Experiment 1 demonstrates that trichome density is inducible in M. guttatus (within generation plasticity) and that there is genetic variation for the extent of induction (Fig. 1, Table 2). Perhaps the most striking result of this study is from Experiment 2, which shows that trichome induction can be maternally transmitted and that there is genetic variation in the capacity for maternal transmission. While genetic variation for induced chemical defences has been demonstrated (Agrawal et al., 1999c), previous studies of trichome induction have usually failed to show genetic variation. However, this may be due to the fact that most of these experiments were not designed to detect it. Genetic variation is a prerequisite for evolution of a trait, and therefore should be a focus in studies assessing constitutive and induced defence.
The genetic variance demonstrated in this study is inter-populational. The RILs are derived from a cross between two divergent populations, the trichome dense PR and trichome depauperate IM. Not only does the IM genotype produce no trichomes constitutively, it does not respond to damage either within or between generations (Figs 1 and 3). Herbivory levels are low in the natural IM population, and it is hypothesized that they escape herbivory through rapid development. Rapid development has likely evolved as a method of drought escape in the IM population (Franks et al., 2007; Holeski, 2007). The absence of trichomes in the IM population may be due to the fact that they are costly to produce or simply because selective pressures have not necessitated them.
In contrast, the PR population plants constitutively produce trichomes and greatly increase production in response to damage (Figs 1 and 3). Induction of plant defences (chemical or physical) has been shown to reduce subsequent herbivory in a number of field studies (Denno et al., 1995 and references therein). In the present study, plants responded to leaf punctures that mimic the effect of chewing insects. Herbivores can emit chemicals that prevent or delay signaling pathways in the plant responsible for induction response (Agrawal, 1998, 1999, 2000; Schultz & Appel, 2004). However, other studies have found induction in defence traits with insect damage but no induction with simulated damage (Agrawal, 1998). Thus, the induction response to particular forms of simulated or insect damage seems to often be species/ herbivore specific (Agrawal, 1998, 1999, 2000, 2005; Traw & Dawson, 2002).
Despite the disadvantage of using simulated insect damage in these experiments, insofar as the experimental results are not directly applicable to the natural world, the use of simulated damage has several advantages. First, I was able to precisely control the quantity and location of tissue removed among plants in the treatment category. In contrast, the use of multiple insect herbivores across plants has potential to be less precise (Tiffin & Inouye, 2000; Inouye & Tiffin, 2003; but see Lehtilä, 2003). Secondly, in plant species such as M. guttatus, where many populations experience a number of generalist, chewing-insect herbivores (Holeski, 2007), simulated damage can be used to evaluate the effects of plant tissue loss independent of chewing actions or salivary compounds particular to a single chewing herbivore species (Inouye & Tiffin, 2003).
Plasticity and defence against herbivory
I found no relationship between the level of constitutive trichome production and the capacity for induction across Recombinant Inbred Lines (Fig. 2b). This result is inconsistent with the prediction of optimal defence theory, but is not incompatible with models of generalist/specialist herbivory trade-off. Optimal defence theory (Rhoades, 1979; Zangerl & Bazzaz, 1992; Zangerl & Rutledge, 1996) predicts patterns of defence based on the costs and benefits of defence, as well as the probability of attack. Assuming that defence carries a cost (because of the allocation of resources to the development of the trait that would otherwise be used for growth or reproduction), a negative correlation is expected between plant constitutive and induced levels of defence. To maximize resource use, plants that frequently and predictably experience herbivory are predicted to have high constitutive levels of defence and low levels of induction. Populations subject to infrequent herbivory are predicted to have low constitutive levels of defence and higher levels of induction.
Although induction in response to damage may be dependent on a number of factors including host plant species, herbivore species and type of damage (simulated vs. insect herbivore), data from other studies utilizing both mechanical damage (Zangerl & Berenbaum, 1990; Lewinsohn et al., 1991) and herbivore damage (Traw, 2002) support the predictions of optimal defence theory. In contrast, my results do not support the predictions of optimal defence theory (Figs 1 and 3), nor do the results of other studies using various plant species and types of damage (Brody & Karban, 1992; Agrawal et al., 1999c; Havill & Raffa, 1999; Alpert & Simms, 2002). However, because of the limitations of simulated damage in its direct relevance to the natural world, I would advocate the use of these results as preliminary evidence, rather than conclusive rejection of optimal defence theory in its applicability to M. guttatus.
The generalist/specialist trade-off model suggests that variation in levels of constitutive and induced resistance in a particular trait is maintained because of differential effects of the trait on different herbivores. For example, some specialist herbivores feed with increased frequency on plants with high levels of induction (Chambliss & Jones, 1966; Da Costa & Jones, 1971; van Dam & Hare, 1998; Agrawal et al., 1999c; Holeski, 2007). My results indicate such variation in constitutive and induced trichomes: trichome induction cannot be predicted by levels of constitutive trichomes (Fig. 2b). While this is consistent with the generalist/specialist trade-off model, generalist/specialist herbivores could not be attributed as a causal factor of the observed pattern without more information. Variation in levels of constitutive and induced resistance may also be explained in part by complex interactions between the resistance trait and insect pollinators or other mutualists (Agrawal & Karban, 1999).
In several plant species, resistance to herbivores changes as plants develop, although the direction of this change is variable (Price et al., 1987; Kearsley & Whitham, 1989; Karban & Thaler, 1999). Several species have increased chemical or physical resistance in the juvenile stage relative to the adult (Price et al., 1987; Kearsley & Whitham, 1989), whereas others have increased adult resistance relative to their juvenile condition (Karban & Thaler, 1999). Juvenile true leaves and adult true leaves usually have different patterns of cellular differentiation and are anatomically and biochemically different (Poethig, 1997; Mauricio, 2005; Donaldson et al., 2006; Rehill et al., 2006). Here, I show that constitutive fifth leaf (adult stage) trichome density is significantly higher than second leaf (juvenile stage) constitutive trichome density, indicating that at least insofar as trichomes affect herbivory, resistance is higher in adult M. guttatus plants relative to juveniles (Fig. 2a).
Experiment 2 demonstrates transgenerational induction of trichomes (Table 3). There is also genetic variation in the capacity for this response. A well-documented example of this general phenomenon is passive acquired immunity in human infants. During pregnancy, the mother passes antibodies through the placenta to the infant, so that the infant has high levels of antibodies at birth (Saji et al., 1999). Although the specific mechanism differs between the transgenerational induction in plants seen here and human acquired immunity, the general effect is similar: progeny have increased resistance against common enemies in a particular environment before they have actually encountered them.
Transgenerational induction has been shown in only one other plant species (Agrawal et al., 1999a; Agrawal, 2001, 2002). Agrawal performed a series of experiments with wild radish (R. raphanistrum), and demonstrated that both caterpillar (Pieris rapae) damage, and jasmonic acid treatment to maternal plants increased progeny resistance relative to control plants. Hydroxylated glucosinulate concentration increased in the progeny of damaged plants, whereas other classes of glucosinulates declined in concentation in these progeny. In another experiment with the same plant and herbivore species, indole glucosinulates were induced significantly by maternal damage (Agrawal et al., 1999c). In these experiments, Agrawal found no genetic variation for transgenerational glucosinulate induction, but did find genetic variation for transgenerational trichome induction (Agrawal et al., 1999c; Agrawal, 2001).
The mechanism for transgenerational induction is not known for either R. raphanistrum or M. guttatus. Epigenetic inheritance allows organisms to respond to a particular environment through changes in gene expression (Jaenisch & Bird, 2003; Rakyan & Whitelaw, 2003). Possible mechanisms include post-translational modification of DNA or proteins through processes such as demethylation and the effects of such processes could persist across more than one progeny generation (Jaenisch & Bird, 2003; Rakyan & Whitelaw, 2003). The genomic sequence of M. guttatus, in combination with extensive collection of candidate genes for trichome development in Arabidopsis and Antirrhinum, may allow mechanistic studies of transgenerational plasticity in this system.
In a review of almost 200 phytophagous insect species pair-wise interactions, Denno et al. (1995) found that host plants mediated competitive interactions more frequently than did natural enemies, physical factors, or interspecific competition. Which herbivores are affected by a particular induction response is determined by the lag time between plant damage and defence induction, the length of time that the defence is expressed in its induced/heightened form, and by the life history of the herbivore(s). This combination of factors creates potentially complex influences of induction on arthropod community structure (Moran & Whitham, 1990; Dalin & Bjorkman, 2003; Van Zandt & Agrawal, 2004a,b; Agrawal, 2005; McGuire & Johnson, 2006).
Transgenerational induction of defence introduces a new consideration to studies of plant defence and host-mediated competitive interactions. If densities of a particular herbivore species are consistent across seasons and maternal plants experience heavy damage, inducing higher constitutive defences provides offspring with a fitness advantage. If herbivore communities vary across seasons, transgenerational induction of a defence trait could provide plant-mediated indirect competition, albeit with a longer time lag than within-generation defence trait induction. Transgenerational induction could thus alter herbivore community dynamics and species interactions on a plant genotype across seasons, in a manner similar to within generation induction.