Plant–herbivore interactions: Experimental demonstration of genetic variability in plant–plant signalling

Abstract Plant–herbivore interactions mediated by plant–plant signalling have been documented in different species but its within‐species variability has hardly been quantified. Here, we tested if herbivore foraging activity on plants was influenced by a prior contact with a damaged plant and if the effect of such plant–plant signalling was variable across 113 natural genotypes of Arabidopsis thaliana. We filmed the activity of the generalist herbivore Cornu aspersum during 1 h on two plants differing only in a prior contact with a damaged plant or not. We recorded each snails' first choice, and measured its first duration on a plant, the proportion of time spent on both plants and leaf consumption. Overall, plant–plant signalling modified the foraging activity of herbivores in A. thaliana. On average, snails spent more time and consumed more of plants that experienced a prior contact with a damaged plant. However, the effects of plant–plant signalling on snail behaviour was variable: depending on genotype identity, plant–plant signalling made undamaged plants more repellant or attractive to snails. Genome‐wide associations revealed that genes related to stress coping ability and jasmonate pathway were associated to this variation. Together, our findings highlight the adaptive significance of plant–plant signalling for plant–herbivore interactions.


| INTRODUC TI ON
Plants adjust their defences in response to cues from neighbours that are being attacked by herbivores (Agrawal, 2000;Bouwmeester et al., 2019;Heil & Adame-Álvarez, 2010;Heil & Karban, 2010;Karban et al., 2000Karban et al., , 2003Moreira & Abdala-Roberts, 2019;Moreira et al., 2018). Interactions between damaged plants and their neighbours are generally associated with increased levels of plant defence. This phenomenon appears widespread across biomes, growth forms and phylogeny (Karban, 2021;Karban et al., 2014). The effect of plant-plant signalling on plant-herbivore interactions is often considered adaptive, as it is expected to be naturally selected to reduce plant biomass loss, leading to increased individual fitness (Heil & Karban, 2010). However, the extent to which this process is genetically variable, which is a prerequisite for natural selection to act (Falconer et al., 1996), has hardly been examined so far.
Plant-plant signalling can have prompt effects on herbivore foraging activity. Contact with a damaged plant can prime defences, which allows faster response in case of future herbivore attack (Engelberth et al., 2004;Heil & Bueno, 2007;Zhang et al., 2020). atability of receiver plants when herbivores feed on them (Divekar et al., 2022;Himanen et al., 2010;Sugimoto et al., 2014). Both the nature and concentration of volatiles emitted after herbivore damage are highly variable within and among species (Chen et al., 2003;Degen et al., 2004;Dudareva et al., 2006;Kalske et al., 2019;Müller et al., 2020;Snoeren et al., 2010). Thus, plant-plant signalling can potentially evolve in few generations, depending on the herbivory pressure existing within populations (Kalske et al., 2019). Moreover, plant-plant signalling can be costly for plant fitness depending on the environmental conditions, and therefore be associated with adaptive trade-offs. For instance, Benevenuto et al. (2020) recently showed that simulated attacks of herbivore on bilberry plants reduced leaf consumption on conspecific neighbours at low and mid altitude, but had no effect at high altitude, indicating potential costs of induced defences in stressful environments.
The highly cosmopolitan plant Arabidopsis thaliana (L.) is a powerful model to investigate intraspecific variability in plant-plant signalling and plant-herbivore interactions. A. thaliana is a small annual plant inhabiting contrasting environments and climates all over the world (1001Genomes Consortium, 2016Hoffmann, 2002;Lee et al., 2017).
An international effort of sampling made available fully-sequenced genomes for hundreds of natural genotypes (hereafter 'accessions') (1001Genomes Consortium, 2016. This open-access database is a great opportunity to study the genetic basis of phenotypic variation. For instance, genomic variation between accessions of A. thaliana showed evidence of local adaptation to climate (Clauw et al., 2022;Dittberner et al., 2018;Exposito-Alonso, 2020;Lasky et al., 2014;Vasseur et al., 2018). Moreover, natural accessions of A. thaliana expressed contrasting values of constitutive defences against herbivores, notably in glucosinolate concentration in leaf tissues (Brachi et al., 2015;Gloss et al., 2017;Kerwin et al., 2015). Snoeren et al. (2010) showed that nine accessions of A. thaliana emitted different blends of volatile molecules in response to herbivore damage (see also Savchenko et al., 2013). Other studies showed that the Col-0 accession increased gene expression associated with resistance functions after being exposed to particular volatile molecules such as aldehyde or monoterpenes (Kishimoto et al., 2005;Savchenko et al., 2013). Surprisingly though, the effect of a prior contact with herbivore-induced signal on the foraging activity of an herbivore have been poorly investigated in this model species. The natural variation of the response of such herbivore-induced volatiles within A. thaliana is also an opportunity to understand its genetic bases through genome-wide association studies (GWAS). Such approaches are powerful tools to link complex phenotypes to genetic variation (Francisco et al., 2016).
Studies on plant-plant interactions in response to herbivory often used field-based approaches (Benevenuto et al., 2020;Fowler & Lawton, 1985;Karban et al., 2000Karban et al., , 2003Karban et al., , 2016Karban & Baxter, 2001;Morrell & Kessler, 2017;. Such an approach was associated with a dependence between the environment of the emitter and of the receiver. In other words, the signals emitted by the damaged plants may directly repel herbivores, providing indirect defences to neighbours (as discussed in Karban et al., 2014). Another aspect of field-based approach is that leaf consumption on receiver plants of herbivore-induced signals was mainly measured over long period of time (days, weeks or years) (Agrawal, 2000;Benevenuto et al., 2020;Fowler & Lawton, 1985;Karban et al., 2000Karban et al., , 2003Karban et al., , 2016Morrell & Kessler, 2017;Rhoades, 1983). How short-term foraging activity of herbivores on plants is impacted when plants had a prior contact with a damaged neighbour may be crucial to understand plant-herbivore interaction mediated by plant-plant signalling (Morrell & Kessler, 2017). Here, we built an innovative experimental design (ii) if plant-herbivore interaction mediated by a prior contact varies across natural accessions of A. thaliana? (iii) whether this potential variation is explained by allelic diversity between plant accessions? 2 | MATERIAL S AND ME THODS

| Plant material
We randomly selected 113 natural accessions of A. thaliana from the 1001 Genome germplasm and thus entirely sequenced (1001Genomes Consortium, 2016. Seeds were obtained from NASC and were then reproduced in greenhouse (CEFE, Montpellier, France).
We used garden snail (C. aspersum), a generalist herbivore, whose geographical range overlap the majority of the European distribution of A. thaliana. Snails used in this study were raised in an organic farm (L'escargot du Riberal, L. Sanchez, Tautavel, France). Snails were fed with industrial and specialized flour at the farm and were starved during 1 week in a cold room before the experiment.

| Growth conditions
Seeds were sown at soil surface of 5-cm pots (80 mL) previously filled with peat substrate (Neuhaus N2) and placed in greenhouse (Montpellier, France) in October 2020. We used eight consecutive sowings consisting of four seeds of the 113 accessions (n = 3616 pots in total), randomly arranged in eight blocks, one block corresponding to one sowing date. All blocks were rotated on a daily basis.
All pots were sub-irrigated with water treated by reverse osmosis to field-capacity every 2-4 days. The temperature was set at 20°C during the day and 16°C during the night for the full duration of the experiment. Sciarid flies were observed 5 weeks after germination.
We spread water-diluted larvicide on the surface of all the pots (Vectobac WG, Edialux, France). Due to germination issues, 600 pots were removed. The 3016 remaining pots were undamaged by sciarid flies and represented between 20-32 individuals per accession.

| Measurements
A binomial variable 'First choice' was scored 1 to every arena where the R plant was chosen first by the snail, and 0 elsewhere, i.e. to every arena where N1 plant was chosen first (Table 1). First duration was measured as the time spent by the snail on the first plant (Table 1). We measured the total duration that the snail spent on each individual plant during 1 h. The proportion of time was then calculated as the duration on one plant divided by the sum of the durations on the two plants in the arena (Table 1). We estimated a 'leaf consumption index' from the pictures taken after the passage of the snail (Table 1) The initial rosette area of each plant was estimated from the picture taken before the passage of the snail (ImageJ; version 1.53c, Schneider et al., 2012). We then calculated the 'rosette area difference', as the difference between initial rosette areas of the R plant and the N1 plant in each arena.
TA B L E 1 List of response variables and their observed range of variation.

| Statistical analyses
All analyses were performed with R (version 4.0.5). We tested if a prior contact to a clipped plant and rosette area difference had an effect on each response variables separately. We used quasibinomial models to test these effects on the first choice, the proportion of time spent on R, and the proportion of leaf consumption index (link function 'probit'). We used quasibinomial models instead of binomial model to correct for overdispersion. Quasibinomial models artificially increase standard errors, and are more conservative than binomial models (Dunn & Smyth, 2018). Effect on first duration was tested using a linear model, after log 10 -transformation to fit normality assumptions. Intercepts and slopes estimated by models were used to quantify the effects of prior contact and total leaf area difference between plants, respectively. For example, for the vari-  (Lenth et al., 2018). Correlations between mean responses of accessions were assessed with the Kendal's statistic.

| Quantitative genetic analysis
We tested for allelic effects on the four response variables through genome-wide association studies using the online platform EasyGWAS (https://easyg was.ethz.ch/, Grimm et al., 2017). Marginal means of accessions, as calculated above, were implemented. We used the EMMAX algorithm, filtering for minimum alleles frequency at 5% and including the two first axes of a PCA performed on SNPs to correct for population structure. Significance thresholds were as-

| Genotypic and allelic effects of attractivity of plants exposed to a damaged neighbour
The probability to choose R plants first varied a lot between plant accessions, ranging from 12.5% (1/8) to 100% (8/8) across accessions. Two SNPs were associated with the first choice of snails on R plants (Figure 3). No gene was found at the first SNP on the second chromosome, but this SNP was upstream to AT2G19800, a gene that encodes a myo-inositol oxygenase (MIOX). MIOX gene family is involved in oxidative stress tolerance (Munir et al., 2020). On the F I G U R E 2 Prior contact and rosette area difference effects on snail foraging activity. Estimates and standard errors of the coefficients of models are represented here on a probit scale. *p < 0.05, **p < 0.01, ***p < 0.001.
fourth chromosome, a significant SNP was located in AT4G27080, a gene that encodes a disulphide isomerase-like (PDIL) protein (Table S1) However, no SNP position along the genome was significantly associated with the first duration spent on a plant (Figure 3, Table S2).
We  (Table S4), a protein involved in stress tolerance (Luciński et al., 2011). The effects and the frequencies of minor alleles for every significant SNP are detailed in Table S5.

| DISCUSS ION
For the last decades, consistent progress has been made in the characterization of plant-plant signalling and its effect on herbivore foraging activity (Agrawal, 2000;Dicke & Baldwin, 2010;Heil & Karban, 2010;Karban, 2021). Within-species variation has been however overlooked, despite its central place in local adaptation.
Here, we showed that a prior contact with a damaged plant had, on average, a negative effect on A. thaliana's defence against a generalist herbivore, as snails spent significantly more time and ate more on receiver plants. We also found that this response is highly variable between accessions, and that allelic diversity underlies this variation.

F I G U R E 3
Genome-wide association of the four response variables. GWAS were performed with EMMAX across 113 natural accessions on 1,922,479 SNPs. Each dot is a SNP located on chromosomes 1, 3 and 5 (grey) and chromosomes 2 and 4 (black). Bonferroni threshold at 0.05 and 0.1 was represented with red solid and dashed lines, respectively.
Chr. 1 C hr. A contact with a damaged plant led to higher attractivity of A. thaliana for snails, in terms of time spent on plants and of leaf consumption. Our results confirmed that plant-herbivore interactions can be partly regulated by plant-plant signalling (Dicke & Baldwin, 2010;Karban et al., 2014). However, our results contradict the general idea that plant-plant signalling has been selected to promote deterrence to herbivore attack (see Karban et al., 2014;Karban, 2021; for reviews). Snails may be attracted to plants previously exposed to damaged neighbours for many reasons as developed in the following paragraphs.
First, odour and taste of the emitter damaged plant can have been deposited on the neighbouring plants (Matsui, 2016). Snails possess the ability to sense odour and to discriminate them in a saturated context (Shannon et al., 2016). In the arena, the plant previously exposed to a damaged neighbour, i.e., receiver plant, can thus exhibit a different olfactory signal compared with the plant exposed to an undamaged neighbour. Indeed, snails significantly preferred damaged plants compared with undamaged plants (Appendix S1).
Attractiveness of damaged plants has been shown in numerous species of insect herbivores (Bolter et al., 1997;Carroll et al., 2006;Harari et al., 1994;Loughrin et al., 1995). Life-style of herbivores, such as aggregating behaviour are aspects that are likely to be part of the attractiveness of damaged plants and of their neighbours (Bolter et al., 1997). The search for mating opportunities may explain such attraction for damaged plants by conspecifics (Arab et al., 2007;Kalberer et al., 2001). In case of uncertainty of feeding, a choice on a sub-optimal but detectable plant may also be the better choice for a slow-foraging herbivore (Carroll et al., 2006).

Second, interactions between two individuals of A. thaliana
can also be maladaptive or addressed to species other than snails.
To our knowledge, only one case of increased attraction to herbivores in response to plant-plant signalling has been reported (Zhang et al., 2019). In this example, tomato fruits infested by the whitefly Bemisia tabacito expressed a defence against pathogens instead of herbivores. Neighbours of such manipulated plants also expressed unappropriated defences and thus were more attractive to whiteflies (Zhang et al., 2019). In addition, recent studies on Baccharis salicifolia showed that the volatiles emitted consecutively to herbivore damages, and their effects on neighbours was highly specialized to herbivore species (Moreira et al., 2018; and see Moreira & Abdala-Roberts, 2019 for a review). As we simulated herbivore attack by clipping the leaves of A. thaliana, induced defences may not be specifically addressed to snails. Numerous studies showed that herbivoreinduced volatiles attract parasitoid insects in plant species, including A. thaliana that represent indirect defences to herbivores (Baldwin et al., 2002;Dicke & Baldwin, 2010;Girling et al., 2006;Kessler & Baldwin, 2001;Turlings et al., 1990;Turlings & Erb, 2018;Van Poecke et al., 2001). Such indirect defence through parasitoid attraction may be involved in our study but would require additional manipulations.
Third, the biased foraging activity of snails on plants that experienced a previous contact with a damaged plant was accentuated for accessions that originated from high latitudes and from environment with low mean annual temperatures ( Figure S3). Long-standing and highly-debated hypotheses on latitudinal gradient in herbivory pressure may be associated to such latitudinally structured genotypic variability (Anstett et al., 2016). Our results echoed the study of Benevenuto et al. (2020) that showed that plant-plant signalling effect on plant defence induction against herbivores decreased with altitude in wild populations of bilberries (Vaccinium myrtillus L.). Costs involved in the induction of defences against herbivores by plantplant signalling may overtake benefits for individual from harsh, stressful environments (Strauss et al., 2002). Such results highlight the necessity of studies of intraspecific diversity along environmental gradients of plant-plant signalling and plant-herbivore interactions to better understand their evolution.
A key aspect of modern biology is to understand how complex phenotypes are genetically controlled (Francisco et al., 2016).
Genome-wide association studies are powerful and fast tools for gene hunting (Brachi et al., 2015;Francisco et al., 2016;Gloss et al., 2017). The completely-sequenced genomes of thousands of natural accessions of A. thaliana is then the best material for such genotype-to-phenotype studies (1001Genomes Consortium, 2016. Here, we found that the effect of previous plant exposure to a damaged neighbour on snail behaviour relied on several SNPs along the genome. Most of them were associated with regions with unknown functions, thus opening the way to the exploration of new pathways involved in plant defences (Tables S1-S4). Other significant SNPs were related to plant ability to cope with stress. For instance, AT2G43110 is involved in the regulation of drought tolerance (Noman et al., 2019). AT3G03380 increases stress protection of the photosystem II in A. thaliana (Luciński et al., 2011). Interestingly, we found that variation in leaf consumption was associated with the AT3G10050 gene, which encodes the first enzyme in the synthetic pathway of isoleucine, an activator of jasmonic acid (Staswick & Tiryaki, 2004) that is a well-known molecule involved in plant-plant and plant-herbivore interactions (Ray et al., 2019). Although these genes remain to be validated, for instance using mutant phenotype expressions, they are promising candidates to better understand the mechanisms involved in plant defences against herbivore mediated by plant-plant interactions. If validated, such genetic variation underlying plantplant interactions for induced defence against herbivore may rise opportunities for plant breeding in an agricultural context. For example, the push-pull strategy used in hundreds of sub-Saharan fields, consisting in mixing attractive and repulsive species within a field, may enhanced both 'pull' and 'push' compartments with artificial selection on alleles of interest (Bruce, 2010;Guerrieri, 2016;Pickett & Khan, 2016;Turlings & Erb, 2018). A breeding for such plant-plant interactions within species may also enhance yield stability in varietal mixtures (Pélissier et al., 2021).

| CON CLUS ION
How plant-plant interactions could evolve in response to herbivory is highly debated (Dicke & Baldwin, 2010;Heil, 2014;Heil & Adame-Álvarez, 2010;Heil & Karban, 2010;Karban, 2021;Karban et al., 2014). Discussions were substantial on the advantage for an attacked plant to produce a signal that warns and helps potential competitors (Agrawal, 2000;Heil & Adame-Álvarez, 2010;Heil & Karban, 2010). At the opposite, the advantages for a plant receptor of such signal were often considered as adaptive, in the case of increased resistance to herbivory after being in contact with a damaged plant. Theoretical expectations on the evolution of an increased attractivity to herbivores when receiving cues from attacked plants are missing in the literature (as discussed in . Promoting continuous movement of herbivores between attractive plants may limit biomass consumption at the group level (Morrell & Kessler, 2017

ACK N OWLED G M ENTS
We are very grateful to Ana Elkhaïm and Maëva Tremblay for their help on measurements. We also thank the technical plat- grant ERC-StG-2020-949843).

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in Data.INRAE at https://doi.org/10.57745/ QN7MMM.