Dissecting root-specific non-conventional effector-triggered immunity and its role in defense and shifting the microbiome

Abstract

Plant roots are extremely heterogeneous and non-photosynthetic tissue that is drastically different from leaves. Pioneering studies have revealed the existence of pattern-triggered immune responses in roots; however, whether roots mount tissue-specific effector-triggered immunity (ETI) is a fundamental question in plant immunity. By utilizing inducible effector-expressing lines, we systematically characterize non-conventional physiological and molecular responses during ETI activation in roots. Root ETI seems to be weaker than leaf ETI, evidenced by restricted cell death in the transition zone and fewer number of differentially expressed genes (DEGs). Root ETI also shows much higher DEG overlap with Pep1-triggered immunity than with Flg22-triggered responses, which is the opposite trend in leaves. We find that both PEPR1/2 receptors and BIK1, WRKY42 are required for induced root ETI responses. By comparing root responses to wild type and the T3SS mutant of Ralstonia solanacearum in the Nd-1 ecotype, we confirm that natural root ETI responses show similarities to induced ETI. Our work reveals tissue-specific features of root ETI and provides new insights into engineering root disease resistance in agriculture.

Roots, unlike leaves, are continuously exposed to a complex soil microbiome. Here the authors reveal root-specific effector-triggered immune responses distinct from those in leaves.

Introduction

Plant roots are essential for water and nutrient uptake, shaping beneficial root-associated microbiota^1–3 and forming symbiotic relationships with arbuscular mycorrhizal (AM) fungi and rhizobia^{4, 5}. Roots also possess several distinctive features compared with leaves. For example, roots are constitutively exposed to far more complex soil microbiota than aerial tissues⁶; lacking functional chloroplasts and therefore function as heterotrophic organ; and has highly heterogeneous cellular compositions, including diverse cell types and developmental states⁷. These characteristics suggest that roots may have evolved organ-specific immune regulatory strategies to interact with complex soil microbiota, “tolerant” high microbial loads, and maintain survival. However, over the past decades, most studies in plant immunity have primarily focused on leaf tissues⁸. Understanding the similarities and differences between shoot and root immune regulation therefore represents one of the most fundamental questions in plant immunology and may provide important guidance for future genetic and microbiome engineering strategies aimed at controlling soil-borne pathogens while promoting beneficial microbial associations.

Plants have evolved a two-layered immune system consisting of extracellular pattern-triggered immunity (PTI) and intracellular effector-triggered immunity (ETI)^9–11. Pioneering studies have established the existence of PTI responses to Flg22 (flagellin peptide 22) and phytocytokine Pep1 signaling in roots¹². However, whether roots also mount distinct and tissue-specific ETI responses remains largely unknown. Plants perceive microbe-associated molecular patterns (MAMPs) and damage-associated molecular patterns (DAMPs) through cell-surface pattern recognition receptors (PRRs), including receptor-like kinases (RLKs) and receptor-like proteins (RLPs)¹³. Upon ligand perception, RLKs interact with co-receptor BAK1, whereas most RLPs associate with both SOBIR1 and BAK1 to transduce immune signaling to intracellular receptor-like cytoplasmic kinases (RLCKs)^14–16. In addition, plants employ intracellular Nucleotide-binding Leucine-rich Repeat (NLR) receptors to sense pathogen-derived effectors and initiate ETI responses. Based on their N-terminal domains, NLRs are generally classified into TIR-NLR (TNL), CC-NLR (CNL), and RPW8-NLR (RNL), which contain Toll/Interleukin-1 receptor (TIR), Coiled-Coils (CC), and Resistance to Powdery Mildew 8 (RPW8) domains, respectively¹⁷. Upon activation, CNLs can assemble into resistosomes that form calcium-permeable channels at the plasma membrane (PM) or endoplasmic reticulum (ER), thereby promoting cell death^18–20. By contrast, although TNLs can also form oligomerized resistosomes, they function as NADase that generate nucleotide-based signaling molecules and promote the interaction between EDS1/PAD4 or EDS1/SAG101 with downstream helper NLRs, typically RNLs^{21, 22}.

Both PTI and ETI share several downstream immune outputs, including reactive oxygen species (ROS) production²³, mitogen-activated protein kinase (MAPK) cascade activation²⁴, callose deposition, and transcriptome reprogramming^25–27. Increasing evidence has suggested that PTI and ETI also utilize overlapping downstream signaling components. For example, mutations in PTI receptors or co-receptors can substantially compromise ETI responses^{26, 27}. Moreover, BIK1, a key RLCK downstream of PRRs, is also required for effective ETI signaling²⁶. In addition, the cell-surface leucine-rich repeat receptor kinase SOBIR1 can directly regulate the activity of ETI signaling components EDS1/PAD4 activity, thereby providing a molecular link between PTI and ETI signaling pathways²⁸.

Diverse evidence indicates that plant roots still be able to initiate intracellular effector triggered immunity, while we still lack systematic characterization of root ETI responses and the potential tissue specific molecular and physiological responses. For instance, the dominant I2 gene in tomato (Solanum lycopersicum) encodes a CNL protein that recognizes an effector from Fusarium oxysporum^{29, 30}. Similarly, the tomato RLP Ve1 recognizes the Ave1 effector from Verticillium dahliae, providing resistance against Verticillium wilt^31–33. In potato (Solanum tuberosum), NLR genes Resistance to P. infestans (Rpi) recognize effectors such as AVR2 from the destructive oomycete Phytophthora infestans and confer resistance to late blight disease^{34, 35}. Beyond fungal and oomycete pathogens, distinct R genes conferring resistance to root-knot nematodes (Meloidogyne spp.) have been identified across diverse crops, including Gpa2 in potato, Mi in tomato, Me in pepper (Capsicum annuum), and Rgh1 in soybean (Glycine max)^{36, 37}. In contrast to these pathogenic interactions, beneficial rhizobial strains also have T3SS proteins that release the effector proteins nodulation outer proteins (Nops) to suppress plant defense and promote nitrogen-fixing nodule symbiosis^38–40. Several dominant genes (Rj2, Rj3, Rj4, and Rfg1) from soybean, restrict nodulation with specific rhizobial strains and this process may be controlled by root ETI^41–43.

Considering the substantial differences between roots and leaves, the potential root-specific molecular and physiological responses of ETI signaling and their relevance remain poorly defined. Previous studies have revealed major differences between leaf and root immune responses. Firstly, root immunity exhibits distinct crosstalk with immune hormone signaling pathways. For example, although the immune hormone salicylic acid (SA) plays a substantial role in leaf immune responses to MAMPs, Flg22-induced immune gene activation in roots occurs independently of the SA receptor NPR1 and instead depends on the ethylene signaling component EIN3⁴⁴. In addition, roots are more sensitive to the endogenous phytocytokine Pep1 than Flg22, and Pep1 is also classified as a DAMP^{45, 46}. This might because roots are constitutively exposed to much higher microbial loads in soil and thus have evolved to be less sensitive to MAMPs under normal growth conditions, but can be more sensitive to potential damage upon the detection of DAMPs. Finally, roots display pronounced cell type-specific immune responses. For instance, the canonical immune marker genes FRK1 and PER5 are specifically induced by co-incidence of damage and Flg22 specifically in the meristematic zone⁴⁷. Our recent single-nuclei RNA-seq analyses further suggested that phytotoxin- and triterpene-associated metabolic genes are preferentially induced in root maturation-zone cells in response to the soil-borne pathogen Ralstonia solanacearum⁴⁸.

In this study, we provide the first piece of evidence that plant roots can mount a non-conventional ETI response with similar physiological outputs but overall weaker responses, which also show tissue and cell type specific responses. Moreover, activation of ETI in roots can largely reshape the root microbiome. Our work comprehensively revealed the landscape of root specific ETI responses and provided genetic components required for ETI responses in roots. Our work provides a new direction for engineering root health through root-specific ETI signaling in the context of complex soil microbiome in agriculture.

Results

ETI activation induces cell type specific cell death in transition zone

During bacterial invasion in nature, it is not easy to clearly study ETI response without activation of the PTI response. To address this, previous studies elegantly utilized the inducible expression transgenic lines to express different avirulent pathogen effectors^{26, 27}, which can trigger ETI responses without perturbations from PTI elicitors. We utilized inducible expression lines: AvrRpm1 (Col-0/DEX::avrRpm1) and AvrRpp4 (Col-0/Est::avrRpp4), in which the expression of AvrRpm1 and AvrRpp4 can be driven by dexamethasone (DEX) /estradiol (Est) inducible promoter^{17, 49–51}. AvrRpm1 and AvrRpp4 can be recognized by the CC-NLR receptor RPM1 and the TIR-NLR receptor RPP4, respectively.

Considering that most ETI receptors can trigger a hypersensitive response (cell death phenotype), we then investigated whether ETI activation also triggers cell death in roots. To more precisely assess cell death in different root cell types, we monitored the viability of cells in ETI-activated roots through live-cell imaging after staining with both fluorescein diacetate (FDA) and propidium iodide (PI). Fluorescein derived from FDA hydrolysis accumulates in the cytoplasm of viable cells, indicating intact plasma membrane (PM) integrity, whereas PI specifically stains nuclear DNA in cells with damaged membranes^{18, 52–54}. To more accurately study ETI-induced cell death in roots, we monitored AvrRpm1-expressing cells over a time course (2-24 h). Surprisingly, we found that the nuclei of most root cells were not stained by PI during 24 h of DEX treatment in the Col-0/DEX::avrRpm1 line, and only the nuclei of root cells from transition zone showed significant PI staining first observed at 12 h and more severe at 24 h DEX treatment (Fig. 1a, b, Supplementary Fig. 1). FDA staining shows the opposite pattern, labeling most live cells except cortex and potentially endodermal cells in the transition zone (Supplementary Fig. 2). These results suggest that AvrRpm1-activated ETI triggers localized cell death mainly in the root transition zone. Cell death in the transition zone was largely blocked in the rpm1/DEX::avrRpm1 line, supporting that AvrRpm1-activated cell death in roots is receptor-dependent (Fig. 1c and Supplementary Fig. 3). Interestingly, we found that the early DZ and TZ of DEX-treated roots also exhibited PI staining signal within the stele (Supplementary Fig. 1a), indicating that endodermal barrier cells might also be damaged upon ETI activation.

Fig. 1. AvrRpm1-activated ETI induces cell death in the root transition zone and inhibits the root growth.

a Schematic of different developmental zones in a 6-day-old Arabidopsis thaliana root. MZ meristematic zone, TZ transition zone, E-DZ early differentiation zone, L-DZ late differentiation zone. b AvrRpm1-activated ETI specifically induced cell death in the transition zone. Images correspond to the zones indicated in (a). The white triangle represents the position of meristem. The white dashed rectangle indicates the region where fluorescence intensity was measured in the MZ. c rpm1 mutation can block cell death in Col-0/DEX::avrRpm1 root transition zone within 24 h. d Callose deposition in Col-0/DEX::avrRpm1 root (up) and quantitative analysis (bottom). n = 5, 5, 6, 5 from left to right. e The relative expression of PTI responsive genes induced by Flg22, Pep1 and DEX in roots. n = 3 biological replicates. Each biological replicate was pooled from 3 to 4 roots. f AvrRpm1-triggered ETI induced stronger MPK phosphorylation in roots than in shoots. The numbers indicate molecular weight markers, with units in kDa. g Quantitative analysis of root growth monitored by a commercial high-throughput infrared imaging platform. n = 3 biological replicates. The statistical analyses were performed at 30 h (Mock vs Flg22, P = 0.858; Mock vs Pep1, P = 0.008; Mock vs DEX, P = 0.023; *P value < 0.05, **P value < 0.01, ns, not significant). h The growth of plant was inhibited in activation of ETI after 3 days. i–j The number of cells in the meristematic cortex after different treatments in Col-0/DEX::avrRpm1 and rpm1/DEX::avrRpm1. For i, n = 5, 5, 6, 10 from left to right. For j, n = 6 biological replicates. Box plots show the median (horizontal bar), 25th (bottoms of boxes), and 75th (tops of boxes) quartile range (QR). Error bars represent mean ± SEM for (d, e, and g). For  d, e, g, i, and j, statistical significance was assessed by two-sided t test. All the experiments were repeated three times with similar trends.

To confirm whether ETI-induced cell-type-specific cell death in the transition zone is common for different types of ETI receptors, we also investigated the cell death phenotype in another inducible effector-expressing line, Col-0/Est::avrRpp4²⁷. This effector is recognized by the TIR-NLR receptor RPP4, which differs from the CC-NLR RPM1. Consistently, we found that 24 h of estradiol treatment also induced cell death predominantly in the transition zone of the Col-0/Est::avrRpp4 line (Supplementary Fig. 4a, b). Our findings suggest that both TIR-NLRs and CC-NLRs can trigger ETI-associated cell death in roots, which occurs predominantly in the transition zone.

ETI induction in roots triggers canonical immune responses and inhibits root growth

Considering that PTI and ETI share similarities in the downstream immune outputs⁵⁵, we next investigated whether ETI can trigger canonical physiological immune responses in roots, including callose deposition, the expression of PTI responsive genes and MAPK activation. We also included other well characterized elicitors as control groups, including the typical MAMP molecule Flg22 and the DAMP molecule Pep1. Similar to Flg22 and Pep1 treatments, AvrRpm1-triggered ETI elicited callose deposition in roots (Fig. 1d). We found that AvrRpm1-induced ETI activated the expression of PTI responsive genes such as FRK1, CYP71A12 and MYB51 (Fig. 1e). At the same time, we constructed three PTI-responsive marker lines (pCYP71A12::YFP-NLS/DEX::avrRpm1, pPER5::NLS-3×mVENUS/DEX::avrRpm1 and pWRKY11::YFP-NLS/DEX::avrRpm1)^{44, 47} to investigate whether ETI activation induced cell type specific spatial expression patterns. Our results showed that CYP71A12 was strongly induced upon ETI activation, whereas PER5 and WRKY11 exhibited only slight induction (Supplementary Fig. 5). Furthermore, the phosphorylation of MPKs was activated by AvrRpm1-induced ETI in roots at 16 h of DEX induction (before visible cell death in shoots at 24 h) (Fig. 1f).

A major difference between leaf and root immune response is that Pep1 triggers much stronger immune responses and growth inhibition than Flg22 in roots but not in leaves⁴⁵. We therefore compared the effects of ETI, Pep1, and Flg22 on root growth inhibition. Utilizing a high-accuracy time-lapse imaging and phenotypic analyzing system, we analyzed the dynamic effects of ETI, Pep1, and Flg22 on root growth. We found that Pep1 triggered much stronger growth inhibition than Flg22 during 30 h of treatment, and ETI had a similar effect to Pep1 (Fig. 1g, h and Supplementary Fig. 6). We further quantified the number of root meristem cells and found that AvrRpm1-activated ETI suppressed root growth by decreasing root meristem cells numbers (Fig. 1i and Supplementary Fig. 1b)⁵⁶. Consistent with cell death activation, this inhibition was fully rescued in rpm1/DEX::avrRpm1 (Fig. 1j). Our work suggests that AvrRpm1-induced ETI triggers strong root growth inhibition similar to that induced by Pep1 and that both responses are much stronger than Flg22 in roots.

Differences between ETI response in roots and shoots

We further characterized the transcriptome landscape changes during ETI activation in roots and the potential differences with shoot ETI responses. We used the Col-0/DEX::avrRpm1 line to induce ETI activation in our previously developed 48-well plate system, which enabled us to harvest shoots (entire rosette) and root samples separately (Fig. 2a). After 16 h of induction, fewer ETI induced DEGs were identified in the roots (2452) than in the shoots (4356) (Fig. 2b). We generated a Venn diagram to illustrate shared and tissue specific ETI responses (Fig. 2b).

Fig. 2. Tissue-specific transcriptional reprogramming during ETI in roots and shoots.

a Schematic diagram showing the plant growth and sampling setup for mRNA-seq. b Venn diagram of DEGs induced by DEX treatment (ETI activation) in roots and shoots. c GO enrichment analysis of tissue-specific DEGs. Heatmaps showing the expression patterns of shared genes induced both in root and shoot CRK genes, d; root-specific ETI induced genes auxin related, e and shoot-specific ETI induced genes (vacuolar protein sorting-associated genes), f.

We performed GO enrichment analysis to further capture the potential functions of the shared and tissue-specific ETI responses. As expected, the shared genes were associated with GO terms related to immune regulation and cell death, including the genes Cysteine-rich receptor-like kinases (CRK)⁵⁷ (Fig. 2c, d). Notably, root-specific ETI responses (971 genes, Fig. 2b) were enriched in auxin responses, including well-characterized auxin biosynthesis and signaling genes, YUC1, SAURs, IAAs (Fig. 2c, e)^58–61. Furthermore, shoot-specific ETI responses were enriched in diverse protein and vesicle transportation related functions^62–64 (Fig. 2c, f). Thus, shoot ETI needs broad protein and vesicle transportation systems for extracellular antimicrobial secretion processes, as supported by evidence that a vesicle trafficking regulator MIN7 plays critical role in plant immunity^{65, 66}. The data comprehensively characterized ETI responses in roots and the shared and distinct responses relative to shoot ETI.

ETI responses in roots show higher overlap with Pep1-triggered responses than with Flg22 responses

One of the latest breakthroughs in the plant immune field is that plant membrane and intracellular immune receptors share substantial signaling overlap^{26, 27, 67}. We thus also compared the overlap between ETI responses and other PTI responses in both root and shoot (Fig. 3a, b). We systematically analyzed the differentially expressed genes (DEGs) after Flg22, Pep1, and DEX treatment in roots and obtained DEGs for Flg22 and Pep1 treatment in shoots from previous studies²⁵. Consistent with the findings that shoot ETI shows significant overlap with Flg22-triggered PTI, 44.7% of ETI-induced genes in shoots overlapped with Flg22-triggered DEGs in shoots (Fig. 3a, c). However, only 26.4% of the ETI genes were shared with Pep1 treatment in shoots, which was less than the overlap between ETI and Flg22-induced immunity (Fig. 3a, c). Strikingly, we observed the opposite trend in roots, where 42.8% of the ETI-induced genes overlapped with the Pep1-induced DEGs, which was much higher than 16.3% of the ETI genes overlapped with Flg22 in roots (Fig. 3b, c). These results suggest that ETI shares more overlap with Flg22 responses in shoots, while it shares more similarity to Pep1 responses in roots, which is reminiscent of a previous report that Pep1 is a stronger elicitor in roots than Flg22⁴⁵. It also may be due to the low expression level of FLS2 in roots⁴⁷. We performed GO enrichment analysis to gain functional insights into ETI- and Pep1- induced immune responses in roots. Both the shared DEGs and Pep1 specific genes shared exactly the same enriched GO functions, which supports the finding that Pep1 induced functional transcriptome changes were also broadly shared in ETI responses in roots (Fig. 3d). However, ETI specifically upregulated genes with a few additional GO functions, such as those related to flavonoids and other metabolic processes (Fig. 3d).

Fig. 3. Comparisons between ETI responses and Flg22, Pep1- triggered immune responses.

Venn diagram of DEGs induced by Flg22, Pep1, and DEX (ETI) in shoots (a) and roots (b). c The numbers of overlapped DEGs between ETI and other immune responses in shoots and roots. 44.7% and 26.4% represent the overlapped genes between ETI and Flg22- or Pep1- induced DEGs in shoots. 16.3% and 42.8% represents the corresponding ratios in roots. d GO enrichment analysis of the DEGs from different cluster in (b). e, f Heatmaps showing the distinct expression patterns of RLPs and RLKs.

It is also worth noting that although ETI and Pep1 signaling share substantial transcriptome overlap and similar functions in regulating root cell death, there are still a large number of Pep1-specific (1164, 49.5% of total Pep1 induced genes) and DEX-specific genes (1363, 55.6% of total DEX induced genes). To examine the potential differences between Pep1 and ETI, we analyzed transcriptome changes at the individual gene level to increase resolution. We noticed that many immune receptor genes were present among the ETI- or Pep1-specific genes. For example, CERK1 was upregulated after Pep1 treatment, whereas EDS1 and RLP23 were present only during ETI induction (Supplementary Fig. 7). We thus comprehensively analyzed the expression levels of all differentially expressed membrane immune receptor genes (RLPs and RLKs) (Fig. 3e, f). As we speculated, ETI induction, Pep1, and Flg22 treatments triggered almost exclusive expression patterns of those membrane immune receptors. This finding indicates that although ETI and Pep1 trigger largely shared functional transcriptome changes, they induce different expression patterns of membrane localized immune receptor repertoires.

PEPRs are required for some downstream physiological responses of root ETI

Our results suggested that activation of both Pep1-triggered immunity and ETI led to similar phenotypic outcomes, such as localized cell death in the same region and drastic root growth inhibition (Fig. 1). Consistent with these phenotypic similarities, transcriptome analysis revealed that approximately half of the upregulated genes overlapped between ETI and Pep1-triggered immunity (Fig. 3b), raising the possibility that Pep1 signaling might be required for some ETI downstream responses. Given that ETI activation in shoots is known to rely on PTI receptors such as FLS2, EFR, and CERK²⁶, we next investigated whether the Pep1 receptor PEPRs also contribute to root ETI responses.

To test this hypothesis, we generated the pepr1pepr2/DEX::avrRpm1 line. We found that the pepr1pepr2 mutant almost completely blocked ETI-associated cell death in the root transition zone (TZ) at 24 h after DEX induction (Fig. 4a). To further characterize the role of PEPR signaling in ETI-associated root growth inhibition, we utilized a high-accuracy time-lapse imaging system to monitor dynamic root growth following treatment with either Pep1 or DEX in Col-0/DEX::avrRpm1 and pepr1pepr2/DEX::avrRpm1. Root growth in Col-0/DEX::avrRpm1 plants was inhibited by both Pep1 and DEX treatments, whereas no significant growth inhibition was observed in pepr1pepr2 or pepr1pepr2/DEX::avrRpm1 plants within 30 h of treatment (Fig. 4b, c). In addition to cell death, callose deposition was significantly reduced in pepr1pepr2/DEX::avrRpm1 roots treated with DEX (Fig. 4d, e). Together, these findings support the idea that PEPR signaling contributes to multiple downstream responses associated with root ETI activation.

Fig. 4. Requirement of PEPR proteins for ETI responses in roots.

a ETI-induced cell death was blocked in pepr1pepr2 in root transition zone. The right part shows the quantitative analysis of PI signal intensities in root transition zone. We observed no cell death in 12/20 pepr1pepr2/DEX::avrRpm1 plants, with weak cell death in the remaining 8/20. For Col-0/DEX::avrRpm1, n = 6, 5, 6; for pepr1pepr2, n = 6, 6, 5; for pepr1pepr2/DEX::avrRpm1, n = 5, 4, 5 from left to right. The experiment was repeated three independent times with similar trends. Statistical significance was assessed by two-sided t-test. Box plots show the median (horizontal bar), 25th (bottoms of boxes), and 75th (tops of boxes) quartile range (QR). b The phenotype of 6-day-old plants treated by different immune treatments for thirty hours. c Root growth inhibition kinetics measured using a high-throughput imaging platform. n = 5 biological replicates. Statistical analyses were performed at 30 h using two-sided t test (For Col-0/DEX::avrRpm1, Mock vs Pep1, P = 0.0023, Mock vs DEX, P = 0.025; for pepr1pepr2, Mock vs Pep1, P = 0.892, Mock vs DEX, P = 0.514; for pepr1pepr2/DEX::avrRpm1, Mock vs Pep1, P = 0.192, Mock vs DEX, P = 0.186; *P value < 0.05, **P value < 0.01, ns, not significant). d The callose deposition patterns in Col-0/DEX::avrRpm1 and pepr1pepr2 /DEX::avrRpm1 roots treated with DEX and Pep1. e Quantitative analysis of the callose deposition fluorescence intensity. n = 5, 6, 6, 5, 8, 8 from left to right. Data were analyzed by two-sided t test with correction for multiple comparisons by the Holm‑Šídák method. For all above the graphs, experiments were repeated three times with similar trends and error bars represented mean ± SEM.

Interestingly, ETI-triggered cell death in leaves was independent of PEPR1 and PEPR2, as the pepr1pepr2/DEX::avrRpm1 plants exhibited cell death levels comparable to those of Col-0/DEX::avrRpm1 plants three days after induction (Supplementary Fig. 8). This observation is consistent with previous studies showing that PEPR-mediated signaling is dispensable for ETI-induced cell death in leaf tissue⁶⁸, suggesting a potential distinction between shoot and root ETI responses.

We next asked whether ETI activation enhances endogenous Pep1 production, which could potentially explain the requirement of PEPR signaling during root ETI responses. To address this question, we generated the transcriptional reporter line pPROPEP1::PROPEP1-YFP/DEX::avrRpm1 to monitor PROPEP1 expression during ETI activation. We observed that PROPEP1 was constitutively expressed in the stele of all root zones except the MZ. However, neither transcription induction nor detectable protein accumulation was enhanced following ETI activation (Supplementary Fig. 9). These results suggest that artificial ETI activation does not stimulate PROPEP1 accumulation at either the transcript or protein level, despite the strong requirement for PEPR signaling in several downstream ETI-associated responses.

Since PEPR signaling only partially accounted for root ETI-associated responses, we next sought to identify PEPR-independent ETI-responsive genes. We analyzed the 274 genes exclusively induced in pepr1pepr2/Dex::avrRpm1 roots. This gene set included: (1) classic immune regulators, such as RIN4 and RPP5, together with HR-like lesion-inducing proteins; (2) Casparian strip-associated factors (CASP5 and CASP-like proteins) and (3) genes involved in secondary metabolism. Collectively, these findings reveal a PEPR1/2-independent root defense program that couples structural barrier reinforcement with immune and metabolic responses during ETI activation (Fig. 5a and Supplementary Fig. 10).

Fig. 5. Characterization the molecular basis underlying PEPR-dependent ETI responses in roots.

a Overlap among genes up-regulated in Col-0/DEX::avrRpm1 and pepr1pepr2/DEX::avrRpm1 roots. b GO enrichment analysis of the 664 DEGs up-regulated by DEX both in shoot and root and dependent on PEPRs protein. c Heat-map of immune regulators and WRKYs genes induced by DEX both in root and shoot. d, e The cell death was largely blocked by the bik1 and wrky42 mutants and the corresponding quantification. n = 4, 6, 3, 4, 3, 3, 3, 5 from left to right. Experiments were repeated three times with similar trends. f The callose deposition was slightly decreased in bik1/DEX::avrRpm1 and wrky42/DEX::avrRpm1 lines. g The relative expression of WRKY9/42 after infection by GMI1000 in Col-0. n = 3 biological replicates, each dot represents an independent sample pooled from 3 to 4 roots. h Disease symptoms 7 days post-GMI1000 infection. i Disease index quantification at 5 days (For Col-0, n = 5; for bak1bkk1, n = 5; for bik1, n = 10; for wrky42-1, n = 9; for wrky42-2, n = 6 biological replicates. The experiment was repeated three times with similar trends). For all the above graphs, data were presented as mean values ± SEM and were analyzed by two-sided t test with correction for multiple comparisons by the Holm‑Šídák method.

PEPRs are required for certain transcriptome changes of root ETI

To investigate the molecular basis underlying PEPR-mediated regulation of root ETI responses, we performed transcriptome analysis using the pepr1pepr2/DEX::avrRpm1 line. Through RNA-seq and Venn diagram analyses, we identified 1481 genes induced upon ETI activation in both shoots and roots (Fig. 5a). Among these ETI-responsive genes, 664 genes were not significantly induced in the pepr1pepr2/DEX::avrRpm1 roots (Fig. 5a), indicating that their induction is dependent on PEPR signaling during root ETI activation. Gene ontology analysis revealed that these PEPR-dependent genes were enriched in stress, wounding, and immune-related biological processes, such as BIK1, RBOHD, ADR1, and some WRKY transcription factors (Fig. 5b, c). These findings suggest that PEPR signaling contributes to a substantial transcriptional branch associated with downstream root ETI responses.

A critical mechanism underlying ETI-PTI signaling crosstalk in leaves is mediated through ETI-induced activation of PTI-associated membrane receptor-like kinases (RLKs, e.g., BAK1, CERK1, FLS2) and receptor-like cytoplasmic kinase (RLCKs: BIK1, PBLs)^{26, 27}, and mutation of the critical upstream RLCK protein BIK1 can largely abolish ETI-mediated resistance in leaves²⁶. To delve into PEPR-BIK1-mediated regulation of root ETI responses, we further analyzed genes co-expressed with BIK1 using our recently published single-cell dataset related to root-pathogen interactions⁴⁸. This high-resolution cell type specific dataset is powerful for accurately identifying the co-expression patterns and potential biological correlations among functional genes⁶⁹. Our single-cell-based co-expression analysis identified a cluster of WRKY genes that were significantly co-expressed with BIK1 (Supplementary Table 1). Among these candidates, WRKY9 and WRKY42 displayed prominent root-specific induction during ETI activation and their induction was largely abolished in pepr1pepr2/DEX::avrRpm1 roots as evidenced by our RNA-seq analysis (Fig. 5c).

To test whether BIK1 and WRKY42 are required for some downstream regulation of induced ETI responses, we obtained bik1/DEX::avrRpm1 and wrky42/DEX::avrRpm1 lines. We found that cell death was largely blocked at 16 h and 24 h after ETI induction in both bik1/DEX::avrRpm1 and wrky42/DEX::avrRpm1 (Fig. 5d, e and Supplementary Fig. 11a, b). Additionally, callose deposition was slightly reduced in these mutant lines following DEX treatment compared with Col-0/DEX::avrRpm1 (Fig. 5f and Supplementary Fig. 11c). Consistent with a disrupted immune response, the bik1 mutant exhibited no Pep1-induced cell death and swelling in MZ upon Pep1 treatment within 24 h or 36 h, and also partially dampened Pep1-mediated root growth inhibition after 5 days of treatment (Supplementary Fig. 12). Together, these findings support that BIK1- and WRKY42-associated signaling contributes to multiple downstream responses associated with root ETI activation. Finally, we examined whether WRKY42 contributes to resistance against R. solanacearum GMI1000 infection. We found that the expression of WRKY42 was induced by GMI1000 infection (Fig. 5g), and two independent wrky42 mutant lines displayed enhanced susceptibility to GMI1000 infection (Fig. 5h, i). Although this phenotype may reflect broader immune functions beyond ETI signaling alone, these results further support a role for WRKY42 in root immune responses against R. solanacearum.

Moreover, artificially and ectopically overexpressing effector proteins in roots might trigger side effects different from natural root ETI. To determine whether these genes are induced during natural ETI, we thus established a natural ETI interaction system using the Arabidopsis ecotype Nd-1, which carries the RRS1 immune receptor capable of recognizing the PopP2 effector from R. solanacearum GMI1000 and thereby activating ETI upon infection^{70, 71}. As an ETI-deficient control, we used the T3SS-defective mutant GMI1000 (ΔHrcV), which is unable to secrete effectors and therefore cannot activate ETI⁷². We found that natural root ETI also triggered cell death in the transition zone in a T3SS-dependent manner (Supplementary Fig. 13a). This suggests that our artificial effector induction system can reflect some natural root ETI responses. Both BIK1 and WRKY42 genes were indeed activated by GMI1000 in Nd-1 roots after 24 h of infection, but the activation was much weaker in Col-0 roots or under GMI1000 (ΔHrcV) infection (neither can initiate ETI) (Supplementary Fig. 13b, c). The expression of MYB51 was also more strongly induced upon GMI1000 infection in Nd-1 than that in Col-0 or GMI1000 (ΔHrcV) infection (Supplementary Fig. 13b, d). Our data demonstrate that BIK1, WRKY42, and MYB51 are induced by natural root ETI activation, a pattern similar to that observed in our effector-inducible expression system (Fig. 1e and Fig. 5c).

ETI activation shifts microbiome composition and increases microbiome load

The plant immune system is a key driving force for shaping the root microbiome structure, while a previous study focused only on the effects of PTI receptor mutants on microbiome changes⁶⁵. In this study, we used a spike-in-based quantitative microbiota profiling (QMP) approach to investigate the role of ETI in shifting the root microbiome structure⁴. We surveyed the root microbiome composition during ETI activation and used Pep1 as a comparison group. Principal coordinate analysis (PCoA) based on Bray-Curtis dissimilarities was performed for Col-0/DEX::avrRpm1 plants treated with H₂O, Pep1, and DEX. Our results revealed that there was a significant separation among the three treatments (Fig. 6a; R² = 0.3318, P = 0.001). Interestingly, we found that both ETI induction and Pep1 treatment increased the overall microbial load, which might be due to increased cell death or cell membrane damage, which enhances the release of more nutrients into the rhizosphere (Fig. 6b). This phenotype was also observed in wheat roots infected by root rot disease⁷³. An overall increased microbiome load might also enhance the ability of the microbiome to antagonize pathogen invasion in the rhizosphere.

Fig. 6. Quantitative microbiome analysis of ETI and Pep1 induced microbiome changes.

a PCoA (Bray-Curtis dissimilarities) of root microbiome changes after the DEX (ETI) or Pep1 treatment in natural soil. b Absolute abundance of the top 10 bacterial families in root microbiomes. c Biomarker families of ETI induced microbiome changes analyzed by LEfSe analysis. d STAMP analysis of differentially abundant bacterial families in the root microbiome shifted by ETI and Pep1 treatment. Left plots: mean abundance; right plots: error bars showed difference in mean abundance with 95% confidence intervals. n = 5 biological replicates. P values were determined by two-sided Welch’s t test.

We used Linear discriminant analysis Effect Size (LEfSe) to detect the biomarker species for each treatment group and found that AvrRpm1-triggered ETI activation enriched several bacterial families, like Azospirillaceae and Caulobacteraceae (Fig. 6b, c). STAMP based differential abundant taxa analysis supports the finding that ETI increases the abundance of Caulobacteraceae (Fig. 6d). The previous report showed that Azospirillales especially Azospirillum help plants fix nitrogen from the soil and improve the root growth and development^{74, 75}. Our results reveal that both Pep1 treatment and ETI activation have significant effects on both the root microbiome structure and total microbial load.

Discussion

Although soil-borne root pathogens cause substantial agricultural losses worldwide, our understanding of root immune responses and potential tissue-specific immune regulatory mechanisms is limited. Through extensive genetic and molecular approaches, we comprehensively revealed the tissue-specific non-conventional ETI responses in roots and reached several key conclusions: (a) root ETI has potential tissue specific functional changes to potentially better adapt to rhizosphere environment. For instance, diverse protein and vesicle transportation related functions are induced during shoot ETI but not in root ETI. This might be because roots do not have soft spongy mesophyll tissue and a large apoplastic zone; thus, fewer vesicle transport processes are needed to combat bacterial infection in apoplastic zone. Root ETI also specifically activates auxin responses, as auxin is dominant in inhibiting root growth; (b) overall, root ETI seems weaker than shoot ETI, because only cells from the root transition zone undergo cell death upon ETI activation, and there are fewer DEGs compared to those in shoot ETI; (c) root ETI activated genes show higher overlap with Pep1 than Flg22 triggered immunity. This is reminiscent of the previous reports that Pep1 plays a more important role in root immune responses than MAMP molecules such as Flg22⁶⁷. (d) We further confirmed that PEPR1/2, BIK1 and WRKY42 are required for some downstream responses of root ETI, including localized cell death, callose deposition and immune related transcriptome reprogramming. (e) Root ETI activation can reshape the microbiome composition and increase the microbial load in roots.

The predominant difference between ETI responses in roots and leaves is that only a limited subset of root cells (mainly those in the transition zone) can initiate cell death upon ETI induction. In contrast, no cell-type-specific cell death has been reported in leaves to date. One of the most drastic structural differences between roots and leaves is that both primary and secondary roots are long cylindrical structures. If the maturation zone of the primary root is damaged by ETI-induced cell death, then the plant will lose all or most functional roots even if other parts are still alive. In contrast, localized cell death at the transition zone clearly benefits the whole root systems, as it will only block the function of a small area of young root meristem cells. Moreover, root meristem cells do not contain well-developed casparian strips and have less callose deposition, which are more susceptible to soil-borne pathogens and thus localized cell death at the transition zone will help block bacterial invasion through meristem zone. ETI induced transition cells specific cell death is very similar to the cell death induced by Pep1 reported previously⁵⁴. Considering that exogenous Pep1 treatment for the whole root (which is evenly treatment without concerns about cell type specific expressions) triggers cell swelling and cell death in TZ, which is partially auxin dependent⁵⁴, it is more likely that cells in the TZ are fast-growing and under quick differentiation, and thus more vulnerable to immune or auxin signaling perturbations. For now, why NLRs proteins cannot trigger extensive cell death in most root cells requires further explorations.

Our study carries certain limitations due to the use of an artificial inducible system. We thus compared natural root ETI responses using Nd-1 ecotype, which recognizes GMI1000 as an avirulent pathogen (Supplementary Fig. 13). Data from this system suggest that natural root ETI responses exhibit similar readouts compared with our effector induction system. In addition, simultaneously expressing AvrRpm1 in all cell types (including shoots) would compromise root cell-type-specific regulation. Therefore, it would be valuable to generate a single-cell dataset for root ETI responses, similar to our previous single nuclei RNA-seq study⁴⁸. Meanwhile, as root-shoot grafting experiments were technically challenging, we cannot rule out the possibility that the immune responses observed in roots are driven by systemic immune signals transmitted from leaves.

We found that the most significantly enriched GO term for root specific ETI was the auxin response. This indicates that immune induced auxin signaling is the key downstream difference between leaf and root immunity. This finding is consistent with the fact that Pep1 inhibits root growth predominantly through auxin pathway⁵⁴ and the expression of auxin-related genes is upregulated during R. solanacearum infection in different plant species^76–78. This trend was also observed in our recent snRNA-seq study⁴⁸. Moreover, auxin signaling has been repeatedly reported to affect plant resistance, but whether auxin positively or negatively regulates plant immunity is still controversial. It has been reported that diverse auxin signaling mutants are more susceptible to the necrotrophic pathogen Alternaria brassicicola, which suggests that auxin might positively regulate resistance to some pathogens⁷⁹. However, other studies have shown that auxin-signaling mutants or chemical inhibition of polar auxin transport could confer increased pathogen resistance to Fusarium oxysporum⁸⁰. A PIN auxin efflux transporter in tomato causes enhanced resistance to the soil pathogen R. solanacearum⁷⁷. These findings further support the hypothesis that auxin signaling might play a previously uncharacterized role in combating soil pathogens (especially vascular pathogens). However, considering the dynamic auxin polar transport and cell type specific auxin responses, the mode of action of auxin mediated root defense regulation requires further exploration. In addition, recent studies have suggested that auxin signaling is required for compatible interactions with beneficial commensal microbes in roots⁸¹. In the future, delving into the molecular linkage between root immunity and auxin signaling would be a promising direction for dissecting potential root specific immune regulatory mechanisms, and engineering beneficial root microbiome interactions.

Our work characterized potential root specific ETI responses. We also found that PTI components like PEPRs and BIK1, WRKY42 are required for some downstream responses of ETI. Moreover, our quantitative microbiome profiling study further confirmed that root ETI reshaped the microbiome structure and enhanced microbial load. Considering the increasing threat of soil borne pathogens worldwide due to global climate changes^{82, 83}, further mechanistic understanding of root-specific ETI regulation, interactions with Pep1 signaling pathways, as well as the subsequent consequences for root defense and beneficial microbiome changes, would pave a new direction for regulating root health in the context of the complex soil microbiome in agriculture.

Methods

Plant materials and growth conditions

Col-0/DEX::avrRpm1 and Col-0/Est::avrRpp4 were gifts from Prof. Pingtao Ding²⁷. The pepr1pepr2 line was a gift from Prof. Hongwei Guo. The bik1 mutant was a gift from Prof. Jianfeng Li. The pPER5::NLS-3×mVENUS line was a gift from Prof. Feng Zhou⁴⁷. The rpm1 (SALK_146601), wrky42-1 (SALK_121674C), wrky42-2 (SALK_203446C) mutants were obtained from ABRC. rpm1/DEX::avrRpm1, pepr1pepr2/DEX::avrRpm1, bik1/DEX::avrRpm1, wrky42-1/DEX::avrRpm1 and wrky42-2/DEX::avrRpm1 were generated by crossing rpm1, pepr1pepr2⁸⁴, bik1, wrky42-1 and wrky42-2 with Col-0/DEX::avrRpm1 respectively. pPER5::NLS-3×mVENUS/DEX::avrRpm1 was crossed by pPER5::NLS-3×mVENUS and Col-0/DEX::avrRpm1. pCYP71A12::YFP-NLS/DEX::avrRpm1, pWRKY11::YFP-NLS/DEX::avrRpm1 and pPROPEP1::PROPEP1-YFP/DEX::avrRpm1 were established through genetic transformation. These promoter sequences were referenced from the three publications^{44, 47, 85}.

All the seeds were surface sterilized with 75% ethanol and 20% bleach for 2 min respectively, then washed five times with sterilized water. The seeds were germinated on MS medium supplemented with 1% sucrose and 0.7% agar. 7-day-old seedlings were transplanted into the soil under the 10 h /14 h (light/dark) conditions at 22°C. For 48-well plate culture, sterilized seeds were soaked in 0.1% agar for three days at 4°C. Seeds were germinated on mesh in MS liquid medium with 2% sucrose in 48-well plate. The plates were incubated in a growth chamber (Fujian JIUPO BPC500H-WL) under a 16-h light/8-hour dark photoperiod with a light intensity of 100 μmol m^–2 s^–1. After 10 days, the medium was changed to fresh MS liquid medium without sucrose for more than two days before inoculation.

Confocal settings and image processing

For confocal imaging, seeds were germinated on solid MS medium with 1% sucrose and 1.2% agar for 6 days, and then were transferred to vertical MS solid medium with 1 μM Flg22, 100 nM Pep1, or 30 μM DEX for 24 h. For Col-0/Est::avrRpp4 seedlings, it shows stronger cell death phenotype when treated in liquid MS medium. Thus, 6-day-old Col-0/Est::avrRpp4 seedlings were transferred to MS liquid medium with 1 μM Flg22, 100 nM Pep1, or 50 μM Est for 24 h.

For the R. solanacearum infection assay, 6-day-old Col-0 and Nd-1 seedlings were transferred to vertical MS solid medium without sucrose and inoculated by applying 5 µL of the R. solanacearum suspension (OD₆₀₀ = 0.05) to each root for 24 h.

The roots were subsequently stained with 1% propidium iodide or 10 μg/mL FDA for strictly 5 min. We selected four distinct root zones, the MZ, TZ, Early-DZ and Late-DZ from each seedling. Confocal imaging was performed using an LSM880 microscope (Zeiss). Pictures were taken under 20× objectives, 1024 × 1024 pixel. To ensure consistency in quantitative analysis of PI fluorescence intensity, all confocal images were acquired under identical parameter settings, including laser power, photomultiplier, offset, zoomfactor and resolution. For the PI signal, the excitation wavelength was 561 nm. For the YFP and FDA signal, the excitation wavelength was 488 nm.

We used average PI or FDA signal density to reflect the overall dye staining intensity in four different areas (MZ, TZ, Early-DZ, Late-DZ). For MZ zone, since different groups have drastic variations of cell numbers, we thus set our “area of interest” to be a smaller but same size area (dash box in Fig. 1b) of MZ close to QC (quiescent center). For all other zones, we calculated the Raw Integrated Density of the whole image (1024×1024 pixel). Microscope parameters were calibrated using mock treated roots until the quiescent center (QC) and cortex cell became visibly distinguishable in the MZ. All treatments were set a threshold value before quantification, for example, the threshold values of PI and FDA were 100, the threshold values of YFP and DAPI (callose) were 20 and 50.

RNA extraction and qRT-PCR analysis of gene expression

Col-0/DEX::avrRpm1 seedlings were cultured in 48-well plates for 12 days, after which 1 μM Flg22, 100 nM Pep1, or 30 μM DEX was added, followed by cocultured for 2 h, 2 h and 16 h, respectively. Then, the root samples were collected and frozen in liquid nitrogen. Total RNA extraction was performed with reagents from Vazyme.

Col-0 and Nd-1 seedlings were cultured in 48-well plates (MS medium containing 2% sucrose) for 10 days, and then transferred to sucrose-free MS medium for 2 days. After that plants were inoculated with GMI1000 or GMI1000(ΔHrcV) at a final concentration of OD₆₀₀ = 0.05 and incubated for 24 h. Root samples were then collected and frozen in liquid nitrogen. Total RNA extraction was performed with reagents from Vazyme.

1 μg total RNA was used for reverse transcription with the reagent HiScript IV All-in-One Ultra RT SuperMix from Vazyme. Real-time qPCR analysis was carried out with Taq Pro Universal SYBR qPCR Master Mix (Vazyme). Three technical repeats were performed for each sample. Normalization was based on the plant gene EIF4A, which acted as the reference gene. The primer sequences used are listed in Supplementary Table 2.

RNA-seq sample preparation and data analysis

For the RNA-seq experiments, the plants were treated with 1 μM Flg22, 100 nM Pep1 or 30 μM DEX in 48-well plates and Col-0/DEX::avrRpm1 roots were collected as described above. The shoots of Col-0/DEX::avrRpm1 plants were cut and harvested from 48-well plates treated with 30 μM DEX. The pepr1pepr2/DEX::avrRpm1 seedling roots were treated with 30 μM DEX in 48-well plates and collected. Total RNA was extracted. mRNA purification, library construction, and RNA sequencing were performed by BGI (Beijing Genomics Institute). The libraries were subjected to 150 bp paired-end sequencing on the DNBseq platform. Sequence data were quantified by Salmon (v0.14.1)⁸⁶ against the TAIR10 Arabidopsis reference genome and normalized as transcripts per million (TPM). DESeq2 (v1.44.0)⁸⁷ was used to identify differentially expressed genes (DEGs) across treatment groups (adjusted P < 0.05 and |log₂ FC| > 1). Functional enrichment analyses (Gene Ontology enrichment) were performed with the clusterProfiler (v4.12.6)⁸⁸ package in R (v4.4.1). P values were calculated using one-sided hypergeometric tests and adjusted P values were calculated using the Benjamini-Hochberg (BH) correction.

Callose quantification

Seeds were germinated and grown on MS solid medium containing 1% sucrose for 10 days, then transferred to MS liquid medium for approximately 10 h, followed by treatment with 1 μM Flg22, 100 nM Pep1 or 30 μM DEX for 24 h. Plants were fixed and destained in 1:3 acetic acid/ethanol overnight, after that plant material was incubated for 30 min in 150 mM K₂HPO₄ and 0.01% aniline blue (staining solution) for 2 h^{44, 89}. The seedlings were then collected for imaging. Images were taken with 10× objectives on a ZEISS 880 confocal microscope with a DAPI filter (excitation, 340 nm; emission, 488 nm). The intensity of the callose dots was calculated by ImageJ software, for at least six individual plants in one biological replicate and at least three biological replicates.

MAPK activity assay

For shoot tissue collection, 12-day-old seedlings grown in 48-well plates were treated with 1 μM Flg22 for 15 min or 30 μM DEX for 16 h. Shoots were then harvested and frozen in liquid nitrogen. For root tissue collection, 12-day-old seedlings grown on MS solid medium were excised at the root-shoot junction, and the roots were incubated in liquid MS medium for 8 h prior to treatment with 1 μM Flg22 (15 min) or 30 μM DEX (16 h). Root samples were subsequently harvested and frozen in liquid nitrogen. Total protein was extracted with 2×SDS loading buffer and immunoblot analysis with an anti-p44/42 antibody was used to assess the phosphorylation of MAPKs (Cell Signaling; no. 4370S; 1:2,000 dilution). HSP90 protein was used as the reference protein.

Elicitor and chemical treatments

The Arabidopsis thaliana Plant Elicitor Peptide1, Pep1, and Flg22 oligo peptide were synthesized by Sangon Biotech (https://www.sangon.com). The peptides were dissolved in deionized water and stored as a 1 mM stock solution. Dexamethasone and estradiol were dissolved in DMSO and stored as 300 mM and 1 M stock solution.

Time-lapse imaging and quantitative measurement of root length

6-day-old transgenic Col-0/DEX::avrRpm1 plants were transferred onto MS solid medium supplemented with 1 μM Flg22, 100 nM Pep1 and 10 μM DEX and then docked onto a commercial seedling phenotyping platform vertically and subjected to 30 min interval imaging⁹⁰. pepr1pepr2/DEX::avrRpm1 and pepr1pepr2 plants were treated with 100 nM Pep1 and 10 μM DEX under the same culture conditions, and then imaged by the same method. The root length and curvature were measured with Image J software.

R. solanacearum resistance assay in the soil

Seeds of Col-0, bak1bkk1, bik1, wrky42-1 (SALK_121674C), and wrky42-2 (SALK_203446C) were surface sterilized and germinated on the MS medium for 7 days. Seedlings were then transplanted into autoclaved peat pellets and cultivated in a controlled greenhouse under 12-h light/12-h dark photoperiod at 22 °C with 60% relative humidity for three weeks.

Before inoculation, a horizontal cut (1 cm above the pellet bottom) was made through each peat pellet to induce localized root injury. The peat pellets were soaked in a R. solanacearum GMI1000 suspension (OD₆₀₀ = 0.2) until complete saturation ( ~ 30 min), followed by draining of the excess bacterial solution. Disease progression was monitored daily by recording wilting symptoms and survival rates until all the seedlings succumbed to infection. The scoring criteria followed the previously established method⁹¹.

Analysis of BIK1 co-expressedgenes

To investigate BIK1 co-expression networks, we calculated Pearson correlation coefficients between its expression profile and all other genes based on our recently published single-cell dataset related to root-pathogen interactions. Significance was determined after Bonferroni correction for multiple testing, with genes exhibiting adjusted p values < 0.05 classified as significantly co-expressed.

Root microbial sample collection

The seeds of the transgenic plant Col-0/DEX::AvrRpm1 were generated on MS medium with 1% sucrose, and the 6-day-old seedlings were subsequently transplanted into pots containing 1/4 native soil and 3/4 substrate⁹². Four seedlings from each pot and 16 pots were used for each treatment. All the pots were randomly placed in a greenhouse and treated about four weeks later. Add 5 mL 10 μM DEX, 100 nM Pep1 (with an equal volume of DMSO as DEX) or ddH₂O (with an equal volume of DMSO as DEX) was added to the base of the rosette leaves, respectively and the process was repeated one more time after 24 h. Root samples were sampled at  three days after first time treatment (the leaves treated with DEX showed slightly  wilting).

To collect the microbial communities that colonized the root, the plant roots were cautiously separated from the adhering soil and washed with deionized water in 50 mL Falcon tubes five times, after which the excess water was removed with sterile filter paper. These samples were promptly placed into 2 mL tubes and frozen in liquid nitrogen. The frozen roots were stored at –80 until DNA extraction. Each sample contained three to four roots derived from different pots as a biological replicate. The samples in the tubes were crushed by shaking at 45 Hz for 30 s, three times under the liquid nitrogen conditions. A total of 100-200 μL powder was weighed and used for DNA extraction by a Power Soil DNA Isolation Kit (Qiagen, Germany).

Analysis of the microbial sequence

For bacterial amplicon sequencing, genomic DNA (gDNA) was extracted from samples meeting quantity thresholds (≥20 ng/μL). Variable regions V3-V4 of the prokaryotic 16S rRNA gene were amplified via PCR with the universal primers 341F (5’-ACTCCTACGGGAGGCAGCA-3’) and 806 R (5’-GGACTACHVGGGTWTCTAAT-3’). Libraries were constructed and subjected to paired-end sequencing (2 × 250 bp) on the Illumina NovaSeq 6000 platform through a commercial service provider (Shanghai Genesky Biotechnologies Co., Ltd).

Raw FASTQ files were preprocessed with fastp v0.14.1 for adapter trimming, quality filtering, and removal of low-complexity reads⁹³. Sequence analysis was performed in QIIME2 v2022.2, where amplicon sequence variants (ASVs) were resolved via DADA2 denoising^{94, 95}. Taxonomic classification was assigned against the SILVA 138 reference database, followed by rigorous filtering to exclude mitochondrial, chloroplast-derived, archaeal, and unclassified sequences⁹⁶. Microbial community quantification in root samples adhered to established bioinformatics workflows⁴.

Reporting summary

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

Author contributions

Y.S. and X.T. designed the research. X.T. and K.G. contributed to analysis of RNA-seq. X.T. and P.M. contributed to analysis of quantitative microbiota profiling. X.T. and T.S. contributed to genetic construction and phenotypic analysis. X.T. contributed to microscope imaging and high-accurate time-lapse imaging. Y.Z contributed GMI1000 mutant and helpful discussion. Y.S., X.X., W.Y., and X.T. analyzed the data and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Frank Takken and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

The research was supported by National Natural Science Foundation of China (32270286); National Key Research and Development Program of China (Grant no. 2025YFA0921700), and the Shenzhen Science and Technology Program (Grant No. RCYX20231211090408015 and ZDSYS20230626091659010); Guangdong Innovative and Entrepreneurial Research Team Program 2023ZT10Y013.

Data availability

The RNA-seq data and quantitative microbiota profiling data generated in this study have been deposited in the China National Center for Bioinformation database and are publicly accessible with accession code CRA027840 [PRJCA042975; https://www.cncb.ac.cn/search/specific?dbId=bioproject&q=PRJCA042975] and CRA027799 [PRJCA043020; https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA043020]. Source data are provided with this paper.

Code availability

The analysis pipelines are available via GitHub at https://github.com/Song3-lab/Root_ETI [https://doi.org/10.5281/zenodo.21423264].

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76375-w.