Host tolerance does not impose selection on natural enemies


  • Etzel Garrido Espinosa,

    1. Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, Ap. Postal 70-275, CP 04510, México Distrito Federal, México
    Search for more papers by this author
  • Juan Fornoni

    1. Departamento de Ecología Evolutiva, Instituto de Ecología, Universidad Nacional Autónoma de México, Ap. Postal 70-275, CP 04510, México Distrito Federal, México
    Search for more papers by this author

Author for correspondence: Juan Fornoni Tel: +52 55 5622 9005 Fax: +52 55 5616 1976 Email:


  • • Coevolution between hosts and their natural enemies is believed to operate through the evolution of resistance traits. Although the importance of tolerance to natural enemies as an alternative defensive strategy has been recognized, there is still no consensus about the possible role of host tolerance in the evolutionary outcome of the interaction.
  • • Here, using bioassay experiments, we tested the hypothesis that variation in host tolerance among selected plant genotypes could impose a selection pressure upon a specialist herbivore.
  • • Tolerance did not affect herbivore larvae survival, weight gain, efficiency of food consumption, total food consumption, developmental time and adult mass. These results therefore do not support the hypothesis that host tolerance could affect natural enemy performance. However, resistance did negatively affect herbivore larva survival. Genetic variation in herbivore larva survival was detected, thus suggesting the potential for a coevolutionary response.
  • • Our results indicate that host tolerance would reduce opportunities for a coevolutionary response by the natural enemies of the host. Contrary to predictions from previous models, our results suggest that host tolerance may constitute an evolutionarily unstable defensive strategy.


Coevolutionary theory applied to antagonistic interactions (plant–herbivore, plant–pathogen, predator–prey or host–parasite) has been developed under the assumption that interacting species exert reciprocal negative genetic effects (Janzen, 1980; Futuyma, 1998). Occurrence of coevolutionary responses among species requires two conditions to be fulfilled: (1) each species involved in an interaction must affect the performance of the other species with which it interacts, and (2) genetic variation in those traits involved in the interaction must exist for a response to natural selection to occur. The validity of these conditions has been well supported by empirical work indicating that hosts have evolved resistance traits that negatively affect the performance of their natural enemies (Rausher, 1996; Agrawal, 2001). These resistance traits were shown to constitute the selective pressure upon enemies to evolve counter-resistance traits (Berenbaum & Zangerl, 1998; Geffeney et al., 2002; Ratza et al., 2002; Thrall & Burdon, 2003; Allen et al., 2004).

More recently, there has been recognition that hosts also defend themselves through mechanisms of tolerance that reduce or buffer fitness losses after damage (Strauss & Agrawal, 1999; Jokela et al., 2000; Stinchcombe & Rausher, 2002; Fornoni et al., 2003a, 2004b) or infection (Roy & Kirchner, 2000; Hansen & Koella, 2003; Restif & Koella, 2003, 2004). Thus, while host resistance reduces the probability of being damaged or infected, host tolerance reduces the negative impact of damage on fitness (i.e. reduces the cost of damage or virulence). Although several studies have advocated the presence of fitness benefits and costs of tolerance (Stowe et al., 2000; Koskela et al., 2002; Stinchcombe & Rausher, 2002; Stinchcombe, 2002a; Fornoni et al., 2004b), no study has determined whether the defensive strategy of tolerance results in a different coevolutionary response from the response expected to be produced by resistance (discussed in Stinchcombe, 2002b).

Theoretical models have recently incorporated the evolution of both tolerance and resistance as alternative defensive strategies. These studies assume that, unlike resistance, tolerance does not exert negative effects upon enemy development and performance (Roy & Kirchner, 2000; Tiffin, 2000; Stinchcombe, 2002b; Fornoni et al., 2004a). In other words, tolerance may not constitute a selective pressure upon natural enemies (Stinchcombe, 2002b). In contrast, the first attempt to model a coevolutionary dynamics with tolerance and resistance as host defense mechanisms predicted that tolerance could produce a different pattern of evolution of enemy traits from that expected for resistance (Restif & Koella, 2003). Therefore, that tolerance can constitute a selective pressure on natural enemies is still an untested assumption of previous theoretical models (Tiffin, 2000; Stinchcombe, 2002b; Restif & Koella, 2003; Fornoni et al., 2004a). If higher levels of tolerance correspond to higher levels of host quality (i.e. more nitrogen content associated with compensatory photosynthesis in plants) (Stinchcombe, 2002b; but see Gassmann, 2004), tolerant hosts may select for higher levels of infection or consumption among natural enemies (Restif & Koella, 2003). Under this condition, an association between tolerance and the enemy traits involved in the interaction must exist (see Stinchcombe, 2002b). Although there is evidence of a negative correlation between resistance level and damage (reviewed in Marquis, 1992), no study has ever determined whether an association between tolerance and consumption by natural enemies exists (Stinchcombe, 2002b).

In this study, we tested the hypothesis that host tolerance can impose a selective pressure on the natural enemies of the host. Specifically, we used bioassay experiments with a plant–herbivore system involving the annual plant Datura stramonium and its specialist leaf beetle Lema trilineata to determine whether plant tolerance directly influences herbivore larva survival, weight gain, efficiency of food consumption, total food consumption, developmental time and adult mass. Finally, the presence of genetic variation and correlations among herbivore traits was also examined in order to explore the potential for an evolutionary response by the herbivore.

Materials and Methods

Datura stramonium L. (Solanaceae) is the host of the folivorous leaf beetle Lema trilineata (Olivier) (Chrysomelidae) from Mexico to Canada. All the larval stages of this herbivore occur on the leaf tissue of the host plant, and can consume almost 100% of individual plants (J. Núñez-Farfán, UNAM, Mexico City, Mexico, pers. comm.). Herbivore damage can reduce plant fitness (Núñez-Farfán & Dirzo, 1994). D. stramonium contains tropane alkaloids and foliar trichomes that function as components of resistance against herbivory (Shonle & Bergelson, 2000; Valverde et al., 2001). Plant material used in the present study was gathered from a population of D. stramonium in Central Mexico (18°N, 99°W), for which the existence of additive genetic variation and genotypic selection acting on tolerance and resistance to folivorous herbivores had previously been demonstrated under natural field conditions (Fornoni et al., 2003b, 2004b). From this data set, two groups of host lines were selected, each represented by four genotypes (full-sibs). These two groups showed significant differences in tolerance (F7,51 = 5.55; P < 0.0001) but similar levels of resistance (F7,59 = 0.98; P = 0.4480), as estimated from damage under natural field conditions (Fornoni et al., 2003b). For the purpose of the present study, our choice of host lines reduced by approximately 50 times the variation in resistance relative to that in tolerance (CVresistance = 2.48% and CVtolerance = 104.68%). This manipulation ensured the absence of a correlation between tolerance and resistance (r = –0.1228; P = 0.7932) and increased the power to detect an effect of tolerance. Using this plant genetic material, two bioassay experiments were performed during 2002–2003 under laboratory conditions. Thirty seeds from each host line were sown in the glasshouse of the Ecology Institute [Universidad Nacional Autónoma de México (UNAM), Mexico City, Mexico] to obtain six to eight adult plants per line planted in 4-l pots filled with potting soil. This procedure was repeated for each of the bioassay experiments described below.

Experiment 1

This experiment was designed to determine whether tolerance background (low- vs high-tolerance host lines) affects herbivore larva survival, weight gain, efficiency of food consumption, total food consumption, developmental time, and adult mass. Also, we wished to determine whether this effect relies on the expression of a plastic (after damage) or a constitutive (before damage) component. Because of the late germination rate, only seven host lines were used in this experiment. Before flowering, all the leaves of half of the plant replicates from each host line were artificially defoliated to 50%, following the methodology used by Fornoni & Núñez-Farfán (2000). After the phenostage at which the defoliation treatment was applied, subsequent leaves produced by defoliated and nondefoliated plants were used to feed an experimental population of the herbivore L. trilineata. This phenostage (flowering stage) corresponds to the time at which L. trilineata starts consuming its host plant in the field. Although lines had a priori reduced variation in resistance, this defensive trait was also measured using the reciprocal of the proportional amount of damage inflicted by individual larvae upon each host line. The experimental population of L. trilineata was composed of a sample of 621 individuals collected from a population located in the El Pedregal de San Angel Preserve in the Valley of Mexico (19°N, 99°W). Larvae were taken to the laboratory and reared at 25°C with a 12 : 12 h photoperiod. Each larva was reared in a 250-cm3 plastic pot filled with 80 cm3 of sterilized soil and covered with a mesh. Pots were watered (10 ml) every 2 d until the start of the pupal stage. Each larva was fed using one of the plant replicates of each host line. Fresh leaf squares of 5 cm2 in each pot were replaced every 2 d and stored for estimation of the amount of leaf tissue consumed by the larvae. This area of leaf tissue was similar in terms of mass among host lines, as no significant differences were previously detected in specific leaf weight (F7,395 = 1.63; P = 0.1228). The experiment was continued for 24 d, until all adults had emerged.

During the experiment, the following variables were measured: herbivore larva survival (days to death), weight gain, efficiency of food consumption, total food consumption, developmental time and adult mass. Weight gain (g d−1) was estimated as the proportional increment in mass between the second and fourth larval stages relative to the number of days between the two larval stages. This time interval corresponds to the larval stages with the highest rate of consumption (EGE & JF, unpublished data). Because of the absence of differences among host lines in specific leaf weight (see above), the efficiency of food consumption (g cm−2) was estimated as the ratio of weight gain relative to the amount of consumed leaf area. After 3 wk at 25°C, dry leaf area consumed was calculated by adding together the leaf areas consumed from all the 5-cm2 squares of leaf tissue given to each larva. For each leaf tissue square, leaf area consumed was measured using Digital Image Analysis Systems (WinDias Basic; Delta-T Devices Ltd, Cambridge, UK). Total food consumption (cm2) was estimated by adding together the total leaf areas consumed by each larva. Developmental time corresponds to the number of days the larva takes to reach the adult stage. Adults were weighted (g) as an estimate of size at maturity. Leaf area consumed during the first larval stage was used to estimate host line resistance to the experimental population of the herbivore. Host line resistance was estimated as the average of the proportion of leaf area consumed by each larva.

As we used fewer plants than the number of larva replicates assigned to each combination of tolerance level and defoliation treatment, the results were analyzed as a split plot design following the model: herbivore performance = tolerance + defoliation treatment (tolerance) + resistance + error (Crawley, 1993, pp. 51–52). Survival analysis was performed following the Cox regression model (Cox, 1972). For the other response variables, analyses were performed with the analysis of variance (ANOVA) (type III SS) option in proc glm (SAS, 1999). Weight gain was square root transformed, developmental time was transformed as the inverse of the square root, and efficiency of food consumption was log-transformed to improve normality. The ANOVA for efficiency of food consumption was performed without including resistance as a covariable as damage was used to estimate the dependent variable. For all the variables except survival, the total number of larvae included in the analysis was reduced because of mortality.

Experiment 2

This experiment was designed to estimate the presence of genetic variation in the herbivore population for the same traits as measured in experiment 1. For this experiment, a similar set of host plant replicates to that used in experiment 1 were grown in the glasshouse and used to feed herbivores from an experimental population of 709 larvae (31 maternal half-sib families × 22.87 ± 4.20 eggs per family) obtained from the same site as used previously. Larvae were randomly assigned to one of 20 blocks within the laboratory and maintained under the same conditions as described above. The experiment was continued for 33 d, until all adults emerged. Except for the analysis of genetic variation in survival, larval mortality reduced to 15 the number of families that had sufficient numbers of replicates (8–18 larvae per family) for the analysis of genetic variation on continuous and meristic variables. For these analyses, herbivore genotype was considered a random factor, and resistance was included as a covariable. The analyses of the continuous and meristic variables were performed with the ANOVA (type III SS) option in proc glm (SAS, 1999). Phenotypic and genotypic correlations were estimated among all pairs of variables using the Pearson correlation coefficient in jmp (SAS, 1995). Only the variables that showed genetic variation were included in the correlation analysis. Mean herbivore larva survival was only included in the estimation of genotypic correlations.


The findings of the present study indicate that tolerance and resistance exert different effects upon herbivore larva survival (Table 1). No evidence of an effect of tolerance (before and after defoliation) on herbivore larva survival was detected (Table 1). Mean survival time (± standard error) was 8.15 ± 0.58 and 9.85 ± 0.55 d for insects grown on low- and high-tolerance hosts, respectively. The survival analysis had a power of 0.90 (holding α = 0.05) to detect a difference in mean survival time greater than 1.7 d between levels of tolerance (Collet, 2003, p. 300). Despite the previous absence of differences in resistance among selected host lines at their site of origin, significant differences in resistance were detected with our experimental herbivore population (F6,614 = 10.82; P < 0.0001). After the bioassay experiment, the estimation of plant resistance increased the coefficient of variation in resistance from 2.4 to 8.4%. Surprisingly, this small amount of variation was negatively associated with herbivore larva survival (r = –0.96; P = 0.0003, N = 7) (Fig. 1). Tolerance level, defoliation treatment and resistance did not affect the other herbivore characters measured [degrees of freedom are 1, 2, 1 and 122 for tolerance, defoliation (tolerance), resistance and error source of variation, respectively; all F < 3.1022; all P > 0.0870]. Power analyses indicated that, if the true sizes of the treatment effects are as small as those estimated in this experiment, we would have needed sample sizes 30–2968 times larger (depending on the trait) to obtain statistical significance (holding α = 0.05 and 1-β = 0.80). Given that our experiment was contrived to enhance any possible effect of tolerance, our nonsignificant results for insect survival, growth and performance give reasonable confidence that the corresponding differences are likely to be very small.

Table 1.  Results of the χ2 survival analysis using the proportional hazard model
 d.f.χ2 likelihood-ratioP
  1. N = 621.

  2. d.f., degrees of freedom.

Tolerance1  0.002  0.9599
Defoliation (tolerance)2  2.658  0.2647
Resistance1104.107< 0.0001
Figure 1.

Negative correlation between plant resistance and herbivore larva survival. +, Plant families with high tolerance levels; –, plant families with low tolerance levels. No evidence of a negative correlation between resistance and tolerance was detected (r = 0.49; P = 0.2562).

Genetic variation was detected for herbivore larva survival (inline image = 64.72; P = 0.0002), developmental time (F14,122 = 2.80; P = 0.0234) and adult mass (F14,122 = 2.36; P = 0.0453). Correlation analyses among herbivore characters revealed a positive phenotypic correlation between adult mass and developmental time (r = 0.25; P = 0.0006). No genetic correlations between variables were detected. Although this result suggests that selection imposed by host resistance would not affect the evolution of other characters besides survival, it should be treated with caution given the sample size (N = 15) available for these analyses.


The results obtained in the present study support the hypothesis proposed by Stinchcombe (2002b) that host tolerance could relax the selective pressure on natural enemies. Specifically, plant tolerance did not affect herbivore larva survival, weight gain, efficiency of food consumption, total food consumption, developmental time and adult mass. As these herbivore traits are usually affected by plant quality (Scriber & Slansky, 1981; Weilbull, 1987; Taylor, 1989; Moran, 1992; Wheeler & Halpern, 1999; Awmack & Leather, 2002), our results suggest that tolerance mechanisms may not necessarily be related to plant quality (Stinchcombe, 2002b). Instead, a negative correlation between plant resistance and herbivore larva survival was detected. In addition, the presence of genetic variation in survival among herbivore families suggests the potential for a coevolutionary response.

Two points should be considered in the interpretation of the absence of an effect of tolerance on herbivore performance. First, a genotype × environment interaction in the phenotypic expression of tolerance could have reduced the differences in tolerance among selected host lines, increasing the probability of detecting a nonsignificant effect. Hence, further effort should be devoted to examining the possible existence of an effect of tolerance on herbivore performance in the field. Secondly, as we were not able to determine whether genetic variation in the plasticity of herbivore traits existed across tolerance levels, we cannot rule out the possibility that host tolerance could affect herbivore performance. If genetic variation in plasticity of insect traits is present, host tolerance would select an evolutionary change in the herbivore.

Our results suggest that, for the D. stramoniumL. trilineata system, the host defensive strategy is expected to have contrasting effects on herbivores and hence on the coevolutionary process; while resistance would promote a coevolutionary response, tolerance would not. Although this idea has been previously proposed (Rosenthal & Kotanen, 1994; Rausher, 2001; Stinchcombe, 2002b; Fornoni et al., 2004b), its evolutionary implications have attracted little attention.

Jokela et al.'s (2000) model predicted that the adaptive value of tolerance would increase when natural enemies become locally adapted to the level of resistance of their host population (i.e. when the host receives an increasing amount of damage). Their study described the conditions under which a tolerant mutant could invade a population but did not explore whether tolerance could be evolutionarily stable. Other models have suggested that the conditions for invasion and fixation of tolerant mutants depend on the relative values of the costs and benefits of tolerance (Tiffin, 2000; Fornoni et al., 2004b). Therefore, previous models considered tolerance and resistance as alternative strategies in terms of fitness and also assumed that tolerance levels were proportional to the amount of damage. Based on these assumptions, these models predict that host tolerance would be an evolutionarily stable defensive strategy. While some evidence supports the expectation that tolerance and resistance could function as alternative mechanisms of defense (van der Meijden et al., 1988; Valverde et al., 2003), there is no empirical evidence supporting the assumption that host tolerance could be a linear function of the amount of damage.

Our results suggest that, if tolerance does not affect enemy consumption negatively, an increase in the enemy load would possibly increase the amount of damage, thus reducing the host capacity for tolerance. Although the exact shape of the relation between tolerance and damage has not been deeply examined, it is reasonable to expect that under low levels of damage tolerance would increase (Hutha et al., 2003; see del-Val & Crawley, 2005). As damage increases, tolerance will finally reach its maximum and any further increase in the amount of damage will reduce the benefits of tolerance because of internal/external constrains. Recent studies have indicated that hosts probably face limits on their maximum tolerance because of resource limitation (Fornoni et al., 2004a) and/or physiological and morphological constraints (Hochwender et al., 2000). Thus, the adaptive value of tolerance may be related to variation in the enemy population size. This possible association could explain temporal and spatial fluctuations in selection of host tolerance and the presence of intermediate levels of tolerance observed in natural populations.


K. Boege, C. Cordero, A. Córdoba-Aguilar, C. Dominguez, C. Martorel, R. Miller, J. Núñez-Farfán, M. Vallejo and two anonymous reviewers made useful suggestions that improved the manuscript. Financial support was provided by PAPIIT IN 226305-3 and CONACYT 42031/A-1. A scholarship from DGEP (UNAM) and CONACyT to EGE is acknowledged.