Carlos Bernstein, Université Claude Bernard-Lyon 1, Biométrie et Biologie Evolutive UMR 5558, 43 Bd du 11 Novembre 1918, 69622 Villeurbanne Cedex, France. E-mail: email@example.com
Abstract 1. In two different treatments, groups of healthy hosts (Ephestia kuehniella) or hosts parasitised by Venturia canescens competed for a limited amount of food. The larva to adult survival in each group, as a function of the initial number of hosts and treatment, was fitted to the generalised Beverton and Holt and generalised Ricker survival functions, and a number of life-history traits of the parasitoids was measured.
2. Intraspecific competition was scramble-like, and the parasitised hosts were less susceptible to competition than were their healthy counterparts.
3. For both the healthy and the parasitised hosts, the number of larvae surviving to adulthood gave a good fit to both the generalised Beverton and Holt and generalised Ricker models, but the values of all the parameters differed between the two treatments.
4. Parasitoid size, egg load, and adult survival time decreased significantly with the initial host number.
5. Previous theoretical work suggests that both lower susceptibility to competition by parasitised hosts and scramble competition contribute to the dynamical instability of host–parasitoid systems. Changes registered in life-history traits may also affect host–parasitoid dynamics. These changes have not yet been incorporated into host–parasitoid models.
The study of how hosts and their parasitoids influence each other's life-history traits has received an increasing amount of attention in recent years (see bibliography below). On the one hand, unravelling these reciprocal influences is important for understanding host–parasitoid relationships. On the other hand, changes in life-history traits could have an influence on host–parasitoid dynamics.
Hosts and parasitoids form complex biological entities marked by reciprocal influences, and still more so when the parasitoids are koinobionts (i.e. parasitoids that allow hosts to continue to grow in size after parasitism). The conditions of development of the hosts may affect those of the parasitoids. For instance, intraspecific competition between hosts for food and other resources may influence host size. Large hosts may provide more resources for parasitoid larvae, and thus enhance the future parasitoid reproductive success (Godfray, 1994). For example, in the parasitoid Venturia canescens (Gravenhorst) (Hymenoptera: Ichneumonidae), the host's body mass influences parasitoid size (Harvey et al., 1994; Harvey & Thompson, 1995; Harvey & Vet, 1997) and development time (Harvey & Vet, 1997). Parasitoid size is assumed to affect reproductive success through variations in fecundity, life expectancy, survival, and searching efficiency (Salt, 1940; King, 1987; Bai et al., 1992; Honek, 1993; Visser, 1994; Kazmer & Luck, 1995; West et al., 1996).
Parasitism has been shown to alter host physiology (Jones & Lewis, 1971; Lawrence, 1986, 1988; Slansky, 1986; Hegazi et al., 1988) and life-history traits, due to the injection of different chemical factors when female parasitoids attack the hosts (Jones & Lewis, 1971; Dover & Vinson, 1990; Strand & Dover, 1991), or secretions produced by larval parasitoids (Schopf, 1984; Lawrence, 1986). For instance, parasitism has been shown to reduce the rate of weight gain (Strand et al., 1988; Ohnuma & Kainoh, 1992; Harvey, 1996), to affect development time (Smilowitz & Iwantsch, 1973; Harvey, 1996; Harvey et al., 1996), and to alter host behaviour (Reed et al., 1996). Parasitoid attacks on an insect population result in a mixture of healthy and parasitised hosts that compete for the same resources. Therefore, given that host death before parasitoid emergence entails the death of the parasitoid, intraspecific host competition also involves indirect intraspecific competition between parasitoid larvae and indirect host–parasitoid interspecific competition. Density-dependent mortality of parasitised hosts then implies parasitoid mortality, which in turn depends on the density of the host. Differences in the effect of intraspecific competition on unparasitised and parasitised hosts may be a common feature of host–parasitoid systems. An example is given by Prévost (1985) and Wajnberg et al. (1985), who showed that, while the mortality of unparasitised Drosophila melanogaster Meigen (Diptera: Drosophilidae) larvae is directly density-dependent, the mortality of larvae parasitised by Leptopilina boulardi Barbotin, Carton & Kelner-Pillaut (Hymenoptera: Cynipidae) is inversely density-dependent.
These results suggest that intraspecific host competition may influence the life-history traits of parasitoids, with important consequences for the dynamics of the system. This, in turn, may affect the strength of intraspecific competition between hosts. Such influences can only be unravelled by combining experimental and theoretical work.
The work reported here had two aims: (1) To study the strength of intraspecific competition and possible differences in susceptibility to competition between unparasitised and parasitised hosts, in the system formed by Venturia canescens and one of its hosts, the pyralid moth Ephestia kuehniella (Zeller) (Lepidoptera: Pyralidae). (2) To study the effects of intraspecific host competition on parasitoid fitness at different host densities. This was estimated by the effect of host density on parasitoid survival to adulthood, adult size, egg load, and adult survival in the absence of food. These factors have not been incorporated into models or documented experimentally.
Materials and methods
Cultures and biological details
Venturia canescens is a solitary, koinobiont, larval parasitoid of pyralid moths, which is known to be either thelytokous, mainly (but not exclusively) in granaries and mills, or arrhenotokous, under field conditions ( Beukeboom et al., 1999 ). It attacks the hosts in their second to fifth instar but the development of the parasitoid larva is arrested at its first stage until the host reaches its fifth (final) instar. The fifth-instar host larva continues its development but the development of the parasitoid first reduces then halts the growth of the host larva ( Harvey, 1996 ). When the healthy host-larva development is complete, pupation occurs, followed by the emergence of adult moths. The parasitised host dies, the parasitoid pupates, and the adult wasp emerges a few days later.
The conditions of this experiment: thelytokous V. canescens, host species (Ephestia kuehniella), host environment and food (semolina), and high host densities were akin to those occurring in granaries and mills.
The parasitoids used originated from a thelytokous strain collected in 1985 in a granary near Oxford, U.K. The host came from a mass rearing of Ephestia kuehniella in Antibes (I.N.R.A.), France. The hosts were reared on semolina and maintained in a constant environment of 24 ± 1 °C, 75 ± 5% RH and LD 14:10 h. The parasitoid strain was reared under the same conditions.
In this experiment, healthy or parasitised host larvae (see below for a precise definition) competed for food. For both types, mortality during the pupal stage and at emergence was low (C. Bernstein, pers. obs.) and density-independent, which meant that the overall effect of competition on larval survival could be assessed by comparing the initial number of host larvae with the number of adults (moths or wasps) that emerged. The influence of host competition for resources on the life-history traits of the emerging parasitoids (size, egg load, adult survival in the absence of food) and on the larva-to-adult survival of parasitised and non-parasitised hosts was assessed in a design with two factors: host number and parasitism (the treatment, i.e. whether or not hosts had been subjected to parasitoid attack).
Different numbers (see below) of third-instar 21-day-old host larvae were introduced into cylindrical boxes (diameter 80 mm, height 50 mm) filled with 10 g of semolina. A circular hole in the lid (diameter 30 mm) covered by netting allowed free gas exchange with the environment.
Two (or more) boxes were prepared for each host number. In one box, the hosts were left unparasitised (healthy hosts). Those in the other box were almost all parasitised (henceforth parasitised hosts) through the introduction of 10 parasitoids into each box (see below). At the time of their introduction, the parasitoids were 1 day old (post-emergence). They were taken from the stock culture on the day of their emergence and kept in groups for 24 h in small cages with 50% diluted honey and no hosts, then placed in a box with the hosts for 3 days. This ensured that almost all the larvae were attacked (so that, on average, adult parasitoids represented 93.3 ± 6.3 SD% of the total number of adults collected in these boxes). Throughout the experiment, the boxes were kept at 25 °C, 75 ± 5% RH, and LD 14:10 h, the same conditions at which the stock cultures were maintained.
The boxes were inspected hourly from 09.00 to 18.00 hours (i.e. the photophase) to register the time of emergence of the adult hosts and parasitoids (emergences outside this period were rare). Approximately half of the emerged parasitoids (randomly chosen) were introduced individually into glass tubes (75 mm long, 10 mm in diameter) with a cotton wool plug and no food. These tubes were inspected hourly between 09.00 and 18.00 hours to register the time of death of the parasitoids (for further details, see below).
Some of the remaining parasitoids (chosen at random) were dissected and the number of mature eggs (which were easy to recognise by their shape) in the ovarioles of one of the ovaries and the corresponding oviduct, henceforth termed egg load, were counted under a microscope. The length of the left hind tibia was also measured under the microscope as an estimate of adult body size (Harvey & Vet, 1997).
Preparing boxes with high host numbers was so time consuming that this precluded preparing all the boxes on the same day or fully randomising the host numbers assigned to each particular day. The experiments were therefore carried out twice (in 1993 and 1994). In 1993 the initial host numbers were 10, 25, 50, 150, 200, 250, 300, 350, and 400; in 1994 they were 25, 50, 100, 150, 200, 250, 300, 350, 400, 450, and 500. Although host density can reach very high values in mills and granaries (C. Bernstein, pers. obs.), the highest densities used here are probably exceptional. These densities were used because they give a more accurate estimation of the shape of the larva-to-adult survival curve. One replicate was performed per host number and treatment (parasitised or healthy), with only two exceptions: in 1994, in order to increase the precision of the model for low numbers of emerging parasitoids, the boxes with 25 and 50 parasitised individuals were replicated three and two times respectively. The experimental conditions were controlled and uniform and all the boxes were prepared over a relatively short time span (15 days in 1993, 30 days in 1994). On this basis, it was presumed that the possible differences between the 2 years would be larger than those within either year. The influence of the differences between years in the different life-history traits was tested for all the relationships studied.
Larva-to-adult survival of healthy and parasitised hosts The aims of this analysis were to find out the position of intraspecific Ephestia kuehniella competition on the contest–scramble spectrum (i.e. the abruptness of density dependence; Getz, 1996), the influence of parasitism on competition between host larvae, and how well the two models found in the literature fitted the data for the number of adults emerging from larvae at different densities. The generalised Ricker model (Pomerantz et al., 1980; Getz, 1996) has already been used in the context of host–parasitoid interactions (Bernstein, 1986), but in the generalised Beverton and Holt model (Maynard Smith & Slatkin, 1973; Getz, 1996), the distinction between contest and scramble competition is more clear-cut (Bellows, 1981). It is important to compare the two models in view of their use in modelling host–parasitoid relationships.
This analysis implied the fit of the data to complex, non-linear functions. Including the effect of the year in this analysis would have meant dealing with 12-parameter models. To avoid this, the analysis was performed in two steps. First, the experimental results were fitted to both models, then the differences between years in the residuals of the regressions were tested using a Kruskal–Wallis test.
where A is the number of emerging adults, h is the initial number of hosts, α is the slope of the A–h relationship for low h values, K is the carrying capacity, and θ controls the position in the contest–scramble spectrum. θ < 1 corresponds to contest-like competition; larger values indicate more abrupt, scramble-like, density dependence. The sub-index i takes the values 0, H, or P, depending on whether a single value is used independently of the treatment or whether the values differ for healthy and parasitised hosts. See Bellows (1981) and Getz (1996) for two thorough discussions of the behaviour of these models.
The models were expressed in terms of the proportion of larvae reaching adulthood (A/h), and fitted to the observed proportions by a maximum-likelihood procedure (using the R freeware statistical package, http://cran.at.r-project.org/), assuming a binomial distribution of residuals. First the data were fitted to a maximal model (six different parameters), then to models in which one parameter at a time took the same value for both the healthy and the parasitised hosts. The contribution of the differences between parameters to the regression was tested by means of a likelihood-ratio test.
Parasitoid adult size and egg load The aim of this and the following analysis was to assess the influence of larval competition on parasitoid life-history traits, and only the data from the parasitised larvae were taken into account.
Adult parasitoid hind tibia lengths were initially fitted to the model
where length is the length of the tibia, l is the initial number of larvae, l2 allows for non-linearities in the relationship between length and the initial number of hosts, year is a factor taking into account the year when the replicates were prepared, and ‘×’ denotes the interaction. Inspection of the data suggested that the relationship of tibia length to initial host number was linear, and that fitting a third-order polynomial was unnecessary.
The number of mature eggs per ovary was fitted to a model that took into account the year of the experiment, and both the direct influence of initial number of larvae and the indirect influence of the latter through adult size:
where eggs is the number of mature eggs per ovary (other symbols as above). The regression was performed as a generalised linear model with a Poisson distribution of residuals and a log link. To calculate the contribution of each of them, the variables were introduced to the regression in the order shown in eqn 3. In 1993, poor conservation of stored samples meant that for some host numbers, these data were unavailable.
Parasitoid adult survival Parasitoid adult survival was analysed as failure time data (Kalbfleisch & Prentice, 1989). Emergence and death registered between the last observation of a day and the first observation of the following morning were considered as censored.
The data were fitted to a linear model and to a Weibull distribution using the GLIM® statistical package (Crawley, 1993). In order to proceed to the simplification of the model, and to assess the statistical significance of the different independent variables, the data were fitted to a linear model assuming a Poisson distribution of residuals and using the logarithm of the failure time, multiplied by the shape parameter of the Weibull distribution, as an offset (Crawley, 1993). For adult survival, the death rate λ(t) (other symbols as above) was fitted initially to the model
As l3 did not make any significant contribution to the regression (see below), higher power terms were ignored.
Survival of healthy and parasitised hosts to adult stage
The result of fitting the number of emerging adults (hosts or hosts and parasitoids, depending on the treatment, and pooled over the 2 years), as a function of the initial number of host larvae, to the generalised Ricker model (eqn 1a) is shown in Fig. 1. The lines in the graph show the fitted model. The values of the parameters for both models are shown in Table 1. The values of all the parameters were highly significant (t values), with those for the healthy hosts differing significantly from those for the parasitised hosts. According to the Akaike information criterion (Akaike, 1974), the generalised Ricker model fitted the data marginally better than the generalised Beverton and Holt model (AIC = 450.73 and AIC = 457.64 respectively). For both models, the θ values were larger than unity (suggesting scramble competition) and the carrying capacities were larger for the parasitised hosts (suggesting lower competition than between the healthy hosts). The residuals did not differ between the two years in either model (Kruskal–Wallis test).
Table 1. Parameter values resulting from fitting the generalised Ricker and the generalised Beverton and Holt models to the experimental data on the proportion of larvae surviving to adulthood (moths for healthy hosts, wasps and moths for hosts attacked by parasitoids). The regression was performed using a maximum-likelihood procedure, assuming a binomial distribution of residuals. The likelihood ratios correspond to comparisons between the models including different parameter values for each host type or a common value. Likelihood ratios follow a χ 2 distribution, in this case with a single degree of freedom.
Adult size The lengths of the parasitoids' hind tibias decreased significantly with increasing host number (Fig. 2). The variable l2 (alone or in interaction with year) made no significant contribution to the regression. Differences between years, both in intercept and slope, were highly significant (F1,402 = 94.19, P < 0.001; F1,402 = 15.37, P < 0.001 respectively). The slopes of the two regression lines were significantly different from zero (t = 7.35 and t = 8.39 respectively, both P < 0.001).
Egg load The overall influence (i.e. direct and indirect through the influence of adult size) of host number on the number of mature eggs, for the 2 years, is shown in Fig. 3. No interaction between variables made a significant contribution to the regression. Both the indirect and direct influences of the initial number of larvae were highly significant but, when introduced first, the former made a more important contribution to the regression (χ2 = 135.5, P < 0.001 and = 11.95, P < 0.001 respectively). Differences in intercept between years were highly significant (χ2 = 32.73, P < 0.001).
Adult survival The median survival times and the fitted distributions for both years are shown in Fig. 4. The minimal adequate model was log[λ(t)] = year + l2 + l2 × year. Differences between years, both in the intercept and the slope of l2, were highly significant (χ2= 58.5, P < 0.001 and χ2 = 24.89, P < 0.001 respectively).
The results show that the survival to adulthood of Ephestia kuehniella, a host of Venturia canescens, is highly dependent on the initial number of individuals competing for food and on whether or not the hosts are healthy or parasitised. A number of life-history traits of the parasitoid (size, egg load, and adult survival in the absence of food) also depend on host density. Classically, parasitoids are classified either as regulators or conformers (Lawrence, 1986, 1990). While the former are assumed to regulate host metabolism and development, the latter are assumed to be completely dependent on the host's physiology. The results present a closer and more reciprocal relationship between the two partners: the parasitoids altered host susceptibility to competition, while the hosts determined critical parasitoid life-history traits such as egg load and adult survival (see also Harvey, 1996). The healthy hosts treatment corresponds to simple intraspecific competition. With the parasitised hosts treatment, the situation is much more complex: intraspecific host competition implies indirectly intraspecific parasitoid and host–parasitoid competition. The general features of this process were captured well by relatively simple models.
As already stated, the amount of time it took to prepare the experiments meant that it was not possible to use a fully randomised design. The experiments were therefore carried out twice, in successive years. For one trait (larva-to-adult survival), the year of the replicate did not have any influence on the results. For the other traits (size, egg load, and adult survival), there were differences between years, the 1993 animals being larger and having larger egg loads, but the general shape of the regressions did not differ. This suggests that the general, qualitative nature of the relationships did not depend on the particular sample.
The results presented in Fig. 1 show that beyond a density threshold the mortality rate for healthy E. kuehniella accelerates, leading to a humped survival curve that is typical of scramble competition. The K values and Fig. 1 show that in the present case the parasitised hosts were less susceptible to intraspecific competition. Prévost (1985) and Wajnberg et al. (1985) documented a case where density dependence changes from direct to inverse depending on whether the host is healthy or parasitised. The influence that parasitoids have on metabolism, food intake, and the development of their hosts (Jones & Lewis, 1971; Vinson & Iwantsch, 1980; Lawrence, 1986, 1988; Slansky, 1986; Beckage & Riddiford, 1993) might explain these differences. More specifically, Harvey (1996) has shown that Galleria mellonella (Linnaeus) (Lepidoptera: Pyralidae) and E. kuehniella parasitised by V. canescens grow at lower rates and for shorter times than their healthy conspecifics. Lower metabolic requirements might explain the lower susceptibility to competition. Similar results were obtained by Strand et al. (1988) and Ohnuma and Kainoh (1992) for other host–parasitoid systems. In a review, Slansky (1986) suggested that solitary parasitoids often reduce feeding and growth rates of their hosts. If so, the differences in susceptibility to competition shown by these results might be a relatively common feature of host-koinobiont solitary-parasitoid relationships.
This work considers situations in which either healthy hosts or (almost exclusively) parasitised hosts competed for food on their own. This mimics more extreme conditions than those found in natural situations, where both types of host would share the same resources. Consequently, the results of this work do not allow precise predictions of the number of hosts and parasitoids that would reach adulthood in a mixed population. Nevertheless, it can be suggested that in these conditions, the number of animals of the two types reaching adulthood would be intermediary to those presented in Fig. 1. For instance, at least in conditions of food shortage Ephestia kuehniella resorts to cannibalism (C. Bernstein, pers. obs.). Furthermore, Reed et al. (1996) observed that in a mixed population, Plodia interpunctella Hübner (Lepidoptera: Pyralidae) parasitised by V. canescens were more often the victims of cannibalism than their healthy counterparts. The consequence would be a decrease in the number of parasitoids reaching adulthood and, as a consequence of the transfer of resources from one type of host to the other, an increase in the number of hosts reaching adulthood.
Adult parasitoid size (as estimated by hind tibia length) decreased with the initial number of hosts competing (Fig. 2). Most probably, competition led to a decrease in host size and consequently reduced the quantity of resources available for the development of the parasitoid larvae. In a related system formed by V. canescens and P. interpunctella, Harvey et al. (1995) showed that adult parasitoid size decreased with increasing periods of host starvation, starved hosts being smaller than those fed ad libitum. Harvey et al. (1994), Harvey and Thompson (1995), and Harvey and Vet (1997) have shown that when V. canescens parasitises P. interpunctella, G. mellonella, E. kuehniella, or Corcyra cephalonica Stainton (Lepidoptera: Pyralidae), the size of the adult parasitoids depends strongly on the hosts' body mass and size.
The initial egg load and the adult survival rate of V. canescens decreased with the number of parasitised hosts competing for resources. This is not surprising, as both these life-history traits depend, at least partly, on adult size. Indeed, when egg load was regressed on both adult size and the initial number of larvae, when introduced first, adult size made the most important contribution to the regression. Venturia canescens is weakly synovigenic (Jervis et al., 2001). Adult females emerge with a non-negligible number of mature eggs (Fig. 3) and, if provided with food, they have a steady rate of maturation of new eggs (Trudeau & Gordon, 1989). The classical positive correlation between egg load and adult body size that has been observed in insects (Honek, 1993) also applies to V. canescens (Harvey et al., 1994; Harvey & Thompson, 1995). In V. canescens, the increase in adult survival rate with parasitoid size has been documented by Harvey et al. (1994).
The results also show that survival to adulthood in the E. kuehniella–V. canescens system gives a good fit to the generalised Beverton and Holt (Maynard Smith & Slatkin, 1973; Getz, 1996) and generalised Ricker (Pomerantz et al., 1980; Thomas et al., 1980; Bernstein, 1986) models, which provides an additional argument for their use in host–parasitoid models (see also Bellows, 1981). Intraspecific competition turned out to be scramble-like (θ ≥ 1.5) and parasitised hosts were less susceptible to intraspecific competition, as shown by the K values. For both models, the θ values differed for the healthy and the parasitised hosts, but the large differences in K and the correlation between parameters make it hazardous to speculate about the biological meaning of these differences. Theoretical work (Bernstein, 1986; Tuda & Iwasa, 1998) suggests that both lower susceptibility to competition by parasitised hosts and scramble competition contribute to the instability of the system. It is not yet clear to what extent these predictions are robust and independent of the modelling framework used. As stressed by Getz (1996), much more attention should be paid to the manner in which density dependence takes effect in populations.
In summary, this work shows that in the system formed by Venturia canescens and one of its hosts, the life-history traits of both partners are highly dependent on the initial number of hosts competing for food and on the partners' reciprocal influences. Some of these relationships might have been suggested by previous results, but the present work offers the first experimental evidence of their actual existence. Changes in life-history traits as a result of population fluctuations may influence the dynamics of the system. These influences have so far received little attention in host–parasitoid theory.
We would like to thank Muriel Ney Nifle, Thierry Spataro, Fréderic Menu, Xavier Fauvergue, Laurent Lapchin, Jerôme Casas, Gerard Driessen and Eric Wajnberg for their comments on the manuscript as well as Eric Wajnberg and Daniel Chessel for their help with the statistical analysis. Jeff Harvey, an anonymous referee and a member of the editorial board of Ecological Entomology made many suggestions that helped to improve this paper.