Phenotypic plasticity allows organisms to modify their biochemical, physiological, or morphological traits to survive in changing environments (Schlichting, 1986). While phenotypic plasticity has mostly been studied within the lifespan of organisms, there is increasing evidence that life history experiences of individuals can influence traits in their progeny. These include simple direct maternal effects such as nutrient provisioning to progeny. However, some transgenerational effects can persist over multiple generations and involve heritable epigenetic changes (Lämke and Bäurle, 2017; Bošković and Rando, 2018). These epigenetic responses have the potential to provide adaptive benefits to progeny, thereby enhancing evolutionary fitness of the parents. When facing environmental changes, organisms can adopt various transgenerational strategies to optimize their fitness. When environments change frequently and are unpredictable, parents may adopt a bet-hedging strategy to increase the variability within their progeny (Crean and Marshall, 2009). By contrast, when environments undergo directional and stable changes, which present a more predictable cue about future environmental conditions, parents could enhance reproductive fitness by transmitting specific adaptive traits to their progeny (Marshall and Uller, 2007; Proulx and Teotonio, 2017).

Transgenerational responses to stress have been reported in both plants and animals, ranging from maladaptive pathological effects of environmental pollutants to adaptive immunological traits that increase disease resistance (Holeski et al., 2012; Rasmann et al., 2012; Perez and Lehner, 2019; Tetreau et al., 2019). In plants, the latter response has been referred to as “transgenerational acquired resistance” or “transgenerational induced resistance (t-IR),” which is typically based on a sensitization, or “priming,” of the immune system, mediating a faster and/or stronger immune response (Wilkinson et al., 2019). We have previously demonstrated that bacterial speck disease, caused by the hemi-biotrophic pathogen Pseudomonas syringae pv. tomato (Pst), results in t-IR that can be maintained over two stress-free generations in the self-fertilizing annual plant Arabidopsis (Arabidopsis thaliana) (Luna et al., 2012; Stassen et al., 2018). Although the exact epigenetic mechanisms underpinning t-IR are still under investigation, the induction and/or transmission of the response requires DNA demethylation at transposable elements and is associated with genome-wide changes in DNA methylation (Luna and Ton, 2012; Lopez Sanchez et al., 2016; Stassen et al., 2018; Furci et al., 2019). These results are supported by numerous other studies that have reported transgenerational changes in DNA methylation in response to environmental stress (Lämke and Bäurle, 2017; Wilkinson et al., 2019).

Evolutionary models predict that parental effects on specific traits can act as an adaptive mechanism to increase fitness in changeable environments (Leimar and McNamara, 2015; Pigeault et al., 2016; Proulx and Teotonio, 2017). However, despite numerous reports of transgenerational effects of stress in plants, there is still controversy over whether these responses are adaptive (Uller et al., 2013; Burggren, 2015; Crisp et al., 2016). Transgenerational phenotypic responses to light and water availability have been shown to provide improved fitness when the environments of parents and progeny are matched (Galloway and Etterson, 2007; Herman et al., 2012). However, when parent and progeny environments are mismatched, transgenerational effects can be unfavorable, which may explain why many epigenetic modifications are erased during sexual reproduction (Iwasaki and Paszkowski, 2014; Crisp et al., 2016; Gehring, 2019).

In the case of plant defense against herbivores and disease, the associated costs on host plants are often assumed to have driven the evolution of inducible defenses (Zangerl, 2003; Karasov et al., 2017). Costs associated with induced defense include direct allocation costs arising from the biosynthesis of defensive proteins and secondary metabolites (Züst et al., 2011). Priming of inducible defense provides a mechanism to optimize the cost–benefit trade-offs of defense, by preserving inducible defenses in a state of readiness for augmented responses to future infection without incurring direct costs of maintaining active defense (Karasov et al., 2017; Wilkinson et al., 2019). As well as direct allocation costs, ecological costs are evident when defenses targeted against pathogens or herbivores have negative consequences for interactions with beneficial microbes (Ruiz-Lozano et al., 1999) or pollinators (Strauss et al., 1999). Ecological costs also arise when increased resistance against one form of biotic attack is associated with increased susceptibility against another (Vos et al., 2013). In the case of pathogen-mediated t-IR, progeny from P. syringae-infected Arabidopsis expressed t-IR against another biotrophic pathogen, Hyaloperonospora arabidopsidis (Hpa), but the same progeny showed enhanced susceptibility to the necrotrophic fungus Alternaria brassicicola (Luna et al., 2012). Similarly, progeny from spider mite-infested Arabidopsis were primed to resist spider mites and aphids but suffered increased susceptibility to P. syringae (Singh et al., 2017). These results suggest that t-IR carries both benefits and costs in terms of resistance in matched and mismatched environments, respectively. However, a comprehensive understanding about the reciprocal costs and benefits of t-IR in terms of changes in (a)biotic stress resistance is lacking. Similarly, the relationship between stress severity experienced by the parents, on the one hand, and intensity and transgenerational stability of t-IR, on the other hand, remains unknown. Hence, a better understanding of the dose-dependent effects of (a)biotic stress on the costs and benefits of t-IR is required to understand the ecological significance of this epigenetic plant response.

In considering the evolution of adaptive parental effects, various studies have highlighted the importance of full factorial experimental designs to simultaneously assess the associated costs and benefits (Marshall and Uller, 2007; Bonduriansky et al., 2012; Uller et al., 2013; Burgess and Marshall, 2014; Tetreau et al., 2019). However, to date, most studies in plants have focused on the underpinning mechanisms and have overlooked the evolutionary and ecological significance of t-IR. Here, we employ a full factorial reciprocal experimental design to systematically test the costs and benefits of t-IR within a single experimental framework. We have examined the specificity of t-IR by quantifying impacts of three parental stresses at different levels of intensity in both matched and mismatched progeny environments. The results provide evidence that plants can adjust the strength and/or durability of t-IR based on their ability to interpret the reliability of predictive cues from their environment.

Transgenerational phenotypic plasticity offers a strategy for plants and animals to maximize fitness in variable environments and is more likely to emerge when the same type of environmental stress occurs regularly (Tricker, 2015; Proulx and Teotonio, 2017). Under such conditions, stress-exposed plants can optimize fitness either by maximizing their own immediate performance to the detriment of their progeny (“selfish parental effects”), or by modifying progeny traits to provide enhanced performance in the altered environment (Marshall and Uller, 2007). The latter strategy can take form in either a diversified bet-hedging strategy, or a more deterministic provision of specific adaptive traits, such as (transgenerational) pathogen-IR, which is tailored to the parental environment (Marshall and Uller, 2007; Crean and Marshall, 2009; Proulx and Teotonio, 2017). Not only do evolutionary models predict that transgenerational phenotypic plasticity is likely to evolve in fluctuating environments (Leimar and McNamara, 2015; Pigeault et al., 2016; Proulx and Teotonio, 2017), the model developed by Proulx and Teotonio (2017) suggests that deterministic parental effects provide increased fitness over a wider range of environmental parameters than a randomizing bet-hedging strategy. Such models assume that parental effects are specific to the eliciting stress, involve cost–benefit trade-offs, and that parents have access to reliable predictive cues.

Our study employed a full factorial design to address these predictions in relation to transgenerational responses of Arabidopsis to different (a)biotic stresses. Most previous reports have not, or only partially, addressed the specificity of t-IR, because they only tested t-IR upon a single parental stress and/or in matched environments. By contrast, the experiential design of our study (Figure 1) allowed us to examine t-IR phenotypes in progeny from parents exposed to different stresses in both matched and mismatched environments and in a reciprocal fashion. In the case of disease stress, we found strong evidence that t-IR is specific. Disease by biotrophic Pst bacteria mediates t-IR against taxonomically unrelated Hpa with a similar biotrophic lifestyle (Figures 3B,C) but fails to protect against necrotrophic Pc or abiotic salt stress (Figures 6A–C). Similarly, disease by Pc mediated t-IR against the same necrotrophic fungus (Figure 4), but not against biotrophic Hpa and abiotic salt stress (Figures 6C,D). While we found specificity of t-IR responses to biotic stress, other studies have provided examples of both specific and non-specific t-IR, especially in response to abiotic stress. Progeny of tobacco plants infected with tobacco mosaic virus displayed t-IR against both biotrophic and necrotrophic pathogens, as well as the genotoxic agent methyl methane sulfonate (MMS; Kathiria et al., 2010), while exposure of Arabidopsis to different heavy metals caused t-IR not only against the same heavy metals, but also against sodium chloride and MMS (Rahavi et al., 2011).

By contrast, we found no evidence for increased salt tolerance in F1 or F2 progeny from salt-exposed parents (Figure 5), even though F1 progeny from salt-exposed parents showed non-specific t-IR against biotrophic and necrotrophic pathogens (Figures 6E,F). Several previous studies have demonstrated t-IR against salinity (Boyko et al., 2010; Wibowo et al., 2016). However, the parental plants used in these studies were allowed to recover from the salt stress by transplanting them to normal soil. By contrast, the parental plants in our experiments were not transplanted and showed progressive loss of seed production and viability at increasing levels of salinity (Figures 2B,C), which indicates that they failed to recover from the stress. Accordingly, we propose that parental recovery from salt stress is essential for salt-mediated t-IR in matched environments. This hypothesis is supported by modeling, which shows that t-IR in invertebrates emerges at intermediate levels of disease stress, but not when there are more severe impacts resulting in mortality (Pigeault et al., 2016). By contrast, the pathogen-inoculated parental plants in our study were able to prevent negative impacts on reproductive fitness in our experiments (Figures 2B,C), presumablyby effective deployment of within-generation immune responses and subsequent compensatory growth during reproduction that maintained seed production. Similar compensatory responses were observed in Arabidopsis exposed to wounding, shading, or chilling stress during vegetative development (Lampei, 2019). Under these conditions, t-IR represents an ecological benefit for progeny in matched environment. One might also speculate that the observed differences t-IR effectiveness in our study reflect the nature of the eliciting stresses. Pathogen populations tend to oscillate naturally, and fluctuating environments are consistent with the evolution of adaptive parental effects (Leimar and McNamara, 2015; Proulx and Teotonio, 2017). In contrast, salinity would more likely be long term as opposed to a variable stress in natural environments, favoring genetic, rather than epigenetic adaptation.

The fact that pathogen-mediated t-IR is inducible and reversible in the absence of stress (Figures 3, 4) implies that the response is associated with costs (Vos et al., 2013; Karasov et al., 2017). The costs of defense can manifest themselves either as direct allocation costs, where investment in defense reduces resources available for growth and reproduction, or as ecological costs, when defense responses impact wider plant–environment interactions. Priming of defense has likely arisen as a mechanism to optimize cost–benefit trade-offs of induced resistance responses. In response to a primary exposure to stress, which indicates possible re-exposure in the future, plants prime defenses for more rapid and/or stronger responses (Wilkinson et al., 2019). Accordingly, naïve plants that have not been exposed to stress avoid costs completely by not deploying priming, whereas plants pre-exposed to the stress gain greater benefit from a more effective inducible defense response. This cost–benefit optimization by priming has previously been confirmed experimentally for within-generation priming (van Hulten et al., 2006), but the same rationale applies equally to between generation priming by t-IR (Karasov et al., 2017). Previous work has identified transgenerational impacts of parental stress on vegetative and reproductive development (e.g., Rahavi et al., 2011; Suter and Widmer, 2013; Groot et al., 2016), but it remains unclear to what extent these changes influence fitness. While we did not observe consistent effects on plant growth or seed set in F1 and F2 progeny from disease-exposed plants (data not shown), the reciprocal design of our experiments reveals ecological costs arising from increased susceptibility to other stresses (Figure 6). Antagonism between plant defense pathways against biotrophic pathogens, necrotrophic pathogens, and abiotic stress is well documented (Koornneef and Pieterse, 2008; Pieterse et al., 2012), and transgenerational persistence of these effects has been reported previously (Luna et al., 2012; Singh et al., 2017). Accordingly, we propose that negative cross-talk between defense pathways imposes a major ecological cost on t-IR responses to pathogens.

Central to the provision of adaptive transgenerational traits is the ability to make accurate and reliable predictions about future progeny environments. While this aspect has been emphasized in both evolutionary theory and modeling of parental effects (Burgess and Marshall, 2014; Leimar and McNamara, 2015), it has rarely been addressed experimentally. Rahavi and Kovalchuk (2013) tested the effect on homologous recombination frequency (HRF) and leaf size, of varying the timing and duration of exposure to several different abiotic stresses in Arabidopsis plants and their progeny. Overall, there was little evidence for dose-dependency, except that recombination events occurred earlier during the development of parent plants as doses of UV-C irradiation increased. HRF in progeny plants was not related to the duration of stress experienced by parents and t-IR against stress was not tested in these experiments. Other plant studies addressing this concept applied the same stress repeatedly over multiple generations. For instance, analysis of plants exposed to heavy metal stress for up to five successive generations indicated increasing tolerance as the number of previous generations experiencing stress increased (Rahavi et al., 2011). In one of the most comprehensive studies of this type, Groot et al. (2016) found complex interactions between parental (P), grandparental (GP), and great-grandparental (GGP) salt stress in Arabidopsis. When the stress was applied to only one generation, P effects were typically stronger than GP and GGP effects. For treatments over multiple generations, the impacts of GP and GGP stress were additive to P treatments for some traits, but antagonistic for others. In our study, varying levels of three different stresses were applied within one generation, providing a straightforward design to assess whether parents can distinguish stress severities and adjust the transgenerational response accordingly. Our pathogen treatments resulted in dose-dependent impacts on relative growth rate during the treatment period (Figure 2A), indicating that Arabidopsis perceives these stresses in a dose-dependent manner. Furthermore, analysis of the transgenerational stability of t-IR provided evidence for a dose-dependent relationship with parental disease severity. Although F1 populations from both Pst- and Pc-exposed parents expressed t-IR to statistically similar levels across stress levels (Figures 3A,B, 4A), these t-IR responses only persisted over a stress-free generation when elicited by the highest stress levels (Figures 3C, 4B). Furthermore, in mismatched environments, there was a dose-dependent effect on the costs t-IR by pathogens: both salt sensitivity of F1 progeny from Pst-infected parents and Hpa susceptibility of F1 progeny from Pc-infected parents correlated with the severity of parental disease stress treatment (Figures 6B,C). Overall, these results support our hypothesis that plants perceive disease severity as a predictive cue to adjust t-IR investment.

Collectively, our study demonstrates that parental investment in t-IR by pathogen stress provides benefits in matched environments and costs in mismatched environments. This stress-specific t-IR is dependent on the intensity of the stress experienced by the parents, which holds predictive value for future progeny environments. Accordingly, our findings are consistent with the evolutionary prediction that t-IR by disease stress from pathogens is an adaptive trait. In laboratory experiments, true measures of evolutionary fitness are difficult to estimate, since plants are grown in controlled environments in the absence of competition from non-stressed plants and other external factors, such as seasonal changes in temperature and photoperiod that can influence reproductive development. In one of the most convincing cases of adaptive parental effects in plants, Galloway and Etterson (2007) used field-based studies to demonstrate adaptive transgenerational plasticity in response to the light environment. It will now be of interest to extend our laboratory experiments by undertaking ecological field studies to seek support for the concept of t-IR as an adaptive transgenerational effect in nature.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

JT and MR conceived the project. JT and AL designed and supervised the experiments. AL, DP-P, and LF performed bioassays. JT performed statistical analyses. MR, AL, and JT wrote the manuscript. All authors reviewed and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.