How should toxic secondary metabolites be distributed between the leaves of a fast-growing plant to minimize the impact of herbivory?


†Author to whom correspondence should be addressed. E-mail:


  • 1This paper considers the possibility that the key determinant of leaf age feeding preferences in foliovores is not the concentration of either nutrients or secondary metabolites, but is the ratio of the two.
  • 2In some fast-growing plants, nitrogen is most heavily defended by defensive toxins in the young leaves. This empowers the use of a simple model of leaf-age preference, based on the conflict between maximizing nutrient intake and minimizing toxicosis.
  • 3Young leaves are particularly valuable, not only because they lock up nitrogen, but also because their assimilative value is high. We calculate the loss of value due to herbivory and find that, if the herbivore is moderately intolerant of toxicity and feeds selectively on its predicted optimal leaf age, the costs of damage are greatly reduced. However, efficient toxin distribution protects the plant only if it grows rapidly, so that the well protected young foliage retains a high value.
  • 4The trends are reconcilable with observed leaf-age preferences of both polyphagous and oligophagous species. There is, as yet, little empirical evidence to substantiate the model, but it may be useful for future studies to focus on the toxin : nutrient ratio as a potentially important determinant of feeding preferences.


In the most general terms, many studies assume that the feeding preferences of herbivores are governed by one of two criteria: (1) the need to maximize the intake of some limiting nutrient, often considered to be nitrogen (Thomas 1987; Jacobsen 1992); (2) the need to avoid high concentrations of potentially toxic secondary chemicals (Lincoln & Couvet 1989; Newman, Hascom & Kerfoot 1992; Stamp & Bowers 1994). However, both factors are seldom considered simultaneously, probably because the way in which herbivores evaluate the importance of the differing effects is not known.

Despite this, the nutrient and secondary chemical hypotheses could be unified very simply using a basic physiological rule (Lambdon & Hassall 2001). In this treatment, both polyphagous and oligophagous herbivores are subject to two constraints on their maximum theoretical food intake, Fmax. First, toxicity becomes limiting when the accumulated quantity of toxin in the body (the dose) exceeds a critical threshold, Dmax, and overwhelms the body's ability to buffer its effects. Below the threshold, the costs of the toxin are much reduced. Thus if T is the concentration of toxin in the food:

image(eqn 1 )

Second, meal size can also be limited by gut capacity (Cg), expressed as the maximum weight of foliage that can be consumed. Two conditions may prevail:

  • 1Fmax < Cg, toxicity is limiting to food intake;
  • 2Fmax > Cg, toxicity does not affect feeding behaviour because meals are limited to subtoxic quantities by gut capacity. Fmax cannot exceed Cg.

The two constraints on feeding dictate very different relationships between the maximum nutrient intake, Imax, and leaf age. Imax is the product of nutrient concentration, N, and food intake (N × F). When limited by gut capacity, nutritional benefit is directly proportional to N. However, when limited by toxicity, Imax is inversely proportional to the T : N ratio:

image(eqn 2 )

Hence studies of this nature should perhaps consider the ratio of nutrients to secondary chemicals as a key determinant of feeding preference, rather than either component in isolation.

Herbivores may respond to a chemical not only as a result of the toxic effect it elicits, but also as a secondary cue, either alone or as part of a complex signature (Renwick 1989). However, their responses to leaves of different ages within a plant are likely to be governed principally by the toxicity constraints. In these circumstances, selection decisions are simplified because the chemical signature is usually much less variable between ages than between individuals, although total concentrations may vary considerably over the life span.

Unfortunately, despite the many studies of herbivore leaf-age preference, few have measured secondary chemicals and nutrient content in synchrony. In a study on the distribution of glucosinolates and nitrogen in Brassica napus L. and Brassica rapa L. (Lambdon et al. 2003), it was found that the ratio of the two steadily decreased with leaf age. This would ensure that the valuable, younger foliage is better protected than the older, less photosynthetically efficient foliage. Comparable studies are few, but those on Senecio jacobaea (de Boer 1999); Lycopersicon esculentum (Hoffland et al. 2000); and three pioneer Salicaceous species (Ikonen 2002) appeared to show similar declining ratios. In Cynoglossum officinale (Boraginaceae), pyrollizidine alkaloids can be 490-fold less concentrated in the oldest leaf than in the youngest, while the corresponding N concentrations differ by only 130-fold (van Dam et al. 1995). Such evidence, although limited, offers the intriguing possibility that this type of distribution may be a conserved trait, although much more work would be required before such a statement could be supported. Here we illustrate how potentially valuable this defensive strategy could be by outlining some of the feeding patterns predicted by the use of the criterion and the theoretical benefits of these patterns to the plant, and discuss how the predictions compare with observed trends.

Materials and methods

Changes in leaf status will be based on an idealized model system, although parameters have been estimated from data on the distribution of N and glucosinolates in Brassica from the study of Lambdon et al. (2003). Concentration–time curves have been fitted with continuous functions (Fig. 1). They show an exponential decline as the leaf develops, slowing at maturity, and a further decline as the leaf begins to senesce, when remaining recyclable metabolites are withdrawn to be reallocated (Porter et al. 1991). Digestible nitrogen will differ slightly from the curve shown: many herbivores are incapable of ingesting nitrogen contained within structural components of the leaf (e.g. lignins). It is assumed that these form a relatively constant total biomass throughout the life span (Marschner 1995), equivalent to 30% of that in a mature leaf.

Figure 1.

Predicted changes in leaf chemistry with age: (a) total nitrogen (digestible N shaded); (b) total glucosinolates.

As an approximation, photosynthetic output can be estimated as the product of nitrogen content, the saturated nitrogen-use efficiency (sPNUE) and photon flux density (PDF) (Hikosaka et al. 1999). sPNUE is a measure of photosynthetic CO2 fixed per unit nitrogen in the leaf when no shading effects are present. PDF is the average density of incident radiation which the leaf receives, determined by the shading effect from younger foliage. Logistic equations for these parameters were selected to describe a photosynthesis curve which is broadly consistent with data shown or used by Hartnett & Bazzaz (1984), Iwasa et al. (1996) and Anten et al. (1998).

Harper (1989) predicted the ‘assimilative value’ of a leaf to its parent plant, V, as a function of the total contribution the leaf will make to assimilation during its lifetime. Hartnett & Bazzaz (1984) used a similar function to calculate the same result. For leaves aged A,

image(eqn 3 )

where LA is the probability of a leaf surviving to age A, pA is the photosynthetic output of a leaf aged A, and r is a constant specifying the rate of increase in the plant's foliar biomass. From the photosynthesis curve, VA has been calculated and is displayed in Fig. 2. The relative change in VA is expressed as a proportion of the leaf's assimilative value at foliogenesis. Value reaches a maximum before full expansion, somewhat before the peak of photosynthetic efficiency. Hartnett & Bazzaz (1984), calculated a difference of approximately 7 days between the two maxima.

Figure 2.

Assimilative value of a leaf (VA) as it ages relative to value at foliogenesis.

Herbivory will have an effect on V. To examine this, the term LA can be modified to account for a proportionate decrease in leaf biomass, inline image. The integral of VA with respect to A, for the entire life of the leaf, is a measure of expected lifetime performance. To assess the impact of the herbivore, the integral has been divided by that obtained for a leaf which has not been subject to herbivory. The new index of relative lifetime value, VRL, is a measure of the diminished ‘fitness’ which results.


Influence of toxin and nitrogen distribution on herbivory

A critical factor in the overall pattern of feeding preference is the point at which toxin concentration in the leaf declines sufficiently to cause a switch from Dmax limitation to Cg limitation. This occurs when Dmax/TCg. The implications are shown in Fig. 3 for five values of Dmax where Cg is kept constant. When herbivores have a very low tolerance of toxicity, the switch is to the right of the age range, and thus older leaves are the most favourable (the T : N ratio is lowest here). Very tolerant herbivores undergo a switch to the left of the age range, and can thus gain maximum nutritional benefit from the younger foliage. They also have a large advantage in total consumption during a meal, and should therefore display a strong age preference.

Figure 3.

Predicted nitrogen gain, during a single meal, for herbivores with different tolerances of toxicity: low (a) to high (e).

Because the influence of gut capacity is manifested only to the right of the switch, increasing this parameter (via changes in gut size or food processing speed) at constant Dmax has more restricted effects on food consumption (Fig. 4). A large gut capacity has greatest benefits for those species that favour older foliage.

Figure 4.

Predicted nitrogen intake (Imax) during a single meal for herbivores with different gut capacities: high (a) to low (c).

The optimum leaf age for a particular herbivore is indicated by the following preference statistic:

image(eqn 4 )

where Tmature is the concentration of toxin in mature leaves. Π is not linearly related to leaf age because preference is affected by the rate of change in the T : N ratio. However, as Tmature is the reference point, a Π of 1 indicates a preference for mature foliage. Values <1 indicate preferences for ageing foliage, and values >1 for young foliage (Fig. 5).

Figure 5.

Correspondence between leaf age, T : N ratio and values of the age preference statistic Π, with intervals of Π indicated by the numbered ranges.

Adaptive value of toxin distribution

The importance of the T : N ratio in the above relationship suggests that a plant may be able to exert a strong influence on herbivore feeding preferences by distributing its toxins effectively, thus succeeding in deterring herbivores from the youngest tissues, even though they are nutrient-rich. The benefits gained from restricting herbivory in this way can be assessed by considering the likely impact on photosynthesis. However, herbivores may express leaf-age preferences by adopting different patterns of feeding behaviour, which may have direct fitness consequences. We therefore investigate three patterns of behaviour.

  • 1Unconstrained feeding: the ‘null model’. The herbivore selects at random with no regard for the age of the leaf. All leaves are grazed in equal proportion.
  • 2Constrained feeding: the herbivore selects leaf age at random, but is constrained by toxicity and must stop when the toxin dose exceeds Dmax. Grazing is sustained in proportion to the Imax curve.
  • 3Selective feeding: the herbivore selects only those leaves from which it will gain the most nutritional benefit. Grazing is restricted to those leaves that give a value of Imax within 10% of its optimum.

The impact of grazing on leaf value has been assessed by calculating VRL for the idealized leaf in Fig. 3, assuming that the herbivore is moderately tolerant (Π = 1), and at two levels of grazing severity (10 and 50% removal of leaf nitrogen). The results are described in Fig. 6. Constrained feeding has only slightly less serious consequences than the null model of unconstrained feeding, and the distribution of toxins does not therefore appear to offer a very effective conservation strategy. However, when the herbivore feeds selectively, the reduction in VRL is substantially less. Thus when herbivory is mild, the impact of grazing is relatively inconsequential to the growth capacity of the plant, and even when severe, the limitations imposed by toxicity still allow the leaf to recover more than twice the value that it would if the herbivore was permitted to feed at random.

Figure 6.

Effects of herbivory on the relative lifetime value of a leaf, VRL (taken to be a measure of plant fitness). Of the three patterns of herbivore feeding behaviour that have been investigated, constrained and selective are based on the predicted schedule of herbivore (c) in Fig. 3. Two levels of grazing are shown: (a) mild, 10% removal of nitrogen; (b) severe, 50% removal of N.

The impact of selective herbivores varies with leaf-age preference (Fig. 7). The loss of value is virtually negligible while Π is <0·7, but as it increases, grazing becomes very damaging. When Π is between 2 and 3, grazing is concentrated on leaves at their peak of importance to the plant. In comparison, a constrained herbivore has a moderately damaging impact, irrespective of tolerance.

Figure 7.

Impact of herbivory on plant fitness as a function of herbivore leaf-age preference, Π, at a moderate grazing intensity (30% of nitrogen removed). A measure of fitness is taken to be the relative lifetime assimilative value of a leaf (VRL). Two patterns of herbivore feeding behaviour are displayed: (a) constrained; (b) selective.

In the modelled example, ≈18% of VRL, the relative lifetime value, resides in poorly defended foliage (mature or older). This remains important enough for herbivory to have a significant impact. However, the strategy could be made more effective by skewing the ‘value function’ of the leaf further to the right, so that more of VRL is attributed to the young, well defended stages. Expressed in biological rather than mathematical terms, this corresponds to reinvesting a greater proportion of assimilate into new foliage, thereby ensuring that the best protected leaves play a larger role in perpetuating the growth of the plant. The degree of reinvestment (R) can be assessed as follows:

image(eqn 5 )

When the photosynthetic output is standardized to equal 1, this quotient is proportional to 1/r. In Fig. 8, it can be seen that the proportion of VRL locked in younger leaves varies with R, although cannot be shifted much above 85% (only marginally better than the 82% found in the example). There is clearly little advantage to skewing the value function further in this direction, and there must be strong physiological constraints on the extent of such shifts due to competing demands on new photosynthate, which ensure that only a modest amount can be channelled back into new leaves. The efficiency of the defence declines rapidly when R is small as leaf value is shifted in the opposite direction, towards the poorly protected older leaves. Thus the parameters chosen for this example are close to the optimum for the plant.

Figure 8.

Relationship between a plant's investment in leaf growth and the skewness of assimilative value towards younger leaves. The degree of investment in growth is denoted by R, and is inversely proportional to the intrinsic growth constant. Skewness is denoted by the proportion of assimilative value contributed by young, defended leaves (<30% of the life span). ×, value of R used in all previous analyses.


It is commonly believed that a primary role of acute toxins is to defend their parent plant against herbivory. A certain proportion of photosynthate is invested into the production of these chemicals, and it is important that they should be distributed efficiently, providing maximum defence at a given expenditure. But what constitutes an efficient distribution? This paper suggests how optimal efficiency may be achieved.

In support of the model, Lambdon & Hassall (2001) estimated that the concentration range spanned by toxins in young and old leaves is approximately of the correct order to encompass Dmax thresholds, while Giamoustaris & Mithen (1995) found strong differences in the deterrence of pest species across a comparable range. Provided the T : N ratio remains skewed in favour of the young leaves, the findings are robust to most realistic changes in the growth curves. However, a number of potential limitations must be highlighted:

  • 1In some plant species the chemical signature may vary with leaf age, making the overall interaction more complex.
  • 2Hartnett & Bazzaz's (1984) schedule of ‘assimilative value’ may overestimate the importance of young foliage. Lower leaves are often responsible for the export of carbon to the roots (Meyer & Montgomery 1987), so some retention of them is probably essential. Also, the ability to reallocate nutrients can provide a degree of plasticity (van Dam et al. 1996), so that the loss of a young leaf can be compensated by redirecting further growth to older laminae. In shrubs it has been estimated that juvenile defoliation becomes a serious problem only when more than 40% grazing occurs (Meyer & Montgomery 1987).
  • 3Patterns of damage must be taken into account: diffuse grazing on several leaves may be less harmful than intense defoliation of a given age class (Mauricio, Bowers & Bazzaz 1993).
  • 4The impediment to photosynthesis may be greater than that predicted merely by tissue removal, as herbivory may reduce efficiency over a wide area around the feeding site (Zangerl et al. 2002).
  • 5It is widely known that plants undergo considerable physiological changes in response to grazing which may include enhanced production of inducible chemical defences and mobilization of soluble nutrients (Koritsas, Lewis & Fenwick 1991; Agrawal 1998). These changes would profoundly alter the N and T profiles described in the model example.

In response to some of these limitations, we refer to the Brassica plants studied, where the glucosinolate profile changed very little over time (Lambdon et al. 2003). It is also known that concentrations of most secondary chemicals other than glucosinolates (S-methyl cysteine sulphoxide, phenolics, green leaf volatiles other than hexanols) occur at much lower concentrations and therefore should not affect the overall toxicity values (Lambdon 1998). Similarly, nutritional value may not depend solely on nitrogen, but this is a good index of many other nutrients as the leaves age, and is almost certainly the single most important factor (White 1993).

Ecological implications

It is widely accepted that many folivorous herbivores show specific leaf-age preferences which, in fast-growing plants, often appear to be related to the breadth of the feeding niche. Monophagous or oligophagous species tend to favour younger foliage (Hoy & Shelton 1987; Thomas 1987) than polyphagous ones (Meyer & Montgomery 1987; Woodman & Fernandes 1991). For example, the weevil Neochetina eichhorniae feeds predominantly on the immature foliage of its host plant, water hyacinth Eichhornia crassipes, whereas the polyphagous Arctiid caterpillar Splilosoma virginica suffers high mortality when reared on the same tissues (Center & Wright 1991). Specialist feeders opt to maximize nutrient intake from the young leaves (Crawley 1989; Renwick 1989; DeKogel, van der Hoek & Mollema 1997), which are rich in nitrogen and other nutrients (Center & Wright 1991; Pate, True & Kuo 1991; Karlsson 1994; Boar 1996). Such a pattern is explained adequately by the current model. Species that are very tolerant of toxins (high Dmax) may strongly prefer the young foliage and could potentially be very damaging. Such high levels of tolerance are most likely to occur in oligophagous herbivores (Erickson & Feeny 1974), which often develop specific resistance mechanisms to the prevalent toxin (Lambdon & Hassall 2001).

Although specialists are more likely to prefer young leaves, there are a number of exceptions to this rule (e.g. Martel & Kause 2002). The model offers an explanation: such behaviour may require physiological adaptations associated with tolerance, which could entail substantial metabolic costs (Slansky 1992; Devonshire et al. 1998; McCaffery 1998). If so, the trait would be valuable only where it created a real survival advantage. Particularly, if interspecific competition for the resource is high, then a shift to unexploited age classes may be advantageous. It may also enable an increase in growth rate by enhancing nitrogen consumption, which may be particularly important when the herbivore has a short life cycle which it must complete rapidly.

When under such pressures to maximize intake rates, a ‘selective’ feeding strategy may be adopted whereby only leaves of the optimum age are taken. If the herbivore is only moderately tolerant, this behaviour is beneficial to the plant, which retains a high proportion of its assimilative value and therefore suffers minimal impact on fitness. In Brassica, it has indeed been shown that heavy grazing on mature leaves has little effect on the eventual yield of the crop (McKay et al. 1993), and Ehrlén (1995) found that damage to mature leaves was much less costly than damage to meristems in the everlasting pea, Lathyrus vernus. To generalists sampling an unfamiliar food source where the potency of its toxins is not known, opting for mature leaves may minimize the risk of delayed illness. Specialists feeding on young leaves may unavoidably be much more damaging.

Where relatively little photosynthate is reinvested in growth, a much higher proportion of value is attributable to the vulnerable mature leaves (Fig. 8). Consequently, fast growth is necessary for this pattern of toxin distribution to confer any efficient protective effect. As severe attack on young leaves may still be very costly, this form of defence represents a high-risk strategy. Plants employing it are often difficult to locate, short-lived and highly fecund (Feeny 1976), and may depend partially on these traits to minimize the extent and impact of catastrophic grazing events.

It would appear that, on a theoretical level at least, the T : N ratio model could help explain the relationship between fast-growing plants and their herbivores. We hope that future studies may consider this possibility when interpreting ecological data on feeding preferences.


Thanks to the School of Environmental Sciences, University of East Anglia, for a small maintenance grant during the preparation of this work. Also, thanks to David Lister, Ruth Nightingale, David E. Oatway, Mike Palmer, Ruth Hassall, Ruth Magrath, Jenny Stephenson and Penny Turner for their contributions to the original field work on which the study was based, and to an anonymous referee for valuable comments on the manuscript.