Behavioral Ecology Vol. 14 No. 2: 236-245
© 2003 International Society for Behavioral Ecology
Linking foraging behavior to lifetime reproductive success for an insect parasitoid: adaptation to host distributions
a Laboratory of Entomology, Wageningen University, PO Box 8031, 6700 EH Wageningen, The Netherlands b Biometris, Department of Mathematical and Statistical Methods, Wageningen University, PO Box 100, 6700 AC Wageningen, The Netherlands
Address correspondence to M. Vos, who is now at the Netherlands Institute of Ecology, NIOO-KNAW, CL, Department of Food Web Studies, Rijksstraatweg 6, 3631 AC Nieuwersluis, The Netherlands. E-mail: m.vos{at}nioo.knaw.nl.
Received 10 October 2001; revised 10 June 2002; accepted 29 June 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAppendixReferences |
|---|
European and American populations of the parasitoid Cotesia glomerata show pronounced differences in foraging behavior across plants and leaves. This variation in spatial aspects of foraging behavior was observed about 350 generations after the introduction of C. glomerata from Europe to North America. We used a simulation model to study how these behavioral differences affect lifetime reproductive success in environments that differ in the spatial distribution of hosts. The preferred gregarious host Pieris brassicae occurs in rare large clusters in Europe but is absent in North America. The solitary caterpillars of Pieris rapae are negative binomially distributed across plants during summer in North America, whereas they are Poisson-distributed in Europe, and early and late in the season in North America. Simulations showed that the foraging strategy of American C. glomerata resulted in a higher lifetime reproductive success than did the strategy of European C. glomerata on a Poisson P. rapae distribution, but did not differ on the more clustered negative binomial distribution. American parasitoids spend less time on exploration flights, focusing on the exploitation of P. rapae patches. This suggests that C. glomerata has adapted to the North American environment through the loss of exploration traits necessary for the location of rare clusters of P. brassicae. Lifetime reproductive success of the European strategy was most sensitive to an increase in the giving up time on infested leaves. This behavioral parameter was more than twice as high in the American parasitoids compared with their European conspecifics.
Key words: fitness, hierarchical patches, information use, patch choice, patch leaving, spatial foraging behavior.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAppendixReferences |
|---|
A central assumption in foraging theory is that strategies for the exploitation of patchily distributed resources are closely related to the reproductive success of foraging animals. However, few studies have actually established the link between foraging strategies and lifetime reproductive success. Reproductive success may depend critically on how well a foraging strategy is tuned to the distribution of resources across spatial scales. Many parasitoid species exploit herbivorous insect larvae in plant canopies. These host larvae can occur in hierarchical patches, consisting of plants and leaves. Foraging parasitoids have to make arrival and leaving decisions at both of these scales. Most studies of patch exploitation define a patch at only a single spatial scale and focus on patch-leaving decisions (Charnov, 1976
; Driessen and Bernstein, 1999; Driessen et al., 1995; Haccou et al., 1991; Hemerik et al., 1993; Iwasa et al., 1981; Rodríguez-Gironés and Vásquez, 1997; Vos et al., 1998; Wajnberg et al., 1999, 2000). Arrival decisions and the hierarchical nature of patches are often ignored. However, these are important as they will affect parasitoid travel times, which are a key factor in determining the theoretical optimum for foraging decisions in a certain environment.
In a hierarchically structured environment, parasitoids can choose to travel within or between plants, which implies a choice in the type and magnitude of travel times (Vos, 2001). In optimal foraging models, a performance function is often used that maximizes the long-term rate of offspring production (E), given a certain giving up-rule:
|
|
; also see Driessen and Bernstein, 1999; Stephens and Krebs, 1986). Natural selection is assumed to optimize decisions on Tsearch as a function of Ttravel, where Ttravel is supposed to be imposed by the environment (Iwasa et al., 1981). However, under a hierarchical patch structure, the actual optimization problem does not only involve decisions on Tsearch, depending on Ttravel. Instead, the problem involves the simultaneous optimization of decisions on both search times and travel times. These decisions should depend on the distributions of hosts across plants and leaves, which do impose minimum travel times within and between plants, but also on the costs (energy, time, and wear) and possible benefits (information acquisition) of longer than minimum travel times.
Previous experiments have shown pronounced differences in decisions on travel times and search times between European and American populations of the parasitoid Cotesia glomerata. European parasitoids typically hover and hop across many plants and leaves and have short patch residence times, whereas the American parasitoids fly much less and search more persistently within infested plants (Vos, 2001). C. glomerata travel times are longer between plants than within plants in European parasitoids, but are not different in the American population. European parasitoids tend to hover across several plants before choosing to land and, thus, have longer travel times between plants than do the American parasitoids.
The European wasps have a higher tendency to return to previously visited infested leaves (Vos, 2001). In many theoretical studies (see Charnov, 1976; Iwasa et al., 1981; Rodríguez-Gironés and Vásquez, 1997), the assumption is made that revisits to patches do not occur and that foragers make one leaving decision, that is, final, on each patch. However, parasitoids of our European population of C. glomerata revisit host-infested leaves up to four times, and those of our North American population up to two times, after visits to other patches. This indicates that patch-exploitation can be a process of multistage decision making, with early patch decisions being adjusted during later visits (Vos, 2001). The possibly adaptive value of revisits and decisions to travel within or between plants is intimately tied to the spatial distribution of hosts. Host distribution does impose a certain amount of travel time and will thus affect the time budget of a parasitoid. This will in turn affect how natural selection acts on life-history decisions such as investments in eggs and survival, and affect the proportions of time limitation and egg limitation in a parasitoid population. It is important to know whether a parasitoid population is mostly time-limited: many theoretical studies assume time limitation, which implies that natural selection acts on traits affecting parasitism rates. The distribution of Pieris caterpillars can vary considerably with geographical area, season, and the particular species involved. Parasitoids have to deal with this variation.
Our general aim is to relate the variation in foraging behavior between American and European parasitoids to variation in lifetime reproductive success. We investigate the performance of both foraging strategies on the different host distributions occurring in North America and Europe. Furthermore, we check how lifetime reproductive success changes when we vary the amount of information a parasitoid can have on its environment.
The system
The parasitoid C. glomerata is indigenous to Europe and was introduced into North America from Europe in 1884, about 350 C. glomerata generations ago. In Europe, C. glomerata parasitizes larvae of Pieris brassicae and Pieris rapae. Parasitized larvae are fed on from the inside by immature parasitoids. Successful parasitoid development causes hosts to die and adult parasitoids to emerge. The repertoire of spatial foraging behavior in European C. glomerata seems adapted to the highly gregarious larvae of P. brassicae and is much less efficient on solitarily feeding P. rapae larvae (Wiskerke and Vet, 1994). These parasitoids often leave a patch before finding P. rapae, and waste time with continued searching after having parasitized the single host on a patch. P. brassicae clusters are rare in most years (Pak et al., 1989). A cluster of 7–150 larvae occurs on about one in every 200–300 Brassica plants (calculation based on Castricum L, van Loon JJA, Vos M, unpublished; Pak et al., 1989). This implies that a P. brassicae cluster could be present on about one in every 20–30 Pieris-infested plants (see below). P. rapae occurs at low densities in Europe, with averages around 0.1 P. rapae per plant (Karamaouna F, Vos M, unpublished; Pak et al., 1989), and peaks in averages reaching 1.6 per plant (Pak et al., 1989).
C. glomerata attacks P. rapae in North America, as P. brassicae is absent there. P. rapae occurs at low densities early and late in the season in North America (van Nouhuys S, personal communication). The distribution of P. rapae across plants can appropriately be described with a Poisson distribution at densities below two per plant (Harcourt, 1961). In the middle of the season, North American P. rapae populations reach high densities, having a negative binomial distribution, with means of 3–16 larvae of instar one to three (L1–3) per plant (Harcourt, 1961). Densities on leaves range from one to five (van Nouhuys S, personal communication), versus a typical density of one L1–3 larva per leaf when P. rapae population densities are low (Karamaouna F, Vos M, unpublished). North American C. glomerata have been suggested to have evolved a higher attack rate than that of European C. glomerata on P. rapae (le Masurier and Waage, 1993).
Specific aims and questions
First, we use a simulation model to evaluate the lifetime reproductive success resulting from alternative foraging strategies in two geographically distinct parasitoid populations. The model is based on data from foraging experiments in multiplant environments with European and North American C. glomerata parasitoids and literature/field data on host distributions in Europe and North America. Our model explicitly incorporates the hierarchical structure of patches and parasitoid exploitation patterns within such structure. We specifically ask the following: (1) how is lifetime reproductive success in parasitoids from Europe and North America affected by the distribution of P. rapae hosts across plants and leaves; (2) is the foraging strategy of the American population more successful in the North American environment than the European strategy about 350 generations after the introduction of C. glomerata to North America; and (3) what is the proportion of time-limited animals in European and American parasitoid populations under Poisson and negative binomial host distributions.
Second, we use this model to study parasitoid lifetime reproductive success under different ecological scenarios. In these scenarios, we ask questions on decisions in parasitoids and in hosts. We vary the following: (1) the amount of information available to the parasitoids, (2) host decisions on how to distribute their offspring, and (3) the egg load available to female parasitoids. We specifically ask the following: (4) would the ability to perceive and use information on host infestation levels on plants and/or leaves in the local area lead to an increase in lifetime reproductive success (this is particularly important for the American population that has to deal with a heterogeneous host distribution in the middle of the season); (5) would such an ability have a higher pay-off if it spanned a wider local environment; (6) would the host's decision to always have single eggs on leaves reduce rates of parasitism by North American parasitoids; (7) to which aspects of parasitoid behavior is lifetime reproductive success most sensitive; and (8) how would variation in fecundity further differentiate the foraging success of North American C. glomerata on a Poisson and a negative binomial distribution of hosts?
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAppendixReferences |
|---|
We constructed a model for simulating the foraging behavior of an individual C. glomerata female, during her entire lifetime. She forages in a field of 3025 Brassica oleracea plants (55 plants long and 55 plants wide) and makes patch arrival and leaving decisions both at the spatial scale of plants and leaves. To answer each of our questions, we varied certain characteristics of the basic model: (1) the foraging behavior characteristic for each strain (American or European), (2) host distribution, (3) use of information on host densities, and (4) fecundity (egg load). The behavior of the parasitoid is stochastic. We simulated 100 parasitoid lifetimes for each ecological scenario.
Experimental data
We used the results of two experiments (Vos, 2001) to parameterize the foraging model. We used results of parasitoids from two different populations: one from North America, (Geneva, New York, USA), and one from Europe (Wageningen, The Netherlands). Both strains had been in the laboratory for several generations before being used in the experiments, thus purging maternal effects that might be owing to the original environment. Parasitoids of both strains had been reared on P. rapae, under identical climatic conditions. We will call the Wageningen population "European" and the Geneva population "American," for convenience, and to emphasize that these populations have geographical origins that differ in ecological conditions. In the first experiment, individual C. glomerata females were observed while foraging on a single leaf on a Brassica oleracea plant with host densities of zero, one, or eight P. rapae larvae (N = 22 for each treatment). In the experimental background, four Pieris-infested plants provided alternative patches to go to, in case the parasitoid decided to leave the current patch. This experiment provided data on (1) giving up times on empty leaves (GUTempty), (2) giving up times on infested leaves (GUTinfested), and (3) intervals between ovipositions (IBOs). These are either the time until the first oviposition on a patch, or the time between ovipositions.
In the second experiment, we allowed individual C. glomerata females to forage in a multipatch environment consisting of six B. oleracea plants for 45–60 min. Three of these plants were infested with four P. rapae larvae. The infested plants had two infested leaves on different sides of the plant (e.g., leaves 8 and 11 for a plant with 18 leaves, counting from the ground). Each infested leaf contained two larvae. These P. rapae larvae fed solitarily, thereby causing different feeding damage sites on each infested leaf. We used the same strains as in the first experiment (N = 22 for each parasitoid strain). The multipatch experiment provided data on (1) travel times between plants and within plants, (2) the probabilities to travel between versus within plants, (3) probabilities to fly to empty versus infested plants, (4) probabilities to fly to empty versus infested leaves, (5) the probabilities of revisiting infested leaves, and (6) the same type of data as the single patch experiment (see above).
Literature data
Cass (1960) and Harcourt (1961, 1962a,b) provide extensive data on population densities and distributions of P. rapae in North America. Cass (1960) found averages of 1.9 and 4.3 larvae per cabbage plant, in 2 years, in untreated plots. Harcourt found on average 3–16 (L1–3) P. rapae larvae per cabbage plant, over a 5-year period, in insecticide-free plots, (Harcourt, 1961, 1962a,b). The spatial distribution of hosts across plants can appropriately be described with a Poisson distribution (Equation 2) if random and with a negative binomial distribution (Equation 3) when clumped:
|
|
|
|
. For the variance-mean relationship of a negative binomial distribution, the expectation of the variance is s2 = m + m2/k. Harcourt's estimate of k is 2.91 for the first three instars of P. rapae, which are the most suitable host stages for C. glomerata. Based on Harcourt's (1961) variance-means data (see Harcourt's Figures 1 and 2), we have chosen an average of m = 5 and k = 2.91 as the values we used in our simulations, to create a negative binomial distribution of larvae across plants. The variance-mean relationship of a Poisson distribution gives the expectation of the variance as s2 = m. We estimated the mean density on a plant, as well as the variance across plants, to be about 0.1 for the European distribution, based on the data of Pak et al. (1989), and our own 1997 field data (Karamaouna F, Vos M, unpublished). Harcourt (1961) showed that P. rapae's distribution agrees well with a Poisson distribution at such low densities. We used a mean and variance of 0.1 L1–3 P. rapae per plant to simulate a European field distribution or an American distribution early or late in the season.
The lifetime foraging model
We modeled a realistic spatial environment consisting of plants and leaves, and the behavior of parasitoids within such structure, including revisits to patches and traveling within and between plants. This environment consisted of a field of 55 x 55 plants with a either a Poisson distribution or a negative binomial distribution of P. rapae larvae across plants. Under the Poisson distribution plants contained zero, one, or two hosts. Individual leaves contained zero or one host. Under the negative binomial distribution plants contained 0–13 hosts per plant. Individual infested leaves contained one to five larvae (NBmultiple). Especially to answer question 2), whether a (negative binomial) host distribution with only a single host on each leaf would affect parasitoid lifetime reproductive success, we also simulated the same negative binomial distribution of hosts across plants, where only a single larva per infested leaf was present (NBsingle).
We compared the European and American foraging strategies in environments where only P. rapae is present (as did le Masurier and Waage, 1993). This facilitated a straightforward comparison of the efficiency of both strategies on P. rapae, but note that part of especially the European foraging strategy may be tuned to exploiting the potential presence of P. brassicae in the environment (see discussion). Our simulations start with releasing a parasitoid on the plant in the middle of the field. This parasitoid has either the behavioral strategy of the American strain, or that of the European strain. The foraging parasitoid exploits the plants and leaves in the field subject to (1) a probability of intraplant travel when on an infested plant, (2) a probability of intraplant travel when on an empty plant, and (3) a probability to choose between neighboring infested versus empty plants, either landing randomly with respect to host density on infested plants, or with a probability proportional to host density. The parasitoids can choose to land on one of the eight surrounding plants (in a 3 x 3 plant block), or one of the 24 surrounding plants (in a 5 x 5 plant block), thus allowing parasitoids to make choices at different spatial scales. Parasitoids are also subject to (4) a probability to choose between infested and empty leaves, landing either randomly with respect to host density on infested leaves or with a probability proportional to the density on infested leaves.
Thus, parasitoids can either have no information on host densities, just partial information (only at the plant or leaf level), or "complete" information, (both at the plant and leaf level) on their local environment. In the most simple "default" situation, the parasitoids have no information on surrounding plant and leaf damage levels. They only discriminate between "infested" and "clean," and make landing choices among the eight surrounding plants.
On empty leaves, a parasitoid cannot encounter hosts: it simply "draws" a giving-up time (GUTempty), subsequently draws a travel time and leaves. On infested leaves, a parasitoid may encounter hosts if it searches a sufficient amount of time. To determine whether the parasitoid encounters a host or leaves before such an encounter, both a GUTinfested and IBO are drawn. If the IBO is shorter than the GUTinfested, an oviposition occurs, with a clutch size of 20 eggs (Vos, 2001). This drawing of GUTs and IBOs continues until a GUTinfested is drawn that is shorter than the drawn IBO, or until all hosts on the patch are parasitized, upon which the parasitoid leaves. We have not included superparasitism in our model, although this does admittedly occur in C. glomerata. However, it is relatively rare and involves much less time and eggs than ovipositions in unparasitized hosts. Each parasitoid is allowed to forage an entire lifetime, drawn from an exponential distribution with an average of 86,400 s, (i.e., three foraging days of 8 h; the parasitoids live on average 3 days in a cage in the field [Geervliet, 1997] and are most active between about 0800 h and 1600 h, Vos M, personal observation). We simulated parasitoids with 500, 1000, 1400, 1500, 2000, 2500, or 3000 eggs. C. glomerata female fecundity varies between 500 and 2200 eggs (Laing and Levin, 1982; le Masurier and Waage, 1993; Moiseeva, 1976; Shapiro, 1976; Tagawa, 2000). Our default parasitoid has an egg complement of 1400 eggs. We do not incorporate a trade-off between fecundity and longevity in our simulations, as the necessary experimental data are not available. However, we will show graphs on lifetime reproductive success for the full range of life spans and fecundities.
The above steps are summarized in a flow diagram (Figure 1) that shows the behavioral cycle of the parasitoid. Finally, we show all used parameter values in Table 1, and discuss these in the Appendix. By using these parameters, our simulations provided results on emerging parameters like numbers of ovipositions and visits to infested plants and leaves that were satisfactorily similar to our experimental results (Table 2). Note that both the real behavior in the experiment and the simulations are stochastic. The above check is not a validation but shows we modeled what we intended to model.
|
|
|
For ease of reading we will use the terms parasitoids and strains when presenting the outcome of our simulations. These terms refer to the particular foraging strategies we modeled for the North American and European C. glomerata parasitoids. We will use the terms reproductive success and fitness for the number of eggs laid during the entire lifetime of parasitoids in our simulations.
Statistical analysis
We used N = 100 replicates for each ecological scenario. The results from different ecological scenarios or parasitoid populations were compared using a Kruskal-Wallis test (nonparametric ANOVA). The Kruskal-Wallis test statistic H is distributed approximately as 
2[a–1] df, with a being the number of samples). Subsequently, nonparametric multiple comparisons were performed (Siegel and Castellan, 1988). We used the Mann-Whitney U test (MWU, with large-sample test statistic Z) for several comparisons of two samples. If samples were used in several (n) comparisons, we adjusted the p level to.05/n (Bonferroni correction). We chose to use nonparametric tests, as our data on lifetime reproductive success could not be transformed to approach a normal distribution.
Sensitivity analysis
We tested the sensitivity of lifetime reproductive success to an increase or decrease of 10% in the value of each behavioral parameter. We did this for both the European and the American foraging strategy, in a Poisson environment. The default behavioral parameter values were as in Table 1, whereas fecundity was 1400 eggs. Animals used no information on host densities from a distance and chose to land among the eight surrounding plants. A single set of simulations consisted of 100 parasitoid lifetimes. Sets with (identical) default parameter values were replicated 10 times, for both the American and European foraging strategy, to check how much variation in lifetime reproductive success we can expect between sets owing to stochasticity alone. Sensitivity was analyzed for 10 behavioral parameters. The sets for an increase or decrease of 10% in a single parameter value were compared with the default set, using a two-sided MWU test and a p level of.025. Each default set had been designated to a particular behavioral parameter prior to the simulations. The sensitivity analysis entailed 60 sets of 100 simulations.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAppendixReferences |
|---|
European and American fitness on different host distributions
Host distribution significantly affected lifetime reproductive success for both the European and American C. glomerata foraging strategies (Kruskal-Wallis test, H = 23.2, df = a - 1 = 2, p = 9.0 x 10–6 and H = 20.1, df = a – 1 = 2, p = 4.2 x 10–5).
European parasitoids laid on average 876 eggs in the negative binomial environment with multiple larvae per leaf, 731 eggs in the negative binomial environment with single hosts on leaves, and 523 eggs in the Poisson environment (Figure 2, left graph). Multiple comparisons showed significant fitness differences between the NBmultiple and the Poisson environment, and between the NBsingle and the Poisson environment. There was no significant difference in reproductive success between the NBmultiple and NBsingle environments.
|
American parasitoids laid on average 950 eggs in the negative binomial environment with multiple larvae per leaf, 770 eggs in the negative binomial environment with single hosts on leaves, and 694 eggs in the Poisson environment (Figure 2, right graph). Multiple comparisons revealed that lifetime reproductive success for the American strain was significantly different between the NBmultiple and NBsingle, and between the NBmultiple and Poisson environments. There was no significant difference in reproductive success under the NBsingle and Poisson host distributions.
The American strategy resulted in a higher fitness than the European strategy on the Poisson distribution (MWU, Z = 2.54, p =.011). Reproductive success did not differ between parasitoid strains on the negative binomial distribution (NBmultiple; MWU, Z = 1.46, p =.14).
Incomplete information
We considered several ecological scenarios in which parasitoids either had (and used) no olfactory information on feeding damage levels in their local environment; only partial information, focusing on either plants or leaves; or complete olfactory information on the local environment, both at the level of plants and leaves. The "informed" parasitoids used cues on host density to preferentially land on higher-density patches. In our simulations, "uninformed" parasitoids of both populations landed on plants with an average density of 5.3 hosts and on leaves with on average 2.9 hosts in the negative binomial environment (NBmultiple). Parasitoids visited plants with an average density of 7.0 hosts when informed on plant damage and visited leaves with on average 3.3 hosts when informed on leaf damage. An improved informational state did not result in any significant difference in lifetime reproductive success, neither for the European, nor for the American strain, on any of the three host distributions (Kruskal-Wallis test, all p >.05).
In these simulations, parasitoids perceived the eight surrounding plants as the local environment. We also allowed American parasitoids to perceive 24 instead of eight surrounding plants (a 5 x 5 local environment instead of 3 x 3), both under a Poisson and a negative binomial distribution. This did not lead to a significant change in lifetime reproductive success (MWU, Z = 0.55, p =.58 and MWU, Z = 0.67, p =.50, respectively).
Lifetime reproductive success: variability and sensitivity to behavior
Lifetime reproductive success varied considerably among the 10 default sets of 100 simulations of a parasitoid lifetime, owing to the stochasticity of behavioral decisions. The number of eggs laid by American parasitoids ranged from 548–679, with an average of 630, in these 10 sets of simulations. The European parasitoids laid 438–555 eggs, with an average of 503. Most of the simulations with an increase or decrease in a single behavioral parameter value of 10% fell within the above ranges and did not differ significantly from the default simulation set.
American parasitoids showed a trend for a higher fitness (759 eggs) when the giving up time on infested leaves was decreased with 10 % (MWU, Z = 1.98, p =.0475), as well as a trend for a higher fitness (683 eggs) when the probability to go to infested leaves was increased (MWU, Z = 2.13, p =.0328). European parasitoids had a significantly higher fitness (592 eggs) when the giving up time on infested leaves was increased (MWU, Z = 2.80, p =.0051), and a significantly lower fitness (424 eggs) when the probability to revisit infested leaves was decreased (MWU, Z = 2.34, p =.019).
Time-limited or egg-limited?
Host distribution clearly affected the proportions of time-limited and egg-limited parasitoids in the American and European populations. Of the American parasitoids, 83% were time-limited in the Poisson environment, 50% in the negative binomial environment. In comparison, a somewhat larger percentage of the European parasitoids tended to be time-limited. Of the European strain, 91% was time-limited in the Poisson environment, 61% in the negative binomial environment. These percentages hold for animals with a fecundity of 1400 eggs. Figure 3 shows how the percentage of time-limitation increased with fecundity in the American parasitoid population, in the Poisson and negative binomial (NBmultiple) environments (see Discussion).
|
The parasitoids of both strains made more visits to empty than to (rare) infested leaves in the Poisson environment. The European strategy led on average to 194 visits to empty leaves versus 84 visits (including revisits) to infested leaves during a parasitoid lifetime (MWU, Z = 5.25, p = 1.5 x 10–7). European parasitoids visited on average 60 different plants, of which up to 30 were infested with P. rapae. The American strategy led on average to 105 visits to empty leaves versus 61 visits to infested leaves (including revisits), during a lifetime (MWU, Z = 4.84, p = 1.3 x 10–6), in the Poisson environment. The American parasitoids visited less empty leaves than did their European conspecifics, in the Poisson environment (MWU, Z = 2.80, p =.0050). In the negative binomial environment, the parasitoids of both strains visited more infested than empty leaves. The European strategy led on average to 113 visits to infested leaves and 82 visits to empty leaves during a parasitoid lifetime (MWU, Z = 3.49, p =.00048). The American strategy resulted in means of 50 visits to infested leaves and 23 visits to empty leaves during a lifetime (MWU, Z = 7.02, p = 2.2 x 10–12).
Variation in life span and fecundity
Figure 4 shows parasitoid reproductive success in a negative binomial (NBmultiple) and a Poisson environment depending on life span and fecundity, for the American strategy. The increase in reproductive success with life span is about two times steeper in the negative binomial than in the Poisson environment, for all fecundities. The parasitoids that became egg-limited did so at a younger age in the negative binomial environment. Note that the horizontal series of data points in Figure 4 represent individuals that were egg-limited at the end of their lives, and that the bundle of data points with a positive slope represents individuals that were time-limited at the end of their lives.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAppendixReferences |
|---|
Alternative foraging strategies in C. glomerata: adaptive behavioral differentiation?
The foraging strategy of the North American parasitoid population resulted in a higher lifetime reproductive success than the European strategy on a Poisson distribution of hosts across plants. North American C. glomerata have to deal with such a host distribution during two periods each year: early on, when the population is building up from the first generation of hosts, and later, during the last host generation that will produce overwintering parasitoids. The North American foraging strategy will result in a higher fitness than will the European strategy in the North American environment, where the solitarily feeding host P. rapae is present and the gregariously feeding host P. brassicae is absent. It would in fact be more efficient in Europe as well, with its Poisson distribution of P. rapae, if P. brassicae was never present there. Possibly, the European C. glomerata foraging strategy is more tuned to investing time and energy in exploration flights in order to increase the chance of encountering rare P. brassicae clusters in the European environment (Vos, 2001
). This exploration behavior comes at a cost of encountering less P. rapae hosts than would be possible under more exploitative foraging behavior. The European C. glomerata strategy may be a compromise between conflicting demands on the behavioral repertoire: Time can be invested in (inefficient) searching on P. rapae–infested leaves (see Wiskerke and Vet, 1994) or on traveling to locate P. brassicae. The parasitoids may risk to miss a high-quality cluster of P. brassicae when focusing on the exploitation of P. rapae, and may risk having no offspring at all when focusing entirely on rare P. brassicae clusters.
Host acceptance in C. glomerata reflects this compromise: European parasitoids readily accept P. rapae after development in this host, but often reject it after development in P. brassicae, then only readily accepting P. brassicae as a host (Vos, 2001). European C. glomerata readily accept P. brassicae, even when P. rapae was the natal host. Interestingly, the American population of C. glomerata seems to have lost this trait. It significantly less often accepts P. brassicae (Vos, 2001). When comparing the behavior of European and American C. glomerata, quite a few traits that are useful for C. glomerata when foraging for P. brassicae (see Wiskerke and Vet, 1994) are absent or less frequent in the North American foraging strategy (Vos, 2001). Behavioral differentiation of populations seems to occur more frequently through the loss of (parts of) a behavioral repertoire than through the genesis of novel behavior patterns (Foster, 1999). For example, in a study of more than 25 populations of sticklebacks, all adaptive behavioral differentiation appears to have occurred by loss of ancestral behavior patterns or by shifts in the frequency of their expression (Foster, 1999). A specific pattern in ecotypic differentiation by loss of behavior patterns was repeatedly observed, in 10 populations in geographically disparate lakes (Foster, 1999). We think that the absence of P. brassicae in North America may have resulted in resolving part of a conflict within the generalist behavioral repertoire of C. glomerata. The current American spatial foraging strategy is certainly more adaptive in this environment. On the Poisson distribution of both early and late host generations, its performance is superior to that of the European strategy. The differences between the European and American strains may reflect genetic differentiation owing to 350 generations of selection against traits for dealing with a highly clustered host, in the American environment. However, we cannot formally exclude maternal effects, although these may have been purged in the laboratory (see Materials and Methods), nor can we formally exclude potential infections with microorganisms as a cause of the differences between strains (see Hopper et al., 1993). Crosses could provide a reliable method to determine the genetic contribution to behavioral variation between strains (Hopper et al., 1993).
Very little is known about behavioral variation among C. glomerata strains within Europe or North America. A North American C. glomerata strain from Amherst, Massachusetts, USA, showed flight behavior that was highly similar to that of our New York strain (Vos, 2001), but this strain was less active within patches. Van Nouhuys and Via (1999) showed behavioral differentiation between New York C. glomerata strains from wild and cultivated habitats, but could not show local adaptation to these different environments. To our knowledge no behavioral comparison has been made among European C. glomerata populations.
Host distribution and adaptation, under incomplete information
Previous experiments in a semi-field set-up showed that C. glomerata did not change their time allocation to an empty or low-density P. rapae patch, when an alternative high-quality patch with a P. brassicae cluster was available at a distance of only 60 cm (Vos, 2001). This result indicates that patch-leaving decisions are made locally and are not affected by information from nearby patches. It also suggests that the typical hovering across plants and leaves in European C. glomerata is essential for the location of P. brassicae. In our simulations in a Poisson environment, European C. glomerata visited 60 different plants in a lifetime. Of these plants, up to 30 were Pieris-infested. About 10% of the European parasitoids visited more than 20 different Pieris-infested plants and would thus have a high probability of encountering a P. brassicae cluster during a lifetime.
However, C. glomerata are not only attracted to Pieris-infested Brassica plants in the field. Plant volatiles induced by nonhost herbivores confuse these parasitoids and attract them to plants and leaves without hosts (Geervliet et al., 1996; Vos et al., 2001). Although the parasitoids do not discriminate between host and nonhost-infested plants from a distance, they clearly do discriminate between these patch types once they have landed on a leaf, and adjust their patch times accordingly (Vos et al., 2001). This plant-mediated indirect effect of nonhosts still costs considerable amounts of time, especially when nonhosts are abundant (Vos et al., 2001). In a field situation, in which multiple herbivore-plant complexes will be the norm, the distribution of nonhosts may be as important for lifetime reproductive success as is the distribution of hosts.
C. glomerata does not use information on the concentration of kairomones in a patch to adjust its leaving tendency (Vos et al., 1998). However, the parasitoids might be able to use olfactory information on feeding damage while it is in flight and, thus, preferentially land on patches with high damage levels (see Geervliet et al., 1998a, shown for P. brassicae not for P. rapae). Our simulations did not show a fitness advantage of a preference for landing on relatively high-density patches. This could easily come as a surprise, if one associates high host densities with high patch quality, high within patch efficiency, and shorter times spent traveling than in a scenario with visits to infested patches of random density. However, C. glomerata exploits patches in a way that results in similar efficiencies on densities of one to four hosts on a leaf (Vos et al., 1998). Furthermore, travel times between leaves are relatively short, and intervals between ovipositions are relatively long. This results in similar efficiencies for scenarios with and without use of information on host densities.
The presence of the clustered high-quality host P. brassicae may be the main uncertainty in the life of a European C. glomerata parasitoid. If a female C. glomerata encounters this host early in life, she may learn to specialize on it (Geervliet et al., 1998b; but see Vos, 2001) and use her entire egg complement on one or a few clusters of these larvae. This is likely to occur when P. brassicae is present at peak densities as, for example, in 1982 (see Pak et al., 1989). In most years, only a minority of C. glomerata is likely to encounter a single cluster of P. brassicae, probably after spending quite a lot of time on P. rapae (or even nonhost) patches. Of course, the above uncertainty does not play a role in North America because P. brassicae is consistently absent there.
Distribution as a host strategy
Our simulations showed that lifetime reproductive success in North American C. glomerata would be lower in a negative binomial environment, with single P. rapae larvae per leaf, that is, if adult P. rapae butterflies decided to always only lay single eggs on clean leaves. Under the P. rapae distribution that would result from such behavior, the risk of parasitism would be lower for individual larvae. In fact, P. rapae butterflies do tend to lay single eggs on leaves, on the leeward side of plants (Harcourt, 1961). However, egg-laying decisions may be constrained by flight ability, especially in a windy environment (see Harcourt, 1962b). Butterflies may experience only a few leaves on each plant as suitable for landing, and thus, separate females will tend to lay single eggs on the same leaves. This can result in a situation in which plants may have only one or two infested leaves, each containing up to five larvae (van Nouhuys S, personal communication). P. rapae will face higher rates of parasitism in such a distribution.
Time-limited or egg-limited?
The European strategy will lead to a high proportion of time-limited parasitoids in an environment with a Poisson distribution of P. rapae hosts. The majority of female parasitoids will have eggs in their ovaries at the end of their lives. However, in the European environment, some parasitoids will have a chance to find P. brassicae. They could easily lay their remaining egg complement in a single cluster of this host and thus achieve a relatively high contribution of offspring to the next generation (for parallel parasitoid systems, see Driessen and Hemerik 1992; Ellers 1998; Ellers et al. 2000). As mentioned above, time limitation may depend not only on the temporal and spatial distribution of the host species P. rapae and P. brassicae but also on the distribution of nonhost herbivores that cause C. glomerata to waste time. The American situation is interesting in that host generations follow an alternation of Poisson and negative binomial distributions. Parasitoids are selected to have (relatively) many eggs in the negative binomial environment, but their offspring will emerge in a Poisson environment where it would have paid to invest in longevity. Such fluctuating selection pressures in subsequent generations may lead to the evolution of a trade-off that is not optimal in any of the two environments.
It would be interesting to compare American C. glomerata from northern and southern populations. In more southern areas, a longer growing season supports more generations with a negative binomial distribution per year (e.g., six generations occur around Colombia, Missouri, USA; http://eap.mcgill.ca/CPC_1.htm). In southern populations, selection may lead to higher investments in fecundity than in northern populations.
General conclusions
The spatial distribution of hosts has a significant effect on parasitoid lifetime reproductive success. Natural selection may lead to striking geographic variation in foraging behavior on different host distributions, within a period of 350 parasitoid generations. Constraints on the informational state of individual parasitoids are unimportant with respect to some aspects of environmental variation, but do matter with respect to other environmental characteristics. We showed that the foraging strategy of the parasitoids was robust to a lack of information on host densities, and to the fact that information is only locally available. A lack of information on the presence of the preferred host P. brassicae in Europe may be much more important. The European strategy seems a compromise in which exploration efforts to locate P. brassicae cause reduced parasitism rates of P. rapae. American C. glomerata do not face this dilemma and seem to have adapted to the American P. rapae environment by spending their time and energy on exploitation and not on the costly exploration behavior that is so characteristic of their European conspecifics.
|
Appendix |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAppendixReferences |
|---|
C. glomerata parasitoids adjust their tendency to stay searching in a patch with each oviposition they have (Vos, et al. 1998
). This means that realized patch residence times and efficiencies depend on the intrinsic tendencies of parasitoids to encounter hosts (1/IBO) and leave (1/GUT). In the model, the occurrence of ovipositions during a patch visit depends on draws from two variables that represent these intrinsic tendencies. For convenience, we call these drawn times the intrinsic IBO and the intrinsic GUT. We calculated the intrinsic IBOs and GUTs from the realized IBOs and GUTs that we observed in our experiments. Note that the realized experimental values are not identical to these underlying variables. For example, if the average intrinsic GUT were small relative to the underlying IBO, the IBOs observed in the experiment would only represent the leftmost tail of the distribution of IBOs (because parasitoids mostly leave a patch before the underlying IBO has a chance to become realized). Below we provide a technical description of our method to determine these underlying processes that will result in realized intervals between ovipositions and giving up times.
Realized IBOs and GUTs result from two underlying random processes with different mean times of occurrences. These processes are assumed to be generating theoretical values for IBOs and GUTs after exponential distributions with parameter 
1 and 2 respectively. For a stochastic variable, say z, following an exponential distribution with parameter , the mean and variance are 1/ and (1/)2 (see Equation A1 for the cumulative distribution function):
|
|
X) and E(Y|X Y). It can easily be shown that both these conditional expectations have the same value. The starting integral, some intermediate results and the expected value are given for the realized IBOs in Equation A2:
|
|
1 + 2). Before we were able to estimate the mean values for the theoretical distributions 1/1 and 1/2, it was necessary to look at the fraction of the observations that is a realized giving up time. The complementary fraction is the fraction of realized intervals between ovipositions. In Equation A3, it can be seen that the fraction of realized ovipositions in the data equals 1/(1 + 2). Therefore, we estimated 1/1 as the quotient of the observed mean time for both GUTs and IBOs and the fraction of realized ovipositions:
|
|
|
|
ACKNOWLEDGEMENTS |
|---|
We would like to thank Louise Vet, Bart Nolet, Joop van Lenteren, and our anonymous referees for perceptive comments and useful suggestions.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONAppendixReferences |
|---|
Cass LM, 1960. (No title). Proc Entomol Soc Ont 91:49-52.
Charnov EL, 1976. Optimal foraging, the marginal value theorem. Theor Popul Biol 9:129-136.[CrossRef][ISI][Medline]
Driessen G, Bernstein C, 1999. Patch departure mechanisms and optimal host exploitation in an insect parasitoid. J Anim Ecol 68:445-459.[CrossRef]
Driessen G, Bernstein C, van Alphen JJM, Kacelnik A, 1995. A count-down mechanism for host search in the parasitoid Venturia canescens. J Anim Ecol 64:117-125.[CrossRef]
Driessen G, Hemerik L, 1992. The time and egg budget of Leptopilina clavipes, a parasitoid of larval Drosophila. Ecol Entomol 17:17-27.
Ellers J, 1998. Life-history evolution in the parasitoid Asobara tabida: on the trade-off between reproduction and survival (PhD dissertation). Leiden, The Netherlands: Rijksuniversiteit Leiden.
Ellers J, Sevenster JG, Driessen G, 2000. Egg load evolution in parasitoids. Am Nat 156:650-665.[CrossRef]
Foster SA, 1999. The geography of behavior: an evolutionary perspective. Trends Ecol Evol 14:190-195.[CrossRef][Medline]
Geervliet JBF, 1997. Infochemical use by insect parasitoids in a tritrophic context. Comparison of a generalist and a specialist (PhD dissertation). Wageningen, The Netherlands: Wageningen University.
Geervliet JBF, Ariëns S, Dicke M, Vet LEM, 1998a. Long distance assessment of patch profitability through volatile infochemicals by the parasitoids Cotesia glomerata and C. rubecula (Hymenoptera: Braconidae). Biol Contr 11:113-121.
Geervliet JBF, Vet LEM, Dicke M, 1996. Innate responses of the parasitoids Cotesia glomerata and C. rubecula (Hymenoptera: Braconidae) to volatiles from different plant-herbivore complexes. J Insect Behav 9:525-538.
Geervliet JBF, Vreugdenhil AI, Dicke M, Vet LEM, 1998b. Learning to discriminate between infochemicals from different plant-host complexes by the parasitoids Cotesia glomerata and C. rubecula (Hymenoptera: Braconidae). Ent Exp Appl 86:241-252.[CrossRef]
Haccou P, Vlas SJ de, van Alphen JJM, Visser ME, 1991. Information processing by foragers: effects of intra-patch experience on the leaving tendency of Leptopilina heterotoma. J Anim Ecol 60:93-106.[CrossRef]
Harcourt DG, 1961. Spatial pattern of the imported cabbageworm, Pieris rapae (L.) (Lepidoptera: Pieridae), on cultivated Cruciferae. Can Entomol 93:945-952.
Harcourt DG, 1962a. Biology of caterpillars in eastern Ontario. Proc Entomol Soc Ont 93:61-75.
Harcourt DG, 1962b. Design of a sampling plan for studies on the population dynamics of the imported cabbageworm, Pieris rapae (L.) (Lepidoptera: Pieridae). Can Entomol 94:849-859.
Hemerik L, Driessen G, Haccou P, 1993. Effects of intra-patch experiences on patch time, search time and searching efficiency of the parasitoid Leptopilina clavipes. J Anim Ecol 62:33-44.[CrossRef]
Hopper KR, Roush RT, Powell W, 1993. Management of genetics of biological control introductions. Ann Rev Entomol 38:27-51.
Iwasa Y, Higashi M, Yamamura N, 1981. Prey distribution as a factor determining the choice of optimal foraging strategy. Am Nat 117:710-723.[CrossRef]
Laing JE, Levin DB, 1982. A review of the biology and a bibliography of Apanteles glomeratus (L.) (Hymenoptera: Braconidae). Biocontr News Inform 3:7-23.
le Masurier AD, Waage JK, 1993. A comparison of attack rates in a native and an introduced population of the parasitoid Cotesia glomerata. Biocontr Sci Techn 3:467-474.
Moiseeva TS, 1976. The protective reaction of insects. Zaschita Rastenii 8:21-22.
Pak GA, van Heiningen TG, van Alebeek FAN, Hassan SA, van Lenteren JC, 1989. Experimental inundative releases of different strains of the egg parasite Trichogramma in Brussels sprouts. Neth J Plant Pathol 95:129-142.
Rodríguez-Gironés MA, Vásquez RA, 1997. Density-dependent patch exploitation and acquisition of environmental information. Theor Popul Biol 52:32-42.[Medline]
Shapiro VA, 1976. Apantele: a parasite of the cabbage white butterfly. Zaschita Rastenii 10:17-18.
Siegel S, Castellan NJ, Jr, 1988. Nonparametric statistics for the behavioral sciences. New York: McGraw-Hill.
Stephens DW, Krebs JR, 1986. Foraging theory. Princeton,. New Jersey: Princeton University Press.
Tagawa J, 2000. Sex allocation and clutch size in the gregarious larval endoparasitoid wasp, Cotesia glomerata. Ent Exp Appl 97:193-202.[CrossRef]
Van Nouhuys S, Via S, 1999. Natural selection and genetic differentiation of behaviors between parasitoids from wild and cultivated habitats. Heredity 83:127-137.
Vos M, 2001. Foraging under incomplete information: parasitoid behavior and community dynamics (PhD dissertation). Wageningen, The Netherlands: Wageningen University.
Vos M, Hemerik L, Vet LEM, 1998. Patch exploitation by the parasitoids Cotesia rubecula and Cotesia glomerata in multi-patch environments with different host distributions. J Anim Ecol 67:774-783.[CrossRef]
Vos M, Moreno Berrocal S, Karamaouna F, Hemerik L, Vet LEM, 2001. Plant-mediated indirect effects and the persistence of parasitoid-herbivore communities. Ecol Lett 4:38-45.[CrossRef]
Wajnberg E, Fauvergue X, Pons O, 2000. Patch leaving decision rules and the marginal value theorem: an experimental analysis and a simulation model. Behav Ecol 11:577-586.
Wajnberg E, Rosi MC, Colazza S, 1999. Genetic variation in patch time allocation in a parasitic wasp. J Anim Ecol 68:121-133.[CrossRef]
Wiskerke JSC, Vet LEM, 1994. Foraging for solitarily and gregariously feeding caterpillars: a comparison of two related parasitoid species (Hymenoptera: Braconidae). J Insect Behav 7:585-603.
CiteULike 
Connotea 
Del.icio.us What's this?
| |||||||||||||||||||||||||||||||||||||||||||