Plant–plant communication and community of herbivores on tall goldenrod

Abstract The volatiles from damaged plants induce defense in neighboring plants. The phenomenon is called plant–plant communication, plant talk, or plant eavesdropping. Plant–plant communication has been reported to be stronger between kin plants than genetically far plants in sagebrush. Why do plants distinguish volatiles from kin or genetically far plants? We hypothesize that plants respond only to important conditions; the induced defense is not free of cost for the plant. To clarify the hypothesis, we conducted experiments and investigations using goldenrod of four different genotypes. The arthropod community on tall goldenrods were different among four genotypes. The response to volatiles was stronger from genetically close plants to the emitter than from genetically distant plants from the emitter. The volatiles from each genotype of goldenrods were different; and they were categorized accordingly. Moreover, the arthropod community on each genotype of goldenrods were different. Synthesis: Our results support the hypothesis: Goldenrods respond to volatiles from genetically close plants because they would have similar arthropod species. These results are important clues elucidating adaptive significance of plant–plant communication.

. Such plant-plant communication has been reported in more than 30 plant species so far (Heil & Karban, 2010).
Recent studies suggest that communication among plants can be specific: Sagebrush (Artemisia tridentate) distinguishes volatiles from self-and non-self-genotype. The plants which received volatiles from self-genotype got less damage than the plants which received volatiles from non-self-genotype (Karban & Shiojiri, 2009). Moreover, when they received volatiles from genetically closer individuals, they became more resistant than when they received volatiles from genetically distant individuals (Karban et al., 2013). It has been reported that the similarity in the blend of volatiles is related to genetic similarity (Ishizaki et al., 2012). Goldenrod (Solidago altissima L) also responds to self-genotype stronger than non-self-genotype by volatiles under the low herbivore population (Kalske et al., 2019). Thus, plants may be able to perceive and respond to volatiles that are similar of their own.
Why should such specificity of plant signaling and communication evolve? Induced plant response is thought as one of the plant's strategies to save defense cost (Agrawal et al., 1999). Plant communication, the response to volatiles of damaged neighboring plants to become resistant to herbivore, is one of the induced plant responses.
The merit of plant communication is to be able to induce defense before plants get damage. Kalske et al. (2019) demonstrated that goldenrods that experienced high pressure by herbivory induced resistance in all neighboring conspecifics by volatiles, whereas those experiencing herbivore exclusion induced resistance only in neighbors of the same genotype. Plants would adapt to respond to necessary information. Previous studies indicate that genetically related individuals are similar in leaf chemistry and thus share similar herbivore communities (Kagiya et al., 2018). VOC signals from close relatives could provide accurate information about future herbivory on the receiver plant, whereas VOCs from distantly related individuals may provide misleading information. If there is correlation between similarity of plant genetic and similarity of herbivore community, the receiver plant, tuning into VOC signals from close relatives, is predicted to be more beneficial than that from unrelated individuals.
To test this hypothesis, we conducted these surveys and analysis using tall goldenrods (Solidago altissima) as the first step. (1) Do tall goldenrods become more resistant when the plant receives from volatiles from closer genetic plant than from genetically far plant? (2) Are the volatiles different among genotypes, if tall goldenrods can distinguish among genotypes? (3) Are the arthropod community different among genotypes? Are there any relationships between plant genetic similarity and the herbivore community? Finally, we will discuss the beneficial of plant communication in consideration of these results.

| Study system
Tall goldenrod, Solidago altissima L. (Asteraceae), which was introduced to Japan from North America around 1900, is a dominant and well-studied perennial herb found throughout Japan. Tall goldenrod is host to diverse arthropod communities (Ando et al., 2011). It is rhizomatous and its clones exhibit considerable interclonal genetic variation in many plant traits (Abrahamson & Weis, 1997;Crutsinger et al., 2006Crutsinger et al., , 2008Maddox & Root, 1987).
In early May 2008, rhizomes of tall goldenrod were collected four genotypes from four sites (one genotype per site) 4.5-17.5 km apart in Shiga Prefecture (Table 1). Rhizomes directly attached to one another were considered as the same genotype. We propagated clones of each genotype from rhizome cuttings into 7 cm in a greenhouse in order to prevent herbivores from feeding. Watering as needed, four genotypes were kept in a large cage until our experiments in 2008, 2011, and 2012.

| Herbivore community census
In early May 2008, 10 rhizome cuttings from each of four genotypes (total of 40 ramets) were individually planted in pots (ca.18 cm, hight 20 cm) and were grown in the large cage until late May. All potted plants were then randomly transplanted into an experimental plot in a 6 m × 16 m grid in the common garden.
The field survey was conducted in our study site at the Center for Ecological Research, Kyoto University, in Otsu, Shiga Prefecture, Japan. To examine how herbivorous insects respond to different genotypes, we conducted herbivore community censuses three times in June 2008. Abundance of each herbivorous insect species was recorded. The density of each arthropod species per plant was calculated by averaging the three census data.

| Genetic dissimilarity of tall goldenrods
To assess genetic dissimilarity among four genotypes, we extracted DNA from green leaf tissue of each genotype using the CTAB method (Milligan, 1992). Following protocols of supporting online material in Crutsinger et al. (2006), we assessed genetic variation among four genotypes by using the AFLP (amplified fragment length polymorphisms) technique (Vos et al., 1995). AFLP markers were generated by using four selective primer pairs: EcoRI-AGT and MseI-CTA,  (Nei, 1972(Nei, , 1978, using POPGENE 1.31 (Yeh et al., 1999).

| Volatiles collection and analysis
VOCs from artificially damaged tall goldenrods were collected. We planted five tall goldenrods of each genotype in a laboratory room (16L8D, 24 ± 1°C) for around 1 month. We damaged three leaves of each plant with scissors. VOCs from one damaged plant were collected in a glass container (2 L) using Tenax 60/80 (Gerstel GmbH and quantified by comparing their mass spectra and retention times with those of authentic compounds (see above). Quantification of each compound was carried out on the basis of their GC peak areas and expressed as percentages in the total ion chromatogram.

| Statistical analyses
2.5.1 | Variation in herbivore community among genotypes To examine whether herbivore community differ between the treatments, we used nonmetric multidimensional scaling analysis (NMDS) with the Bray-Curtis dissimilarity coefficients. Points that are close together represent samples that are very similar in community composition, based on the number of species and relative abundance of each species. Then, difference in community compositions of herbivores among plant genotypes was determined using permutational multivariate analysis of variance (PERMANOVA, Anderson, 2001).
Significance was assessed with 999 permutations and the Bray-Curtis dissimilarity. We conducted NMDS and PERMANOVA analysis in MASS and vegan packages of the software R Studio ver.

| Plant volatile compounds relevant to clonal identification
To identify the volatiles compounds that are related to clonal identification, we conducted discriminant analysis (DA) to detect the differences in composition ratio of each compound among genotypes.
However, our volatile profile data included variables whose number (40 compounds) are more than the number of observations (20 individuals) and some of volatile compounds were highly correlated.
These situations did not fulfill the condition of DA. Therefore, before conducting DA we transformed the data using principal component analysis (PCA). This procedure allowed us to perform DA with the variables that are uncorrelated and that their number is less than analyzed individuals (Jombart et al., 2010

| Herbivore community
We recorded five herbivorous insect species in four orders on tall goldenrods in June (Table S1). The herbivore community consisted of one Coleoptera (Erateridae sp.), one Diptera (Agromyzidae sp.), two Hemiptera (Uroleucon nigrotuberculatum, Corythucha marmorata), and one Lepidoptera (Ascotis selenaria). The main leaf chewers were a geometrid moth caterpillar, A. selenaria NMDS analysis of the dissimilarity of herbivore community composition revealed that herbivore community was clearly distinct among four genotypes ( Figure 1; PERMANOVA: R 2 = .19 p = .01). NMDS showed that herbivore communities between genotype A and genotype B were the most similar pairs of the four genotypes.

| Plant resistance after receiving volatiles in the field
Tall goldenrod plants that received volatiles from the same genotype experienced less damage than other plants. In 2011 when the emitter was genotype A, leaf damage was the lowest on genotype A receiver. The greatest damage was found in genotype D with 45% of leaves damaged by herbivores; more than 2 times as high damage as genotype A (Figure 2). In control (2012) The average of damage was 0.10 + 0.02.

| Genetic dissimilarity of tall goldenrods
In 59 loci for four genotypes, the number of polymorphic loci was 39.
Mean genetic distance between genotypes was 0.40 (range: from 0.29 to 0.49). The most similar genetic distance was between genotype A and genotype B (Table 2).

| Relationship between plant genetic dissimilarity and the herbivore community
The Mantel test between all the plants and pairwise Mantel test between genotypes on community dissimilarity × Nei's genetic distance matrix indicated that genetically related genotype pairs have

F I G U R E 1 Nonmetric multidimensional scaling (NMDS)
ordination of herbivore insect communities on four genotypes of tall goldenrods. The herbivore communities were significantly different among genotypes. Each symbol indicates the mean (±SE) of the herbivore community on each genotype F I G U R E 2 Ratio of damaged leaves of goldenrods in each genotypes. Genotype A was as an emitter similar herbivore community (between all the plants: r = .11, p = .01; between genotypes: r = .88, p = .001) (Figure 3).

| Volatiles from four genotypes
The volatiles from tall goldenrods were comprised of 40 compounds including four unidentified ones (Table 3). Because the amount of volatiles were similar, we compared the ratio of compounds among four genotypes.
Volatile profiles from different genotypes were profoundly different, while that of the same genotype were similar. Twenty-five compounds of 40 volatile compounds were emitted from all genotypes, while the others were not found in one or more genotypes (Table 3). Discriminant analysis revealed that first and second discriminant functions explained 79.1% and 15.6% of variance, respectively (Table 4), and showed clear discriminations of genotypes ( Figure 4), with the error rate 0.15. First discriminant function was mainly contributed by PC2 and PC4, and second discriminant function was contributed by PC3 and PC7 (Table 4). First discriminant function which was positively contributed by γ-Gurjunene, unknown 2, and Isoledon discriminated genotype B that emitted those 3 compounds highly (Tables 3 and 5, Figure 4). Second discriminant function discriminated genotype D that emitted less 2β-Pinene and Bicyclo2.2.1heptan-2-ol, and more γ-Terpinene than other genotypes (Tables 3 and 5, Figure 4).

| D ISCUSS I ON
In the field experiments, we showed that tall goldenrod which received volatiles from same genotype got the least damage than the other genotypes. Some plants such as sagebrush (Karban & Shiojiri, 2009), Ambrosia dumosa (Mahall & Callaway, 1996), and Cayratia japonica (Fukano & Yamawo, 2015) can distinguish between self and non-self by volatiles. Our result partially supports previous tall goldenrod study in recognizing the same genotype (Kalske et al., 2019). Kalske et al. (2019) showed that plants induced resistance in the same genotype under lower herbivore pressure and in all genotypes under higher herbivore pressure. On the other hand, the goldenrod which received volatiles from closer genotype got less damages in our experiments. The results suggest that the goldenrod could recognize the volatiles of genetically closer plants. As for whether the induction of plant resistance is limited to closer relatives, it may depend on the history of the degree of herbivore pressure. Herbivore pressure in our field is likely to be lower than in the original habitats with many natural enemies (e.g., the enemyfree hypothesis) (Fukano & Yahara, 2012), so all genotypes did not need to respond to volatiles of damaged leaves in the same way. Kin recognition through volatiles also has been reported in Sagebrush (Karban et al., 2013).
To distinguish volatiles from kin from nonkin in goldenrod, the volatiles should be different among genotype. Actually, volatiles of goldenrod were different between genotypes (Figure 4). Our result of discriminant analysis indicated that a few volatile compounds are associated with clone identification. In our study, γ-Gurjunene, unknown 2, Isoledon, 2β-Pinene, Bicyclo2.2.1heptan-2-ol, and γ-Terpinene were suggested to contribute to genotype identification (Table 4). More study is needed to clarify whether these compounds actually cause clonal distinction. In our analysis of volatile profile, genotype A and genotype B, which are genetically close, showed considerably different volatile profiles. Therefore, we could not find the correlation between similarities of volatiles and genetics. and reproductive phenology affect arthropod community (Johnson & Agrawal, 2005). In tall goldenrods, the herbivorous communities were significantly different among genotype and the community dissimilarity was correlated with genetic distance (Figures 1 and 3). The volatiles must be useful information to the neighbor plant.

TA B L E 2 Genetic dissimilarity of tall goldenrods
They could predict the level of danger from volatiles' information.

TA B L E 3
Means ± SDs of composition ratio (% to total GC peak areas) of each volatile compound detected from each genotype of Solidago altissima supported the hypothesis: goldenrods respond to volatiles from close-genotype plants because they would have similar arthropod species. These results are important clues elucidating adaptive significance of plant-plant communication.

A K N OWLED G M ENTS
We would like to thank Dr. Akane Uesugi (Monash University) for advices on data interpretation. We thank Ms. Satomi Yoshinami and Mr. Akira Matsumoto (Kyoto Univ.) for field assays. This study was supported in part by JSPS Grants-in-Aid (23770020) to K.S. and in party by The Naito Foundation. We thank Dr. Elizabeth for editing a draft of this manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data from this manuscript were archived in the publicly accessible repository Dryad (https://doi.org/10.5061/dryad.80gb5 mkpv).