Rhizophagus irregularis modulates neighbour-primed defences in Medicago truncatula against foliar pathogens via common mycelial network-mediated inter-plant signals

Abstract

Arbuscular mycorrhizal fungi (AMF) are ubiquitous root symbionts that form common mycelial networks (CMN), linking multiple plants underground. CMN are thought to contribute to information exchange between plants and to neighbour-primed defences, but the transcriptomic and metabolomic responses of receiver plants connected through CMN remain poorly understood. In addition, the confounding effects of hyphal damage caused by CMN disconnection have not been clearly resolved. To assess the contribution of CMN integrity to neighbour-primed plant defences, we used the model AMF Rhizophagus irregularis to connect two Medicago truncatula plants and examined the effects of sender wounding and flg22 elicitation on receiver leaf responses and pathogen tolerance. We found that changes in receiver biotic stress and defence signalling pathways depend on CMN-mediated inter-plant signals rather than on damage to the mycelial network. These responses were associated with distinct changes in leaf isoprenoid production, including volatile monoterpene and triterpene compounds. CMN-mediated signals from stressed senders enhanced receiver resistance to Fusarium sporotrichoides while increasing susceptibility to Botrytis cinerea. These findings highlight the role of CMN in inter-plant signalling and pathogen-specific resistance and susceptibility, with implications for understanding plant community defence in natural and agricultural systems.

The legume Medicago truncatula transmits stress signals to neighbours through mycorrhizal network, reshaping leaf defence and metabolism and causing pathogen-specific effects, from greater Fusarium resistance to Botrytis susceptibility.

Introduction

Exchanging danger signals between plants provides an effective strategy for priming the defences of receivers and preparing them for imminent pest or pathogen attacks that have already affected their neighbours. Such inter-plant communication may contribute to enhanced community-level resilience against biotic stress and presents a promising trait for future crop improvements and agricultural engineering. Upon perception of pest and pathogen elicitors such as herbivore eggs¹ or bacterial virulence proteins², plants are known to elevate pathogen resistance in their neighbours via airborne volatile or soil-mediated signals. Furthermore, ubiquitous plant root symbionts, such as arbuscular mycorrhizal fungi (AMF), are capable of inter-connecting root systems from multiple plant hosts, effectively forming underground common mycelial networks (CMN) which are hypothesised to function as an information superhighway in transferring diverse signals from plant-to-plant in response to different sender plant stimuli^3,4. Given that the vast majority of land plants form AMF symbiosis⁵, testing this hypothesis is paramount to understanding the ecological roles and mechanisms by which CMN can contribute to the health of wild plants and crops⁶.

While a few studies^7–9 have implicated CMN for nutrient redistribution among connected plants, functional studies on CMN-mediated inter-plant signals involved in plant defences and stress tolerance have been extremely scarce to date. Moreover, in all previous studies, which used CMN disruption^10,11 to test the hypothesis that CMN mediates stress-induced inter-plant signals, the effect of CMN disruption was confounded by mechanical damage of the symbiotic hyphal network. Therefore, the neighbour plant responses to signals associated with disconnection of inter-plant CMN were not decoupled from potential plant responses resulting from CMN hyphal damage. Another key shortcoming of the previous studies is the lack of a control for inducible sender or donor plant signal. For example, previous studies in tomato¹⁰ and pea¹¹ did not include controls for pathogen-non-infected or aphid-non-infested sender plants, impeding non-biased conclusions about participation of CMN in stress-inducible as opposed to constitutive inter-plant signal exchange.

In the present study, we sought to overcome the limitations of previous studies and test two working hypotheses: whether inter-plant CMN connection, firstly, contributes to distinct receiver responses to stress-induced inter-plant signals and, secondly, mediates neighbour-primed defences against plant pathogens. To decouple the effect of CMN disruption from the symbiotic hyphal network damage, we set an experimental control where the disruption of the mycorrhizal hyphal network occurs without perturbing the inter-plant CMN continuity. We used model AMF Rhizophagus irregularis to establish CMN between two Medicago truncatula plants, then subjected leaves of signal senders to known plant defence elicitors with non-stimulated sender controls and analysed transcriptional and metabolic responses in naïve signal receiver plants. Intriguingly, we discovered CMN-connection-dependent transcriptional changes in receivers, involving strong modulation of multiple plant biotic stress and defence signalling pathways. Moreover, leaf isoprenoid, especially monoterpene and triterpene, production was elevated in receivers with the intact inter-plant CMN compared to the cut CMN, independently of hyphal damage, strongly supporting our first hypothesis. Surprisingly, after the elicitation of sender plants, CMN-connected receiver plants exhibited heightened susceptibility to the fungal pathogen Botrytis cinerea, while simultaneously showing increased resistance to Fusarium sporotrichoides. This finding aligns with our second hypothesis and reveals species-specific effects in neighbour-modulated pathogen susceptibility within CMN-linked plant systems.

Results

Wounding combined with flg22 induces local leaf responses and systemic acquired resistance to Botrytis cinerea

To test the responses of signal receivers to sender stimuli, we first wished to identify stimuli that would induce leaf responses in the signal senders themselves. To this end, we subjected leaves to mechanical wounding by forceps as well as applied a well-known plant immunity elicitor - bacterial flagellin (flg22). Since flg22 and wounding have been well documented to induce different defence responses in leaves of a model plant, Arabidopsis thaliana¹², and are required in combination to activate defence in plant roots¹³, we also decided to test the combination of both wounding and flg22. This would mimic a natural scenario where bacterial infections follow mechanical plant damage by leaf herbivores, for example. Additionally, we also infiltrated a mixture of live plant pathogenic bacteria with a broad host range (Pantoea aglomerans, Pseudomonas syringae pv. tomato, Erwinia rhapontici, Dickeya chrysanthemi) to test for their ability to trigger local responses in the infected leaves. We selected a range of known plant immunity genes as potential markers for local leaf responses, designed primers for their homologues in Medicago truncatula and performed RT-qPCR at 2 h, 4 h and 6 h following the application of the abovementioned stimuli. Leaves infiltrated with flg22 and leaves treated with the combination of wounding and flg22 (WF) displayed strong induction of MtPAD4 already at 2 h, while MtVSP1 expression was strongly induced at 4 h and 6 h following WF treatment (Supplementary Fig. 1).

24 h after stimulation of M. truncatula, we tested the performance of the fungal pathogen Botrytis cinerea on systemic leaves distal from the stimulated leaves. The WF treatment reduced the area of necrotic zones under the Botrytis inoculation site compared to water controls, characteristic of systemic acquired resistance (SAR) (Supplementary Fig. 2). Together with marker gene expression data, the WF treatment appeared to produce early local leaf responses as well as systemic responses that contribute to altered pathogen performance in our experimental system. Therefore, this treatment was chosen as a sender plant trigger for evaluating receiver plant responses in the following experiments.

CMN mediates receiver plant responses to stimulation of signal senders

We grew two Medicago truncatula seedlings in a single pot, each in a separate 30 µm nylon mesh pocket that allowed AMF hyphae and root exudates to pass through while preventing direct root contact. One of the plants in each pot was inoculated with Rhizophagus irregularis spore suspension for seven weeks to establish a CMN between the two plants. In the CMN disconnection treatment (henceforth “cut”), we severed the CMN between plants using a blunt knife. In the CMN “connected” treatment, we preserved CMN connectivity, leaving the space between plants uncut. However, to separate the effects of CMN disruption from mere hyphal damage, we disrupted the extraradical hyphal network at the far edge of the “connected” pot. This approach controlled for potential plant responses to wounded AMF hyphae, allowing us to isolate the effect of CMN continuity - a key aspect missing from previous studies^10,11. Moreover, the nylon mesh pockets also protected plant roots from accidental wounding during the CMN perturbations. Next, we mechanically wounded and applied flg22 to the sender leaves or treated them with water as the control (Fig. 1A).

Fig. 1. CMN connection contributes to wound and flagellin-induced inter-plant responses.

A A duet of Medicago plants were grown in 30 µm mesh pockets (dashed lines) and inoculated with AMF Rhizophagus irregularis to form CMN (red). Positions of the mycorrhizal hyphal network cut are indicated with a blue double-head arrow. Sender plants with the cut-CMN and the CMN connection were stimulated by wounding and flg22 application to measure responses in the receiver plant leaves. B Wounded and flg22 (WF) treated plants were first compared to corresponding H₂O treatments in each CMN treatment, respectively, to identify up- and downregulated responder plant DEGs. Afterwards, the DEGs in CMN-connected pots (red circles) and the cut-CMN (blue circles) were compared. The volcano plot displays log10 of p-adj. and log₂ fold change for each transcript in the (WF) comparison with water controls. Each treatment contained four biological replicates (N = 4). Full list of DEGs, their annotations, fold change and FDR values are available in Supplementary Data 1.

To test whether there is any effect of the position of hyphal network perturbation without induced sender signals, we compared receiver responses to water-treated senders in the pots with “cut” and “connected” CMN. Importantly, there was no significant transcriptomic change detected when the “cut” and “connected” treatments were compared (Supplementary Fig. 3), strongly indicating that, firstly, the hyphal damage was likely applied evenly in both CMN treatments and, secondly, the position of the CMN interference alone does not contribute to any significantly differentially expressed genes (DEGs) in receiver plants. Therefore, this result permits a fair comparison of how the integrity of inter-plant CMN connection contributes to unique receiver responses after sender stimulation by mechanical wounding and flg22 application. To further test this, we stimulated the sender plants and compared receiver responses to non-stimulated senders in the “cut” and “connected” CMN treatments (Fig. 1).

We found that naïve receiver plants display distinct transcriptional responses to sender stimulation (WF vs H2O) when CMN is cut compared to uncut inter-plant connection (Fig. 1B). We identified 1880 up- and 3005 downregulated DEGs that characterise CMN-connection-dependent receiver responses (Fig. 1B). Interestingly, among them were twelve DEGs, including M. truncatula homologue of plant defence gene PHATHOGENESIS RELATED 1 (MtrunR108HiC_038642), which show significant downregulation in the receiver plant connected by CMN but upregulation in the cut-CMN treatment (Supplementary Data 1; tab-D). In contrast, 3 DEGs - transcription factor NAC25 (MtrunR108HiC_019250), peptidoglycan-binding LysM domain protein (MtrunR108HiC_035065) and RLK-Pelle-LRR-XII-1 family receptor kinase (MtrunR108HiC_036113) were upregulated in the receivers with connected CMN but downregulated in the cut treatment (WF vs H2O).

CMN-dependent inter-plant signals transcriptionally remodulate plant defence

Interestingly, gene functions relating to DNA metabolic processes - DNA repair (integration) and phosphotransferase activity - were significantly enriched within the 1880 upregulated DEGs (Supplementary Fig. 4; Supplementary Data 2 tab-A). Functions with the most significant enrichment by downregulated DEGs were related to translation, RNA binding and processing, ribosomal subunit structure and organisation as well as plant stress responses - primarily ROS metabolism, response to stimuli and jasmonic acid (Supplementary Fig. 5; Supplementary Data 2 tab-B).

GO-term annotations were available for around 84% of the 4885 DEGs specific to the uninterrupted CMN treatment (Supplementary Data 2 tab-C), representing a good functional assessment. Complementary, we also performed an independent de novo functional prediction with all M. truncatula mRNA fasta sequences using Mercator ortholog-based function inference. Congruent with our first GO-term analysis, the Mercator enrichment analysis listed protein phosphorylation and phosphotransferase activity to be enriched by the upregulated DEGs (Supplementary Data 2 tab-D) and RNA processing and protein biosynthesis (Supplementary Data 2 tab-E) by the downregulated DEGs.

Intriguingly, after careful inspection of the original bin descriptors from the automated Mercator annotation (Supplementary Data 3 tab-A), it was evident that many of the pre-assigned bins relate to plant defence (Supplementary Data 3 tab-B). Therefore, we adopted a similar approach to the recent study in Arabidopsis¹⁴ and manually curated a complementary “Defence signalling & response” functional bin (Supplementary Data 3 tab-C) for more accurate functional classification within the original hierarchy of de novo prediction by Mercator (Supplementary Data 3 tab-D).

Interestingly, the curated “Defence signalling & response” category contained a higher cumulative transcript fold change of the 4885 DEGs for the CMN-connected compared to the 5683 DEGs specific to the cut-CMN treatment (Fig. 2A) and displayed the strongest WF treatment-induced inter-plant responses alongside protein biosynthesis and photosynthesis in receivers with inter-plant CMN connection (Supplementary Data 3 tab-E). The defence signalling & response in CMN “connected” receivers showed downregulation of enzymes involved in ROS homeostasis, receptor-like kinases and intracellular MAPK and Ca-dependent kinases, as well as the Ub-proteasome pathway (Fig. 2B). Furthermore, genes involved in jasmonic acid synthesis and perception were downregulated compared to the salicylic acid pathway. In addition, receiver plants displayed transcriptional changes in genes encoding specialised metabolites such as terpenoid and flavonoid synthesis.

Fig. 2. CMN connection is required for remodelling of biotic stress responses in inter-plant signal receivers.

A Cumulative transcript fold changes of plant functions encoded by up- (red) and downregulated (green) DEGs display different and specific responses to sender stimuli in the connected and cut-CMN treatments. Visualisation is based on Mercator4 functional annotation via PageMan. The ranked enrichment for each category is provided in Supplementary Data 3 tab-E. B 884 DEGs corresponding to plant biotic stress responses and plant defence signalling display dramatic transcriptional responses in responders with inter-plant CMN connection. Green colour represents downregulated, red – upregulated DEGs with colour intensity proportional to their log₂FC. Each treatment contained four biological replicates (N = 4). List of displayed DEG annotations and individual fold changes is provided in Supplementary Data 1 tab-B. The image file for MapMan visualisation of ‘’Defence signalling & response” pathways is available for download as Supplementary Fig. 6. Mapping file containing bin-assignment to immune receptors, intracellular signalling components and other defence-related functions is available in Supplementary Data 3 tab-D.

We selected a subset of receiver genes showing among the strongest positive or negative responses to sender stimulation in the RNA-seq dataset under either CMN-connected or CMN-cut conditions (Supplementary Data 7) and examined their expression in an independent experiment by RT-qPCR (Supplementary Fig. 10A). Overall, concordance between RNA-seq and RT-qPCR was stronger for the CMN-connected comparison than for the CMN-cut comparison (Supplementary Fig. 10B), supporting the conclusion that receiver transcriptional responses to sender stimulation are CMN dependent, although not all individual genes were reproducibly validated across the repeated experiments. Only PP1 (putative F-box protein interactor) displayed significant induction by wounding and flg22 treatment in the CMN-connected receivers and not in the cut-CMN treatment (Supplementary Fig. 10A), supporting the CMN-dependent signalling effect. No other tested gene showed significant changes when comparing stimulated versus water-treated sender effect on receivers with either cut or connected CMN (Supplementary Fig. 10A). However, several genes displayed similar direction of fold change (Supplementary Fig. 10B) and tendencies for induction or suppression in the repeated experiments. For example, ABP19 (auxin binding protein), PrP4 (proline-rich protein 4), PC (putative endochitinase), NAC47 transcription factor, and DC2 (proline-rich lipid transport protein encoding) displayed a tendency for suppression in receivers with connected CMN, albeit not significant in the repeated experiments. In contrast, LP4 (LRR extension-like), PBP1 (Ca-binding protein), gene encoding putative transmembrane protein (PP2) and premnaspirodiene oxygenase (PPO) did not robustly replicate the treatment responses observed in the RNA-seq experiment (Supplementary Fig. 10).

Nevertheless, an intriguing observation from the RNA-seq experiment is the relatively high number of upregulated and low number of downregulated DEGs which represent exclusive responses in the receivers with cut CMN compared to connected CMN (Fig. 1B). Certain gene functions such as cell division, protein glycosylation, amino acid and nucleotide metabolism appear to be more enriched in the cut treatment (Fig. 2A). Interestingly, when we investigated plant stress and defence-related functions (Supplementary Fig. 7), most of the transcripts from the cut-CMN treatment displayed upregulation in sharp contrast to downregulation in the connected CMN treatment (Fig. 2B). This strongly indicates that the receiver responses are highly influenced by the CMN integrity. Given that plant volatiles were not restricted in either the connected or cut treatment, this finding may suggest that the receiver responses to CMN-mediated signals could interact with responses from inter-plant volatile signals to affect gene regulation via yet unknown mechanisms.

Inter-plant signals modulate volatile and metabolite composition in a CMN-dependent manner

To further test whether inter-plant CMN connection influences the potential infochemical production and specialised metabolite composition in the receiver plant, leaf metabolites and volatiles were analysed using an untargeted metabolomics approach. When Medicago senders were wounded and treated with flg22 (Fig. 3A), CMN-connected responders display significantly reduced production of volatile leucinamide (V2) and increased production of monoterpenoid β-fenchyl alcohol (V5) alongside several other putatively annotated (MSI level 3) leaf compounds such as medicagenic acids (L45; L47), acetylsoyasaponin (L46), genistein 7-O-glucoside (L15), 1H-Indole-3-acetaldehyde (L3), tyrosine (L8) and hex-bayogenin (L42) when compared to the H₂O-treated senders (Fig. 3B). Notably, these compounds were not significantly over- or under-produced in inter-plant signal receivers with the interrupted CMN, suggesting CMN-dependent metabolic changes to neighbour signals. In contrast, leaf metabolites soyasapogenol E (L51) and ethyl-D-glucopyranoside (L9) were significantly upregulated while mannopyranoside compound (L52), chrysoeriol (L19) and tricin GlcAGlcA (L20) were downregulated in receivers with the cut CMN as opposed to connected CMN (Fig. 3B), suggesting that CMN-dependent signals may additionally dampen certain leaf responses to neighbour stress.

Fig. 3. CMN modulates leaf metabolite and volatile production in response to neighbour stress-induced signals.

A Two Medicago plants were grown in 30 µm mesh pockets (dashed lines) and inoculated with AMF Rhizophagus irregularis to form CMN (red). The blue double-headed arrow indicates the position of CMN interruption prior to sender plant wounding and flg22 application to measure responses in the receiver plant leaves. B Volcano plots displaying downregulated or upregulated leaf metabolites (L) and volatile compounds (V) when comparing inter-plant responses to wound and flg22 with H₂O controls in receivers with connected (top) or cut (bottom) CMN. C Hierarchical clustering of colour-coded log₂ fold changes (FC) for all 53 detected leaf metabolites and 27 volatiles. Two clusters of compounds display the largest relative differences in receiver responses to inter-plant stress signals between connected (con) and cut-CMN treatments. The asterisks indicate significant FC based on the volcano plots in (B). Each pairwise comparison is based on at least four biological replicates (N = 4) per treatment. Compound predictions are based on available standards for MSI level 1 or fragmentation spectra and retention times (tR) for MSI levels 2 and 3. Volatile compounds (GC-MS) are labelled by their molecular ion (M+), whereas soluble compounds (HPLC-MS) are labelled by the detected ion m/z. For illustrative purposes, putatively annotated compound or class representative structures from ChemSpider and PubChem databases are displayed. D The network produced by regularised canonical correspondence (rCCA) analysis displays the correlation between normalised CPMS read counts of 4885 DEGs (identified in Fig. 1) and relative metabolite quantities. Only correlations with a threshold >0.7 are displayed and include 4 metabolites (orange). The available homology-based gene function predictions from the MtrunR108_HiC reference genome are mapped on the network along with colour-coded gene expression changes in CMN-connected receivers. The source data for metabolite identification and network generation are provided in the Supplementary Data 4. Analytical standards were used to quantify selected metabolites and validate correspondence to independent samples in Supplementary Fig. 11.

To assess the robustness of the interpretation that CMN connection modulates leaf metabolite accumulation, we re-analysed a subset of metabolites using analytical standards and independent samples. Among the validated metabolites, (S)-homostachydrine (L7), leucine/isoleucine (L10), phenylalanine (L11), tryptophane (L12) and apigenin derivative (L28) showed no significant differences in measured concentrations between the original and the independent experiment in any of the four treatments, while proline (L5), jasmonic acid (L32) and medicagenic acid derivatives (L45, L46, L47) showed similar levels in at least three of the four treatments (Supplementary Fig. 11A). Leaf metabolite levels from the repeated experiments were strongly correlated across all four measured treatments (r = 0.789 for WF connected, r = 0.753 for WF cut, r = 0.726 for H2O connected, r = 0.683 for H2O cut), suggesting that differences between the original and independent experiments were not confounded by a specific sender treatment or CMN status.

While amino acids proline (L5), isoleucine (L10), phenylalanine (L11), and tryptophan (L12) showed similar levels of accumulation within each treatment across the repeated experiments (Supplementary Fig. 11A), they did not display significant responses to sender signals in either connected or cut-CMN receivers (Supplementary Fig. 11B) and showed opposite trends in accumulation between repeated experiments within cut CMN (L7, L10) and connected CMN (L12) (Supplementary Fig. 11C). Apigenin derivatives (L18, L21–24, L28) displayed much reduced accumulation in receivers across all treatments in the repeated experiment compared to the original, and tyrosine (L8) was not detected in the repeated experiment at all (Supplementary Fig. 11A, B). The majority of the tested apigenin compounds did not display significant changes or consistent fold change to neighbour signals in either CMN-connected or cut plants in both repeated experiments (Supplementary Fig. 11B, C). However, sender stimuli significantly decreased apigenin 7-O-diclucoronide (L18) accumulation in receivers from the cut but not the connected CMN treatment (Supplementary Fig. 11B). Similarly, we originally detected accumulation of soyasapogenol E (L51) in the cut CMN relative to the CMN-connected treatment (Fig. 3B, C), but we could not robustly replicate this finding with the independent samples (Supplementary Fig. 11A–C). Therefore, the compounds L8, L51, L21-L23, and L28 may not be considered as robust CMN-dependent plant signalling markers.

Nevertheless, the medicagenic acid (L47) displayed significantly greater accumulation in CMN-connected receivers compared to the cut-CMN treatment (Fig. 3C) in response to WF. This is consistently reproducible with the analytical standards and independent samples as well (Supplementary Fig. 11B). Similarly, the medicagenic acid derivative (L45) also display similar trend for higher induction in CMN-connected receivers compared to cut CMN (Supplementary Fig. 11B, C), albeit being significant only in the original measurement (Supplementary Fig. 11B). Together, the independent experiment corroborates the finding that inter-plant CMN connection modulates the magnitude of medicagenic acid (in particular L47) accumulation in receivers and strongly implicates these pentacyclic triterpenoids as markers for CMN-dependent responses.

All leaf samples for metabolomics were collected from the same plants as the samples for RNA-seq, enabling investigation of whether there is correlated change in leaf metabolites and defence gene expression using the set of 4885 DEGs specific to wounding and flg22 response in CMN-connected receivers. Surprisingly, proline (L5) and soyasapogenol E (L51) demonstrated the strongest correlation with a number of genes involved in plant defence responses (Fig. 3D), including plant surface LRR Ser/Thr kinase receptors, including the flagellin receptor FLS2, PATHOGENESIS RELATED 4, E3 Ub ligases PUB22 and PUB23, ethylene and gibberellin response factors as well as several transcription factor families, especially MYB-related and GRAS. Additionally, aquaporin gene NIP2 expression was strongly correlated with production of ethyl-D-glucopyranoside (L9) and HO-α-muurolene (V24) (Fig. 3D).

CMN-mediated inter-plant signals induce acquired susceptibility to Botrytis cinerea but resistance to Fusarium sporotrichoides

Since we identified CMN-dependent changes in receiver plant biotic stress genes and leaf metabolites with reported functions in plant–pathogen interactions, we wished to test whether stressed sender plants modulate resistance to known plant pathogens in CMN-connected receivers. We evaluated leaf infections by two plant generalist fungal pathogens - Botrytis cinerea and Fusarium sporotrichoides. Importantly, when the inter-plant CMN connection was cut, there was no significant difference in receiver lesion size upon B. cinerea or F. sporotrichoides infection between sender treatment of WF and water (Fig. 4B). Interestingly, when we added the generalist pathogen F. sporotrichoides to naïve receiver plants, the necrotic lesions were significantly smaller when sender plants were wounded and treated with flagellin (neighbour-primed) compared to water-treated senders (Fig. 4A), suggesting inter-plant signal-induced resistance to the pathogen attack. However, when we exposed CMN-connected naïve receiver plants to another pathogen B. cinerea, the necrotic lesions were significantly larger when sender plants were wounded and treated with flagellin compared to water-treated senders (Fig. 4A). Surprisingly, this effect was the opposite to intra-plant signals induced by wounding and flagellin treatment of distal leaves (SAR) which originally reduced lesion size in B. cinerea infected distal leaves (Supplementary Fig. 2). This suggests that intra- and inter-plant signalling induced by wounding and flagellin may lead to different outcomes, possibly due to distinct underlying mechanisms. Our findings underscore the important role of the intra-plant CMN connection in how stress-induced signals affect pathogen resistance in neighbouring plants. Additionally, we highlight that stress-induced inter-plant signals in mycorrhiza-connected plants can result in varying levels of tolerance to different fungal pathogens in the receivers.

Fig. 4. Stress-induced inter-plant signals modulate resistance to fungal pathogens in CMN-connected plants.

A Size of necrotic lesions on Medicago truncatula leaves 72 h after inoculation with Botrytis cinerea and Fusarium sporotichoides in CMN-connected plants. Wounding and flagellin (WF) treatment of sender plants induces resistance to F. sporotrichoides (F = 21.355; p = 2.629e-05) but susceptibility to B. cinerea (F = 5.6603; p = 0.021). B Size of necrotic lesions on Medicago truncatula leaves 72 h after inoculation with Botrytis cinerea and Fusarium sporotichoides in receivers with cut CMN. WF treatment of sender plants does not significantly alter receiver plant lesion size upon F. sporotrichoides (F = 2.089; p = 0.155) or B. cinerea (F = 0.3496; p = 0.556) infection compared to receivers next to the H₂O-treated senders. Experiments in (A and B) were performed independently of each other. Pathogen assays in each graph were repeated 3 times independently and represented as pooled data. Line indicates the median, boxes represent the interquartile range (IQR), whiskers represent the variance within 1.5× IQR. Detailed comparison of all 3 individual experiments for each panel along with normality tests and repeated measures analysis is provided in Supplementary Fig. 8. Each treatment contained ≥8 biological replicates (N ≥ 8) (receiver plants) per repeated experiment. Raw data are available in Supplementary Data 5.

Discussion

For a few decades, CMN has been hypothesised to function as an inter-plant signalling conduit^3,4,6,15. However, to date our mechanistic understanding of the molecular responses to stress-induced signals in mycorrhizae-interconnected plants has been limited by very few empirical studies. Importantly, the interruption of CMN in tomato¹⁰ or pea¹¹ plants was confounded by physical damage to symbiotic hyphae. Moreover, the few existing studies were all based on targeted gene expression or enzyme activity measurements^10,16, rather than characterising global responses of receiver plants to reveal the key pathways involved in the CMN-mediated neighbour priming against pathogens in an unbiased way. In this study, our experimental design allowed us to separate the effects of CMN disconnection from damage to the symbiotic hyphal network (Supplementary Fig. 3).

While in vitro studies have reported the capacity for reestablishment of hyphal connections after hyphal cutting¹⁷, the rate of hyphal growth, estimated to be around 10–20 µm/h in vitro¹⁸ or 1–3 mm/d on soil¹⁹ would not exceed the distance between nylon mesh in our experiment and is unlikely to confound the experimental results¹⁷. Furthermore, the reported hyphal length density estimates on soil^19,20 reflect cumulative length of multidirectional branching rather than growth along a single axis. The presence of obstacles, such as soil particles, would induce hyphal branching or rerouting around the obstacle²¹, thus increasing time to reconnect the CMN in our experimental system.

Using an untargeted approach, we have identified a unique profile of volatiles, leaf metabolites, and transcriptional responses in receiver plants that share an AMF connection with biotic stress-stimulated signal senders, compared to plants with severed AMF connections (Figs. 1 and 3). Furthermore, this manifested in altered plant susceptibility to pathogenic fungi Botrytis cinerea and Fusarium sporotrichoides (Fig. 4). Together, our study implicates CMN as important players for defence signal exchange. To establish this as a general phenomenon within plant communities in nature and agricultural systems, future research should expand the experimental scope to include diverse AMF and plant taxa and investigate the communality of plant regulatory and metabolic pathways that are modulated by neighbour signals. Furthermore, extended effects of CMN-mediated inter-plant signals on nodulation and plant interaction with plant growth-promoting or disease-suppressing²² bacteria are another important avenue for future research in legume and non-legume crops and ultimate harnessing of CMN functions for agriculture.

The majority of the putatively annotated (MSI level 2–3) receiver leaf compounds responding to CMN-mediated neighbour signals are implicated in plant defence during biotic stress. One of the important key metabolic changes highlighted was the increased production of triterpenoid saponins (including hex-bayogenin (L42), medicagenic acids and their derivatives L45–L47) (Fig. 3B). Interestingly, medicagenic acids, bayogenin and other saponins are well-known important legume triterpenoids that act as phytoanticipins (preformed defence compounds) and phytoalexins (induced defence compounds), contributing to defences against pathogens by perturbing microbial membrane integrity^23–27. Additionally, levels of numerous monoterpene leaf volatiles were among the most notable differences between receivers in the connected and cut treatments (Fig. 3C). Monoterpenoid α-muurolene, along with other terpenoids from endophytes of Catharanthus roseus, are known to display antifungal properties²⁸. These compounds display inter-plant stress-induced production in CMN-connected receivers but not in the receivers with cut CMN.

Interestingly, receiver plants with intact AMF connection displayed increased production of 1H-Indole-3-acetaldehyde (IAAld) (L3) – an intermediate for auxin synthesis. As reviewed²⁹, auxins have important roles in plant-microbe interactions. Curiously, P. syringae strain DC3000 produces IAA as a plant virulence factor from IAAld using dehydrogenase AldA in order to suppress salicylic acid-mediated defences in A. thaliana³⁰. Furthermore, many auxin efflux proteins were significantly upregulated in the connected CMN treatment (Supplementary Data 6 tab-A). Among them was the vacuolar auxin transporter WALLS ARE THIN1 (WAT1)-related gene MtrunR108HiC_014019. AtWAT1 is also known to be involved in cell wall thickening³¹. Since CMN-connected plants also displayed transcriptional changes in cell wall components as well as auxin synthesis and perception (Fig. 2), we hypothesise that auxin homeostasis may in part contribute to cell wall remodelling and specialised metabolite production in defence against pathogens in CMN-connected Medicago plants.

The integrity of CMN was not only responsible for increased metabolite production but also suppressed specific metabolites in response to neighbour signals. Certain triterpenes such as ethyl-D-glucopyranoside (L9) or soyasapogenol E (L51) demonstrated more pronounced induction in the cut-CMN treatment as opposed to the connected treatment, suggesting that CMN-mediated inter-plant signals may dampen the inter-plant effects of other CMN-independent signals. Soyasapogenol E levels were negatively correlated with expression of several ethylene response factors and members of MYB and GRAS transcription factor gene families (Fig. 3D), which were well represented among the 4885 DEGs specific to the connected treatment (Supplementary Data 6 tab-A) and may indicate their regulatory roles in triterpenoid production. CYP450s and UDP-glycosyltransferases (UGTs) are primarily involved in mevalonate-dependent isoprenoid biosynthesis, including monoterpenes, sesquiterpenes and triterpenes³². We found sets of up- and downregulated CYP450 and UGT gene transcripts that are specific to the cut and connected treatments (Supplementary Data 6 tab-B, C) and could be involved in the observed CMN-dependent differences in metabolite production. Lastly, leaf proline content was significantly correlated with expression patterns of several genes involved in plant defences, including flg22 receptor FLS2 and several E3 Ub ligases (Fig. 3D). Proline and its derivatives, pyrroline-5-carboxylate, are involved in plant responses to abiotic and biotic stress, including pathogen defence and mitigation of oxidative damage by scavenging reactive oxygen species (ROS)^33,34. While the relative changes in proline content after sender stimulation were not significantly different between the cut and connected treatments (Fig. 3C), genes involved in ROS production and scavenging were significantly enriched among the 4885 DEGs specific to the connected treatment (Supplementary Fig. 5).

CMN-transmitted inter-plant stress signals have been previously considered to provide fitness benefits to the receiver by providing increased pathogen resistance¹⁰ and recruitment of natural enemies¹¹. However, our data on differential inter-plant M. truncatula responses to B. cinerea and F. sporotrichoides infection clearly demonstrated that the outcome may not always be positive for the receiver. We emphasise that the potential beneficial effects from eavesdropping on conspecific signals via CMN must be holistically studied in the context of different pest and pathogen species as well as host genotypes⁶. Intriguingly, neighbour-modulated susceptibility (NMS) has been documented in rice and durum wheat growing next to pathogen-free conspecifics of different genotypes and appears to be mediated by inter-plant signals in soil³⁵. In particular, the study showed that wheat growing next to a different conspecific genotype displays increased susceptibility to hemibiotroph Zymoseptoria tritici but reduced susceptibility to biotroph Puccinia triticina, suggesting different production or perception of signals by different plant genotypes. Similarly, Magnaporthe oryzae showed fewer lesions on rice neighbouring a different conspecific genotype compared to an isogenic cultivar, while another hemibiotroph, Bipolaris oryzae, demonstrated more lesions³⁵. Arabidopsis plants exposed to elicitors in insect eggs are able to produce inter-plant signals that induce broad-spectrum resistance against the bacterial pathogen Pseudomonas syringae¹ and the fungus Botrytis cinerea³⁶, suggesting that certain plant-to-plant signals may be potentially more conserved and less specific to different pathogen species. This further highlights the need to test a range of different elicitors on senders and evaluate responses in responders of different genotypes and species in future experiments.

However, in our foundational experiments presented herein, we wished to use stimuli or elicitors that would give consistent and reproducible PAMP or DAMP responses without the added complexity and potential variability of using live herbivore or pathogen infections. For example, insect oral secretions are known to contain additional chemical elicitors that are recognised by plant HAMP receptors or deliver insect effector proteins that suppress plant defence responses³⁷. Similarly, microbial pathogens produce effectors that either suppress PTI responses in plants or induce effector-triggered immunity (ETI)³⁸, thus introducing mechanistic complexity for the generation of systemic and inter-plant signals in the senders. Furthermore, different feeding guilds of herbivores, such as cell-content feeding, leaf or root chewing, stem or leaf boring, phloem or xylem sucking, as well as the range of pathogens with biotrophic to necrotrophic lifestyles and diverse effector repertoires, offer an extended scope for future investigations on plant-to-plant responses to these diverse interactions.

Neighbour-induced susceptibility or resistance would provide opposing selective pressures for the evolution of specific signal generation and perception mechanisms along with receiver responses. For example, volatiles are known for their roles in pollinator as well as parasitoid attraction. Since Medicago truncatula is primarily a self-pollinator, it is unlikely that the altered receiver VOC profiles would affect plant reproductive fitness directly but may be important in other outcrossing species. Interestingly, volatile monoterpenes (e.g., penene, campene, sabinene) have been shown to induce inter-plant systemic acquired resistance (SAR) against Pseudomonas² and could potentially relay SAR signals from plant-to-plant³⁹. While we did not find changes in the same monoterpene compounds², it would be intriguing to test whether VOCs could further propagate and amplify CMN-mediated inter-plant signals and thus affect plant defences at a community level. Moreover, VOC and specialised metabolite composition is very diverse across angiosperms. For example, the lineage-specific number and expansions of terpene synthase genes across angiosperms suggest functional divergence in terpene structure and regulation in different ecological conditions and plant genotypes^40–42. Therefore, metabolic effects of CMN-mediated inter-plant stress signals could have diverse effects in crops under field conditions.

Conclusion

Receiver plants’ biotic stress responses and defence signalling pathways can be uniquely shaped by inter-plant signals transmitted via CMN and manifest in differential disease outcomes. These findings underscore the potential pivotal role of CMN in modulating plant immunity at a community scale, revealing a sophisticated and nuanced mechanism of pathogen-specific susceptibility and resistance. This insight prompts further investigation into the mechanisms of CMN-mediated inter-plant signalling, which could be transformative for understanding plant defence dynamics in complex natural ecosystems and for optimising disease management strategies in agriculture.

Materials and methods

Plant growth

30 µm pore size autoclavable PA6-grade nylon mesh (JPP140 from Technical textiles, Ekobalta, Latvia) was used to make 11 cm wide and 10 cm deep pockets which were filled with autoclaved soil:sand mixture (3:1 v/v). Soil (gardening substrate with black earth) was purchased from Biolan Baltic OÜ (pH: 6,0; EC: 25 mS/m; N~100 mg/L; P~60 mg/L; K~200 mg/L; peat fraction <20 mm). Quartz sand (grain size ≤0.5 mm) obtained from SACRET^®, Latvia. Two substrate-filled mesh pockets were placed into 1 L black plastic pot (11 × 11 × 12 cm L × W × H) and surrounded with the surplus soil:sand mixture to prevent them from touching each other or the edge of the pot. Medicago truncatula R108 seeds were scarified with a sterile scalpel and germinated on wet Whatman filter paper for 2–3 days before planting onto the soil:sand mixture within each mesh pocket. The images of the setup are provided in the Supplementary Fig. 9. Plants were grown in growth cabinet (SANYO MLR-351H, Japan) at 20 °C, 50%RH, 16 h day/8 h night cycle using a set of twelve 6500 K (3200 lm) and three 2700 K (3350 lm) fluorescent lightbulbs (T8 MASTER TLD G13 Philips, 36 W) to cover the entire PAR spectrum with 300 mmol m² s⁻¹ light intensity. Each pot was watered with 200 mL 0.5× Hoagland nutrient solution once a week.

AMF cultivation and inoculation

Rhizophagus irregularis (DAOM 197198) was cultivated on Daucus carota hairy root cultures^43–45, grown on M-media (1% sucrose, 0.4% Gelrite gellan gum) at 25 °C in darkness. 6-month-old cultures were processed for R. irregularis spore collection using 10 mM citrate buffer (pH = 6) to dissolve spore-containing solid media and capture spores within 40 µm cell stainer (SPL Lifesciences) and resuspend in sterile dH₂O. Spore density was determined microscopically. Suspension volume containing 200 spores was applied to one of the seedlings in the pot. Plants were grown for 7 weeks to establish CMN between the sender and receiver plants. The colonisation status of the receiver plant was verified using the ink-vinegar staining method⁴⁶ after the experiment.

Treatment of CMN and sender plants

7-weeks post AMF inoculation all experimental pots were divided into two groups: (1) uninterrupted CMN (sender-receiver plants remain connected via CMN) where a blunt knife was inserted into the soil along the edge of the pot as a control and (2) a group of cut CMN where a blunt knife was inserted into the soil between the mesh pockets to disrupt the CMN connection between signal sender and receiver plants as depicted in Supplementary Fig. 9. Three days after this manipulation, sender Medicago plants were subjected to mechanical wounding by applying three squeezes with metal forceps to each leaflet of two fully expanded trifoliate compound leaves, followed by the application of a 5 μL drop of synthetic 1 μM bacterial flagellin (flg22, MedChemExpress, Sweden) on each forceps-wounded leaflet. 5 μL of dH₂O without mechanical wounding was used as the control treatment. Receiver plants were left untreated and collected 3 days after the stimulation of the sender plants.

RNA extraction and transcriptome sequencing

Each treatment included four biological replicates, defined as four independent RNA-seq libraries derived from four separate receiver plants. For each library, three separate trifoliate leaves from a single 8-week-old receiver plant were pooled and harvested in liquid N₂, homogenised with pestle and mortar for RNA extraction using ReliaPrep^TM RNA Miniprep System (Promega, cat. Z6012). RNA quantification was performed with Qubit™ RNA BR Assay Kit and Agilent Bioanalyzer RNA 6000 Pico Kit. Samples from 4 different plants were processed as biological replicates for each treatment. A total of 400 ng of each RNA sample (RIN ≥ 7, OD_260/280 between 1.9 and 2.1, OD_260/230 between 1.5 and 2.0) was processed for library preparation using MGIEasy RNA Directional Library Prep Set following the manufacturer’s instructions. QIAseq FastSelect –rRNA Plant Kit (QIAGEN) was used to deplete rRNA prior to the library fragmentation step. Qubit^® dsDNA HS Assay and Agilent HS DNA Kit were used to determine library concentration and length (average 326 nt) prior to multiplexing and circularisation. The final pooled libraries (1.94 ng/µL) were sequenced on a DNB G400RS high-throughput sequencing machine (MGI) using an FLC PE150 flow cell.

Raw data processing and DEG analysis

Read filtering and adaptor removal were performed with fastp 0.20.0 (default parameters; read trimming >100 bp). STAR 2.7.10a⁴⁷ was used to map the reads against the Medicago truncatula R108 reference genome⁴⁸ available at https://medicago.toulouse.inrae.fr/MtrunR108_HiC (INRA, Toulouse). SortMeRNA version 4.3.6⁴⁹ was used to quantify and filter rRNS using the smr_v4.3_default_db database. DEG analysis was performed with limma (Bioconductor)⁵⁰. Low-count features were filtered using the function filterByExpr implemented in edgeR (v4.6.2). Voom with quality weights was used to transform the data prior to linear modelling with limma. Features with FDR < 0.05 were regarded as statistically significant and used for downstream selection of CMN-connection-specific inter-plant responses.

Functional enrichment analysis

Enrichment of GO biological and molecular functions for DEGs was calculated using agriGO v2.0⁵¹ Fisher exact test (FDR) with MtrunR108_HiC_functional annotation file as reference and visualised with enrichment bubble SRplots. To independently validate our enrichment analysis, we performed complementary enrichment analysis with the de novo gene function predictions obtained with the Mercator4 v2.0 tool⁵². To this end, MtrunR108_HiC.1.0.fasta_features (all genomic mRNA nt sequences) were used as input for the gene orthology search (Supplementary Data 3; tab-A). The automated function descriptors were manually checked and complementary assigned to the “Defence signalling and response” bin (Supplementary Data 3; tab-B). Since a distinct and comprehensive “plant stress and defence” function category was originally absent from the automated functional assignment (Supplementary Data 3; tab-A), we extracted the corresponding gene identifiers and de novo compiled the “plant stress and defence” bin (BINCODE 80) (Supplementary Data 3; tab-C). This list was then combined with the original list of Mercator4 functions from tab-A to produce an updated input (mapping) file for functional enrichment analysis, available in Supplementary Data 3; tab-D. Subsequent visualisation of DEG functions was performed with MapMan v3.7⁵³ and PageMan⁵⁴ functional annotation tools. Statistical analysis on DEG enrichment for MapMan pathways was performed using the built-in Wilcoxon rank test and Benjamini–Hochberg (BH) p-value correction⁵⁵. The graphical overview of plant defence pathways displayed in Fig. 2 was drawn manually in Inkscape and loaded as a separate pathway image file into MapMan (available for public download as .jpeg in Supplementary Fig. 6 and reusable by citing this manuscript).

RT-qPCR

\To verify the expression data from the RNA-seq dataset, we performed RT-qPCR on selected genes from independent samples. Test genes were selected based on the largest FC difference among connected and cut treatments (Supplementary Data 7 tab-A). Reference gene candidates were selected based on the lowest fold change across all treatments (Supplementary Data 7 tab-B). M. truncatula gene-specific primers (Supplementary Data 7 tab-C) were designed in NCBI Primer BLAST for 100–300 nt amplicons spanning the exon-intron junction where possible. cDNA was synthesised with Maxima First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) and qPCR performed with Maxima SYBR Green/ROX qPCR Master Mix (2X) (Thermo Fisher Scientific) on ViiA™ 7 Real-Time PCR System (Applied Biosystems) and analysed with the native QuantStudio 7 Pro 1.6.1 software. Primer efficiency (E%) was calculated using the formula $\left(\frac{-1}{{10}^{\mathrm{Slope}}-1}\right)\times 100$, with slope being obtained by plotting the log₁₀ values of 10-fold serial dilutions from 1 ng to 1ag of purified cDNA (reported in Supplementary Data 7 tab-C). The expression of each test gene was normalised by the geometric mean of two selected reference genes (MtEXO70 and MtBP28) using the formula $\mathrm{relative\; expression}=\frac{\sqrt{E_{ref1}^{Ct\left(ref1\right)}{xE}_{ref2}^{Ct\left(ref2\right)}}}{E_{test}^{Ct\left(test\right)}}$, where E_ref – primer efficiency of reference gene, E_test – primer efficiency of test gene, Ct(ref) – Ct value of reference genes, Ct(test) – Ct value of the test gene.

Leaf metabolic analysis with LC-HRMS

Metabolome was analysed from the same leaf samples as the plant transcriptome by aliquoting 200 mg of the frozen homogenised leaf tissues and resuspending in 1.5 mL of 70% methanol, vortexed for 1 min and left for maceration for 5 days at 4 °C in darkness. The obtained extract was filtered through a 0.45 µm filter and injected into the Agilent 1290 Infinity II series HPLC system combined with an Agilent 6530 qTOF MS system (Agilent Technologies, Deutschland GmbH, Waldbronn, Germany). Zorbax Eclipse Plus C18 Rapid Resolution HD (2.1 × 150 mm, 1.8 μm particle size) column was used at a flow rate of 0.3 mL/min. The column oven was set at 45 °C, and the sample injection volume was 3 μL with 40 s needle wash (70% methanol). The mobile phase consisted of a combination of A (0.1% formic acid in water) and B (0.1% formic acid in acetonitrile). Gradient elution programme was used as follows: Initial 5%B, 0–6 min 20%B, 6–20 min 40%B, 20–23 min 80%B, 23–28 min 95%B, 28–33 min 95%B, 33–39 min 5%B. UV/VIS spectra were recorded at 280 nm and 330 nm. The adjusted operating parameters of the mass spectrometer were set as follows: Fragmentation: 70 V; Gas temp: 325 °C; Drying gas 10 L/min; Nebuliser 20 psi; Sheath gas temp 400 °C; Sheath gas flow 12 L/min. Electrospray ionisation (ESI) was used as a source operating in positive mode. The internal reference masses of 121.050873 m/z and 922.009798 m/z (G1969-85001 ES-TOF Reference Mass Solution Kit, Agilent Technologies & Supelco) were used for all analyses of the samples. The Agilent MassHunter Qualitative analysis 10.0 data acquisition software was applied to analyse LC-MS data using accurate mass (MS¹), isotope/adduct pattern consistency, retention time, and database matching (e.g., METLIN, LIPID MAPS®, Phenol-Explorer, LOTUS, FooDB). No MS/MS fragmentation spectra (MS²) were acquired for this screen. Acceptance criteria for putative annotations were mass error ≤5 ppm (lock-mass corrected) and consistency of isotope/adduct patterns; where applicable, in-source fragments were considered due to the fragmentor voltage settings.

Analytical standards (L-proline, jasmonoyl-isoleucine, medicagenic acid, apigenin, soyasapogenol B) were used for absolute quantification of matching compounds and their derivatives. External calibration was confirmed at MSI Level 1 by co-injection RT match (ΔRT ≤ 3%) and HR m/z for the dominant ESI(+) adduct (±5 ppm). Quantification used external solvent-based calibration (7 levels; 3 replicates per level; linear fit with 1/x weighting when required). No internal standards or matrix-matched/standard-addition approaches were applied. Primary stock solutions were prepared at approx. concentration 0.1 mg mL⁻¹ in methanol except jasmonoyl-isoleucine in ethanol, then serially diluted in methanol to the working calibration levels. Validation data are summarised in Supplementary Data 4 tab-B. Quantification is reported for the analytes with available authentic standards; for additional metabolites lacking standards and MS/MS, results are reported as putative annotations (MSI level 3/4) and, where applicable, semi-quantitatively as equivalents of the closest available standard (as specified in Supplementary Data 4). This represents a limited-scope validation (blank check, external calibration, LOD/LOQ by S/N, intra-day precision by replicate injections); matrix effects were not assessed. Measured concentrations (µg/g FW) with independent samples are reported in Supplementary Data 4 tab-C. The correspondence of the repeated experiment and the experiment reported in Fig. 3 is displayed in Supplementary Fig. 11.

Plant volatile analysis

Volatile compound analysis was performed using the same leaf samples as the plant transcriptome analysis. The samples were ground into a fine powder using liquid nitrogen, and 100 mg of the homogenised leaf tissue was aliquoted into clean 20 mL headspace vials. A saturated NaCl solution was added to prevent enzyme reactions, and the vials were sealed with Agilent headspace silica gel caps. Each vial was then heated for 20 min at 60 °C with an agitator cycle of 30 s on and 15 s off. Samples were injected into an Agilent Technologies 7820 A gas chromatograph coupled with an Agilent 5977B mass selective detector (MSD) using a Gerstel MPS Autosampler. The syringe temperature was set to 130 °C, and the injection volume was 2500 μL. A polar CP-Wax 52CB capillary column (50 m × 0.32 mm, 0.20 µm film thickness) with polyethylene glycol was used. Helium (He) served as the carrier gas with a split ratio of 1:20 and a flow rate of 1.2 mL/min. The temperature programme started at 60 °C, increased at a rate of 10 °C/min to 250 °C, and was held for 3 min. The injector temperature was set at 260 °C. Mass spectra were recorded at 70 eV within an m/z range of 50–500. The ion source temperature was maintained at 230 °C. Compound identification was based on retention indices (determined using a homologous series of C5-C24 n-alkanes) and mass spectral comparison with the NIST (National Institute of Standards and Technology) MS Search 2.2 library. Data acquisition and analysis were performed using Agilent MassHunter Qualitative Analysis 10.0 software.

Target volatiles (β-myrcene, linalool, β-fenchyl alcohol) were confirmed at MSI Level 1 by retention index (vs C5-C24 n-alkanes) and EI spectral matches to authentic standards. Primary stocks (1 mg mL⁻¹ in methanol) were used to prepare calibration standards: each 20 mL HS vial contained 4.7 mL of 30% (w/v) NaCl and 0.3 mL of the appropriate standard solution; working levels were obtained by serial dilution to five concentrations (100–3000 ng mL⁻¹) with three replicate vials per level. Quantification used external calibration (linear regression; 1/x weighting when required). Validation outputs are reported in Supplementary Data 4 tab-E.

Correlation of gene expression and metabolite data

Regularised canonical correspondence analysis (rCCA) was used with the mixOmics R package (available from Bioconductor; http://www.mixOmics.org) in RStudio. R script provided in Supplementary Data 4 tab-D. Correlations >0.7 were visualised via Cytoscape v3.10.3.

Plant–pathogen assays

Fusarium sporotrichoides (Fs) and Botrytis cinerea (Bc) were grown on 0.5× PDA at 20 °C in darkness for 4 weeks. Mycelial plugs, obtained with a 7.5 mm diam. cork borer, were placed on three different trifoliate leaves. Each leaf was placed in a 5.5 cm diam. Petri dish with water-saturated filter paper to maintain moisture level. The edge of the Petri dish had a pre-cut slit for the leaf petiole to remain attached to the plant throughout the experiment. Plants were placed in shade at 20 °C for 72 h. Agar plugs were removed before leaf photography. ImageJ (Fiji) was used to measure the area of necrotic lesions.

Statistics and reproducibility

mRNA and MS data were analysed from 4 independent libraries per treatment derived from individual replicate plants. RT-qPCR and metabolite accumulation analysis with synthetic standards were measured in two repeated experiments. Plant pathoassays were conducted with at least 8 individual plants and 3 repeated experiments. Data distribution is plotted as boxplots. Datapoints correspond to individual biological replicates. Fold change and FDR estimates used in Figs. 1 and 2 were generated with the Bioconductor package limma. Treatment differences in metabolite accumulation were analysed using two-sample t-tests (or Welch’s t-test for unequal variance) in MetaboAnalyst, followed by Benjamini–Hochberg FDR correction for multiple testing. Lesion sizes from individual experiments were analysed using parametric two-tailed t-tests or non-parametric U-tests where normality was violated. The pooled data from 3 experiments in Fig. 4 were analysed using a linear mixed model with repeated experiments as random factors with the lme4 package in R.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Z.O.: conceptualisation and experimental design (1), data curation (2), formal analysis (3), funding acquisition (4), investigation (5), methodology (6), project administration (7), reagent acquisition and lab resource planning (8), student supervision (10), data validation (11), statistical analysis and visualisation (12), writing (13), reviewing and editing the original manuscript (14). I.N.: investigation – metabolic analysis (5), reviewing the original manuscript (14). E.V.: formal analysis (3) of pathogen data, investigation – plant–pathogen assays (5), reviewing the original manuscript (14). D.P.: investigation – experiment setup, RNA extractions, RT-qPCR (5), reviewing the original manuscript (14). A.K.: investigation – experiment setup, RNA extractions, RT-qPCR (5), reviewing the original manuscript (14). D.S.: investigation – qPCR validation (5). A.A.: investigation – root staining for AMF colonisation (5). K.T.B.: investigation – experiment setup, RNA extractions (5), reviewing the original manuscript (14). I.S.: data curation (2), formal analysis (3) of RNA-seq data, reviewing the original manuscript (14). S.J.L.: conceptualisation and experimental design (1), methodology (6), reviewing and editing the original manuscript (14)

Peer review

Peer review information

Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editor: David Favero. A peer review file is available.

Funding

This work was supported by the Latvian Council of Science Fundamental and Applied Research grant lzp-2021/1-0056 awarded to Z.O. D.S. was funded by the LU EARF PhD grant 1.1.1.8/1/24/I/003.

Data availability

mRNA-seq raw data (fastq) are publicly available in NCBI with GEO number GSE325469. Raw MS data are open access via MetaboLights with reference number MTBLS14821. Source data for Fig. 4 is available in Supplementary Data 5. All other data are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10727-6.