment (Supplementary Fig. S12). Interestingly, the biomass of Phleum pratense exposed to the live bacterial NatComs correlated strongly with the biomass of plants exposed to the metabolites only (Fig. 5B), suggesting that for this plant species, metabolites are likely responsible for the observed bacterial NatCom effects on plant growth. The growth of Cichorium intybus and Trifolium pratense was more variable and there was no consistent correlation between performance with live and sterilized inoculum (Supplementary Fig. S13), indicating species-specific mechanisms or the involvement of factors beyond purified metabolites. Discussion Our results show that neighbor plants and their soil legacies strongly reduce the influence of the host plant on shaping its own root-associated microbial community, challenging the prevailing view that host plant identity is the dominant driver influencing the root-associated microbiome [32–35]. While previous studies have also shown that a neighbor plant can influence the root microbiome of a focal plant [14–18], our study expands on this by examining four different plant species across three functional groups, revealing the extent of these interactions. A key finding of our study is that, under the experimental conditions we tested, neighboring plants directly influence the composition of the bacterial community associated with the roots of the focal plant, while the microbial soil legacy effect of other plants is the key factor driving the assembly of fungal communities associated with the roots of the same focal plant. The relative importance of host identity, neighbor effects, and soil legacies may vary with spatial scale, particularly in natural systems where root systems are larger, soil heterogeneity is greater, and interactions accumulate over longer timescales. Nevertheless, as we also show in this study, root-associated microbiomes are closely linked with host plant growth, these findings may have implications for sustainable agriculture and microbiome engineering. Importantly, our work shows, for these grassland species that are common in Europe, that direct effects of neighbor plants are more important than the indirect effects that occur via soil legacies in influencing plant growth via their influence on bacteria. These belowground neighbor-plant effects have received increased attention in the context of intercropping [36–38]. If our results can be extrapolated to field situations, they suggest that when the aim is to change bacterial communities for a particular plant species, intercropping is more likely to be effective than selecting particular crop rotations, while the opposite appears to be true for fungal community changes. A broadly accepted concept in the field of soil biology is that fungi grow much slower than bacteria [39, 40], and as a result, that fungi are more stable and less influenced by other environmental factors, while bacterial communities respond more rapidly to such changes [41–43]. A previous study with grassland species reported that bacterial soil legacies faded away rapidly, and that bacterial communities in the soil were influenced strongly by the new plant that was growing in the soil, while soil fungal legacies were detectable for months after planting the current species, and less affected by this current plant species [13]. In our study, we show that fungal communities associated with the roots of a focal plant are more influenced by the soil legacy than by the neighboring plant growing alongside the focal plant. This finding highlights that soil is likely the main reservoir from which focal plants recruit root-associated fungi, even when they grow together with plants of different species. In our study, neighbor plants enriched or depleted specific microbes in the roots of different plant species. Similarly, we observed that roots of different plant species growing in soils with a particular legacy also contained the same microbes. While neighbor-plant effects on the microbiome of a focal plant can thus be both direct and indirect, via the soil legacy, we found that there was virtually no overlap in the indicator taxa that were enriched or depleted directly by neighbor species and indirectly via soil legacies left in the soil by these plant species. These results provide evidence, in a different way, that neighbor-plant effects and soil-legacy effects of a plant on the root-associated microbiome of a plant belonging to another species fundamentally differ. A plausible explanation for this lack of overlap is that direct and indirect neighbor effects operate through distinct mechanisms and timescales. Direct neighbor effects occur during simultaneous plant growth and may reflect short-term processes such as root-root contact, microbial spillover between adjacent root systems [18, 20], or changes in root exudation driven by plant-plant competition [44]. In contrast, soil-legacy effects reflect longer-term conditioning of the soil environment by a preceding plant, consistent with plant-soil feedback processes that alter the abundance and composition of microbial propagules or nutrient availability [13]. Further work should reveal in more detail the consequences of these direct and indirect effects of neighboring plants on the root-associated microbiome of a focal plant, but our study highlights that in microbiome research these effects of other plants cannot be ignored. Importantly, our results also highlight that the influence of a neighbor plant is not random. The consistent effect of a specific neighboring plant on the abundance of certain taxonomic groups in the roots of a set of different focal plants could result from direct transmission through root contact [18], or from the production of specific root exudates that favor or deter certain microbial communities [45]. Several mechanisms can drive these specific effects, and the exact causes will likely differ for each plant-plant and plant-soil legacy combination. For example, specific chemical compounds in root exudates can selectively influence the assembly of root-associated microbiota [46] as microbes differ greatly in their tolerance to and preference for specialized exudate compounds [47]. While multiple mechanisms are likely involved, and their relative importance may differ among plant-plant and plant-soil legacy combinations, our findings highlight that both direct and indirect effects of neighboring plants should be explicitly considered in studies of root-associated microbiomes. The microbes that were consistently enriched or depleted by a specific neighbor plant species or its soil microbial legacy are associated with a wide range of putative functions. For example, having Cichorium intybus as a neighbor enriched the relative abundance in the roots of the other three focal plant species of the fungus Arthrobotrys which has been described as a nematode-trapping fungus in the rhizosphere of plants [48]. The same plant species enriched bacteria such as Massillia and Variovorax which are described in the literature as plant growth-promoting bacteria [2, 49, 50], and the relative abundance of the bacteria Haliangium, known for phosphorus solubilization [51], and Caulobacter known for pathogen suppression [52, 53]. Similarly, as neighbor plants, both Phleum pratense and Trifolium pratense enriched the relative abundance of Rhizobacter and Steroidobacter, which are described to be involved in nitrogen metabolism [54]. The fungal genus Cladorrhinum was enriched when plants were grown in soil with a legacy of the plant Phleum pratense. This fungal genus has been documented for its potential for the biocontrol of fungal phytopathogens and for the promotion of plant growth [55]. On the other hand, the relative abundance of the fungal genus Dactylonectria declined when plants were grown in soil with a legacy of Phleum pratense. This fungal genus has been described as a plant pathogen which causes black root rot [56]. These associations provide qualitative indications that neighbor plants and soil legacies can influence the prevalence of microbial taxa that are linked in the literature to beneficial or detrimental functions. However, because functional assignments are based on taxonomic identity rather than direct functional measurements, these interpretations should be considered as hypothesis-generating rather than definitive evidence of functional outcomes. Nevertheless, our results suggest that in agriculture, selecting the right neighbor plant or creating the right soil legacy can be used as a way of microbiome engineering to promote beneficial microbes in the rhizosphere of a focal plant or to suppress detrimental ones in a targeted manner. Our study indicates that compounds produced by culturable bacterial communities associated to the roots of plants can also potentially influence plant growth. The UV-sterilized solutions collected from the bacterial NatComs obtained from the roots of each focal plant species, which presumably only contain metabolite solutions, greatly affected the growth of that focal plant. Notably, for the grass species Phleum pratense, plant biomass produced when grown with specific live bacterial NatComs highly correlated with the biomass produced when plants were exposed to the metabolites produced by these bacterial communities. This correspondence is consistent with a role for bacterial metabolites in mediating plant growth responses [57, 58], and suggests that metabolites produced by these bacterial NatComs may contribute to the observed growth effects for this plant species. It is important to note that NatComs represent a culturable subset of the natural microbiome, and that the UV-treatment may incompletely eliminate viable cells or resistant propagules and may alter or degrade light-sensitive metabolites. Even though our findings indicate that metabolites are likely one of several interacting mechanisms through which bacterial communities influence host-plant performance, future work isolating and characterizing dominant bacterial strains and their metabolic products will be necessary to disentangle the relative contributions of specific taxa and compounds to plant growth effects. In summary, our study demonstrates that the composition of root-associated microbiomes is shaped not only by the focal plant itself, but also by the presence of neighboring plants both directly, by growing together with the focal plant, and indirectly, through soil legacies. In our experimental context, microbiome changes influenced the growth of the focal plant. Interestingly, root-associated bacterial communities were primarily influenced by the direct effects of neighboring plants, whereas fungal communities were shaped primarily by soil microbial legacies. Our study highlights the importance of accounting for both direct neighbor effects and soil legacies to realistically understand the drivers of root-associated microbiome assembly and their functional impact on plant growth. Methods First, we created soils with legacies of four plant species (Phase I). Subsequently, we carried out an experiment to quantify the relative contribution of soil legacies and neighbors in shaping the microbiome of four focal species (Phase II). From plants grown with different neighbors and soil legacies, we then cultured root-associated microbial communities and tested the effects of the cultured NatComs on plant performance (Phase III; Fig. 1 and Supplementary Fig. S1). Finally, we collected metabolites from the bacterial NatComs, sterilized them and tested their effects on plant performance. Creating soil legacies (Phase I) Soil legacies were created by growing four plant species in monocultures in pots. Plants were grown in a climate-controlled growth chamber with artificial lighting (light regime 16/8, light/dark, with 21/16 ℃ during light/dark). Forty-two 2-L pots (13 × 13 cm) were filled for each plant species (4 species × 42 replicates = 168 pots) with 1600 g of soil originally collected from a grassland near Lange Dreef, Driebergen, The Netherlands (52° 02′ N, 5° 16′ E). The soil is characterized as holtpodzol, sandy loam (84% sand, 11% silt, 2% clay, ~ 3% organic matter, 5.9 pH, 1,151.3 mg/kg total N, 2.7 mg/kg total P, 91.0 mg/kg total K; analyses by Eurofins Analytico Milieu B.V., Barneveld, The Netherlands, using in-house method [59]. To each pot, 50 ml of Hoagland solution was added. To increase microbial diversity, the soils were enriched with a microbial slurry of 1:2 (soil: water ratio) with soil extracted from a multispecies grassland site near Wassenaarse weg in Leiden, The Netherlands. Before applying, the slurry was sieved through a 500 µm mesh and then 50 ml of this sieved slurry was added to each pot. Soil legacies of four grassland species, belonging to three different functional groups commonly used in production grasslands—Phleum pratense (PP, grass), Cichorium intybus (CI, forb), Trifolium pratense (TP, legume), and Trifolium repens (TR, legume)—were created by planting four 1-week-old seedlings in each pot (Fig. 1C and Supplementary Fig. S1). Seedlings were grown from seeds that had been surface-sterilized for 20 min using a 5% sodium hypochlorite solution, followed by rinsing with demineralized water. The seeds were germinated on sterile glass beads in the same climate-controlled chamber. After eight weeks, plants were harvested, and roots were removed from the soil. The replicated pots from each monoculture were used to create five independent soil replicates, each containing homogenized soil from 9 pots, resulting in 20 soils. Soil moisture was determined gravimetrically by weighing soil samples before and after oven-drying at 105 ℃ for each replicate soil sample. Soil pH, ammonium, nitrate and available phosphorus were assessed for each replicate soil using the potassium chloride [60] or calcium chloride extraction method [61] and are presented in supplementary Table S11. Examining soil legacy and neighbor effects (Phase II) The conditioned soil from each of the five replicates per plant species was divided over seven pots (9 × 9 cm, filled with soil equivalent to 550 g dry soil). Each of these seven pots was then planted with two one-week-old seedlings. Pots contained either (i) a conspecific pair (two seedlings of the same species) or (ii) a mixed-species pair (two seedlings of different species). In mixed-species pots, one of the two seedlings always matched the species used to create the soil legacy, whereas the second seedling belonged to a different species. In total there were 140 pots [4 plant species × (4 monocultures + 3 different mixtures) × 5 replicates] (Fig. 1C and Supplementary Fig. S1). The plants were grown under the conditions described above. The position of the pots in the chamber was randomized each week. Plants were watered three times per week with equal amounts of demineralized water and the final soil moisture level in each pot was reset to 15% of soil moisture once a week using a balance. Seedlings that emerged from the soil were removed daily. Four weeks after transplanting, 50 ml of Hoagland solution was added to each pot to prevent nutrient deficiency. The climate chamber settings were as described above. After eight weeks all plants were harvested. The roots were rinsed and cleaned by submerging in water. As each pot contained two plants (two plants belonging to the same species or a mixture of two plants belonging to different species), the roots of both plants were carefully separated from each other in water. After disentangling of roots, fresh root biomass was determined and subsamples from randomly selected washed root fragments of each plant in each pot were collected and stored at − 80 ℃ prior to molecular analysis. For the remaining root and shoot material fresh weight was determined again, and after oven-drying at 40 °C dry weight was determined. The tiny amount of biomass of the root fragments was added to the total root dry weight, using the fresh weight:dry weight ratio of the remaining root biomass. To further investigate the effects of root microbial communities on the growth of plants of the same species, but without the presence of neighboring plants or soil legacies, after washing, a subset of the roots of plants harvested at the end of Phase II was kept and culturable bacterial communities were cultured from these roots (NatComs). Root subsamples from the five replicates per treatment were combined per species to obtain one pooled cultured community, resulting in a total of 36 NatComs. Although the intended experimental design included ten samples per plant species for NatComs isolation for four plant species, a total of four samples were lost during processing prior to culturing. Specifically, one sample was lost for Phleum pratense and three samples were lost for Trifolium repens. Pooled roots per species per treatment were ground with a mortar and pestle in 2 ml physiological salt solution. Then 50 μL of a 10–2 dilution was inoculated in 1.5 ml of 10% Tryptic soy broth (TSB) supplemented with nystatin to inhibit fungal growth in 24-well plates and incubated at 27 ℃, 180 rpm for 14 days to create the communities (NatComs). Fungal NatComs were cultured in 10% potato dextrose broth (PDB) supplemented with chloramphenicol to inhibit bacterial growth. Control wells without inoculation were included on each plate to monitor contamination. All bacterial NatComs were stored in 20% glycerol at − 80 °C, while fungal NatComs were stored in 80% glycerol at − 80 °C. Testing the effects of bacterial NatComs on plant growth (Phase III) To test the consequences of variation in root-associated microbiomes for plant growth, the effect of the bacterial NatComs on focal plant growth was determined. For the four plant species, seedlings were inoculated with bacterial NatComs that were isolated from plants belonging to the same species (Fig. 1C). Seedlings were prepared as described in Phase I. The bacterial communities (NatComs) were revived from glycerol and grown in 10% TSB by inoculating 10 µL of the cultures into 30 mL medium and incubating at 27 ℃ while shaking at 180 rpm. After 7 days, 2 ml of each NatCom was centrifuged, and the pellet was collected for DNA sequencing (stored at − 20 °C until further processing). The remaining volume for each NatCom was then centrifuged to obtain the pellet for inoculation of the seedlings. The pellet was resuspended in 30 ml of physiological salt solution creating the NatCom inoculum. To inoculate the plants, the roots of five one-week-old seedlings were submerged in one of the NatComs’ inocula for one hour and then transferred to tissue culture tubes filled with 35 g of sterile soil. Control seedlings were treated identically but inoculated with the same volume of physiological saline solution without microbial biomass. Finally, the remaining inoculum or control solution was added to the tubes after planting the seedlings. Seedlings were grown in a growth cabinet (light regime 16/8, light/dark, with 21/17 ℃ during light/dark). After 25 days of growth, plants were harvested, dried at 40 °C and dry weight was determined. We also carried out a similar NatComs experiment for fungal communities that were isolated from the roots of plants from Phase II. The methods and results of this fungal NatComs experiment are presented in the Supplementary information (Supplementary methods). Testing the effects of metabolite solutions produced by bacterial NatComs on plant growth To investigate a potential mechanism explaining the effects of bacterial NatComs on plant growth, we assessed the effects of metabolite-containing solutions produced by bacterial NatComs on focal plant performance. We selected three of the four plant species for this study (Phleum pratense, grass, Cichorium intybus, forb, and Trifolium pratense, legume). Seedlings were treated with sterile metabolite solutions produced by bacterial NatComs that were isolated from plants of the same species (Fig. 1B). We sterilized the NatComs supernatants with UV after centrifuging at 4700×g for 15 min. Control seedlings received the same volume of sterile Milli-Q water and were handled identically. The effectiveness of the UV treatment was assessed by plating aliquots of the treated supernatant on growth medium. No bacterial growth was observed within the first 48 h, whereas limited bacterial growth occurred after prolonged incubation (> 48 h), indicating a strong, but not complete elimination of viable bacteria. We acknowledge that the UV treatment substantially reduced but did not completely eliminate viable bacteria and that this treatment may alter UV-sensitive compounds. Hence this experiment was designed as an exploratory assay to assess whether soluble products associated with bacterial NatComs could contribute to plant growth responses. The detailed method for setting up the experiment and metabolic profiles measurement and analysis is presented in the Supplementary information (Supplementary methods). DNA extraction, sequencing, and bioinformatics DNA from 20 homogenized soil samples at the end of Phase I (4 species × 5 replicates) was extracted from 250 mg of soil using the DNeasy PowerSoil pro kit (Qiagen Inc., Hilden, Germany). DNA from roots from the end of Phase II (120 samples) was extracted from approximately 200 mg of fine roots using the PowerPlant DNA Isolation Kit (Qiagen, Germany). DNA from the 36 samples of NatComs used in Phase III was extracted using the Blood & Tissue Kit (Qiagen, Germany). Bacterial and fungal community composition was determined by amplifying the v4 region of the 16S SSU rRNA and the ITS2 genetic markers, respectively, using the primers 515 F (5′-GTGYCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACNVGGGTWTCTAAT-3′) [62] for bacteria, and ITS4 (TCCTCCGCTTATTGATATGC) and ITS3 mix (GCATCGATGAAGAACGCAGC) [63] for fungi. Bacteria and fungi were sequenced for soil and root samples, while only bacteria were sequenced for the NatComs. Bacterial and fungal raw sequences were processed using the DADA2 pipeline implemented in R [64]. Taxonomic assignment was done with the RDP classifier using its default settings (minimum bootstrap confidence threshold = 50) [65] and using the databases SILVA v138.1 [66] for bacteria, and the UNITE v982 for fungi [67]. Because our focus was on genotypic similarities among samples, we used amplicon sequence variants (ASVs) for both bacterial and fungal analyses, although OTU-based approaches may provide better resolution at the species level [68]. Sequencing processing and removal of non-targeted hits (i.e., chloroplast, mitochondrial and plant for the bacterial dataset, and selecting only fungi for the fungal dataset) resulted in 35,759 ASVs for bacteria and 6608 ASVs for fungi. Because the fungal primers used have limited ability to detect arbuscular mycorrhizal fungi (AMF), AMF were likely underrepresented or absent from the fungal sequencing data and were therefore not evaluated as part of the fungal community analyses. To account for large differences in read numbers between different compartments (e.g., soil, root and NatComs) after removing non-targeted hits, we processed filtering and rarefaction for the bacterial and fungal data within each compartment. For the soil dataset, after filtering out singletons and AVSs with a relative abundance less than 0.05% per sample for both fungi and bacteria, the bacterial and fungal datasets were rarefied to 25,000 and 5912 reads per sample respectively (Supplementary Fig. S14A), resulting in 1361 and 920 ASVs, respectively. For the root datasets, only ASVs present in more than two samples and with a relative abundance higher than 0.01% for bacteria and 0.05% for fungi per sample were kept. The bacterial dataset was rarefied to 2035 reads per sample (Supplementary Fig. S14B) but four samples with lower sequencing depth were later added, resulting in 2380 ASVs. Samples in the fungal dataset were rarefied to 509 reads (Supplementary Fig. S14B), resulting in 1739 ASVs. The NatComs dataset was rarefied to 33,328 reads (Supplementary Fig. S14C), resulting in 195 ASVs. Data analysis We first investigated the effects of four plant species on creating specific soil bacterial and fungal legacies using a PERMANOVA model based on Bray-Curtis distances in the ‘vegan’ R package [69]. We then determined the effects of focal species, soil legacy, neighbor, and soil legacy + neighbor on shaping the root-associated bacterial and fungal communities of the four focal species (Supplementary Fig. S15). We tested four models per microbial group (bacteria or fungi) using the rarefied datasets with a redundancy analysis (RDA) using the vegan R package [69]. (i) To test for focal effects, the response variable was the total dataset pruned to the microbial communities obtained from the focal species growing in its own soil legacy and in the presence of a neighbor of the same species (12 samples: each plant species growing with another plant of the same plant in soil conditioned by itself; three replicates), and the explanatory variable was focal identity. (ii) Soil legacy effects were tested on the pruned datasets (48 samples: 4 focals growing with neighbors with the same identity as the focal in four soil legacies; 3 replicates) with the explanatory variables: focal identity and soil legacy. (iii) Neighbor effects were tested on the pruned datasets (48 samples: four focal species with 4 different neighbor species in focal soil; 3 replicates) with the explanatory variables: focal identity and neighbor identity. The combined effects of soil legacies and neighbors were tested by using the full dataset (120 samples: 4 focals with four different neighbors in 4 different soils; 3 replicates) with the explanatory variables: focal identity, soil legacy, and neighbor identity. Individual contribution of each factor in explaining the structure of microbial communities was derived from the varpart function in the vegan R package using the adjusted R2. Subsequently, we investigated the effects of soil legacies and neighbors on shaping the root-associated bacterial and fungal communities of each focal plant using RDA and variance partitioning analysis (VPA) based on the dataset of 30 samples per focal plant. For each focal plant, the response variable was the bacterial or fungal community associated with its roots, while the explanatory variables were the soil legacy origin and neighbor identity. Finally, to determine the effects of soil legacy type (focal grown in focal legacy soil versus neighbor plants’ legacy soil) and neighbor type (focal grown together with another same plant species versus a different plant species) on the root-associated bacterial and fungal communities of each of the focal plants, we performed RDA and VPA with soil legacy and neighbor type as the explanatory variables and the couple identity as a conditional effect. Principal components analysis (PcoA) with Bray–Curtis dissimilarities on Hellinger-transformed counts was conducted to visualize these effects on the root-associated microbial communities of the four focal plant species. To determine how each specific soil legacy, neighbor plant and the combination of soil legacy and neighbor plant changes the root-associated microbial community of a focal plant, we first calculated Bray–Curtis dissimilarities between all root samples using the vegdist function in the vegan package [69]. For the focal effect, we determined the Bray–Curtis dissimilarities among three samples from each focal species, each growing with a conspecific neighbor in its own soil legacy. For the soil legacy effect, we compared each replicate of a focal plant growing in its own soil with three replicates of the focal plant growing in other soil legacies and then calculated the average Bray–Curtis dissimilarity. This resulted in three independent values for each specific soil legacy. For the neighbor effect, we compared each replicate of a focal plant growing with another conspecific neighbor in its own legacy soil with three replicates of the focal plant growing with a heterospecific neighbor plant in its own legacy soil, and then calculated the average Bray–Curtis dissimilarity. This resulted in three independent values for each specific neighbor plant. For the combined effects of neighbors and soil legacies, we compared each replicate of a focal plant growing with another conspecific neighbor in its own legacy soil with three replicates of the focal plant growing with a heterospecific neighbor plant in the soil of the neighbors and then calculated the average Bray–Curtis dissimilarity. This resulted in three independent values for each combined neighbor and soil legacy. We first used a t-test to investigate whether growing in a specific soil, with a specific neighbor or in a specific soil significantly changed microbial community similarity compared to the similarity amongst focal plants grown together with another plant of the same species in conspecific soil (focal effect). These results are shown in Supplementary Fig. S3. To investigate whether the four different effect types (focal, soil, neighbor and soil + neighbor) resulted in different microbial communities, we first grouped samples based on their effect type, which resulted in nine samples for each of the soil, neighbor and soil + neighbor effects (three biological replicates * three specific conditions for each effect type, e.g., Supplementary Fig. S16). We then used a one-way ANOVA model, with effect types as the explanatory variable. Planned comparisons were then used to compare soil, neighbor, and combined effects against the focal baseline within the same ANOVA model. Planned comparisons were interpreted only when the overall ANOVA was significant. To identify the microbial taxa that were consistently depleted or enriched by soil legacies and neighbors, indicator species analysis using the indicspecies R package [70] was performed for each parameter individually at the focal species level. To calculate the indicator taxa of the soil legacies for each neighbor species, we compared the monocultures of the other three species growing in that neighbor soil against all the monocultures of the other three species growing in all soils except the neighbor soil. To obtain the indicator taxa of the neighbor effect for each neighbor species, we compared that neighbor growing with the three other species in its own soil against all pairs of plants (monocultures and two-species combinations) in all soils except the neighbor soil (Supplementary Fig. S7). To ensure that indicator taxa were both sufficiently frequent and representative of the contrasted group, after evaluation of indicator species at a significance level of 0.05, these taxa were further considered indicators if present in at least 50% of the samples per contrasted group, reflecting a minimum level of fidelity. This filtering step reduced the influence of rare taxa with significant but low-occurrence patterns. To obtain potential functions of each indicator species, putative functional information for bacterial indicator taxa was inferred using FAPROTAX [71] based on their taxonomic classification, primarily at the genus level when available. Functional guilds for fungal indicator taxa were assigned using FungalTraits [72], also primarily at the genus level. Both functions were complemented by targeted literature searches. To estimate the relative importance of soil-legacy effects and neighbor effects on focal-plant shoot biomass in Phase II, we performed a two-way ANOVA with soil-legacy effect and neighbor-identity effect as main factors using the aov function in the Base R package. To determine how much of the variation in the microbial communities associated with the roots of the focal plant was explained by plant biomass of that focal plant, we performed an RDA analysis using the rda function in the vegan R package [69] with focal-plant shoot biomass as the explanatory variable and neighbor identity and soil legacy as covariances. A PERMANOVA model was constructed to investigate the effects of focal host plant identity on the community compositions of the bacterial NatComs using Bray–Curtis distance in vegan [69]. We created a heatmap visualization of the 30 most abundant ASVs in the isolated NatComs. The effects of the NatComs on the biomass of each host plant were tested using one-way ANOVA, in which shoot biomass of inoculated plants (Phase III) was treated as the response variable and the identity of the inoculated bacterial NatCom was included as the explanatory variable. The proportion of variance explained was calculated as the ratio of the sum of squares of the explanatory factor to the total sum of squares (R2 = SS_effect/SS_total) for each species. For all data, normality assumptions were tested using QQ-plots and a Shapiro–Wilk test, and homogeneity of variances was tested using a Levene’s test. All statistical analyses were performed in R (v4.1.0, http://www.r-project.org/). Supplementary Information Supplementary Material 1: Supplementary methods, Supplementary Figures and Supplementary Tables. Acknowledgements We thank Karin Van Der Veen-Van Wijk, Xiangyu Liu and Yuhui Li for help in planting and harvesting the experiments. Mireadili Kuerban was funded by a China Scholarship Council (CSC) grant (no. 202106350028). Authors’ contributions M.K., S.I.F.G. and T.M.B. conceived and designed the experiments. M.K. carried out experiments. O.E. prosed metabolic profiles and analyzed metabolic data. M.K. analyzed the data with the assistance of S.I.F.G. and T.M.B. M.K., S.I.F.G. and T.M.B. wrote the manuscript. All authors approved the manuscript. Funding This research did not receive any funding. Data availability The raw sequence data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1193320. The data are publicly available at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1193320/. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1: Supplementary methods, Supplementary Figures and Supplementary Tables. Data Availability Statement The raw sequence data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1193320. The data are publicly available at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1193320/.