The glutamate receptor‐like GLR2.7 modulates insect egg‐induced defense responses in Arabidopsis

Summary

• Upon perception of insect eggs, Arabidopsis thaliana activates a generic immune response that culminates in cell death (hypersensitive‐like response (HR‐like)). While this response can subsequently impact egg survival, the molecular mechanisms are poorly understood.

• Through a genome‐wide association study (GWAS), we identified the amino acid‐gated calcium channel GLUTAMATE‐LIKE RECEPTOR2.7 (GLR2.7) as an important gene controlling the extent of HR‐like and accumulation of salicylic acid (SA) in response to egg extract of Pieris brassicae. Analysis of natural polymorphisms showed that two major haplotypes segregate at the species‐wide level and suggests that balancing selection acts at this locus.

• Insect oviposition triggered a long‐lasting localized cytosolic calcium accumulation that depended on GLR2.7 and was linked with egg‐associated glutamate (Glu).

• We propose that Glu‐activated GLR2.7 is involved in egg perception and early immune responses.

Introduction

Plants face various biotic threats in nature and have evolved an efficient immune system that relies on the perception of enemy‐associated molecules and activation of defenses. In particular, the study of plant–insect interactions has been the focus of intense research in the past decades, and a wealth of information on signaling steps and molecular players has accumulated (Erb & Reymond, 2019). By contrast, knowledge on how plants respond to the perception of seemingly inert insect eggs at the molecular level is still limited. Oviposition triggers defense responses in different plant species after recognition of egg‐associated molecules (Hilker & Fatouros, 2015, 2016; Reymond, 2022). They include the production of ovicidal substances, neoplasm formation, egg‐crushing tissue outgrowth, or attraction of egg parasitoids (Reymond, 2013; Hilker & Fatouros, 2015). In plants of the Brassicales order, an efficient defense response is a localized cell death underneath eggs or at the site of egg extract (EE) application, a process named hypersensitive‐like response (HR‐like) based on the resemblance to the effector‐triggered immunity (ETI) response triggered by adapted pathogens (Reymond, 2013; Hilker & Fatouros, 2016). Several studies have reported that HR‐like symptoms induced by eggs from lepidopteran Pieridae species can result in egg desiccation, increased egg mortality, higher egg parasitism, and lower subsequent larval performance (Shapiro & DeVay, 1987; Pashalidou et al., 2013; Fatouros et al., 2014; Griese et al., 2017). Furthermore, the intensity of Pieris brassicae egg‐induced HR‐like and the extent of defense gene induction was shown to vary between or within Brassicaceae species (Griese et al., 2021; Groux et al., 2021; Bassetti et al., 2022, 2024; Caarls et al., 2023), indicating that this trait is under genetic control. Indeed, three loci containing cell‐surface receptor‐like kinases (RLK), TIR‐NBS‐LRR (TNL) intracellular receptors, and genes involved in innate immunity were associated with HR‐like in Brassica rapa (Bassetti et al., 2022). A single locus containing a cluster of TNLs was shown to control HR‐like in Brassica nigra (Bassetti et al., 2024). However, Arabidopsis symptoms in response to oviposition are primarily chlorosis or mild cell death, reminiscent of the pattern‐triggered immunity (PTI), an innate immunity response that follows detection of microbial patterns. By contrast, B. nigra displays a much more intense and egg‐killing response that extends beyond the oviposition site and can be qualified as a typical ETI that may have evolved to counteract yet‐to‐be‐discovered egg‐derived effectors (Caarls et al., 2023; Bassetti et al., 2024).

Signaling of egg‐triggered immunity in different species is still poorly understood, but the accumulation of reactive oxygen species (ROS) together with salicylic acid (SA) has been frequently observed in plants expressing HR‐like (Little et al., 2007; Bruessow et al., 2010; Bittner et al., 2017; Bonnet et al., 2017; Geuss et al., 2017; Lortzing et al., 2019; Caarls et al., 2023). In Arabidopsis, it was shown that treatment with P. brassicae eggs or EEs activates PTI, including early induction of defense genes, and that SA‐dependent signaling and biosynthesis are required for HR‐like induction (Little et al., 2007; Gouhier‐Darimont et al., 2013). Phosphatidylcholines (PCs) in eggs were found to activate these immune responses via the LecRK‐I.1 homolog LecRK‐I.8 (Stahl et al., 2020), a receptor that was previously identified as a significant component of egg perception (Gouhier‐Darimont et al., 2019). Interestingly, the lecrk‐I.1 mutant was also impaired in egg‐induced SA accumulation and defense gene expression, suggesting a combined role for these RLKs (Groux et al., 2021). Also, PCs were detected in egg‐associated secretions from different herbivores, including P. brassicae, and induced PTI responses (Lortzing et al., 2024). Nonlipidic and nonproteinaceous components in the P. brassicae egg glue induce ROS, callose, ethylene, and HR‐like in B. rapa and B. nigra, but their exact composition and perception are still unknown (Caarls et al., 2023). An annexin (ANN)‐like protein in egg‐associated secretion of the sawfly Diprion pini is responsible for the emission of parasitoid‐attracting volatiles in pine, but perception and signaling steps are unknown (Hundacker et al., 2022). With the accumulating evidence that hallmarks of PTI are conserved between different plants and against different eggs (Lortzing et al., 2019; Lortzing et al., 2024), the current model is that, like with microbial pathogens, recognition of conserved insect egg‐associated molecular patterns (EAMPs) leads to a PTI that can trigger defense against further herbivory (Valsamakis et al., 2020) or pathogens (Hilfiker et al., 2014; Alfonso et al., 2021). Some plants that experience more regular oviposition by adapted herbivores may in addition develop ETI to respond to specific effectors (Griese et al., 2021; Bassetti et al., 2024).

Elevation of cytosolic calcium (Ca²⁺) concentration is a prominent feature of PTI (Köster et al., 2022). Similarly, the rise in cytosolic Ca²⁺ plays a role in defense against herbivores through activation of different calcium channels, including cyclic nucleotide‐gated ion channels, glutamate receptor‐like proteins (GLRs), two‐pore channels and ANNs (Mousavi et al., 2013; Lenglet et al., 2017; Vincent et al., 2017; Erb & Reymond, 2019; Meena et al., 2019; Malabarba et al., 2021). For instance, Arabidopsis GLR3.3 and GLR3.6 are glutamate (Glu)‐gated Ca²⁺ channels that participate in long‐distance wound‐induced plant defenses (Mousavi et al., 2013; Toyota et al., 2018; Alfieri et al., 2019). By contrast, Ca²⁺ signaling in response to oviposition remains poorly characterized. Indirect evidence comes from the observation that eggs from the spider mite Tetranychus urticae or from P. brassicae induce the expression of Ca²⁺‐related genes in Arabidopsis (Stahl et al., 2020; Ojeda‐Martinez et al., 2021). However, whether Ca²⁺ accumulation is linked to egg‐induced HR‐like and the nature of calcium channels potentially involved are still unknown.

Exploiting the power of genome‐wide association studies (GWAS) to identify loci underlying genetic variation in Arabidopsis (Atwell et al., 2010), we recently explored the response of Arabidopsis accessions to P. brassicae EE treatment, which tends to amplify and accelerate the natural oviposition symptoms. Using a robust symptom score ranging from no response to localized chlorosis and cell death, we revealed two loci explaining most of the variation observed, one of them containing LecRK‐I.1 (Groux et al., 2021). We validated the locus by showing that a lecrk‐I.1 knockout mutant displays decreased HR‐like response after EE treatment and found that two main haplotypes explain part of the variation in this response between Arabidopsis accessions. We also identified potential signatures of balancing selection at this gene, suggesting that it may be important ecologically (Groux et al., 2021). Interestingly, a LecRK‐I.1 ortholog was found in a locus associated with HR‐like in B. rapa (Bassetti et al., 2022).

Here, we investigated the nature of the second locus identified in the GWAS and report that GLR2.7, which is part of a small clade 2 GLR cluster located on this locus, plays a significant role in triggering Ca²⁺ accumulation and HR‐like underneath the eggs.

Materials and Methods

Plant materials and growth conditions

Arabidopsis thaliana (L.) Heynh. plants were grown in growth chambers in short day conditions (10 h light, 100 μmol m⁻² s⁻¹, 22°C, 65% relative humidity) and were 4 to 5 wk old at the time of treatment. For Ca²⁺ measurements, YELLOW CHAMELEON 3.6 (YC3.6) plants were grown in vitro. Seeds were surface‐sterilized and then plated on ½‐strength Murashige & Skoog medium (½MS), 0.3% sucrose, and 0.8% agar (w/v). Plates were placed in short day conditions for 10 d, and seedlings were transferred to water in 96‐well plates the day before the experiment. Nicotiana benthamiana Domin. seeds were sown directly on soil and cultivated under long day conditions (16 h light, 25°C, 65% relative humidity). A population of the Large White butterfly, Pieris brassicae L., was maintained on Brassica oleracea var. gemmifera DC (Reymond et al., 2000).

Arabidopsis mutant lines

GLR2.7 (At2g29120) was mutated in Arabidopsis by a CRISPR‐Cas9 approach. Two single‐guide RNAs, sg1 (cc_GLR_7b_rev) and sg2 (cc_GLR_7a_fw), were inserted into the pDE42‐Cas9, generating pDEGLR2.7. This construct was used to induce a 424‐bp deletion in the GLR2.7 gene (Supporting Information Fig. S1a). After plant transformation, the presence of a deletion was verified by polymerase chain reaction (PCR), using primers CCglr2.7fw2 and CCglr2.7rv2 flanking the deletion and amplifying a 242‐bp fragment. To generate the beta‐glucuronidase (GUS) reporter line GLR2.7p:NLS‐GFP‐GUS, a 2107‐bp promoter fragment of GLR2.7 was amplified using the primers GLR2.7pfw and GLR2.7prv. The amplified sequence was cloned upstream of the NLS‐GUS coding sequence in the pMK7S*NFm14GW (VIB, Ghent) vector via the Gateway cloning system. Final expression clones were inserted in Agrobacterium tumefaciens strain GV3101 for transformation of Columbia (Col‐0) Arabidopsis by floral dipping.

Arabidopsis accessions were categorized as ‘weak’ or ‘strong’ depending on whether their symptom score in response to EE was in the lower, respectively higher, quintile (Groux et al., 2021; Fig. 1; Table S1). Complementation of ‘weak’ accessions was performed using the GLR2.7 genomic sequences from ‘strong’ accessions. GLR2.7 sequences from Tomegap‐2 (N76250) and Lz‐0 (N28482) were amplified and cloned into the plasmid pGreen229mVenus, using the primers GLR2.7‐clon‐fw and GLR2.7‐clon‐rv, and digested with EcoRV and BamHI. pGreen229mVenus_GLR2.7^Tomegap‐2 and pGreen229mVenus_GLR2.7^Lz‐0 were agroinfiltrated into ‘weak’ accessions, Ull2‐3 (N78817) and Ren‐1 (N77210), respectively. In addition, Ull2‐3 was complemented with the GLR2.7 genomic sequences from Ren‐1 (pK7_Ren1GLR2.7) and Ull2‐3 (pK7_Ull23GLR2.7), using the Gateway cloning with primers GLR2.7‐clon‐Fw2 and GLR2.7‐5UTR‐Rv for the promoters, and primers GLR2.7‐ATG‐Fw and GLR2.7‐clon‐Rv for the GLR2.7 gene sequences.

Fig. 1.

Genome‐wide mapping of insect egg‐induced cell death. (a) Manhattan plot of genome‐wide association study (GWAS) mapping for symptom score after 5 d of Pieris brassicae egg extract (EE) treatment of Arabidopsis plants using an accelerated mixed‐model. Full imputed genotypes for all 295 accessions were used for mapping. Chromosomes are displayed in different colors, and the horizontal dashed line indicates the Bonferroni‐corrected significance threshold at α = 0.05. (b) Local association plot of the GLUTAMATE‐LIKE RECEPTOR2.7 (GLR2.7) locus using the 250 K genotype data for symptom score. The x‐axis represents genomic position on Chromosome 2, and color boxes indicate genes. Linkage disequilibrium (LD) with the most significant single nucleotide polymorphism (SNP) is indicated by a color scale. The dashed line indicates the Bonferroni corrected significance threshold at α = 0.05. (c) Representative images of EE‐treated (dashed circles) leaves with symptoms used for scoring accessions. Shown below are zoomed images of the site of EE treatment.

Arabidopsis lines (MCGf and MCGf_glr2.7) used for calcium quantification were obtained by transformation of Col‐0 and glr2.7 with the plasmid pB7_UBQ10_CGf. pB7_UBQ10_CGf was obtained after amplification of the region of interest of TCTPpro_UBQ10p_mCherry_GCaMP6f (Weigand et al., 2021) using the primers attB1_Fragment_FOR and GCaMP6‐rv and cloned into pB7m24GW3 (VIB, Ghent, Belgium) under the ubiquitin‐10 (UBQ10) promoter using the Gateway system. The resulting expression construct (pB7_UBQ10_CGf) was inserted in A. tumefaciens GV3101 for transformation of Arabidopsis by floral dipping.

The SA‐biosynthesis mutant sid2‐1 was obtained from C. Nawrath (University of Lausanne). T‐DNA insertion lines for glr2.7 (SALK_121990), glr2.8 (SALK_111659) and glr2.9 (SALK_125496) were obtained from the NASC stock center. The CRISPR‐Cas9‐generated triple mutant glr2.7/2.8/2.9 was obtained from C. Zipfel (University of Zürich) and described previously (Bjornson et al., 2021).

Calcium and glutamate reporter lines

Ca²⁺ fluxes were monitored using Arabidopsis GCaMP6f and YC3.6 reporter lines. For experiments measuring calcium in glr2.7, we used the lines MCGf and MCGf_glr2.7. The ratiometric CGf line (mCherry fused to GCaMP6f) was provided by A. Costa (Weigand et al., 2021). The YC3.6 in Col‐0 and glr2.7/2.8/2.9 background was obtained from C. Zipfel (University of Zurich). The 35S::CHIB‐iGluSnFR Glu reporter line (Toyota et al., 2018) was obtained from C. Faulkner (John Innes Center, Norwich, UK).

Oviposition and treatment with EE

Plants were placed in a tent containing c. 20 P. brassicae butterflies for a maximum of 2 h and then placed in a growth chamber in plastic boxes until hatching of the eggs. Control plants were kept in the same conditions without butterflies. P. brassicae eggs were collected and crushed with a pestle in Eppendorf tubes. After centrifugation (15 000  g , 3 min), the supernatant (‘egg extract’) was collected and stored at −20°C. For each plant, two leaves were treated with 2 μl of EE. A total of four plants were used for each experiment. After 3 to 6 d, depending on the experiment, EE was gently removed with a scalpel blade and treated leaves were stored in liquid nitrogen. Untreated plants were used as controls. For calcium measurement using YC3.6, plants were grown in vitro and used at 11 d. EE was added to 96‐well plates containing seedlings at a final dilution of 1/150.

Glutamate treatment

L‐glutamic acid (G1251; Sigma‐Aldrich), ≥ 99% (high‐performance liquid chromatography (HPLC)) was prepared at a 58 mM stock solution in water. For experiments with 4‐ to 5‐wk‐sold plants, a 2 μl drop of 10 mM Glu, diluted in a control solution (0.05% Silwet L‐77), was applied to the abaxial leaf surface.

Genome‐wide association mapping and haplotype analysis

GWAS analysis of Arabidopsis response to insect eggs was described recently (Groux et al., 2021). Briefly, a set of 295 accessions from the HapMap panel (Horton et al., 2014) was used. For each accession, three leaves from three to six plants were treated with EE diluted 1 : 1 with deionized water. Treated plants were left in the growth chamber for an additional 5 d until phenotyping. After 5 d, treated leaves were removed with forceps, symptoms were scored from 0 to 4, which are arbitrary units corresponding to the intensity of the response to EE (Groux et al., 2021; Fig. 1; Table S1), and leaves were frozen in liquid nitrogen for further SA quantification. Pools of 30 accessions were phenotyped every week. GWAS mapping, haplotype analysis, and population‐wide cladogram were described previously (Groux et al., 2021). Haplotype networks were built using the ‘ape’ and ‘pegas’ R packages on the full GLR2.7 gene from 122 sequenced accessions. Sequences were obtained from SALK 1001 genome browser (http://signal.salk.edu/atg1001/3.0/gebrowser.php), and sequences were aligned using Multiple Alignment using Fast Fourier Transform for further analysis.

Cell death measurement

For visualization of cell death, EE was gently removed. For all experiments, cell death was quantified by red light measurements in Arabidopsis leaf disks at 6 d posttreatment using a Hidex microplate reader (excitation at 650 nm and emission at 680 nm; Landeo Villanueva et al., 2021).

Salicylic acid quantifications

SA quantification was performed using the bacterial biosensor Acinetobacter sp. ADPWH using a Hidex microplate reader, according to the published protocols (Huang et al., 2006; Stahl et al., 2020; Groux et al., 2021).

Gene expression analysis

Analysis of gene expression by real‐time quantitative PCR was described previously (Groux et al., 2021). For analysis of GLR2.7 expression in different Arabidopsis accessions, primers were designed to match GLR2.7 sequences from all accessions. A list of all primers used in this study is found in Table S2.

Transient transformation of N. benthamiana and protein visualization

To visualize and determine the subcellular localization of the GLR2.7 protein in planta, the construct 5′ untranslated region (UTR)‐coding sequence of GLR2.7^Col‐0‐Venus‐3'UTR was synthesized by GenScript, amplified using the primers GLR2.7‐5UTR‐Fw and GLR2.7‐clon‐Rv, and cloned into pB7m34GW (VIB) under the ubiquitin‐10 (UBQ10) promoter using the Gateway system. The resulting expression construct (pB7UBQ10_GLR2.7_Venus) was inserted in A. tumefaciens GV3101 for leaf infiltration in 4‐wk‐old N. benthamiana plants. Epidermal cells of infiltrated leaves were observed 2 d postinfiltration. To visualize the plasma membrane, an aqueous solution of 10 μM FM4‐64 (SymaptoRed Reagent; Sigma‐Aldrich, 574 799, ≥ 99%, HPLC) was infiltrated into the target leaf area 15 min before imaging.

Fluorescence microscopy

For analysis of GCaMP6f and iGluSnFR reporter lines, imaging of leaf surfaces was conducted using an SMZ18 stereomicroscope equipped with an ORCA‐Flash4.0 camera and eGFP and mCherry emission/excitation filter sets.

For GLR2.7‐mVenus protein localization, fluorescence imaging of epidermal cells was performed using a Leica Stellaris confocal laser scanning microscope. Images were acquired with a 20×/0.75 dry objective at zoom factor 5.

Excitation and detection windows were set as follows: eGFP (470 nm, 510–523 nm), mVenus (514 nm, 520–550 nm), and mCherry (561 nm, 605–650 nm). All images were processed using the imagej software (v.2.14.0/1.54f).

Amino acid quantification

To quantify amino acids released from P. brassicae eggs, butterflies were allowed to oviposit on filter paper as described previously (Stahl et al., 2020). Eggs and filter papers were collected separately after 1 d. Samples were immediately flash‐frozen in liquid nitrogen and stored at −80°C until analysis. Extraction buffer (EtOH/H₂O, 50 : 50, 0.1% HCO₂H) was added to each sample at 1 : 10, w/v. Samples were centrifuged for 10 min at 10 400 g at 10°C, and supernatants were separated and evaporated at 45°C under vacuum. The residues were reconstituted with ultra‐pure water, resulting in a 10‐fold concentration (e.g. 200 μl supernatant reconstituted in 20 μl water). Free amino acids were quantified using the AccQ‐Tag Ultra Derivatization Kit (Waters, Milford, MA, USA) according to the manufacturer's instructions. Amino acid content was analyzed by liquid chromatography mass spectrometry (LC‐MS) in positive mode with single‐ion recording and quantified by comparison with a mixture of the corresponding standards (Meier et al., 2024).

Cytosolic calcium and apoplastic glutamate analyses

GCaMP6f and iGluSnFR reporter lines were treated with EE or Glu on the abaxial side of the leaf. Images were acquired at 0, 10, 30, and 60 min as well as 24, 48, and 72 h posttreatment. Fluorescence intensity in the treated area was quantified using the imagej2 software (v.2.14.0/1.54f), and the ratio R = GFP/mCherry was calculated. The relative change of fluorescence was then determined as (R–R ₀)/R ₀, where R ₀ is the ratio value at time 0. For background correction, Arabidopsis Col‐0 wild‐type (WT) leaves were treated with EE and imaged at the same time points as the reporter lines. The low autofluorescence signal detected in GFP and mCherry channels in EE‐treated WT leaves was quantified and subtracted from the fluorescence measurements of reporter lines.

For YC3.6 seedlings, plants were placed in black 96‐well plates containing 150 μl of distilled water and kept in the dark for 12 h. The following day, the water was carefully replaced without disturbing the seedlings; the signal acquisition and the data analysis were performed as described in Bjornson et al. (2021).

Data analysis

Statistical analyses were performed using graphpad prism 10 (v.10.1.1). GWAS mapping and subsequent analysis of the data obtained were performed with the R software v.3.6.

Results

GWAS mapping and identification of associated loci

To investigate the genetic basis of insect egg‐induced HR‐like responses, we performed a GWAS on a world‐wide set of 295 Arabidopsis accessions. Symptom score (from 0 to 4; Fig. 1) and total SA levels were quantified in all accessions after 5 d of EE treatment and were used for mapping (Table S1). Initial mapping using an imputed genotype matrix for 2029 accessions (Togninalli et al., 2018) revealed a highly significant peak associated with symptom score and total SA levels (−log₁₀(P) = 15.96 and 10.64, respectively) spanning over 10 Kb on Chromosome 2 (Figs 1, S2 for SA) and one marker reaching significance on Chromosome 3 (−log₁₀(P) = 7.59) that is only associated with symptom score. The description of this second locus was previously reported and revealed that the receptor kinase LecRK‐I.1 (At3g45330) contributes significantly to insect egg‐induced responses (Groux et al., 2021). As the genotype matrix used was produced by imputing missing genotypes based on a subset of accessions that were sequenced, we used the 250 K single nucleotide polymorphism (SNP) genotype data that were available for all used accessions for further investigation of specific markers (Horton et al., 2012). The three most significantly associated SNPs occurred within the coding region of GLR2.7 for both phenotypes, while two additional markers were located in the promoter of the gene (Figs 1, S2). Interestingly, this locus contains two other clade 2 GLR members, GLR2.8 and GLR2.9, which are the closest homologs of GLR2.7 (Roy & Mukherjee, 2017). To further explore whether variation within GLR2.7 could be causal for variation in HR‐like responses and SA accumulation, we examined linkage disequilibrium (LD) between the top SNP and the other markers in a 50‐Kb window around the variant. We found that the most significantly associated SNPs in both scans were in very high LD with other SNPs located in the gene body of GLR2.7, while LD decayed rapidly around the gene for other markers (Figs 1b, S2). Both SNP1 and SNP2 appeared to be in very high LD with SNP3, suggesting that this gene might be involved in cell death induction and SA accumulation upon insect egg perception.

As mentioned, SA is necessary to induce cell death upon EE perception (Gouhier‐Darimont et al., 2013). Interestingly, we found that total induced SA and symptom score were moderately correlated (r = 0.40) in the entire mapping panel (Fig. S2). This indicates that SA does not fully explain the strength of egg‐induced HR‐like responses and that other signaling components might also contribute, which is consistent with the partial reduction in cell death observed in the SA‐biosynthesis mutant sid2‐1 (Gouhier‐Darimont et al., 2013). However, this could also be consistent with SA working in a more qualitative way, as almost all accessions showed some degree of SA accumulation in response to EE treatment, regardless of the degree of symptoms (Table S1).

Validation of the GLR2.7 locus

To verify that the GLR2.7 locus identified in the GWAS was responsible for the observed phenotype, we first selected accessions that showed a highly contrasting response to EE treatment. Ren‐1 and Ull2‐3 were among the 10 accessions with the lowest score (< 0.06) and were considered ‘weak’ accessions, whereas Lz‐0 and Tomegap‐2 were among the top 20 accessions with the highest symptom score (> 3.00) and were considered ‘strong’ accessions (see the Materials and Methods section and Table S1). Then, we introduced the GLR2.7 genomic sequence from strong accessions into weak accessions. Strikingly, when GLR2.7 from Lz‐0 and Tomegap‐2 was introduced into Ren‐1 and Ull2‐3, respectively, each complemented accession showed an HR‐like that was significantly higher than in the weak accession and as high as in the strong accession (Fig. 2a,d). Also, the EE‐induced SA level in Ull2‐3 complemented with GLR2.7 ^Tomegap‐2 was significantly higher than in Ull2‐3 (Fig. 2b). However, GLR2.7 ^Tomegap‐2 did not fully restore SA accumulation in Ull2‐3 and this effect was not observed for GLR2.7 ^Lz‐0 in Ren‐1, suggesting a partial contribution of GLR2.7 for this response.

Fig. 2.

Arabidopsis GLUTAMATE‐LIKE RECEPTOR2.7 (GLR2.7) locus is responsible for quantitative variation in Pieris brassicae egg extract (EE)‐induced responses. Complementation of Ren‐1 and Ull2‐3 ‘weak’ Arabidopsis thaliana accessions with GLR2.7 from Lz‐0 and Tomegap‐2 (Tmgp‐2) ‘strong’ accessions leads to enhanced cell death (a, d), salicylic acid (SA) accumulation (b), and H₂0₂ accumulation (c, d) in response to EE treatment. (a) Cell death quantification after 6 d of EE treatment was measured by red light fluorescence. Mean ± SE of one biological replicate is shown (n = 10–12), this experiment was repeated twice with similar results. (b) SA quantification after 3 d of EE treatment. Mean ± SE of three biological replicates is shown (n = 4 per experiment). (c) H₂0₂ accumulation measured by 3,3′‐Diaminobenzidine (DAB) staining after 3 d of EE treatment. Mean ± SE of one biological replicate is shown; this experiment was repeated twice with similar results. (d) Representative pictures of cell death and H₂O₂ accumulation (DAB staining), respectively after 6 and 3 d of EE treatment. Statistical differences within each group of weak, complemented, and strong accessions are indicated by lowercase and upper‐case letters (ANOVA followed by Tukey's HSD). For clarity, weak accessions are depicted by green bars, strong accessions by red bars and complemented accessions by brown bars. RFU, relative fluorescence unit; ROS, reactive oxygen species.

Given the known correlation between egg‐induced HR‐like and ROS accumulation (Little et al., 2007; Bittner et al., 2017; Geuss et al., 2017), we then measured EE‐induced H₂O₂ in the complemented accessions. Again, introducing GLR2.7 from strong into weak accessions significantly enhanced ROS production (Fig. 2c,d).

In addition, we complemented the weak Ull2‐3 with GLR2.7 from the weak accessions Ull2‐3 or Ren‐1, and from the strong accession Tomegap‐2 as a positive control. Only GLR2.7 ^Tomegap‐2 was able to increase HR‐like when introduced in Ull2‐3, indicating that this effect is specifically due to the GLR2.7 haplotype (Fig. S3). Collectively, these data confirm that the GLR2.7 locus is responsible for the observed variation in EE‐induced HR‐like and SA content between Arabidopsis accessions.

To further investigate the role of the GLR2.7 locus in the quantitative response to EE, we generated a GLR2.7 ^Col‐0 knockout by CRISPR‐Cas9. In this glr2.7 mutant, no measurable expression could be detected (Fig. S1b). Importantly, glr2.7 displayed significantly lower EE‐induced HR‐like, SA accumulation, and ROS production compared with Col‐0 (Fig. 3a–c), indicating that the modulation of these responses depends on a functional protein. Expression of the known egg‐responsive genes PR1 and SAG13 (Stahl et al., 2020) was also attenuated in glr2.7 (Fig. 3d).

Fig. 3.

Arabidopsis GLUTAMATE‐LIKE RECEPTOR2.7 (GLR2.7) modulates responses to egg extract (EE). (a) Cell death quantification after 6 d of Pieris brassicae EE treatment was measured by red light fluorescence. Mean ± SE of one biological replicate is shown (n = 12). This experiment was repeated twice with similar results. (b) Salicylic acid (SA) quantification after 3 d of EE treatment. Mean ± SE of three biological replicates is shown (n = 4 per experiment). (c) H₂0₂ accumulation measured by DAB staining after 3 d of EE treatment. Mean ± SE of one biological replicate is shown (n = 10–13). This experiment was repeated twice with similar results. (d) Expression of egg‐induced defense marker genes after 3 d of EE treatment. Transcript levels were monitored by real‐time quantitative polymerase chain reaction and normalized to the reference gene SAND. Mean ± SE of three technical replicates is shown. This experiment was repeated twice with similar results. Letters denote statistical differences (ANOVA followed by Tukey's HSD). ROS, reactive oxygen species.

In summary, we identify GLR2.7 as a modulator of egg‐induced defense responses. This protein contributes to the quantitative variation in HR‐like and SA levels between accessions, together with LecRK‐I.1 for HR‐like (Groux et al., 2021).

Different haplotypes of GLR2.7 segregate in Arabidopsis populations

A deeper analysis of the GLR2.7 locus using the imputed genotype matrix for 2029 accessions revealed different haplotypes segregating in the population. By looking at all significantly associated markers within the GLR2.7 gene sequence, we found 49 SNPs spanning the entire coding sequence and promoter, and most of them showed a highly significant association with symptom score (−log₁₀(P) > 10; Fig. 4a). As with the lower density SNP data, we found that all markers within the GLR2.7 coding region were in very high LD (> 0.8). The fact that many markers are genetically linked again supports the potential existence of distinct haplotypes segregating at this locus. Overall, out of 49 significantly associated SNPs, 16 were found to induce nonsynonymous changes in the protein sequence. Regarding SNP1‐3, SNP1 is located in an intron, SNP2 leads to a synonymous amino acid change, and SNP3 to the nonsynonymous substitution V448D (Tables S1, S3).

Fig. 4.

Local association and haplotype analysis of the Arabidopsis GLUTAMATE‐LIKE RECEPTOR2.7 (GLR2.7) locus. (a) Local association plot of the GLR2.7 locus using the full imputed genotype data. The x‐axis represents genomic position on Chromosome 2. Linkage disequilibrium (LD) with the most significant SNP is indicated by a color scale. The dashed line indicates the Bonferroni‐corrected significance threshold at α = 0.05. (b) The three selected SNPs from Fig. 1 define two major haplotypes (only haplotypes containing at least two accessions are shown). Min and Max values are indicated, as well as the median (horizontal line). The box represents 50% of the values. The whiskers extend from the box to the minimal and maximal values. Box and whisker plots of total salicylic acid (SA) and symptom score after 5 d of egg extract (EE) treatment are shown. Significant differences are indicated (Student's t‐test, ***, P < 0.01). Col‐0, Columbia.

To evaluate the population‐wide genetic structure of this locus, we used the three significant SNPs (SNP1‐SNP3) identified in Fig. 1 to search for GLR2.7 haplotypes. In most accessions, SNP1 was either a T or an A, SNP2 either a C or a T, and SNP3 either an A or a T (Table S1). Strikingly, we found that SNP1‐SNP3 define two main haplotypes within the entire mapping population, one having TCA alleles for SNP1‐SNP2‐SNP3 and associated with a low symptom score (weak HR‐like) and another one having ATT alleles and associated with a high symptom score (strong HR‐like; Fig. 4b). We thereafter named these two haplotypes GLR2.7 ^ATT and GLR2.7 ^TCA, respectively. By contrast, other allelic series were either not present or contained a very low number of accessions and were therefore not further considered (Fig. 5a). Interestingly, Col‐0 possesses the GLR2.7 ^ATT haplotype (Fig. 4b), as well as Lz‐0 and Tomegap‐2 mentioned above. Altogether, our data indicate the existence of alleles at the GLR2.7 locus that modulates the intensity of cell death and contributes partially to SA accumulation upon treatment with P. brassicae EE. We also noticed that, overall, the 49 significant SNPs found in the GLR2.7 locus tend to have only two variants that correspond to either the TCA or the ATT haplotype (Table S1).

Fig. 5.

Various signatures of balancing selection at the Arabidopsis GLUTAMATE‐LIKE RECEPTOR2.7 (GLR2.7) locus. (a) Frequency of the haplotypes within a world‐wide set of Arabidopsis accessions. (b) Cladogram constructed from a genome‐wide kinship matrix of the 295 accessions used in this study. The outermost circle indicates haplotype. (c) Haplotype network of the GLR2.7 sequence in the 122 sequenced accessions. Circles represent haplotypes, and lines represent the number of stepwise mutations separating two haplotypes. The two colored areas delimit sequences belonging to the haplotypes defined using genome‐wide association study (GWAS) data. (d) Sliding window analysis of Tajima's D, Wu and Li's D and F statistics along the GLR2.7 coding region using a window size of 200 bp and a step size of 25 bp. A subset of 125 accessions for which available full genome sequences were used for this analysis. For Wu and Li's statistical tests, Arabidopsis lyrata was used as an outgroup. The dashed lines indicate significance threshold at P < 0.05. The gene structure of GLR2.7 is shown below.

Signatures of balancing selection at the GLR2.7 locus

Genes that confer herbivore or pathogen resistance are advantageous in the presence of pests but are usually costly or detrimental to maintain in their absence (Vila‐Aiub et al., 2011; Van Velzen & Etienne, 2015). Balancing selection is a process through which multiple alleles are kept at intermediate frequency in a population, a process that is known to shape disease resistance loci in plants (Bakker et al., 2006; Huard‐Chauveau et al., 2013). Loci under balancing selection display highly diverged alleles, and haplotype distribution is widespread across the geographical range of the considered species. As shown by the haplotype frequency distribution at the GLR2.7 locus, two haplotypes defined by SNP1‐3 are present at intermediate frequencies, while all others are found in very few individuals (Fig. 5a).

To explore whether the distribution of GLR2.7 haplotypes was caused by geographical or phylogenetic factors, we constructed a cladogram representing genome‐wide distance between accessions by using the kinship matrix used for the GWAS mapping. Clearly, both GLR2.7^ATT and GLR2.7^TCA haplotypes are homogeneously distributed among all populations of Arabidopsis, indicating that they are not confined by phylogenetic or geographical proximity (Figs 5b, S4). To further analyze the genetic structure of GLR2.7, we used gene sequences from 122 accessions with full genome available (Table S1) to build a haplotype network (Fig. 5c). These data show that the various alleles at this locus are structured into two deeply divergent haplotypes, separated by at least 12 mutations, corresponding to the ones defined previously. Interestingly, the GLR2.7 ^TCA haplotype displays high intra‐haplotype divergence as shown by the presence of rare alleles. This is potentially indicative of relaxed selection or drift occurring in some accessions. The broad geographical presence of both haplotypes and their deep divergence suggests that they might be ancient polymorphisms and have been conserved through selection. Several statistical tests – such as Tajima's D and Fu and Li's F or D – have been described to test the hypothesis that a sequence evolves neutrally by comparing the amount of variation observed and the variation expected for a given sequence. Deviations from neutrality are identified by either positive or negative test values, indicating balancing or purifying selection, respectively. Tajima's D is known to be sensitive to the species demography, and we therefore also computed the Fu and Li's D and F statistics (Tajima, 1989; Fu & Li, 1993), which incorporate an outgroup sequence (here Arabidopsis lyrata) to better discriminate signatures of selection from demographic history. All three tests displayed positive values on most parts of the GLR2.7 gene sequence (Fig. 5d) with some stretches reaching significance, therefore clearly indicating that this locus does not evolve neutrally. Altogether, these results show that two deeply diverged haplotypes of GLR2.7 segregate in natural Arabidopsis populations, and we provide evidence that balancing selection may be maintaining variation at this locus.

As LecRK‐I.1, the other locus identified in our GWAS experiment, also displays signatures of balancing selection with two dominant haplotypes (Groux et al., 2021), we looked at the distribution of strong and weak haplotypes from both genes between accessions. Interestingly, although they are not linked genetically (LD, r ² = 0.025), an additive effect of haplotypes on HR‐like could be observed. Indeed, accessions harboring a strong haplotype for both genes (GLR2.7 ^ATT and LecRK‐I.1 ^TACAA) showed a significantly higher symptom score than accessions harboring a weak haplotype (GLR2.7 ^TCA and LecRK‐I.1 ^CGTGC), whereas accessions having a mixed combination displayed an intermediate phenotype (Fig. S5). On the contrary, EE‐induced SA was only correlated with the GLR2.7 haplotype (Fig. S5), as expected by the lack of association at the LecRK‐I.1 locus for this trait (Fig. S2).

GLR2.7 expression

In order to explore how GLR2.7 could contribute to defense responses, we first measured its expression after EE treatment. Expression of GLR2.7 was significantly induced by EE, and this induction was dependent on SA, as shown by a lack of induction in sid2‐1 (Fig. 6a). Additionally, analysis of previously published microarray data showed that GLR2.7 expression is induced after natural oviposition by P. brassicae (Little et al., 2007). In line with these findings, we showed using a GLR2.7::NLS‐GFP‐GUS reporter line that GLR2.7 expression is strongly induced at the site of P. brassicae oviposition or EE treatment (Fig. 6b).

Fig. 6.

Expression of Arabidopsis GLUTAMATE‐LIKE RECEPTOR2.7 (GLR2.7) is induced by Pieris brassicae eggs. (a) GLR2.7 expression in response to egg extract (EE) in Columbia (Col‐0) and salicylic acid deficient line sid2‐1. Transcript levels were monitored 3 d after treatment by real‐time quantitative polymerase chain reaction and normalized to the reference gene SAND. Mean ± SE of three technical replicates is shown. This experiment was repeated twice with similar results. Letters denote statistical differences (ANOVA followed by Tukey's HSD). (b) Beta‐glucuronidase (GUS) staining of GLR2.7::NLS‐GFP‐GUS line 3 d after P. brassicae oviposition or EE treatment. (c) Subcellular localization of transiently expressed UBQ10::GLR2.7‐Venus and the plasma membrane marker FM4‐64 in Nicotiana benthamiana leaves. Bars, 20 μm.

To study the protein cellular localization, we expressed a Venus‐tagged GLR2.7 under the strong UBQ10 promoter in Nicotiana benthamiana leaves. Most of the Venus signal was localized to the plasma membrane, as indicated by colocalization with the plasma membrane marker FM4‐64 (Fig. 6c). We, however, also noted signal at the nucleus and in cytosolic projections, which might be due to overexpression or indicate that GLR2.7 is not exclusively targeted to the plasma membrane.

Then, to assess whether GLR2.7 expression was associated with symptom scores measured in the GWAS analysis, we selected 40 accessions from the mapping population, 21 with a weak (< 0.55) and 18 with a strong symptom score (> 2.55; Table S1), and we assessed GLR2.7 expression in response to P. brassicae EE treatment. We observed variable basal and inducible levels of expression between accessions (Fig. S6; Table S1). Although basal expression was significantly higher in strong accessions, this difference was no longer significant after EE treatment (Fig. S6b). However, when looking at the GLR2.7 haplotype in these 40 accessions, both basal and EE‐inducible GLR2.7 expression were significantly higher in accessions harboring the GLR2.7 ^ATT strong haplotype (Fig. S6b). Altogether, these findings point to a potential contribution of GLR2.7 expression to the difference in egg‐induced responses between accessions.

Although GLR2.7 is located next to GLR2.8 and GLR2.9, we did not detect significant SNPs associated with HR‐like or SA levels in the two homologous genes. We anyhow observed a significant induction of the three genes in response to EE treatment, pointing to a potential additive role for GLR2.8 and GLR2.9 (Fig. S7a). However, single glr2.8 and glr2.9 knockout lines only showed a mild and nonsignificant reduction of EE‐induced cell death compared with the other lines, contrary to the reduction observed in glr2.7 and glr2.7/2.8/2.9 lines compared with Col‐0 (Fig. S7b).

GLR2.7 regulates egg‐induced cytosolic Ca²⁺ accumulation

Members of the GLR gene family are amino acid‐gated Ca²⁺‐permeable channels that play a role in diverse physiological processes, including defense (Grenzi et al., 2022; Simon et al., 2023). Seminal studies have demonstrated the involvement of Arabidopsis clade 3 GLR 3.3 and GLR 3.6 in insect‐ and wound‐induced systemic electrical and Ca²⁺ signaling (Mousavi et al., 2013; Toyota et al., 2018; Gao et al., 2023). We thus tested whether eggs or EE application trigger cytosolic Ca²⁺ influx, potentially indicative of GLR function. We used a ratiometric Ca²⁺ reporter line CGf, which contains a mCherry reference domain fused to the intensiometric Ca²⁺ reporter GCaMP6f (Weigand et al., 2021). Strikingly, we observed a green fluorescent signal under P. brassicae eggs or at the site of EE application, indicating a localized cytosolic Ca²⁺ accumulation (Fig. 7a). A time‐course analysis showed that EE triggers a fast and long‐lasting cytosolic Ca²⁺ increase, with an early peak at 10 min followed by a gradual increase up to 72 h after application (Fig. 7b).

Fig. 7.

Clade 2 Arabidopsis glutamate receptor‐like proteins (GLRs) participate in egg‐induced cytosolic calcium (Ca²⁺) accumulation. (a) Visualization of cytosolic Ca²⁺ on the leaf adaxial side using Arabidopsis reporter GCaMP6f (ratiometric CGf line) 3 d after Pieris brassicae oviposition or egg extract (EE) treatment (upper panels). Corresponding pictures of the abaxial side are shown in the lower panels. (b) Time course of cytosolic calcium (Ca²⁺) accumulation after EE treatment in GCaMP6f plants. CTL, untreated. Mean ± SE of two biological replicates is shown (n = 6 per experiment). Values represents the ratio (R) of green fluorescence (GCaMP6f) to magenta fluorescence (mCherry; proportional to the Ca²⁺ concentration), normalized to the initial ratio (R ₀). (c) Time course of cytosolic Ca²⁺ accumulation after EE treatment in Columbia (Col‐0) or glr2.7 containing GCaMP6f (MCGf and MCGf_glr2.7 lines). Mean ± SE of one biological replicate is shown (n = 8). This experiment was repeated once with similar results. (d) Cytosolic Ca²⁺ quantification 10 min and 72 h after EE treatment. Values are mean ± SE from two independent biological replicates (n = 8 per experiment). Letters denote statistical differences (ANOVA followed by Tukey's HSD).

To specifically assess the contribution of GLR2.7, we introduced the CGf ratiometric reporter in Col‐0 and glr2.7. In the mutant line, Ca²⁺ influx was significantly reduced compared with Col‐0, pointing to a prominent role of GLR2.7 in EE‐induced Ca²⁺ accumulation (Fig. 7c,d).

We also took advantage of a genetically encoded YC3.6 Ca²⁺ indicator line with a large deletion in the GLR2.7‐GLR2.9 genomic region (Bjornson et al., 2021). Compared with YC3.6 Col‐0, the YC3.6 glr2.7/2.8/2.9 triple‐mutant line accumulated significantly less Ca²⁺ after EE treatment, confirming the involvement of GLR2.7 in this response (Fig. S8).

Glu is one of the known ligands for plant GLRs (Alfieri et al., 2019), and apoplastic Glu was shown to trigger systemic Ca²⁺ accumulation and defense responses (Toyota et al., 2018). We thus reasoned that insect eggs or egg‐associated secretions may contain and release Glu to activate plasma‐membrane localized GLRs. Strikingly, we measured 41.3 ± 11.9 nmol mg⁻¹ of free Glu in eggs and 7.3 ± 1.6 nmol mg⁻¹ on filter paper 1 d after P. brassicae oviposition, indicating that this amino acid either diffuses out of the egg or is present in egg‐associated secretions, and thus reaches the leaf surface during natural egg laying (Fig. 8a). We also found that, besides Glu, the majority of amino acids present in eggs are not detected on filter paper, with the exception of proline and, to a lesser extent, histidine and aspartic acid (Fig. S9). Then, to see whether exogenous Glu can enter the apoplastic space, we used the apoplastic Glu reporter line 35S::CHIB‐iGluSnFR (Toyota et al., 2018). Upon oviposition or EE treatment, we could clearly detect a fluorescent signal under the eggs or at the site of treatment, strongly suggesting the presence of apoplastic Glu (Fig. 8b). As a positive control, external application of 10 mM Glu, which is in the range found in EE, similarly induced a fluorescent signal at the site of treatment (Fig. S10a).

Fig. 8.

Egg‐derived glutamate (Glu) diffuses to Arabidopsis leaf surface and external Glu triggers glutamate receptor‐like protein (GLR)‐dependent calcium (Ca²⁺) influx. (a) Glu quantification in Pieris brassicae eggs and on filter paper (FP) below the eggs. Values are mean ± SE measured 1 d after oviposition (n = 3–7) (b) Visualization of apoplastic glutamate (red arrowheads) 3 d after oviposition or egg extract (EE) treatment using Arabidopsis reporter iGluSnFR (upper panels). Corresponding pictures of the leaf abaxial side are shown in the lower panels. Trichomes display constitutive Glu accumulation. (c) Time course of apoplastic Glu accumulation after EE treatment. Fluorescent signal (F) was measured in iGluSnFR and normalized to the value at time 0 (F ₀). Mean ± SE of two biological replicates is shown (n = 3 per experiment). (d) Time course of cytosolic Ca²⁺ accumulation after 10 mM Glu treatment in GCaMP6f plants. CTL, control solution (0.05% Silwet L‐77). Mean ± SE of two biological replicates is shown (n = 6 per experiment).

Also, quantification of the fluorescence signal over time indicated a fast and long‐lasting accumulation of apoplastic Glu in response to EE treatment (Fig. 8c). We then showed that Glu treatment triggers cytosolic Ca²⁺ increase at the site of application (Fig. S10b) and that the temporal accumulation (Fig. 8d) is similar to the one observed with EE treatment (Fig. 7). Finally, like with EE treatment (Fig. S8), the YC3.6 glr2.7/2.8/2.9 triple mutant line accumulated significantly less Ca²⁺ after Glu treatment than YC3.6 Col‐0 (Fig. S10c,d).

Altogether, these findings thus identify GLR2.7 as a key component mediating egg‐induced defense responses through Ca²⁺ signaling and point to a potential activation via the presence of Glu at the oviposition site.

Discussion

The use of natural variation helps reveal unprecedented levels of detail in the interactions between plants and insects by studying traits whose genetic variation has been shaped through selection and drift. The understanding of plant responses to insect eggs is still scarce at the molecular level but could provide useful tools for pest control (Tamiru et al., 2015; Fatouros et al., 2016). Here, we showed that mild cell death or chlorosis triggered by egg perception is associated with a new locus in Arabidopsis.

It is quite fascinating to observe such extensive variation in the response to P. brassicae eggs and EE in natural Arabidopsis accessions. Together with many phenotypes previously described, such as cell death, ROS, and SA accumulation, gene expression, and the induction of systemic acquired resistance (SAR; Little et al., 2007; Bruessow et al., 2010; Gouhier‐Darimont et al., 2013, 2019; Hilfiker et al., 2014), this observation further supports the findings that EE gives a similar, yet accelerated, response to real P. brassicae eggs (Little et al., 2007) and thus represents a useful tool to dissect this response. It is, however, noteworthy that Arabidopsis accessions develop a range of relatively mild symptoms compared with other Brassicaceae, like B. nigra, that display a strong egg‐killing necrosis (Bonnet et al., 2017; Griese et al., 2020, 2021). Although Arabidopsis and B. nigra trigger cell death, defense gene expression and SA accumulation (Shapiro & DeVay, 1987; Bonnet et al., 2017; Griese et al., 2020; Caarls et al., 2023), these illustrate a generic PTI response that is activated by EAMP perception. The enhanced ETI‐like response of some Brassicaceae species likely illustrates the ongoing arms race between adapted herbivores and resistant plants that have evolved genes to counteract effectors, including the PEK locus that contains TNL receptor genes (Bassetti et al., 2024). However, given recent findings suggesting that PTI and ETI immune pathways are interconnected (Yuan et al., 2021a,b), explaining similarities in downstream responses, future studies should aim at clarifying whether Arabidopsis lacks resistance genes to oviposition.

In this study, we provide clear evidence that the GLR2.7 locus identified by the GWAS mapping modulates egg‐induced PTI. Indeed, the transfer of GLR2.7 from strongly responsive accessions to weakly responsive ones can restore chlorosis, higher SA accumulation and higher ROS production. This finding is corroborated by decreased EE‐induced symptoms and defenses in a glr2.7 knockout line in Col‐0. The large effect associated with the GLR2.7 locus in the accession panel most likely indicates that it is an upstream component of the EE‐induced response, consistent with the strong association observed with total SA levels. Despite being mildly correlated, both traits thus appear to have a partially overlapping genetic structure. Since variation in symptoms has an additional contribution from LecRK‐I.1 (Groux et al., 2021), these findings reveal that the response of Arabidopsis to EE of P. brassicae has a polygenic structure, with two loci accounting for a large proportion of the variation observed. This is consistent with the observation that quantitative disease resistance is usually under the control of multiple genes (Roux et al., 2014; Corwin & Kliebenstein, 2017) and contrasts with the B. nigra HR‐like response that is under the control of the single locus PEK (Bassetti et al., 2024).

Although it is located next to GLR2.8 and GLR2.9, GLR2.7 alone can rescue the phenotype of a weak accession, and EE‐induced Ca²⁺ influx is significantly reduced in glr2.7. In addition, EE‐induced cell death was not affected in glr2.8 and glr2.9 mutants. It is thus unlikely that these two homologs have a substantial contribution to egg‐induced defenses, although it cannot be formally ruled out. Further investigation will be needed to clarify this point. Recently, GLR2.7, GLR2.8 and GLR2.9 were shown to belong to a set of core immunity responsive genes, which were induced by various pathogen‐associated molecular patterns (PAMPs) involved in PTI and, consistently, the glr2.7/2.8/2.9 mutant was more susceptible to bacterial infection (Bjornson et al., 2021). Thus, studying the various roles of each member of this subset of clade 2 GLRs may unveil specific or overlapping functions in response to different biotic stresses.

Interestingly, we observed that the GLR2.7 locus consists mainly of two diverged haplotypes present throughout the entire Arabidopsis population studied. Both haplotypes are found at intermediate frequency independently of any geographic or phylogenetic pattern, strongly suggesting that balancing selection contributes to their long‐term maintenance (Wu et al., 2017). This hypothesis is further supported by the significant presence of positive Tajima's D and Fu and Li's D/F statistics in the coding sequence of GLR2.7. Remarkably, similar signatures of selection were previously observed at immune loci in Arabidopsis (Todesco et al., 2010; Huard‐Chauveau et al., 2013; Karasov et al., 2014) and in the Capsella genus (Koenig et al., 2019). Given that GLR2.7 was identified in a process related to direct defenses against insect eggs, it is conceivable that the presence of a strong haplotype could confer a selective advantage in natural Arabidopsis populations exposed to insect oviposition by promoting stronger HR‐like. The recent finding that GLR2.7 also functions in response to pathogens (Bjornson et al., 2021), together with the absence of any clear geographical separation of the haplotypes, suggests that microenvironment variations in herbivore and/or pathogen pressure may structure this gene. As both GLR2.7 and LecRK‐I.1 show signatures of balancing selection but haplotypes are not linked genetically, further research should aim at understanding the reasons for the maintenance of such genetic polymorphisms in natural populations.

The potential role of GLR2.7 in egg‐induced responses

The family of GLUTAMATE RECEPTOR‐LIKE genes was first described based on their homology to animal ionotropic glutamate receptors (iGluRs), yet phylogenetic analyses indicate that plant and animal GLRs diverged from a common ancestor. As opposed to iGluRs in animals, which are mainly involved in neurotransmission, plant GLRs play roles in various developmental processes and stress responses (Simon et al., 2023). In particular, GLR3 clade members have been previously reported to be involved in pathogen‐ and herbivore‐triggered immunity (Li et al., 2013; Manzoor et al., 2013; Mousavi et al., 2013; Xue et al., 2022), whereas GLR2 members are much less characterized. Plant GLRs are generally regarded as amino acid‐gated Ca²⁺‐permeable channels (Wudick et al., 2018), although in most cases, ligands are not known. In comparison with animal iGluRs, data gathered so far on plant GLRs indicate that they have a broader specificity and that, besides Glu, the ligand binding site can accommodate other amino acids (Alfieri et al., 2019).

Previous transcriptome studies revealed that genes related to Ca²⁺ transport and signaling were upregulated in response to oviposition (Little et al., 2007; Lortzing et al., 2019; Ojeda‐Martinez et al., 2021) but, to our knowledge, there is not yet evidence for Ca²⁺ accumulation. We found that P. brassicae oviposition and EE treatment trigger a localized and long‐lasting cytosolic Ca²⁺ accumulation and that this response was substantially reduced in the glr2.7 mutant. This strongly suggests a link between Ca²⁺ signaling and egg‐induced defense responses and supports the involvement of GLR2.7. Given that the Ca²⁺ influx was not fully abolished in glr2.7 or glr2.7/2.8/2.9 lines, other GLRs or other types of Ca²⁺ channels may contribute to the response and will deserve further investigation. The localization of GLR2.7 to the plasma membrane and a strong upregulation of GLR2.7 expression underneath the eggs provide additional evidence for a role of this channel in the modulation of egg‐induced responses.

Strikingly, we also showed that Glu that originates from P. brassicae eggs or egg‐associated secretions accumulates in the apoplastic space following oviposition or EE treatment. Also, exogenously applied Glu entered rapidly in the apoplastic space and triggered Ca²⁺ accumulation. One plausible explanation is that exogenous Glu could act as a GLR2.7 ligand to activate the channel, although we cannot formally discard the possibility that eggs induce the release of cytosolic Glu in the apoplastic space. This process has been shown to occur following PAMP or pathogen perception and leaf injury (Vatsa et al., 2011; O'Leary et al., 2016; Grenzi et al., 2023). In vitro and in vivo studies showed that L‐Cys, L‐Glu and Gly have strong affinity for GLR3.3 and induce Ca²⁺ responses in roots (Alfieri et al., 2019). Another study on GLR2.9 revealed that Gly is a probable ligand of the channel and that Gly‐mediated Ca²⁺ influx was inhibited by the animal iGluRs antagonist DNQX, suggesting a potentially conserved function in amino acid sensing (Dubos et al., 2003). However, we did not detect Cys nor Gly on filter paper after egg deposition, suggesting that they are unlikely to activate GLR2.7 in the context of oviposition. But, again, whether eggs induce the extracellular accumulation of different amino acids that bind GLR2.7 will clearly deserve further investigation. Also, testing the effect of other potential amino acid ligands on calcium channel activity of GLR2.7 would be interesting.

Intriguingly, out of 49 SNPs significantly associated with symptom score, four are located in the GLR2.7 promoter and 16 lead to amino acid changes to the protein sequence. The significant correlation between haplotype and GLR2.7 expression suggests that SNPs in the promoter may affect the binding of transcription factors or that SNPs in the gene sequence may modulate expression via modification of mRNA secondary structure, miRNA target sites or mRNA stability. This hypothesis will require a systematic analysis of the impact of single or multiple SNPs on GLR2.7 expression. Alternatively, changes in amino acids may alter GLR2.7 function. A structure of Arabidopsis GLR3.4 has identified a ligand‐binding domain for Glu as well as the transmembrane helices important for ion channel assembly (Green et al., 2021). Interestingly, the predicted amino acid substitutions K669L, K670T, E520G, A533T, and V544T are in close proximity to the Glu‐binding domain, whereas K672Q and K673R are within the ion channel. Moreover, the substitution D448V (SNP3) is located in the linker between the activation and ligand‐binding domains. Whether these residues are crucial for ligand‐binding and channel activity will need further studies of biochemically reconstituted GLR2.7 variants in heterologous systems (Alfieri et al., 2019; Green et al., 2021). These findings anyhow suggest that variation in egg‐induced responses may be regulated at different levels.

The discovery of Glu as a potential modulator of egg‐induced defenses, through activation of GLR2.7, adds to our recent finding that PCs in lepidopteran eggs trigger immune responses that partially require LecRK‐I.8 (Stahl et al., 2020). Although no genetic variation was found at the LecRK‐I.8 locus for egg‐induced HR‐like, such variation was identified for the close homolog LecRK‐I.1 (Groux et al., 2021). This suggests a complex regulatory process whereby the potential perception of egg‐derived PC by both cell‐surface receptors leads to a downstream signaling cascade whose strength may depend on LecRK‐I.1 haplotypes. In addition, the modulation of this response by GLR2.7 provides another layer of fine‐tuning, supported by the additive role of GLR2.7 and LecRK‐I.1 haplotypes. The exact mechanism by which GLR2.7‐dependent Ca²⁺ influx modulates PC/LecRKI.1/8‐dependent signaling is currently unknown. The association of GLR2.7 with egg‐induced SA accumulation and the fact that SA regulates downstream HR‐like responses and defense gene expression (Gouhier‐Darimont et al., 2013) suggest a role in the early steps following egg perception. Decoding of Ca²⁺ signals is mediated by Ca²⁺ sensors, including calmodulins, calcineurin B‐like proteins, and calcium‐dependent protein kinases. Research in PTI signaling has revealed a complex and flexible regulation of cellular responses to Ca²⁺, connecting decoders with various transcription factors, including CALMODULIN BINDING PROTEIN 60 g and SYSTEMIC ACQUIRED RESISTANCE DEFICIENT 1 that control SA biosynthesis (Zheng et al., 2015; Xu et al., 2022; Jiang & Ding, 2023).

In conclusion, we have identified GLR2.7 as an important component of the insect egg perception pathway in Arabidopsis, highlighting a novel role for a member of clade 2 GLRs. Future studies should focus on the activation mechanism of GLR2.7 and Ca²⁺ influx in response to oviposition. To place our results in a biological context, future studies should aim at testing the quantitative contribution of GLR2.7 to egg hatching/survival and whether it also impacts further larval performance. It will also be important to monitor oviposition‐induced calcium accumulation in accessions with different GLR2.7 haplotypes and explore the ecological significance of conserved haplotypes at the locus.

Competing interests

None declared.

Author contributions

MM, RG and PR conceived and designed the research. RG, MM, CG‐D, PM and CAMR conducted experiments. MM, RG and PR analyzed the data. PR, RG and MM wrote the manuscript with input from all authors. MM and RG contributed equally to this work.

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.

Supporting information

Fig. S1 CRISPR–Cas9 targeting of Arabidopsis GLR2.7.

Fig. S2 Manhattan plot of GWAS mapping for SA accumulation.

Fig. S3 Complementation of Ull2‐3 weak accession with GLR2.7.

Fig. S4 Geographic distribution of GLR2.7 polymorphisms across Europe.

Fig. S5 Contribution of GLR2.7 and LecRKI.1 haplotypes.

Fig. S6 GLR2.7 expression in accessions with weak or strong symptom scores.

Fig. S7 Role of GLR2.7 homologs.

Fig. S8 Calcium influx in glr2.7/2.8/2.9.

Fig. S9 Amino acid release from Pieris brassicae eggs or egg‐associated secretions.

Fig. S10 Glutamate accumulates in the apoplastic space and triggers Ca²⁺ influx.

Table S1 SNPs in the GLR2.7 locus that are significantly associated with symptom score.

Table S2 List of primers used in this study.

Table S3 Significantly associated substitutions in the Arabidopsis GLR2.7 protein sequence.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

All data that support our findings are available as Table S1.