Plant behavioural ecology: dynamic plasticity in secondary metabolites


K. L. Metlen, Fax: +(406) 243 4184; e-mail:


Behaviour is in part the ability to respond rapidly and reversibly in response to environmental stimuli during the lifetime of an individual. Plants and animals both exhibit behaviour, but plant behaviour is most often examined in the context of morphologically plastic growth. Rapid and reversible secondary metabolite production and release is also a key mechanism by which plants behave. Here, we review plant biochemical plasticity as plant behaviour, and explicitly focus on evidence for responses that display rapid induction, reversibility and ecological relevance. Rapid induction and attenuation of plant secondary metabolites occur as chemically mediated root foraging, plant defence, allelochemistry and to regulate mutualistic relationships. We describe a wealth of information on the induction of various plant biochemical responses to environmental stimuli but found a limited body of literature on the reversibility of induced biochemical responses. Understanding the full cycle of dynamic plasticity in secondary metabolites is an important niche for future research. Biochemical behaviours extend beyond the plant kingdom; however, they clearly illustrate the capacity for plants to behave in ways that closely mirror the classic definitions and research approaches applied to behaviour in animals.

‘While we animals were busy nailing down things like locomotion and consciousness, the plants . . . acquired an array of extraordinary and occasionally diabolical powers by discovering how to synthesize remarkably complicated molecules.’

– Michael Pollan, Botany of Desire


The ability to respond to environmental stimuli in a fashion that is rapid, reversible and ecologically meaningful is the fundamental nature of behavioural ecology. Mobile animals, and particularly those with nervous systems, have received the lion's share of attention from behavioural ecologists, but plants can also respond rapidly to changes in their environment in ways that transcend ontogeny and that might reasonably be called behaviour. Understanding whether and how plants might ‘behave’ is not trivial; by seeking commonalities in biological phenomena, such as behaviour, across taxa, we increase our chances of discovering fundamental organizing principles in ecology and evolution.

A precise definition of behaviour is problematic even for animals that possess complex neural control and, of course, more so for plants. Many early definitions of behaviour included some component of movement, but animals with nervous systems behave without moving. Recent definitions of behaviour have focused on phenotypic plasticity – variation in phenotype for the same genotype caused by environmental stimuli – expressed over the lifetime of an individual (Silvertown & Gordon 1989; Karban 2008). Using this definition, many authors have examined morphological phenotypic plasticity of plants in the context of behaviour (Silvertown & Gordon 1989; Evans & Cain 1995; Silvertown 1998; Fransen, Blijjenberg & de Kroon 1999; Novoplansky 2002; Trewavas 2005; Karban 2008) realized the problems inherent to equating all phenotypic plasticity with behaviour (see Silvertown & Gordon 1989) and argued that behavioural plasticity should be relatively rapid and potentially reversible.

Morphological plasticity clearly fits some definitions of behaviour, but it is often not fully reversible and, in fact, the stimulus (commonly a resource patch or herbivore) can be much more ephemeral than the response (Hutchings & de Kroon 1994; Fransen & de Kroon 2001). For an in-depth review of morphological plasticity as behaviour, see de Kroon in this issue. Here, we attempt to refine ecological perspectives on plant behaviour by focusing explicitly on secondary metabolites that plants produce rapidly to respond to and manipulate their abiotic and biotic environments, and then cease producing once the job is finished.

The conceptual model in Fig. 1 illustrates the potential differences in benefit conferred to a plant exhibiting either biochemical or morphological behaviours in response to a generic stimulus. Maximal benefit to the plant is achieved when the rate of induction is as rapid as possible. The more rapid response should provide the greatest benefit by maintaining a maximum level of a particular behaviour for the time period in which the response can be useful. In other words, if the response is slow and the stimulus is ephemeral, then response may not maximally match the need for the stimulus. This hypothesized difference in rate of induction between morphological and biochemical responses (indicated by the shaded region on the left side of Fig. 1) results in disparate benefits received from the two response paths. For example, the way plants forage for nutrients suggests that physiological and biochemical plasticity may be more responsive and energy efficient than morphological plasticity (Drew & Saker 1978; Caldwell et al. 1987; Jackson, Manwaring & Caldwell 1990; Caldwell, Dudley & Lilieholm 1992; van Vuuren, Robinson & Griffiths 1996; Fransen et al. 1999; Hodge 2004).

Figure 1.

A conceptual model illustrating the differences between behaviours executed by secondary metabolites (solid curve) and morphology (dashed curve). This model describes responses to both the onset and removal of a stimulus. Relative costs and benefits are determined by the time to maximum response or cessation (steepness of curves). The cross-hatched areas represent the benefit that is conferred by the rapidity of induction and relaxation of a relatively rapid biochemical behaviour compared to a relatively slow morphological behaviour.

Once a stimulus is removed, the behaviour elicited by the presence of that stimulus has the potential to become only a construction or maintenance cost. As will become clear below, we know little about rapid relaxation of responses by plants, but as for the induction, more rapid relaxation of responses should also maximize the benefit of a behaviour. As indicated by the shaded region on the right side of Fig. 1, we hypothesize that biochemical behaviours are likely to cease more rapidly than morphological behaviours. Full reversal of some morphological behaviours may be very slow or may not occur during the lifetime of the individual plant, but secondary metabolite production can be curtailed almost immediately.

The ability to rapidly respond to a stimulus becomes yet more advantageous for ephemeral stimuli. There is no benefit (only cost) to a response that develops slowly and only reaches maximum response after the stimuli has ceased. Our model is highly conceptual, but it clearly demonstrates the potential for biochemical behaviours to provide greater benefit to plants, in particular, if the stimulus or the reason for it is short-lived. These differences in response rates have important ecological and evolutionary ramifications as the costs and benefits of plant behaviour can influence competitive outcomes and potentially lifetime fitness.

Plants use biochemicals to rapidly and reversibly respond to environmental stressors, the basis of physiological plant homeostasis and the primary mechanism by which plants acclimate to rapidly changing environmental conditions. This plasticity can take an almost endless variety of forms, e.g. altered ratios of photosynthetic enzymes to adjust to light conditions, suberin production to limit moisture loss from roots, genetic regulation of enzymes in response to resource limitation, or production of secondary metabolites in leaves in response to ultraviolet light. For practical purposes, we have divided these biochemical responses into two areas; ‘primary’ biochemistry, including fundamental physiological and genetic processes, and ‘secondary’ biochemistry. Secondary biochemicals, or metabolites, are loosely defined as organic compounds not directly involved in primary metabolic processes such as photosynthesis, cell respiration, cell division or cell growth. The absence of secondary metabolites generally does not immediately kill plants but can have profound effects on a species' ecology and evolution. Related to the latter, particular secondary metabolites are often restricted to individual species or narrow sets of species within a phylogenetic group, providing the potential for biochemically mediated behaviour almost as diverse as the plant kingdom.

The biochemical basis of metabolism and photosynthesis can have behavioural contexts, but these are common to all life. The ‘behavioural’ aspects of secondary metabolites are certainly not limited to plants, but the prolific diversity of plant secondary metabolites and their species-specificity provides an opportunity to take a somewhat plant-centric take on behaviour. Here, we present evidence for rapid biochemical induction and subsequent attenuation of plant secondary metabolites in chemically mediated root foraging, plant defence, allelochemistry and mutualistic relationships as dynamically plastic plant behaviour. Secondary metabolites clearly allow plants to behave plastically in many ways beyond these four examples, but these processes provide the best analogues we can think of to classic definitions and research approaches for behaviour.


Here, we focus on secondary metabolites exuded by plant roots in response to abiotic rhizosphere conditions. Nutrient deficiencies and toxic metals in soils elicit intense and dynamic secondary metabolite production and exudation from roots that increase the bioavailability of recalcitrant nutrients and neutralize potentially toxic metals in soil. These changes can be rapid and reversible (Fig. 1), and as described in a later section, these also appear to affect the way some species interact with their neighbours.

Soil nutrients are patchily distributed with relatively fertile microsites making up a small proportion of total rhizosphere volume (Chapin 1980; Grime, Crick & Rincon 1986). Root proliferation in nutrient-rich patches (Drew & Saker 1978; Caldwell et al. 1987; Jackson et al. 1990; Fransen, de Kroon & Berendse 2001) is a common strategy for maximizing the uptake of nutrients in a patchy environment, and has been presented as a fundamental form of plant plasticity (de Kroon et al. 2005) and behaviour (Hutchings & de Kroon 1994; Fransen & de Kroon 2001; Borges 2005; Semchenko et al. 2008). Roots are costly structures for plants to build and maintain. Therefore, mechanisms that promote efficient foraging and extraction of limiting soil resources without the creation of living tissue helps temper the costs of morphological root foraging in a heterogeneous soil matrix (Jansen et al. 2006). There are clear metabolic analogues to behaviour in resource acquisition by roots mediated through primary biochemistry (Liu et al. 1998; Dong et al. 1999; Fransen et al. 1999, 2001; Muchhal & Raghothama 1999; Hodge 2004; Misson et al. 2005), but plants also increase the pool of bioavailable nutrients by directly manipulating the soil solution with dynamically released secondary metabolites from their roots (Gardner, Parberry & Barber 1981; Marschner, Römheld & Cakmak 1987; Dinkelaker, Römheld & Marschner 1989; Duff, Sarath & Plaxton 1994; Hinsinger 2001; Neumann & Martinoia 2002; Hodge 2004; Lambers et al. 2006; Shane et al. 2006; Jin et al. 2007; Wang et al. 2007). Research suggests that physiological and biochemical plasticity appear to be more responsive and energy-efficient strategies than morphological plasticity (Drew & Saker 1978; Caldwell et al. 1987, 1992; Jackson et al. 1990; van Vuuren et al. 1996; Fransen et al. 1999; Hodge 2004), potentially supporting the hypothesized disparity between morphological and biochemical behaviours described in Fig. 1. Dynamic release of root exudates in response to mineral deficiencies can be rapidly inducible, reversible and ecologically meaningful, thus providing good examples of plant behaviour, but to our knowledge has not been explored in the context of behavioural ecology (but see Vivanco, this issue).

Deist et al. (1971) observed that some dicots, legumes in particular, have the ability to obtain phosphorus (P) from highly recalcitrant rock phosphate (apatite) but that most monocots do not. Subsequently, many plant species have been shown to access soil nutrients through a number of strictly biochemical manipulations often derived from a wide array of secondary metabolites released into the soil by roots (see reviews by Hinsinger 2001; Dakora & Phillips 2002; Lambers et al. 2006). This is particularly true for soil P, the bioavailability of which can vary widely depending on soil type and is often chemically bound in ways that severely curtail plant uptake. For example, in calcareous grassland soils of western Montana, P is highly limiting to plant growth. Total P is ≈900 µg g−1 soil, but the most biologically available fractions, those that are easily accessible to roots and mycorrhizae, range from 82 to 207 µg g−1 (Gundale, Sutherland & DeLuca 2008). In almost all soils, plants must ‘compete’ for P with positively charged minerals such as iron (Fe) and aluminium (Al) oxides which tightly bind to soil P and make it unavailable to plants. Phosphorus can also be incorporated into many other recalcitrant minerals with different metals such as calcium (Ca), Fe and Al.

Many plant species solve the problem of unavailable P in mineral complexes by releasing biochemicals that cleave the mineral bonds to P, often by binding to the metal (chelation), resulting in biologically available P. In calcareous soils with high pH and limited P, root exudation can also include proton extrusion to decrease soil pH in the rhizospheres which decreases the affinity of Ca for P. Cluster root forming species such as Banksia grandis (giant banksia), Lupinus albus (white lupine), Vicia faba (faba bean) and Caustis blakei (koala fern) exude specific blends of carboxylates (e.g. malate and citrate) as dictated by local soil chemistry (Lambers et al. 2002; Misson et al. 2005; Playsted et al. 2006; Li et al. 2007; Wang et al. 2007, 2008). The large number of different secondary metabolites released by roots results in highly species-specific utilization of the many different ways P is held in the soil. The production of various P-acquiring root exudates by different plant species is well known and there is a growing body of literature describing how plants adjust the content and concentrations of their exudates in response to soil conditions (Hinsinger 2001; Lambers et al. 2002; Wang, Shen & Zhang 2006; Li et al. 2007).

Given the difficulties inherent to simply describing the content of root exudates, many measurements of rapid biochemical responses of roots to external stimuli are remarkably precise. For the most part, such variation has been measured at the scale of days, but some responses initiate and cease within much shorter time periods. For example, Li et al. (2007) monitored the pH of V. faba rhizospheres hourly, and found that when grown in P-poor conditions, this species acidified its rhizosphere with malate and citrate, lowering the pH of agar gel by ∼2 units in as little as 6 h. Malusàet al. (2006) found that 10 days after P deprivation, Phaseolus vulgaris (common bean) root exudates had far higher phenolic concentrations than the exudates of P-supplied control plants, and after 28 days, phenolic concentrations in the root exudates of P-deprived plants were 100 times higher than those of controls.

Aluminium toxicity and P-deficiency responses tend be tightly correlated because Al has a high affinity for P, forming Al–P complexes that are particularly difficult for plants to separate. Dong, Peng & Yan (2004) studied the interplay of Al toxicity and P limitation on Glycine max (soybean) while controlling for pH. They found that P deprivation resulted in malate exudation from roots within a single day of P stress and increasing continuously to a peak at 8 days. Oxalate exudation began within 2 days of P stress. G. max roots exuded citrate in response to Al within 2 days, but without P limitation, oxalate and malate were not exuded. In contrast, Liao et al. (2006) found that P deprivation elicited G. max to exude citrate, malate and oxalate while Al induced malate as well as citrate and malate. Importantly, Liao et al. (2006) found that all of these responses were induced in less than 6 h. Wang et al. (2007) found that both Al addition and P deprivation resulted in citrate exudation from L. albus within 1–2 h of the stimulus.

Iron solubility even in well-aerated soils is far lower than necessary for the optimal growth of plants (Lindsay & Schwab 1982; Hell & Stephan 2003). Plants acquire Fe through a combination of the exudation of protons or the release of organic acids and phenols (Römheld & Marschner 1981a,b, 1986; Schmidt 1999). This response to Fe deficiency can be rapid. Less than 1 day after experimentally initiating Fe deficiency, Trifolium repens (red clover) began to exude Fe(III) chelate reductase, a powerful chelater that enhances Fe uptake in calcareous soils (Zheng et al. 2003). This response was developed further by Jin et al. (2007) who found that Fe deficiency stimulated the secretion of phenolic compounds 2 days after Fe removal from the growth medium, and phenolic secretion reached its highest levels after 8 days. When these phenolics were removed chemically from the growth medium, the concentration of Fe in leaf tissues decreased and plants become chlorotic (Jin et al. 2007).

These examples demonstrate that secondary metabolites can change rhizosphere chemistry rapidly, often within less than a day. However, genetic analyses indicate that plants may respond even more rapidly to the availability of soil nutrients. By monitoring genes that control root exudate responses to nutrient deprivation, Müller et al. (2004) found that genes coding for Arabidopsis response to P deprivation were activated in less than 30 min after placing plants in a P-deficient medium. The exceptional rapidity of root exudate induction is important for conceptually linking the ways plants sense and respond to changes in their environment to behaviour (Fig. 1).

A second step in this conceptual link is measuring whether or not putatively ‘behavioural’ responses are reversible. While this aspect of plant behaviour is not well understood, the particular shapes of the relaxation curves in response to stimulus removal (Fig. 1) could have important implications for behavioural costs and benefits. A fundamental problem with linking many aspects of plant morphological plasticity to behaviour is that the growth of new structures is not fully reversible (Hutchings & de Kroon 1994; Fransen & de Kroon 2001). Rapidly induced leaf or root growth can cease, but the effects of this plastic response reverse only after senescence of the tissues. In contrast, root exudation of secondary metabolites can be induced and reversed rapidly because of the highly ephemeral nature of most chemicals in root exudates.

Neumann & Römheld (1999) demonstrated that P deficiency increased proton release and citric acid accumulation in Triticum aestivum (wheat) and Lycopersicon esculentum (tomato), whereas Cicer arietinum (chickpea) exuded protons, citric acid and malonic acid. These exudates caused rhizosphere pH to drop from 7.0 to 4.0 and 4.9 for L. esculentum and C. arietinum, respectively, but T. aestivum had no effect. When P was provided to previously deprived L. esculentum and C. arietinum, pH returned to ≈7.0 within 3 days, indicating reversibility of biochemical exudation. Müller et al. (2004) grew Arabidopsis in P-limiting conditions and found that genes coding for responses to P deprivation were activated in less than 30 min. When P was added to previously P-deprived plants, most of the induced genetic responses were reversed within 30 min. The speed at which genes induced and suppressed indicated that roots were functioning autonomously of the overall nutrient status of the plant; in other words, roots appeared to sense and respond to their environment.

Aluminium in the rhizosphere of L. albus elicits specific, rapid and reversible secondary metabolite release (Wang et al. 2007). Within 1–2 h of Al addition, roots began to exude citrate, but this response was not induced by another trivalent cation, Lanthanum (La3+). When the same plants were moved from an Al-rich to an Al-free root solution, citrate exudation ceased within 3 h (Wang et al. 2007). Apparently, plants are not only capable of sensing highly specific stimuli in their soil environment but to continuously assess variation in the stimuli and behave accordingly.

Römheld & Marschner (1981a) found remarkable reversibility in the response of Helianthus annus (sunflower) to Fe deficiency. The reducing capacity of roots and associated rates of Fe accumulation in roots and shoots began to increase after less than 2 days of growth in Fe-free hydroponic solution. Root-reducing capacity nearly doubled after 4 days. After Fe was re-supplied, all values returned to normal within 12–18 h. Interestingly, Römheld & Marschner (1981a) went on to measure the behaviour of plants grown in Fe-limiting calcareous soils over time with and without supplemental Fe. They described a ‘rhythmic behaviour’ whereby plants became chlorotic and initiated Fe-deficiency responses. This resulted in increased uptake rates, which would return to normal as the plants became less chlorotic. Iron deficiency responses returned with the symptoms of chlorosis.

Localized dynamic secondary metabolite production within the root system of an individual plant suggests that plants may control initiation and cessation of root exudate production to a high degree. Brassica napus (oilseed rape) roots exude organic acids (mostly malic and citric acid) in response to P deprivation (Hoffland, Findenegg & Nelemans 1989). This exudation occurs near the root tips when the roots are in a homogenous but P-limiting solution. However, when roots were exposed to P bound in highly recalcitrant rock phosphate, exudation shifted to the place where roots contacted the rock phosphate (Hoffland et al. 1992). Although clear evidence for initiation and cessation of this behaviour at one place and time was lacking, such precise spatial control of exudation suggests a strikingly dynamic behaviour in the use of secondary metabolites.

In another remarkable example of qualitative localization of exudate chemistry, Liao et al. (2006) grew G. max in a stratified hydroponic solution with enriched P in the upper rhizosphere and Al contamination in the lower rhizosphere. Within a single root system, citrate and malate were exuded in regions subjected to Al, while citrate, malate and oxalate were exuded in the P-rich region of the rhizosphere. In addition, P-efficient genotypes of G. max foraged more efficiently for P in the upper rhizosphere (greater exudation and uptake) and exuded more malate from roots in the Al-contaminated rhizosphere, particularly when more P was added to the upper rhizosphere. By exuding greater quantities of malate in the Al zone, Al-resistant genotypes appeared to increase the longevity of taproots in the contaminated environment. When the different genotypes were grown in a homogeneous hydroponic system with P and Al, no differences in secondary metabolite exudation or growth were observed between genotypes. This study demonstrates the importance of plant behaviour for dealing with stressful conditions in the rhizosphere and dramatically illustrates the importance of plant biochemical plasticity in a heterogeneous soil environment.

There is some evidence that the dynamic plasticity in secondary metabolites described above has substantial adaptive value. In soils with low concentrations of available P, plant species that show plasticity in their ability to biochemically forage perform better than those that strictly forage via morphological plasticity (Ae et al. 1990; Dechassa et al. 2003; Gahoonia & Nielsen 2003; Hodge 2004). Even within a species, genotypes capable of greater exudation amounts and concentrations consistently show more resistance to P deficiency and toxic metals than genotypes that exude poorly (Ryan, Delhaize & Randall 1995; Dong et al. 2004; Liao et al. 2006). Such intraspecific variation clearly seems to be adaptive but also raises the potential for natural selection to favour genotypes with greater secondary biochemical plasticity in some environments, which may eventually lead to speciation.

Cluster roots provide an excellent example of intraspecific variation in secondary biochemical plasticity, with the potential to contribute to speciation. Cluster roots may be proteoid or dauciform, may occur in many families and may consist of unusually compact branching roots, rootlets and rootlet hairs (Purnell 1960; Lamont 1974). From a secondary metabolite perspective, the different types of cluster roots function similarly. When P is deficient, the typical morphologically plastic responses such as increase in root growth, fine roots and root hairs, or mycorrhizal associations are poorly expressed in cluster-rooted species. Instead, P-deficiency induces extensive cluster root formation (Marschner et al. 1987; Keerthisinghe et al. 1998; Neumann & Römheld 1999; Playsted et al. 2006). Once induced, rootlets of cluster roots grow quickly to a deterministic length and then, mature cluster roots exude a burst of carboxylates (primarily citrate), protons and phosphatases, which lasts approximately 3 days (Marschner et al. 1987; Watt & Evans 1999; Playsted et al. 2006; Shane et al. 2006). This response is affected, both quantitatively and qualitatively, by soil P availability and pH (Wang et al. 2007). Chelation and changes in rhizosphere pH precipitate a pulse of bioavailable P which is taken up by the tightly packed root hairlets (Gardner et al. 1981; Marschner et al. 1987; Dinkelaker, Hengeler & Marschner 1995; Keerthisinghe et al. 1998; Neumann & Römheld 1999; Playsted et al. 2006; Shane et al. 2006).

This complex adaptation has arisen in at least nine different plant families and can be found on six out of seven continents (Purnell 1960; Lamont 1974; Dinkelaker et al. 1995; Skene 1998, 2000; Neumann & Martinoia 2002; Lambers et al. 2006). Despite this wide geographic range, cluster-rooted species consistently occur on and often dominate soils with very low P (<80 mg kg−1) (Dinkelaker et al. 1995; Skene 1998, 2000; Lambers et al. 2006). Wide geographic dispersion, convergent evolution and high fidelity to a specific, exceedingly stressful environment supports the adaptive value of cluster roots, a specialization that happens to rely specifically on manipulating the environment with dynamically regulated secondary metabolites.

In sum, the dynamic manner in which plants biochemically forage for nutrients in the soil provides an excellent example of plant behavioural plasticity. Plants initiate very specific biochemical responses to given rhizosphere conditions quite rapidly. Rapid reversal of the response when the stimulus is gone completes the cycle of plant behavioural response to localized soil conditions. Biochemical behavioural plasticity thereby maximizes nutrient capture and minimizes plant carbon cost, potentially optimizing the curve in Fig. 1.


Herbivore-induced dynamics in plant secondary metabolites

Plant–herbivore interactions can have profound impacts on the lifetime fitness of individual plants (Karban & Baldwin 1997; Agrawal 1998, 1999; Baldwin 1998) and determine the composition of plant communities (McNaughton 1979; Huntly 1991). Plants have adopted many forms of defence in their attempts to discourage predation. Obvious physical defences such as leathery or hairy leaves, thick bark, and thorns or spikes are widespread. Morphological defences, however, are largely irreversible even though they may be plastic (Young & Okello 1998). Plants also utilize secondary metabolites as a form of biochemical defence in both inducible and constitutive forms.

Induced responses to attack differ from constitutive defences in that they are mobilized only after an attack has occurred and are potentially reversible, while constitutive defences are always present. It has largely been argued that induced responses to herbivores represent a significant cost savings for plants – only investing in defence when necessary and allocating resources to growth and reproduction when not under attack (Fagerstrom, Larsson & Tenow 1987; Zangerl 2003). Others have argued that through highly dynamic defences, plants can slow or even avoid the coevolved tolerance of herbivores to defensive compounds (Karban, Agrawal & Mangel 1997; Agrawal & Karban 1999). However, induced defences are not without limitations. Induced defences have a necessary lag time from the moment of induction to the time of effectiveness (see Fig. 1) – thereby allowing a herbivore to potentially cause irreparable damage before the full response can be invoked (Baldwin 1998; Zangerl 2003). In addition, increased plasticity in plants can result in reduced fitness in the absence of herbivory, which suggests that the evolution of inducible defence may be impeded by the genetic cost of being more plastic (Agrawal et al. 2002).

Induced biochemical responses of plants to herbivores have been documented for over 100 species in lab and field experiments, and these comprise highly diverse chemical constituents and mechanisms of delivery (Karban & Baldwin 1997). The induced responses of plants to herbivores have been the subject of many books and reviews (Karban & Myers 1989; Karban & Baldwin 1997; Agrawal & Karban 1999; Karban et al. 1999; Tollrian & Harvell 1999) and we cannot fully review the subject here. Instead, we aim to highlight examples that demonstrate the dynamic plasticity of secondary metabolites in plant–herbivore interactions. As above, we focus on the rapid induction of an ecologically relevant response to a stimulus, and then the rapid relaxation or reversal of the response when the stimulus is removed (Fig. 1).

Despite intense interest in induced responses to herbivores over the last 20 years, there is a surprising lack of information on their relaxation. Relatively few have attempted to characterize the full cycle of induction, peak response and relaxation and, instead, have focused on the mechanism by which responses occur and the rate of induction. This creates a problem for fully understanding the costs of induced defences because rapid reversal would reduce costs whereas slow reversal would increase costs. Underwood (1998) addressed the importance of understanding the full induction cycle to both plant and herbivore population dynamics. In a study of the effect of induced resistance in G. max and Epilachna varivestis (Mexican bean beetle), Underwood utilized larval bioassay experiments to characterize the timing of the full induction cycle in four genotypes of soybean. Results indicated that all four genotypes of G. max mounted a peak response within 3 days, followed by complete relaxation within 15 days post-attack. However, following the induction cycle, all four genotypes of G. max entered a phase of increased susceptibility where E. varivestis showed preference for leaves of previously damaged plants over leaves of undamaged plants for a period of between 15 and 20 days. This result underscores the importance of closely tracking the full cycle of induced responses to herbivores, as the rate of decay can have profound ecological and evolutionary consequences for both the host plant and the herbivore.

The interaction between Betula pubescens (mountain birch) and Epirrita autumnata (autumnal moth) in Fennoscandia has been studied for more than 40 years. Early research showed that trees were capable of inducing resistance to moth attack in two ways: (1) delayed induced resistance (DIR) in which trees responded to herbivores in the years following attack; and (2) rapidly induced resistance (RIR) which was immediate (Haukioja, Suomela & Neuvonen 1985). Bioassays with moth larvae indicated that low-molecular weight phenol content in leaves was crucial to the induced resistance of the birch to moth attacks (Haukioja & Neuvonen 1985).

Phenols represent a large class of secondary metabolites; compounds that are often implicated in herbivore-induced plant defence strategies (Ruuhola & Yang 2006) and are found in large quantities in mountain birch. Phenols, such as chlorogenic acid and tannins, are not particularly good deterrents of autumnal moth larvae (Haukioja 2005). However, when phenols are oxidized into a much more biologically active form, such as semi-quinones and quinones, by peroxidase (PODs) and polyphenoloxidase (PPOs) enzymes, they have a much larger impact on herbivore larvae performance (Felton et al. 1989).

Ruuhola & Yang (2006) demonstrated that mountain birch leaves respond to attack by not only rapidly up-regulating the production of both PODs and PPOs but also simultaneously down-regulating the production of antioxidative enzymes (Ruuhola et al. 2008). This twofold immediate response in enzyme activity results in rapid accumulation (less than 24 h) of compounds that directly affect the herbivore's ability to assimilate amino acids and proteins in the digestive tract. Although the exact time frame for relaxation of the RIR response is not known, Kaitaniemi, Neuvonen & Nyyssonen (1999) demonstrated that DIR in mountain birches relaxes in at the most 2 years, but potentially much more rapidly than that.

Relaxation of induced resistance (RIR and DIR) does not, however, mean that trees are as susceptible to attack in the future as are naïve trees. It has been proposed that plants have an immunological memory much like that of animals. When plants experience a herbivore attack in 1 year, a stronger response is mounted to subsequent attacks in later years (Karban & Niiho 1995). When B. pubescens was exposed to herbivory by E. autumnata 5 years prior, they were able to mount a much stronger response to attack than trees without herbivore experience despite nearly identical leaf chemistry (Ruuhola et al. 2007). This suggests that plants may be capable of ‘learning’ behaviours and retaining lessons learned for future encounters with predators (Trewavas 2005).

Plants also defend themselves via indirect interactions by utilizing volatile secondary metabolites to attract the predators of their herbivores. Research by Turlings, Tumlinson & Lewis (1990) identified an interaction between Zea mays (Iona sweet corn), the generalist herbivore, Spodoptera exigua larvae (beet armyworm) and the parasitic wasp Cotesia marginiventris. In this tri-trophic interaction, when Z. mays is damaged by the feeding of S. exigua larvae, volatile terpenoids are synthesized and released from both damaged and undamaged leaves of the plant that attract female C. marginiventris. The induced production and release of volatile terpenoids are extremely rapid – occurring in just a few hours – and are also rapidly reversed, with signal attenuation occurring over a period of roughly 24 h after attacks end (Turlings & Tumlinson 1992; Turlings et al. 1995).

In addition to responding to herbivory rapidly and reversibly, Z. mays produces a signal that is chemically complex and herbivore specific. Composed of a blend of terpenoids, the signal transmits specific information to the wasp regarding the herbivore host as well as about the host plant. The information contained in this volatile terpenoid signalling allows C. marginiventris to distinguish between closely related host species S. exigua and S. frugiperda feeding on Z. mays (Turlings et al. 1995). Additionally, the signal is not just precise, it is also reliable. Research shows that the signal is only induced when caterpillar damage is detected – not when mechanical damage is sustained. Induction of the signal requires both damage and saliva from S. exigua larvae in order to release predator attracting terpenoids (Turlings et al. 1990). This suggests a level of sophistication and complexity that was previously thought to be the unique domain of animals – complex, multi-trophic behaviours.

Does dynamic secondary metabolite plasticity in response to herbivore attack confer higher fitness to plants? The potential reduction in carbon budget allocated to defensive purposes when a plant employs an ‘only when necessary’ approach should result in improvements in fitness over plants with constitutive defences or no defence at all when subjected to herbivory. But this has proven very difficult to demonstrate empirically (Heil & Baldwin 2002; Cipollini, Purrington & Bergelson 2003), although good examples exist. Agrawal (1998, 1999) showed that early induction of defences in Raphanus sativa (wild radish) reduced subsequent herbivore attacks by as much as 50% and increased seed mass, a lifetime fitness correlate, by 60% when compared with controls. This suggests that plants with induced resistance to herbivores have higher fitness over other less plastic genotypes when herbivores are present. More recent research that takes advantage of advances in molecular techniques and the development of mutant strains (non-inducible, inducible, and constitutive) of plants have shown that induced defences offer high fitness benefits over plants with either constitutive defences or no defences at all (Glawe et al. 2003; Kessler, Halitschke & Baldwin 2004; Zavala & Baldwin 2004; Zavala et al. 2004). Use of these techniques avoids the potential confounding problems of pleiotropic effects of the induction pathways, thereby providing some of the most convincing evidence for the adaptive benefit of induced defences (Zheng & Dicke 2008).

Competition and dynamic plant secondary metabolites

Dynamic biochemical responses in root exudates to nutrient limitation and toxicity in soils are good examples of behaviour, and there is a substantial literature on root–root interactions that are likely to be chemically mediated (Mahall & Callaway 1991, 1992; Callaway 2002; Callaway & Mahall 2007) and how root exudates mediate interactions with other species (Bais et al. 2004). In fact, root-mediated interactions have been suggested as an analogue to territoriality, one of the most important forms of behaviour (Schenck, Mahall & Callaway 1999). However, we know far less about how root exudates are induced by other plants. To our knowledge, the only example, and it is incomplete, involves Centaurea maculosa (spotted knapweed), a European invader in North America, and native species it encounters. C. maculosa exudes the compound (±)-catechin from its roots (Bais et al. 2003; Weir, Bais & Vivanco 2003; Blair et al. 2005; Ridenour et al. 2008), which inhibits the growth of neighbouring competing plants (Bais et al. 2003; Weir et al. 2003, 2006; Callaway et al. 2005; Perry et al. 2005a,b; Thelen et al. 2005; Thorpe 2006; Rudrappa et al. 2007; Inderjit et al. 2008a,b; Simões et al. 2008; but see Blair et al. 2005, 2006 for critiques of the system). (±)-Catechin is a chelator, the addition of which makes P available in soils where it is bound by calcium. Adding activated carbon to the soil, which adsorbs to many charged organic molecules, eliminates the chelating effect of (±)-catechin (De Luca & Callaway unpublished data; also see Thorpe, Archer & DeLuca 2006). Native species vary a great deal in their susceptibility to (±)-catechin (Thorpe 2006). Weir et al. (2006) found that two good competitors with C. maculosa, Lupinus sericeus (lupine) and Gaillardia grandiflora (blanketflower) produced more than an order of magnitude higher levels of oxalate in their root exudates than three poor competitors. They also found that oxalic acid reduces the oxidative damage generated by (±)-catechin. Furthermore, exposure to (±)-catechin increased the exudation of oxalate by G. grandifolia by four times and L. sericeus by 50 times. Interestingly, native grasses are highly associated with L. sericeus in communities invaded by C. maculosa, and field experiments show that L. sericeus indirectly facilitates native grasses in C. maculosa stands. This facilitation was correlated with the presence of oxalic acid in the soil. When oxalic acid was applied to the roots of native grasses, it alleviated the allelopathic effects of (±)-catechin, indicating that root-secreted oxalic acid may act as a chemical facilitator for plant species that do not produce the chemical.

Tharayil et al. (2008) measured the release of 8-hydroxyquinoline, also an allelopathic chemical (Vivanco et al. 2004), from the roots of Centaurea diffusa (diffuse knapweed) in response to iron deficiency. They found a diurnal rhythm in the release of the biochemical and phytotoxic effects but no pattern of induction. It is not known if the presence of competitors might induce the production or release of 8-hydroxyquinoline.

To our knowledge, the (±)-catechin-oxalic acid system provides the only evidence for competitor-induced biochemical interactions in the literature, and there is no indication that the production of (±)-catechin is affected by neighbors, nor is there information about the relaxation of the induced response.

Herbivory can stimulate root exudates that in turn have allelopathic effects on other plants (Cipollini 2002; Thelen et al. 2005). For example, methyl jasmonate (MeJA) is an important secondary metabolite induced and released as a volatile when plants are attacked by insect herbivores and microbial pathogens. Bi et al. (2007) found that the application of MeJA to two different strains of Oryza sativa (rice) enhanced the plant's allelopathic potential. Extracts of the leaves that were treated with MeJA were more inhibitory to the growth of Echinochloa crus-galli (barnyard grass) than leaves of plants not treated with MeJA.

Facilitation and dynamic plant secondary metabolites

We know of no examples of fully dynamic plasticity in secondary metabolites as a mechanism of facilitation. But the release of chemical exudates to acquire soil nutrients may benefit neighbours and, in other context, such release is induced and then relaxes. This warrants some speculation on the potential for such behaviour to facilitate other species. If a species does not exude a chelator that functions well in a certain soil type, it might benefit that species to grow next to a neighbor that does. Many studies have found that when grains such as wheat, sorghum or rice are intercropped or planted in rotation with legumes such as Cajanus cajan (pigeon pea), V. faba, Pisum sativum (field pea) or L. albus, the yield of the grains and their tissue P concentrations increase significantly (Zhang & Li 2003; Zhang et al. 2004; Nuruzzaman et al. 2005; Lambers et al. 2006; Li et al. 2007). This fertilizing effect has been attributed to the chelation of Fe–P complexes by exudates rich in secondary metabolites in Fe-rich soils (Ae et al. 1990; Li et al. 2007) and the exudation of carboxylates and protons by legumes in calcareous soils where P is bound to Ca (Zhang & Li 2003; Zhang et al. 2004; Nuruzzaman et al. 2005; Li et al. 2007).

Zhang & Li (2003) reviewed the processes involved in intercropping and found that facilitation between different crop species may have been related to the shared effects of root exudates. Z. mays improved Fe uptake for intercropped Arachis hypogaea (peanut) plants, V. faba enhanced nitrogen and P uptake by intercropped Z. mays, and C. arietinum (chickpeas) facilitated P uptake by associated T. aestivum. In the last case, Zhang & Li hypothesized that P depletion in C. arietinum rhizospheres induced the release of substances that mobilize organic P.

In an elegant experiment, Li et al. (2007) separated the effects of roots, root exudates and soil types in a V. faba/Z. mays intercropping system. The presence of V. faba increased maize yield the greatest when P was limiting and this effect diminished with increasing exogenous P fertilization. When P was bound in progressively more recalcitrant forms (Ca–P, Fe–P and Al–P), the positive effect of V. faba on Z. mays increased. This effect was tested in the field and in the greenhouse with barriers that blocked roots and root exudates and barriers that blocked roots but not root exudates. In the field, they found that the positive interactions of intercropping occurred belowground. In pots, when no P was added to natural soils, Z. mays plants took up 30% more P when V. faba root exudates were allowed to cross into the Z. mays rhizosphere than when exudates (organic acids and protons) were excluded (Li et al. 2007).

Mutualisms and dynamic plant secondary metabolites

Plants send biochemical signals to their mutualists, and parasites can detect biochemicals released from plants. In the most well described of these, chemical signals between legumes and bacteria form the basis of all plant-N-fixing bacteria mutualisms. Legume roots release different flavonoids and other secondary metabolites which induce the nod gene in N-fixing bacteria allowing the bacteria to enter plant roots and begin the formation of nodules (see review by Cooper 2007). Legumes also release other compounds including as signals to N-fixing symbionts including betaines, aldonic acids, xanthones, simple phenolics and jasmonates, but the purposes of these signals have not been worked out. The presence of appropriate N-fixing bacteria can induce the release of flavonoids (van Brussel et al. 1990). N-fixing bacteria communicate back to the plant through ‘nod factors’ and proteins which alter plant defences allowing the bacteria to infect the root (Bartsev et al. 2004; Cooper 2007). The secondary biochemical signals and induced chemical responses to signals in this system are very well defined (Begum et al. 2001; Morón et al. 2005; Antunes, Rajcan & Goss 2006; Mabood et al. 2006), but to our knowledge, there are no studies that clearly show reversal of the response, and thus, this system offers an incomplete perspective on plant behaviour. Plants also signal their fungal mutualists in a similarly complex manner that involves secondary metabolites exuded from plant roots (Harrison 2005), but little is known about either induction or relaxation of this process.

Many plants participate in a mutualistic relationship with ants; the plants provide food for the ants in the form of extrafloral nectar or other structures and the ants aggressively defend the plant. Herbivores can affect this mutualism by inducing higher production or biochemical constituency of nectar in many different plant taxa (Smith, Lanza & Smith 1990; Swift & Lanza 1993; Wäckers & Wunderlin 1999; Heil & McKey 2003; Ness 2003) and volatile secondary biochemical signals can also induce nectar production in the absence of nectar (Kost & Heil 2005; Heil & Bueno 2007). Importantly, the induction of extrafloral nectar can be reversed rapidly. Nectar produced by Macaranga tanarius showed a marked decrease when herbivores were removed (Heil et al. 2000), indicating a relaxation of the response when the behaviour is not needed.

Pollinator–plant mutualisms can also be tightly controlled by species-specific secondary metabolites in nectar, on pollen or on flowers (Dodson et al. 1969; Williams & Dodson 1972; Rhoades & Bergdahl 1981; Vonderohe 1982; Svensson, Pellmyr & Raguso 2006), but to our knowledge, nothing is known about induction or relaxation responses in the production of these biochemicals.

In sum, highly dynamic secondary metabolite plasticity can strongly influence plant interactions among species and across trophic levels with important ecological implications. Induced biochemical responses to herbivores are some of the most well-documented components of plant behaviour to their biotic environment, but induced biochemical by plant competitors, facilitators or mutualists are much less well documented. Demonstrations of reversibility or relaxation of responses, which is a crucial behavioural component, are surprisingly rare. In one of the more important examples, Underwood (1998) found that the rate of reversibility of induced defences had important ramifications for the evolutionary trajectory of plant–herbivore interactions. This likely holds true for other ecological relationships as well, suggesting that maximal fitness benefits can only be realized if plants are capable of rapid reversibility once a particular stimulus has ceased.


Plant biochemical responses to environmental stimuli are a ubiquitous but often overlooked component of behavioural ecology. Here, we show that plants can behave dynamically in response to environmental stimuli by rapid induction and reversal of secondary metabolite production. This supports our conceptual model of a relatively steep response curve after stimuli for secondary metabolites (Fig. 1). Such ‘behavioural’ aspects of biochemistry are not unique to plants, but the diversity of secondary metabolites and the ubiquity with which they are used by plants suggest that dynamic biochemical plasticity is a primary system for understanding plant behaviour. This is because biochemical responses may be more rapidly induced and fully reversible than morphological plasticity. Rapid induction of secondary metabolites has been shown to benefit plants foraging for nutrients and responding to herbivore attack. Interactions among plants as well as between plants and their symbionts are also mediated by dynamically produced biochemicals.

Despite a number of studies that have investigated various inducible plant biochemical responses to environmental stimuli, there exists a remarkable gap in the current understanding of the reversibility of inducible plant responses. Thus, the shapes of the relaxation curves on the right side of Fig. 1 are hypothetical. To best understand secondary biochemistry in the context of behaviour, we must also understand how plants respond to the removal of a stimulus, not only the nature of the original response. The true costs and fitness consequences of inducible secondary metabolites can only be deduced once the full cycle of these responses and their rates have been elucidated. These fitness consequences are a measure of the adaptive value of inducible secondary metabolites and, therefore, may allow the most significant evaluation of plant responses to environmental cues in the context of behaviour.