As they develop, many plants deploy shifts in anti-herbivore defense allocation due to changing costs and benefits of their defensive traits. Plant defenses are known to be primed or directly induced by herbivore damage within generations, and across generations by long-lasting epigenetic mechanisms. However, little is known about the ontogenetic trajectories of epigenetically inducible defensive traits across generations and their consequences. To help fill this knowledge gap, we conducted a multigenerational experiment to determine whether defense induction in wild radish plants was reflected in chromatin modifications (DNA methylation); we then examined ontogenetic trajectories (seedlings to reproductive plants) of current and transgenerational plasticity in anti-herbivore chemical (glucosinolates) and physical (trichomes) defenses in this species. Herbivory triggered genome methylation both in targeted plants and their offspring. Within one generation, both defenses were highly inducible at the seedling stage but only marginally or non-inducible in reproductive plants. Across generations, herbivory experienced by mother plants caused strong direct induction of physical defenses in their progeny, with effects lasting from seedling to reproductive stages. For chemical defenses, however, this transgenerational induction was evident only in adults. Transgenerational priming was observed in physical defenses both for seedlings and adult plants. Our results show that transgenerational induction and priming in response to herbivore offense differ for physical and chemical defense and change across plant life stages.

Methods

System natural history, experimental layout, plant culturing and herbivore induction

Wild radish, Raphanus sativus L. (Brassicaceae) is a self-incompatible herbaceous annual weed common in Mediterranean climates. Caterpillars of Pieris rapae (Lepidoptera) are specialist herbivores to the Brassicaceae family. Glucosinolates and foliar trichomes have been described as effective resistance traits for R. sativus (1).

We produced an F1 generation of half-sib plants from a non-induced maternal family (P). This F1 generation, termed ‘mother plants’, comprised of plants that were either never exposed to herbivory (naïve mothers), or subjected to herbivory by P. rapae for two weeks when plants were two weeks old (induced mothers). We then grew their progeny (F2), keeping plants naïve or subjecting them to herbivory at two stages of their lifecycle: seedling stage (two-leaf status, about one week after germination) and reproductive stage (after blooming). Thus, we completed a fully factorial design of herbivore induction in mother plants and their progeny to study current and transgenerational phenotypic changes elicited by herbivory (physical and chemical defenses) at two life stages of the progeny (seedlings and adults). The study comprised of 160 plants (76 seedlings and 84 adult plants). Growing conditions were those described in Neylan et al. (11). Distribution of the 160 plants was as follows; 24 adult plants with maternal and current herbivory, 26 adult plants with maternal but no current herbivory, 18 adult plants with no maternal herbivory and with current herbivory, 16 adult plants without maternal or current herbivory. 21 seedlings with maternal and current herbivory, 22 seedlings with maternal but no current herbivory, 14 seedlings with no maternal herbivory and with current herbivory, 19 seedlings without maternal or current herbivory. Chemical defenses were analyzed in 155 plants and physical defenses were analyzed in 157 plants due to missing values and extreme outliers.

The herbivory treatment consisted in placing two caterpillars per plant, for the seedling group when plants had two leaves, and at the adult group when plants were flowering. The caterpillars were at the second instar, and they were left to feed freely for two weeks, until pupation. Plants lived around two months so this means that the herbivory treatment lasted around a quarter of the plants’ lives.

The attack by caterpillars on the plants’ offspring was strong, in many cases almost all leaf tissue was consumed. But the plants were capable of consistently growing new tissue, and consequently no plants were killed by caterpillars. Only fresh, recently expanded leaves were sampled both for seedlings and adult plants. Any possible induction caused by sampling was controlled by sampling all plants across herbivory and non-herbivory treatments before and after treatment. We assume that any possible induction caused by sampling would affect each treatment group similarly, and therefore would not affect comparisons between the groups.

We harvested the plants two weeks after exposure to caterpillars and analyzed the number of trichomes and the glucosinolate concentration, the main physical and chemical defenses in this species, respectively. To quantify trichome density, we sampled two 2.7-cm diameter discs taken with a cork borer from the two most recently expanded leaves of each plant (N = 4 measures per plant). The number of trichomes in digital pictures of the discs was counted with ImageJ analysis software (22).  For chemical defenses, leaf discs were sampled as above and flash frozen with liquid nitrogen, freeze dried, and stored at -80ºC. Glucosinolates were analyzed by chromatography, a protocol indicated in (9), and concentration was expressed as mg/g of dry mass.

Inducibility, transgenerational induction and transgenerational priming

We also studied i) the ontogenetic trajectories of the expression of induced defenses after naïve plants experienced their first herbivory challenge (i.e., inducibility), ii) the ontogenetic trajectories of the maternal effects elicited by herbivory that directly increased constitutive allocation to defenses in their progeny (i.e., transgenerational induction), and iii) the ontogenetic trajectories of the maternal effects elicited by herbivory that primed the defense system of the progeny to produce an increased response when damaged by the same herbivores (i.e., transgenerational priming).

We performed two linear models for physical and chemical defenses. Statistical analyses were performed in RStudio for R version 4.0.2. Models were fitted using the function glm from the lme4 package. Statistical significance of fixed effects was determined by analysis of deviance, Type II. Models included the fixed effects of age (seedlings and adults), maternal herbivory (mother plants that experienced herbivory and naïve mothers), and progeny herbivory (progeny exposed to herbivory and control). Models additionally included all the interactions between effects. Variables were log transformed to achieve normality in the distribution of the residuals.

To test the hypothesis that inducibility of chemical and physical defenses in response to an initial herbivory challenge changes throughout ontogeny, we analyzed the interaction between herbivory and age. A significant herbivory × age interaction would show that the effect of herbivory-induced defense allocation changed with age. To test the hypothesis that transgenerational induction of defenses is contingent on the ontogenetic stage, we examined maternal herbivory × age. For the third hypothesis, that transgenerational priming of defenses changes during ontogeny, we examined maternal herbivory x herbivory x age interaction.  

Methylation analyses

To verify that phenotypic changes were related to chromatin changes, DNA methylation events experienced by individual genotypes were assessed in the progeny at the seedling stage (N = 94). A simplified MSAP method was performed using only primer combinations with the methylation-sensitive HpaII. We collected leaf material from the F2 generation plants before and after exposure to herbivory. We cut one 2.7-cm diameter disc with a cork borer from two leaves as before. Leaf material was kept in dry silica gel until DNA extraction. Epigenetic characterization of individual plants before and after treatment was performed by methylation-sensitive amplification fragment length polymorphism. In this technique, genomic DNA is digested by methylation-sensitive enzymes, providing an epigenetic fingerprint of the plants. Thus, we used an epigenetic fingerprint of each plant before and after induction.

By studying the epigenetic fingerprint before and after induction, we can examine the occurrence of methylation events across the genome during treatment. We compared the chromatin changes occurring in the plants exposed to herbivory and in the naïve plants not attacked by herbivores. HpaII cleaves CCGG sequences when cytosines are not methylated. Cleaving may be impaired when at least one cytosine is hemi-methylated and is inactive if one or both of the cytosines are fully methylated. Thus, within the same genotype, polymorphism of MSAP markers reflects variation in the methylation status of the CCGG sites. A change from presence to absence implies a methylation event in a locus, and a change from absence to presence indicates a de-methylation event.

DNA was isolated using the hexadecyl-trimethyl-ammonium-bromide procedure followed by MSAP fingerprinting. We analyzed 188 samples from 94 plants in the F2 generation using two primer combinations. The selected combinations X-AC/M-AC and X-AC/M-ATC were used for fragment amplification.  MSAP analyses were carried out by Keygene Laboratories (Netherlands).

To test whether herbivory was linked to the probability of occurrence of methylation per locus, we proceeded as follows. Because methylation and de-methylation cannot occur at the same time, we considered non-methylation events only when de-methylation did not occur on the marker. MSAP marker scores for samples were transformed by comparison with the corresponding values (i.e., same marker and plant individual before treatment). MSAP marker scores involving a change from 1 to 0 denoted a methylation event of the marker involved. Only fragments >300 base pairs in size were included, to reduce the potential impact of size homoplasy. A new sample (N = 94 plant individuals) by marker (N = 402) score matrix was obtained, where each element showed whether the sample involved had experienced a methylation event (score = 1) or not (score = 0) in the corresponding marker (or missing when a de-methylation event occurred and therefore a methylation event was not possible).

To test whether herbivory and maternal herbivory were related to chromatin changes, a generalized linear mixed model was fitted to the data. Probability of methylation per marker was analyzed using a generalized linear mixed model in which methylation event was fitted to a binomial distribution link logit, marker and individual were added as random factors, and current and maternal herbivory were added as fixed factors. This analysis was performed for the 94 individuals in F2 and the 402 markers selected. The response variable took a value of 0 when a methylation event did not occur during treatment and a value of 1 when a methylation event occurred during treatment. In cases where loci were already methylated before treatment and therefore methylation during treatment was not possible, we treated them as missing values. The model including the interaction between current and maternal treatments detected no significant interaction and performed worse in terms of AICc.

See README.txt attached.