The chemical interaction between plants and the rhizosphere microbiome

Highlights

• The use of advanced detection methods and molecular tools enables scientists to unravel a plethora of beneficial interactions between plants and their root microbiota.

• Chemical interactions in the rhizosphere involve different mechanisms, such as organic carbon provision, antimicrobial compounds, and microbiota recruitment signals, which help explain how the diversity of root exudate metabolites shapes microbial community assembly.

• Modern high-input agriculture may have diminished the role of this chemical interaction through the use of synthetic inputs (nutrients, pesticides) and modern genotypes.

• Harnessing natural mechanisms for agricultural sustainability becomes increasingly viable, potentially helping agriculture to face mounting challenges from environmental issues, climate change, and rising input costs, which are threatening long-term farming viability.

Abstract

Research into the interaction between plants and the soil microbiota has expanded rapidly and is unravelling a plethora of interactions between plants and their root microbiota. The rhizosphere exhibits remarkable chemical diversity, driven by an evolutionary arms race. Through these chemicals, plants shape the rhizosphere microbiome using different mechanisms: organic carbon provision, antimicrobial compound production, and exudation of microbiota recruitment signals. Modern high-input agriculture may have diminished the role of natural chemical interactions and modern crops may have lost some of the relevant traits. As our understanding of root–rhizosphere interactions grows, harnessing natural mechanisms for agricultural sustainability becomes increasingly viable, potentially helping agriculture to counteract growing challenges from environmental stresses, climate change, and rising input costs.

Research into the interaction between plants and the soil microbiota has expanded rapidly and is unravelling a plethora of interactions between plants and their root microbiota. The rhizosphere exhibits remarkable chemical diversity, driven by an evolutionary arms race. Through these chemicals, plants shape the rhizosphere microbiome using different mechanisms: organic carbon provision, antimicrobial compound production, and exudation of microbiota recruitment signals. Modern high-input agriculture may have diminished the role of natural chemical interactions and modern crops may have lost some of the relevant traits. As our understanding of root–rhizosphere interactions grows, harnessing natural mechanisms for agricultural sustainability becomes increasingly viable, potentially helping agriculture to counteract growing challenges from environmental stresses, climate change, and rising input costs.

Toward a microbiome-supported agriculture

Agriculture is a major consumer of nonrenewable resources and a key driver of pollution and climate change [1]. Chemical pesticides accumulate in soils and waters, while fertiliser runoff causes eutrophication [2]. Fertiliser use also drives CO₂ and N₂O emissions, with nitrogen (N) fertiliser production via the energy-intensive Haber–Bosch process consuming vast amounts of fossil fuel [3]. In addition, microbial activity releases the potent greenhouse gases, NH₃ and N₂O, from N fertiliser [4]. Phosphorus (P) fertiliser production depends on finite rock phosphate, projected to be depleted within 100 years [5]. Pesticide use is rising, reaching 3.7 million tonnes in 2022, double the levels in 1990 [6].

Growing evidence suggests that a better understanding of the interaction between plants and microbes in the soil could contribute to finding urgently needed solutions. Microbes from the soil can enter the root (the root endosphere), colonise the surface of the root (the rhizoplane) or populate the rhizosphere, the thin layer of soil surrounding the roots. The communities of microbes in the former two compartments together constitute the root microbiome, while the latter is called the rhizosphere microbiome. This complex ecosystem hosts a vast number of organisms that plants can partially control using chemistry. In this review, we explore how plants perform this balancing act (recruiting beneficial microbes to the endosphere, rhizoplane, and rhizosphere, and deterring pathogens) through the chemical environment they establish with root-exuded compounds. For insight into the reverse communication from microbes to plants, we refer readers to other reviews [7,8].

Plants allocate 5–30% of photosynthetically fixed carbon to the rhizosphere [9], likely to enhance interactions with root microbiota and optimise growth, development, and reproduction. Evidence shows that root exudate metabolite variation influences microbiota assembly [10., 11., 12., 13., 14.], but a theoretical framework remains lacking. Better insights into the chemical interaction belowground will allow us to harness the beneficial relationships and suppress the harmful ones. This will enable knowledge-based changes in how agriculture is performed, including the use of chemical inputs, how more resilient crops can be bred, and how agricultural soils can be rendered more resilient.

Here, we explore the role of root exudates in plant–microbiota interactions. We discuss their chemical diversity and plasticity, their role in rhizosphere communication, and the methodologies for studying them. We also define different types of chemical interaction and propose future research directions.

How to analyse the chemistry of the rhizosphere

Understanding plant–microbe crosstalk requires characterising, quantifying, and functionally describing root-secreted metabolites. Various plant growth and root exudate collection methods exist, each with trade-offs in reproducibility, throughput, resemblance to soil conditions, sample purity, and extraction suitability (Box 1).

Box 1. Root exudate collection.

The method by which root exudates are collected is vital to obtain a realistic picture of what is happening in the rhizosphere, as discussed in recent reviews [17,105,106]. Here, we focus on some of the most common methods and discuss their advantages and disadvantages.

Hydroponic systems (i.e., where plants are cultivated with their roots in a liquid nutrient solution [107]) have often been used for the collection of root exudates, since it is straightforward to get a large and clean sample. Glass beads can be added for support of the roots [108]. Potential problems with hydroponics arise through hypoxia in the plant roots, unless, for example, aquarium pumps are used to apply oxygen to the system. To circumvent this issue, aeroponics is an alternative in which plant roots are allowed to grow in the air while being moistened regularly every few minutes with nutrient solution via a sprinkler [61]. An advantage of this system is that the roots are aerated, that changes in nutrient availability can be easily achieved, and that exudate and roots can be easily collected. Moreover, the seedlings can be planted in a small volume of sterile or nonsterile substrate, allowing for inoculation with soil microbes, making the system also suitable for microbiome analysis, in addition to gene expression and metabolite analyses (D. Abedini et al., unpublished, 2025).

Root exudates can also be collected from plants grown in soil, sand, or soil/sand mixtures, which has the advantage that they resemble natural conditions more closely compared with liquid media. Some researchers first uproot their plants to clean the roots and subsequently incubate them in deionised water for a few hours to a few days to sample exudates [109]. Arguably, this procedure causes stress and damage to the plants and will miss metabolites that are exuded instantaneously; it also excludes the analysis of gene expression and the microbiome on the same roots from which the exudate is collected. A less intrusive method involves flushing the soil/sand mixture in which the plant is growing with water or nutrient solution and collecting the flow-through [10,110]. This method is nondestructive and allows the recovery of the roots for gene expression and microbiome analyses. Sometimes, a low percentage of organic solvent is added to the water or nutrient solution to improve metabolite recovery [10,110]. However, this influences the root and microbes, making this method unsuitable for sample collection for root gene expression and microbiome profiling.

It is of increasing interest to study microbial isolates and synthetic communities in isolation [111]. Therefore, closed, gnotobiotic environments for the co-cultivation of sterile plants with defined, and potentially genetically modified, microorganisms have become more relevant. One such system is the so-called ecosystem fabrication (EcoFAB), where plants can grow either hydroponically or in a soil matrix, and with or without microorganisms [112]. Exudates can be sampled nondestructively, root systems imaged, and roots easily collected for microbiome analysis [113,114]. The FlowPot system [108,115] also uses sterilised soil, for repopulation with microorganisms, and can be used for nondisruptive exudate collection, and sampling of roots from the same pots for downstream DNA and RNA isolation (K. Wippel, personal communication, 2025). Another system potentially suitable to study the interaction between microbes and the plant root is the Growth and Luminescence Observatory for Roots (GLO-Roots) system, although it was not designed to collect exudate [116]. EcoFABs, FlowPots, and GLO-Roots are complex and challenging to assemble, making them low-throughput strategies.

Over the past few years, microdialysis has been adapted to plant and soil research, either for the collection of rhizosphere metabolites or, vice versa, to assess how the introduction of solutes changes soil chemistry [117]. To collect soil metabolites, deionised water is pumped as a perfusate inside a probe, creating an osmotic pressure difference that drives the uptake of solutes through a semi-permeable membrane from the environment into the probe, from which they can be sampled [117]. This strategy has been used in an agar-based system to investigate metabolite exchange between an ectomycorrhizal fungus and eucalyptus [118], but can also be applied to soils or field settings [119]. Low throughput and small sample volumes may limit applications of this system.

Overall, an advantage of nutrient solution or agar-based growth conditions is the reproducibility and purity of the exudate samples. However, microbial colonisation fails or differs from natural growth environments. Furthermore, soil complexity and characteristics can alter both plant and microbial chemical compounds and make it difficult to reproduce phenotypes and chemistry. Taken together, there are several rhizosphere metabolite collection approaches available, and novel, innovative solutions are being developed to allow for controlled, yet natural-like conditions and collection of multiple sample types from the same plant. The method of choice largely depends on its suitability for the research questions being asked.

Alt-text: Box 1

Metabolomics, the study of small molecules using gas chromatography (GC)-mass spectrometry (MS), liquid chromatography (LC)-MS, and nuclear magnetic resonance (NMR) for untargeted analysis, is the method of choice to analyse the composition of the root exudate. However, analysis of root exudates is challenging due to their chemical diversity, low metabolite abundance, and complex matrix [15., 16., 17.]. Metabolites range from sugars, amino acids, and organic acids to specialised metabolites, produced by both plants and microbes, and sometimes occur in concentrations that remain below the detection threshold of untargeted analysis (see below), complicating their identification and origin. Advances in high-resolution mass spectrometry and bioinformatics have improved metabolomics analysis, yet annotation/structure elucidation of root exudate metabolites remains challenging [17].

Machine learning may help to overcome some of these challenges by predicting molecular properties, such as adduct species, fragmentation patterns, and retention times [18]. Computational tools, such as SIRIUS, MetFrag, MetaboAnnotatoR, and DEREPLICATOR(+), annotate metabolites using LC-MS/MS data [19., 20., 21., 22.]. Nevertheless, there is a lack of appropriate entries in mass spectral databases. Moreover, it is challenging to isolate pure compounds from the highly complex mixtures of metabolites in root exudates for structure elucidation (using NMR), which is reinforced by the sometimes extremely low concentrations of these metabolites.

To improve this, 2D LC separation and time-series retention index standards have been recommended [18]. Stable isotope labelling (e.g., ¹³C-labeled plants or ¹⁵N-labeled microbial substrates) helps distinguish metabolite origins [23], and integration of metabolomics with transcriptomics, genomics, and genome-scale metabolic modelling can also potentially contribute to unravelling metabolite identity and origin (see below) [24,25].

Even though annotation of all the metabolic features in a root exudate is challenging, untargeted metabolomics is instrumental in analysing the chemistry in the rhizosphere of plants. Even without complete annotation, data integration is feasible, allowing preselection of those unknown features that are biologically relevant for further identification. Moreover, even without annotation, metabolomics demonstrates that the root exudate of plants comprises a highly complex mixture of hundreds to thousands of small molecules.

Why are root exudates so complex?

The radiating evolution of plant specialised metabolites is the result of an evolutionary arms race between plants and arthropod, specialist and generalist, herbivores [26] (see Box 2). A similar radiation appears to have occurred belowground, resulting in the complex composition of root exudates, with a large contribution of plant species (and plant family)-specific chemistry (Box 3). This further complicates annotation and necessitates renewed identification of unknown metabolites when investigating the root exudate of a previously unstudied plant species.

Box 2. Analogy with plant–insect interactions.

Just as insects exhibit varying degrees of host specificity, ranging from generalists that feed on a wide variety of plants to specialists that are restricted to a single species or a narrow group, the plant microbiome also harbours a diverse range of microorganisms with varying degrees of host specificity [120,121]. In generalist insect species, consumption of a biochemically diverse diet leads to a canalised, generic metabolic response, aligning with the metabolic generalism hypothesis [122]. The latter study revealed that many diet-specific metabolites, such as those contributing to unique colour, odour, or taste, were not effectively metabolised but instead accumulated in the consumers, potentially to their detriment. By contrast, specialist insects feed on a restricted number of closely related plant species, with similar defence chemicals to which the insect has evolved tolerance, through detoxification (among other mechanisms) [123]. The classical example is the caterpillar of the large cabbage white, Pieris brassicae, which feeds only on the glucosinolate-containing Brassicaceae [123].

Similarly, within the plant microbiome, generalist microbes, living in a variety of different environments, may encounter a diverse array of plant species, each with their unique suite of specialised metabolites [124]. This exposure to a range of plant compounds may lead to a more generalised metabolic response, potentially limiting their ability to efficiently utilise or detoxify specific plant metabolites produced by particular plant species. Generalist microbes have evolved various strategies, such as tolerance mechanisms [125], genome fluidity, biofilm formation, enhanced intraspecies interactions, and rapid growth and dispersal, to deal with a range of different secondary metabolites [126], allowing them to persist in the plant microbiome despite the presence of various antimicrobial compounds. However, continued exposure to these compounds exerts a selective pressure, potentially driving the evolution of specialist microbes.

These specialist microbes have evolved the capacity to catabolise specific plant-produced specialised metabolites. Specialist microbes are often found in association with particular plant families, and have unique enzymes that allow them to catabolise plant family specific, often anti-microbial, specialised metabolites. This allows them to form an intimate and potentially mutually beneficial relationship with their host [31]. For example, specialist microbes, such as Sphingobium spp., have evolved mechanisms to catabolise and metabolise glycoalkaloids, a class of specialised metabolites that serve as potent defence compounds in Solanaceae against generalist microbes by disrupting microbial cell membranes with their detergent-like properties [127]. This adaptation allows specialist microbes to thrive where others cannot [41]. Other examples include microbes that can catabolise benzoxazinoids, cucurbitacins, and other triterpenoids, and, therefore, can colonise the rhizosphere of maize, cucumber, and arabidopsis, respectively [40] (see main text).

Alt-text: Box 2

Box 3. Mutualism in the rhizosphere.

The complexity of root exudates appears to be the result of an evolutionary arms race, a concept that also applies to mutualisms (i.e., interactions in which one species enhances another's fitness). Theory formation on such interactions initially occurred for flower–pollinator interactions [128]. However, when the difference in generation time of partners is significant, mutualism becomes less stable, because faster-evolving partners can adapt more quickly, potentially disrupting the balance [128]. This theory was based on the fact that flowers cannot punish or switch pollinators within a single generation. However, in the rhizosphere, plants are likely to have more control. They can sanction cheating mutualists by reducing resources or releasing antimicrobial compounds, and can recruit alternative partners from the surrounding soil.

When mutualistic interactions persist, certain traits may evolve in ways that either stabilise or diversify the relationship. An example of stabilising evolution is seen in the type 3 secretion system (T3SS). The T3SS, used by both microbial pathogens and mutualists to deliver effectors into host cells, shows much less genetic diversity in mutualistic species, such as Sinorhizobium fredii and Bradyrhizobium japonicum, than in the pathogenic species Pseudomonas syringae [129]. This suggests a classical arms race between hosts and pathogen, while the mutualist–host interaction is under balancing selection.

In the rhizosphere, it would appear that mutualists can easily exploit plant resources without providing benefits, shifting into parasitism. Since natural selection favours individual fitness, mutualism can be evolutionarily unstable [130]. However, stable mutualisms exist, suggesting mechanisms that maintain cooperation [131]. Three proposed models explain this stability: (i) byproduct cooperation; (ii) partner fidelity feedback; and (iii) partner choice [132., 133., 134.]:

In byproduct cooperation, mutualism arises when one organism benefits from another’s byproducts, which are produced independently and at no cost. Examples include plants releasing carbon into the soil, benefiting bacteria, and bacteria producing antibiotics that protect hosts. However, many microbial mutualists incur costs, making this model insufficient to explain all stable mutualisms (e.g., costly N fixation in legume–rhizobia symbiosis) [132].

In partner fidelity feedback, mutualism is maintained when host and symbiont fitness are linked over time, often through vertical transmission. While most root microbiomes are recruited from bulk soil rather than inherited, the soil legacy effect suggests that interactions persist across plant generations [135].

By contrast, in partner choice, hosts selectively favour cooperative microbes over cheaters, exerting control over microbial populations to maintain mutualism.

Marin and Johnson [136] proposed that plants and their microbiomes function as complex adaptive systems (CAS), a concept from economics, anthropology, and biology. Evolutionary theory supports multilevel selection in forming cooperative groups [137]. The ‘law of increasing functional information’ reinforces this, stating that evolving systems: (i) have many interacting components; (ii) form multiple configurations; and (iii) favour beneficial arrangements [138]. The plant–microbiome holobiome meets these criteria, with diverse microbial interactions enhancing plant productivity, survival, and reproduction.

As open, dynamic systems, CAS evolve through host–microbiome–environment interactions [139,140]. This complexity means that reductionist lab and greenhouse studies may not fully predict their behaviour in real ecosystems. Soil microbial diversity increases the likelihood of selecting beneficial microbiome functions that boost the productivity and fitness of the host plant. Rhizosphere microbes depend on plant-derived organic compounds, and their populations rise or fall with host success. Genetically controlled root exudates have a key role in microbiota selection.

Alt-text: Box 3

The complexity of root exudates is nicely illustrated by the strigolactones, with well over 30 different strigolactones reported, produced by a range of plant species [27,28]. This diversification may have evolved through a classical arms race to avoid a negative effect on plant fitness, inflicted by mutualists turning into cheaters, or the use as a cue by parasites such as parasitic plants (and perhaps other pathogens [29]). Alternatively, structural diversification may result in new functionality, and the discovery of a role for certain strigolactones in the recruitment of beneficial bacteria may point in that direction (D. Abedini et al., unpublished data, 2025) [30].

However, not all root exudate metabolites appear to follow this pattern of diversification, with plants producing only a limited number of different benzoxazinoids, as far as currently known [31]. The difference may be caused by their different functions; that is, strigolactones evolved to attract mutualistic partners but were later hijacked by parasites [29], while benzoxazinoids likely evolved to repel pathogens [32] and were co-opted by mutualistic microbes adapted to metabolise them. In addition, benzoxazinoids do not appear to have a signalling function, which perhaps lowers the selection pressure on structural diversification seen in the strigolactones.

In some metabolites, this distinction is less clear. For example, coumarins attract beneficial bacteria that catabolise them [33], as for the benzoxazinoids, but their structural diversity suggests diversification beyond balancing selection. This may stem from their role in defence against fungi and nematodes, with resistance in these pathogens driving diversification [34]. It is conceivable that pathogens overcome benzoxazinoid-mediated defences less effectively compared with coumarin-mediated ones, thus exerting greater selective pressure for diversification on coumarins than on benzoxazinoids. Finally, coumarins have also been suggested to display a signalling role [11], which could explain their larger chemical diversity compared with benzoxazinoids, as seen for the strigolactones.

The root exudate of plants displays a vast chemical diversity. How does this chemical diversity drive the recruitment of different microbes to the root and rhizosphere?

Root exudate drives microbial community composition through different mechanisms

Studies of the effect of root-exuded metabolites on the root microbiome suggest that there are several mechanisms by which exudates shape the root microbial community composition: (i) by providing organic carbon; (ii) by exerting (specific) anti-microbial activity; and (iii) through signalling. Here, we discuss the evidence for these three different mechanisms.

Organic carbon

Plants are the primary source for the provision of reduced carbon to the soil, on which soil microbes depend. Many of the metabolites reported in root exudates represent relatively abundant chemicals [detectable with high-performance liquid chromatography (HPLC), GC-MS, and other forms of untargeted MS] and are relatively simple, primary metabolites, which plants likely exude as a carbon source for their root microbiota (Figure 1). This includes sugars, sugar alcohols, nucleotides, nucleosides, amino acids, organic acids, and fatty acids, among others [35,36]. There are indications that microbial metabolite preferences, which are predictable, to some extent, from the microbial genome sequence [37,38], determine which microbiota colonise the rhizosphere, in dependent of the root exudate composition [35]. The latter authors observed a preference by bacteria that colonise the rhizosphere of Avena barbata, for consumption of aromatic organic acids, such as a.o. nicotinic acid, shikimic acid, and cinnamic acid, which are exuded by this species into the soil.

Figure 1.

Chemical interaction between plants and microbiota in the soil.

The antimicrobial benzoxazinoids, produced by Poaceae, such as maize, and coumarins have a negative effect on some (pathogenic) microbes and increase the abundance of (beneficial) bacteria that can degrade them. Plants also secrete primary metabolites that appear to serve primarily as organic carbon source for root microbiota. Microbiota recruitment signals, such as strigolactones and flavonoids, recruit arbuscular mycorrhizal fungi (AMF) and nitrogen-fixing bacteria, respectively, as well as other beneficial microbes. Created with BioRender (biorender.com).

Antimicrobial compounds

Many of the compounds in the root exudate of plants display antimicrobial activity. For example, members of the Poaceae exude indole-derived benzoxazinoids, which can selectively inhibit the growth of common root-associated bacteria [32] (Figure 1). Benzoxazinoids were demonstrated to shape the rhizosphere microbiome [39] and to favour bacterial strains equipped with a gene cluster for their catabolism, enabling them to thrive on benzoxazinoids as their sole carbon source [31] (Figure 1). This is reminiscent of specialist herbivores, equipped with detoxification machinery for antiherbivorous compounds (Box 2). These antimicrobial root exudate constituents are usually limited to a single plant family, but appear to display commonalities. Antimicrobial triterpenoids in arabidopsis (Arabidopsis thaliana), Cucurbitaceae (cucurbitacins), and Solanaceae (glycoalkaloids) have all been linked to enrichment of specific root microbiota and the capacity of these microbes to catabolise them [10,40., 41., 42.].

Coumarins are a highly diverse class of antimicrobial metabolites that occur in root exudates of both monocotyledonous and dicotyledonous species, with over 700 structures known [43]. Similar to BXs and triterpenoids, many coumarins have antimicrobial activity and a role in shaping a beneficial microbiome [44,45] (Figure 1). Coumarin biosynthesis was shown to be required in arabidopsis to recruit beneficial Pseudomonas spp. that protect the plant (and successive generations grown on the same soil) against downy mildew [46]. Intriguingly, however, chemical analysis demonstrated that the exudation of common arabidopsis coumarins decreased upon downy mildew infection [47]. It is unclear whether this really represents a decrease or is due to modification of the common coumarins that were analysed [47]. Furthermore, there is evidence that microbes, including Pseudomonas spp., can degrade coumarins [48].

Another example is camalexin, an indole-derived secondary metabolite found in selected brassicaceous species, including arabidopsis. Camalexin biosynthesis is triggered by both pathogens and beneficial microbes and has a role in the defence against various pathogens while promoting beneficial interactions with other microbiota [49., 50., 51.].

Microbiota recruitment signals

The third class of metabolites that plants exude to influence the root microbiota community composition, we coin here as ‘microbiota recruitment signals’. These signals are not antimicrobial and are exuded in such low concentrations that they cannot be considered a source of carbon for the microbiota. Instead, they are molecules that give information to, and affect the behaviour of, beneficial microbes, the definition of a signal [52].

A prime example of these microbiota recruitment signals are the signal molecules that initiate the mutualistic relationship between plants and arbuscular mycorrhizal (AM) fungi and N-fixing endosymbionts (Figure 1). The symbiosis with N-fixing rhizobia in legumes is initiated by exudation of flavonoid signals from the roots [53]. For AM fungi, strigolactones are the signal that triggers the symbiotic process [54] (Figure 1). In contrast to the antimicrobial compounds discussed in the preceding text, there is evolutionary conservation of these microbiota recruitment signals: strigolactones are produced by all land plants [27,55] and flavonoids appear to be universal signals for N-fixing microbiota not only in legumes, but also in maize [33,53] (Figure 1).

These microbiota recruitment signals act as true signals, conveying information on the host plant condition to microbiota. Flavonoids indicate N deficiency to rhizobacteria, triggering N-fixing symbiosis in legumes [56]. Strigolactones signal P deficiency to AM fungi, which solubilise phosphate in exchange for plant-provided carbohydrates and lipids [27,54]. The AM fungi–plant symbiosis dates back ~450 million years, as might also strigolactone production [57]. Primitive plants, such as Marchantia, produce a simple strigolactone (bryosymbiol) [58], while higher plants synthesise diverse blends requiring multiple genes [27,59]. Strigolactone diversity may help fine-tune specificity across different mutualistic relationships [60] (Figure 1).

Decades before their discovery as AM fungi signals, strigolactones were identified as cues for root parasitic plants [27]. It is biologically intuitive that parasites evolved to exploit these essential signals [27]. This concept led to investigations of eclepins, such as solanoeclepin A, initially known only as cues for plant-parasitic cyst nematodes, under the hypothesis that they also serve a beneficial role (A. Guerrieri, PhD thesis, University of Amsterdam, 2022) [61]. Indeed, eclepin biosynthesis increases under N deficiency, promoting microbial recruitment to enhance N availability (A. Guerrieri, PhD thesis, University of Amsterdam, 2022; D. Abedini et al., unpublished, 2025). The relatively recent discovery of strigolactones and eclepins as microbiota recruitment signals suggests that they represent only the tip of the rhizosphere signalling iceberg.

Although antimicrobial, (some) coumarins should perhaps be also classified as microbiota recruitment signals. Gene expression analysis in Pseudomonas simiae WCS417 upon exposure to root exudates of wild-type and coumarin biosynthetic mutant f6’h1 demonstrated repression of flagellar biosynthetic genes, suggesting that (one of the) coumarins in the exudate also acts as a signal inducing biofilm formation [33].

The small molecules in the root exudates of plants drive the composition of the root and rhizosphere microbiome through several different mechanisms: (i) plants provide microbes with a mixture of metabolites that they can use as a carbon source. Through the chemical diversity in these metabolites, microbial community composition is shaped via the microbial preference/potential to catabolise these molecules; (ii) plants exude antimicrobial small molecules. These inhibit microbes that are sensitive to their biological activity, and favour those that can catabolise them; (iii) plants secrete microbiome recruitment signals, which stimulate colonisation of the roots and/or rhizosphere by specific microbes. These signals are small molecules that are not antimicrobial and are exuded in very low concentrations, excluding significance as a carbon source.

Challenges for the analysis of microbiota recruitment signals

In contrast to the relatively abundant antimicrobials, microbiota recruitment signals are challenging to study, because they are produced in very low concentrations. Indeed, eco-evolutionary theory predicts that selection pressure in a mutualistic signalling relationship selects for low concentrations of the signal and high sensitivity for the perception in the signal receiver [52,62]. This reduces costs and improves specificity. An antimicrobial-based relation, such as the benzoxazinoids mentioned in the preceding text, selects for insensitivity in the target organisms, thus requiring higher concentrations (or more efficient molecules) to remain effective. Indeed, these antimicrobials can be easily detected with metabolomics, but signal molecules generally remain under the detection threshold of this approach. For example, in a study of maize growing under P deficiency, strigolactones were not detected using metabolomics [63]. However, targeted analysis using multiple reaction monitoring (MRM)-LC-MS/MS allows unambiguous detection of strigolactones under P deficiency in maize root exudate [59,64]. Similarly, under N deficiency, the root exudate composition of tomato dramatically changes, as seen using untargeted LC-MS/MS metabolomics (D. Abedini et al., unpublished, 2025). Furthermore, exudation of the strigolactone solanacol increases, but requires targeted analysis on MRM-LC-MS/MS to be detected (D. Abedini et al., unpublished, 2025).

Plant root exudates contain a large variety of different small molecules that affect microbiome recruitment through different mechanisms. The production of these small molecules is not stable but is affected by environmental conditions, making the root exudate composition a reliable tell-tale for the physiological status of the plant.

Plasticity of the root exudate

The rhizosphere microbiome is crucial for plant health but is not inherited directly. Plants recruit their root microbiota mostly from the soil and this recruitment depends on their root exudate and architecture and the microbial composition of the surrounding bulk soil [35,65]. Root architecture and exudate composition are determined to some extent by host genetics. However, the influence of host genetics on microbiome assembly appears small and is complex to unravel due to the involvement of many genes. Genome-wide association studies can help, as shown in maize, where a locus on chromosome 4 was linked to specific microbes [66]. However, environmental factors often far outweigh genetic effects and, thus, have a larger effect on the root and/or rhizosphere microbiome composition, as seen, for example, in sorghum, where microbiome differences between genotypes decreased under drought stress [67].

Although direct evidence is missing in most studies, we, and many others, postulate that the effect of environmental conditions on microbiome composition is largely mediated through changes in the root exudate [9,14,35,39,68]. This is clearly illustrated by the changes in the microbiome recruitment signal concentrations in the root exudate. Under N and P deficiency the concentrations of flavonoids and strigolactones in the root exudate increase, resulting in the recruitment of N-related bacteria and AM fungi, respectively [53,56,69,70] (Figure 1). Similarly, under aboveground pathogen attack in arabidopsis, coumarin biosynthesis in the roots is affected [11], a concept known as the cry-for-help hypothesis [68,71]. The effect of environmental conditions appears to extend beyond the exudation of individual compounds. Metabolomics analysis demonstrated that the root exudate composition of both wild maize, Teosinte, and cultivated maize, which differed under control conditions, changed and, interestingly, converged, under low P [63].

The changes in the root exudate composition as a result of changes in environmental conditions, in combination with changes in the microbiome, suggest that the two are causal. The challenge is to unravel whether causal pairwise metabolite–microbe relationships exist, although it is possible that the interaction is more complex than that. Nevertheless, several studies have suggested such relationships. Tomato plants infected with the pathogen Ralstonia solanacearum displayed increased exudation of phenolic compounds, particularly caffeic acid [72]. This not only suppressed the growth of the pathogen directly, but also changed the rhizosphere microbiome. In rice, the biosynthesis of diterpenoid phytoalexins (DPs) is induced under biotic stress, which not only protects the plant against foliar microbial pathogens and the root-knot nematode Meloidogyne graminicola, but also affects the microbiome composition, enriching taxa potentially antagonistic to nematodes [73]. In cucumber, inoculation with Fusarium oxysporum f.sp. cucumerinum over several generations resulted in decreased disease incidence (soil suppressiveness). Metabolomics analysis using GC-time of light (TOF)-MS demonstrated an increase in threonic acid and lysine in the root exudate of infected cucumber and a concomitant increase in the recruitment of Bacillus and Sphingomonas that protect cucumber against Fusarium infection [74]. The authors confirmed the effect of these two metabolites on the soil microbiome, and its effect on disease suppression.

Thus, there are many examples demonstrating that the root exudate composition changes because of environmental conditions. This plasticity represents an attractive mechanism by which the plant microbiome may adapt to changing environments to optimise nutrient uptake, defence, and growth. However, unambiguous identification of (causal changes in) metabolite–microbe relationships is challenging, due to limitations in (the sensitivity and comprehensiveness of) metabolomics (see in the preceding text) and data integration challenges (see following text). Furthermore, the confirmation of such causal relationships is not trivial (see following text).

Effect of domestication and breeding

We discussed in the preceding text the pivotal role of the root exudate in the recruitment of the microbiome, and the strong effects that environmental conditions have on this. In modern agriculture, nutrients are, to a large extent, supplied through fertiliser application, and pathogen infection and insect herbivory are suppressed using pesticides. Furthermore, the selection of crop genotypes is usually done under optimal conditions, such as lack of pests and diseases and optimal nutrient input. The latter may have inadvertently resulted in the erosion of plant traits required for mutualism with nutrient-providing microbes. Indeed, under low-input conditions, old barley cultivars had superior biomass production and higher concentrations of leaf macro- and micronutrients compared with modern cultivars [75]. According to the authors, one possible explanation for this is that older barley cultivars have retained the capability of their wild ancestors to synergise with the soil microbiota, enhancing nutrient acquisition in low-input systems. This was demonstrated in another study on wild and domesticated barley [76]. A study in wheat showed that domestication also resulted in changes in the root exudate [77]. We speculate that the latter is likely also true for barley and that this mediates the changes in the barley root microbiome along the domestication axis, as was shown for maize and its ancestor Teosinte [63] (see following text). In rice and soybean, rhizosphere fungal communities, including AM fungi, were shown to be more influenced by crop domestication compared with bacterial communities, and wild varieties had a higher abundance of beneficial fungal symbionts and a lower abundance of pathogens compared with cultivated varieties [78]. The authors speculated that this is likely related to the root exudate and, considering the evidence for the effect of domestication on the root exudate of other crop species, we also consider this likely, although evidence is lacking.

Breeding in maize has resulted in a decrease in the recruitment of N-fixing microbial taxa over a chronosequence of maize cultivars spanning 40 years of breeding, while there is increased recruitment of taxa that contribute to N losses [79]. The same group also compared microbiome recruitment in Teosinte and modern maize and found that modern maize recruits a more diverse microbiome, possibly as a result of weaker ‘filtering’ resulting in recruitment of copiotrophic prokaryotes and pathogenic fungi [80]. However, the authors did not analyse the root exudates of their genotypes. Thus, it will be of interest to do these analyses and see how much of the variation in microbiome recruitment is caused by changes in the root exudate.

In other work on Teosinte and cultivated maize, the root exudate composition and microbiome were shown to be different under control conditions but converging under low P conditions [63]. Both species harboured P-solubilising microbes in their rhizosphere, independent of P availability, and this did not change upon P deficiency. The authors mentioned the importance of AM fungi in P uptake, but did not report data on their presence; furthermore, strigolactones remained under the radar of their MS analysis and were not mentioned [63]. Whereas Brisson et al. only investigated inorganic P solubilisation, two recent studies in tomato and a series of additional crops demonstrated that overall P acquisition is not negatively affected by domestication [81,82]. However, the study in tomato does suggest that wild tomato is more efficient in the recruitment of P-solubilising microbes.

It has been suggested that domestication has resulted in less efficient symbiosis with, and a weakened ability to downregulate, AM fungi, which might shift their symbiosis to parasitism [83]. This might not be due to selection against strigolactones, given that, over the course of domestication, the secretion of strigolactones does not appear to have changed that much: domesticated species still produce and exude a complex blend of strigolactones, at similar levels as their wild ancestors [84].

There are many indications that domestication and breeding have resulted in changes in the root and rhizosphere microbiome of crops compared with their wild ancestors. There are also indications that there are differences in the root exudate composition of crops and their wild ancestors. However, to conclude that this is a casual relationship, and that breeding has resulted in changes in the root exudate, which in turn resulted in changes in the microbiome, is not yet possible. More research will be needed, including advanced data integration approaches.

Use of ‘omics data integration approaches to elucidate rhizosphere chemical interactions

Data-driven integration

To effectively integrate and model the relationships between plant root exudate composition and microbiome assembly, a suitable experimental design to generate (‘omics) data, data fusion, causal inference analysis, and structural causal modelling are required (Figure 2). Experimental design is vital for setting up experiments in such a way that optimal information regarding the underlying mechanisms can be obtained, and requires interaction between plant microbiome biologists and statisticians. Of special importance is to design the experiment in such a way that the data collected can be used to discriminate between alternative mechanisms, for example, to discriminate between a metabolite causing the recruitment of a microbe, or the microbe producing that metabolite. This requires informed selection of treatments, types of ‘omics data, and time-resolved measurements [85]. The selection of the time points in such a design is a challenge because the response of plants to a treatment (gene expression or changes in the root exudate composition) occurs faster than the change in the microbial community composition.

Figure 2.

Schematic overview of effective data integration and modelling requiring four steps.

Step 1: experimental design; Step 2: data fusion; Step 3: causal inference analysis; and Step 4: structural causal modelling. We focus here on the elucidation of the causal relation between changes in plant root gene expression, consequent changes in the root exudate, and changes in the plant root microbiome. In turn, the latter will also affect the former two, but that is disregarded here.

Data fusion comprises an exploratory analysis of the relationships between the ‘omics data collected according to the experimental design, such as transcriptomics, metabolomics, and metabarcoding data (Figure 2), using statistics and machine learning methods [30,86]. The final result of this stage is that insights are obtained into the global relationships between the blocks of data. A new development in this area is the integration of heterogeneous data (i.e., data of different modalities, such as binary and quantitative data). This type of data fusion can also be explicitly combined with information regarding the underlying experimental design [87,88].

After having established a relationship between certain ‘omics data, the next step involves causal inference analysis. Using the established relationship and prior biological knowledge, machine learning tools [89] allow for inferring several alternative topologies, such as how the measurements in the different data blocks (i.e., biological compartments) are related to each other in a causal way (Figure 2). Here, it is vital to use the treatment design setup and the temporal behaviour in the datasets. The result of this stage is a set of possible topologies for the causal relationships.

Finally, structural causal models (SCMs) are formulated on the basis of the alternative topologies [90., 91., 92.] (Figure 2). These models are then fitted simultaneously to all ‘omics data collected, with the treatment design and temporal aspect again being crucial. This fitting process delivers fitting diagnostics (e.g., R² values) for the different models (i.e., based on the different topologies), and these can be used to find the most probable topology. This results in causal relations and yields candidate gene–metabolite–microbe relations for biological validation, but does not answer the question as to the nature of the causal relationship.

Gene expression analysis to support metabolomics

The integration of metabolomics and microbiome data is an important tool, used by many to pinpoint potential chemical interactions between plant hosts and their root microbiota [13,18,23,35,63]. For the integration of such datasets, ideally measurements of the root exudate composition and the root microbiota are done on the same plant. This requires a root exudate collection method that is not invasive and does not affect the microbiome (Box 1) and metabolite annotation, which is challenging. Moreover, the catabolism of certain root exudate metabolites by microbes present in the system could result in a negative instead of a positive correlation between certain microbes and metabolites. Vice versa, the microbiota members also produce metabolites and may modify plant metabolites, further complicating drawing conclusions from correlations.

To address some of these challenges, the analysis of host root gene expression in such experiments is helpful. RNA sequencing (seq) is far more sensitive compared with metabolomics because a transcript multiplication step is included. Indeed, RNAseq can detect the upregulation of the expression of strigolactone biosynthetic genes under P deficiency [93,94], even though metabolomics does not detect the increase in strigolactone concentration, as discussed in the preceding text [63]. This is true to such an extent that RNAseq can be used as a tool to identify thus far unknown strigolactone biosynthetic genes, simply by looking for genes the expression of which is induced under P deficiency with a similar expression profile (across treatments) as known genes from the strigolactone pathway (co-expression analysis) (D. Abedini et al., unpublished, 2025) [59,93,95,96]. The joint use of gene expression and metabolomics data, even though both have their limitations with regard to annotation, can be useful to reduce uncertainty and identify new important genes.

Model-driven integration: genome-scale metabolic modelling

Genome-scale metabolic modelling (GEM) or model-driven integration is a powerful ‘omics-based approach that integrates genomic and biochemical data to predict microbial metabolism and infer chemical interactions within complex communities [97., 98., 99., 100.]. In the rhizosphere, GEM has been used to model 193 bacterial strains from arabidopsis roots, revealing that microbial metabolic traits are phylogenetically structured, with significant functional redundancy. This enables the design of minimal synthetic communities (SynComs) that preserve the metabolic potential of the broader microbiome [99]. By incorporating nutrient conditions and root exudate profiles into these models, researchers have shown that environmental stresses, such as nutrient limitation, can promote cooperative metabolic interactions, enhancing community stability. In a recent study modelling 270 metagenome-assembled genomes, a SynCom was rationally designed to improve crop yield by maintaining key plant growth-promoting traits [98]. A key aspect of these analyses is the quantification of biosynthetic dependencies and complementarity with the Biosynthetic Support Score as a measure of the ability of a host to fulfil the metabolic needs of a microbial partner, and the Metabolic Complementarity Index quantifying mutual support between microbial species. Although many inferred exchanged metabolites remain unidentified, key compound classes, including amino acids, lipids, organic acids, and coenzymes, were found to mediate the interaction between the host and the SynCom. These findings demonstrate the predictive power of GEM in designing microbial consortia that foster plant productivity and, reciprocally, benefit from plant-derived nutrients, offering a systems-level perspective on rhizosphere chemical interactions.

Hybrid modelling

Whereas both data-driven and model-driven approaches have their advantages and disadvantages, a logical next step is the combination of both approaches, coined hybrid modelling. In chemistry, such models are called grey models and have already proven their usefulness. There are several routes to take in hybrid modelling. One route is to combine SCMs with biochemical pathway knowledge of both plant host and microbes, for example using Bayesian SCMs [101]. Bayesian statistics is a well-developed statistical strategy used to combine data with prior knowledge; using this for SCMs of biological systems is a promising new avenue that can enrich the causal relationships found in the data-driven approach with mechanistic underpinning. Another route to take is to combine SCMs with GEMs of both the plant host and the microbiome. This could also be done in a Bayesian context, although this requires developing new methodology. Given the rapid developments in GEMs and the analysis of ‘omics data, such a hybrid methodology, once properly developed, could be very effective. Regardless of the method of data analysis and integration used, conclusions on possible (causal) relationships cannot be taken at face value, and need to be confirmed experimentally.

How to validate hypotheses on gene–metabolite–microbe relations

Several approaches can be used to validate postulated relationships between genes, metabolites, and/or microbes. A promising, relatively straight-forward, intermediate approach, covering the latter two components, is the addition of candidate metabolites to soil and the subsequent analysis of the effect on the microbiome. This approach was used to confirm enrichment, through daidzein application to soil, of legume-associated microbes, as well as enrichment, through glycoalkaloid application, of Sphingomonadaceae, which are enriched in the tomato rhizosphere [102,103]. In a further reduced version of this approach, chemotaxis assays are used to study the effect of a single chemical on the motility of microorganisms. In an assay with the synthetic strigolactone GR24, chemotaxis of a Sphingobium sp. was demonstrated (D. Abedini et al., unpublished, 2025), lending further support to a role of strigolactones in recruitment of this beneficial microbial species (see following section).

An approach to cover all three levels in the relationship is the use of plants with mutations in biosynthetic genes, using existing (transposon) mutant collections, or metabolic engineering, the genetic modification of the biosynthesis of the metabolites of interest. These mutants can be tested for the consequences of the change in their root exudate composition on the interaction of the plant host and its root microbiota.

For example, the hypothesis that triterpenoids have a role in the plant–root microbiome interaction in arabidopsis was investigated using mutants in thalianin, triterpene fatty acid ester (TFAE), and arabidin biosynthesis [40]. This research showed that these mutations altered the root metabolite profile and that these changes, particularly in the thalianin pathway, resulted in distinct root microbiome assemblages compared with wild-type Col-0. Interestingly, one bacterial strain, Pseudomonas sp. A215, was found to catabolise TFAEs into thalianol and palmitic acid, facilitating the proliferation of A215, which utilised the liberated palmitic acid as a carbon source, showing that the distinction between root-exuded organic carbon as a food source and antimicrobial compounds is sometimes vague. Similarly, insight into the role of benzoxazinoids in shaping the root microbiome in maize was obtained using a benzoxazinoid-knockout mutant, bx1 [39]. The authors showed that recruitment of 5% of the wild-type maize microbiome depended on the presence of benzoxazinoids [39].

Evidence that strigolactones also have a role in signalling to other mutualistic microbes, not just AM fungi, was initially obtained in studies using a natural sorghum mutant with different strigolactone composition [104] and later using natural variation as well as rice genotypes with mutations in strigolactone biosynthesis and perception [30]. Recently, a tomato CCD8 RNAi mutant was used to confirm the hypothesis, generated using ‘omics data integration, that the tomato strigolactone, solanacol, is involved in the recruitment of N-fixing bacteria [(D. Abedini et al., unpublished, 2025). Indeed, the root microbiota composition of the strigolactone mutant deviated from that of the wild-type, but only under N deficiency, with potentially N-providing bacteria enriched in the wild-type. One of these bacteria, a Sphingobium species, was isolated and demonstrated to display chemotaxis toward the synthetic strigolactone GR24 and to stimulate plant growth under low N (D. Abedini et al., unpublished, 2025).

So far, studies with mutants have mostly used loss-of-function biosynthetic mutants (with the exception of strigolactone signalling mutants in rice, which displayed higher strigolactone exudation [30]). Metabolic engineering strategies aimed at the specific, localised overexpression of certain metabolites would be an interesting addition to the tool kit to confirm relationships between metabolites and microbes [7]. However, for this strategy to work, tools are needed that have been mostly developed for model plant species. Expansion of these tool sets to non-model plant species is required to improve the understanding of the plant microbiome interaction in wild plant species under natural conditions.

Concluding remarks and future perspectives

Over the past two decades, research into soil-based chemical communication has expanded rapidly. Advances in detection methods and molecular tools have allowed plant scientists to explore the previously hidden world of roots and the soil surrounding them, the rhizosphere. This has revealed a remarkable chemical diversity, often surpassing that found in plant shoots. This complexity likely arises from the interaction of plants with beneficial and pathogenic soil organisms, which drives an evolutionary arms race where plants must balance specificity toward beneficial partners with defence against pathogens.

However, modern high-input agriculture may have diminished the role of this chemical interaction, because nutrients, pest control, and disease resistance are now provided through synthetic inputs. Comparisons between modern crops and their wild ancestors, as well as across different crop genotypes, suggest that parts of the chemical interaction with beneficial microbes have been lost simply because it was no longer essential. As our understanding of root-zone interactions grows, harnessing these natural mechanisms to enhance agricultural sustainability is increasingly within reach. Such approaches are crucial given the growing pressures faced by agricultural practices from environmental problems, changing climate conditions and rising input costs, threatening the long-term viability of current farming systems (also see Outstanding questions).

Outstanding questions.

Does the only recent identification of flavonoids, strigolactones and eclepins as microbiota recruitment signals imply that more of these microbiota recruitment signals have been overlooked?

Do root exudate metabolites shape the microbiome through their antimicrobial effect or by providing carbon to beneficial microbes, or both?

Do primary metabolites in the root exudate contribute to specificity in microbiota recruitment?

How can advanced metabolomics, artificial intelligence, and integrative ‘omics be further optimised to identify metabolites in root exudates, including their origin?

How has modern high-input agriculture altered the chemical interaction in the rhizosphere, and what key traits involved in these interactions might have modern crops lost?

Alt-text: Outstanding questions

Resources

ⁱ https://git.wur.nl/rhizosmash/rhizosmash