The diversity of parasitic insects remains one of the most conspicuous patterns on the planet. The principal factor thought to contribute to differentiation of populations and ultimately speciation is the intimate relationship parasites share with hosts and the potential for disruptive selection associated with using different host species. Traits that generate this diversity have been an intensely debated topic of central importance to the evolution of specialization and maintenance of ecological diversity. A fundamental hypothesis surrounding the evolution of specialization is that no single genotype is uniformly superior in all environments. This “trade-off” hypothesis suggests that negative fitness correlations can lead to specialization on different hosts as alternative stable strategies. In this study we demonstrate a trade-off in the ability of the parasitoid, Aphidius ervi, to maintain a high level of fitness on an ancestral and novel host, which suggests a genetic basis for host utilization that may limit host-range expansion in parasitoids. Furthermore, behavioral evidence suggests mechanisms that could promote specialization through induced host fidelity. Results are discussed in the context of host-affiliated ecological selection as a potential source driving diversification in parasitoid communities and the influence of host species heterogeneity on population differentiation and local adaptation.

Over half of the known species on the planet live parasitic lifestyles (Windsor 1998). The extraordinary diversity of the world's parasites is often attributed to the intimate relationship they share with a limited number of host species, and the potential for disruptive selective pressures associated with different hosts driving ecological divergence (Price 1980; Ackermann and Doebeli 2004). Most parasites are restricted to using a relatively small number of the available hosts within their environment (Fox and Morrow 1981). The evolutionary mechanisms responsible for host-range limitations have been an intensely debated subject of central importance to the evolution of specialization and maintenance of ecological diversity (e.g., Dethier 1954; Krieger et al. 1971; Bernays and Graham 1988; Futuyma and Moreno 1988; Jaenike 1990; Fry 1990, 1996; Joshi and Thompson 1995; Kawecki 1998). A fundamental hypothesis surrounding the evolution of specialization is that no single genotype is uniformly superior in all environments, thereby promoting the development of locally adapted varieties (Falconer 1952). It follows that extensive specialization in the presence of trade-offs would then generate the aforementioned diversity.

Patterns of host use in parasitic arthropods have been the focus of a vast amount of research with the central goal of better understanding how the diversity of insect communities have evolved and how they are maintained in time and space (e.g., Jermy 1984; Futuyma and Moreno 1988; Jaenike 1990; Thompson 1994; Mopper 1996; Stireman et al. 2005). Most insect species that are parasitic on plants or on insect hosts tend to display narrow specialization, in that they only feed on a small fraction of the potential hosts species they encounter (Fry 1996). Even generalist insects, which can be largely polyphagous at the species level, often exhibit limited host species use at the population or community level (Fox and Morrow 1981; Smith et al. 2007).

Host range is influenced by localized genetic variation in both parasite and host populations and whether selection or trade-offs lead to specialization on different host species (Dres and Mallet 2002; Lajeunesse and Forbes 2002; Kawecki and Ebert 2004). Divergent selection associated with the utilization of particular hosts can result in localized genetic structuring of parasite populations that may impede host shifts to alternate species or restrict an expansion in host range. A “host shift” refers to a population forming an association with a novel host, which differs from the standard definition of host-range expansion, which involves a population colonizing a new host with the continued utilization of the previous host (Agosta 2006). Barriers to the integration of novel hosts can be behavioral, or physiological in nature, with adaptations to one host species potentially resulting in trade-offs in the ability to use alternate hosts. Behavioral barriers may include the ability to accurately locate a host in the environment or microhabitat, recognition of potential hosts, and coping with external host defenses (De Moraes et al. 1998; Althoff and Thompson 2001). Physiological barriers, including biochemical or morphological adaptations, can include digestive enzymes necessary to overcome a host's internal defense, or specialized structures required to access a host (Dethier 1954; Vinson and Iwantsch 1980).

A popular theory to explain some of the patterns of host affiliation in phytophagous insects is that traits leading to an increased fitness on one host are detrimental on others (Krieger et al. 1971; Rausher 1983; Futuyma and Moreno 1988; Jaenike 1990; Via 1990; Fry 1996). Referred to as the “trade-off” hypothesis, this theory suggests that negative fitness correlations, caused by the antagonistic pleiotropic action of one or more genes, can lead to specialization on different hosts as alternative stable strategies (Castillo-Chavez et al. 1988; Futuyma and Moreno 1988; Jaenike 1990). Although conceptually appealing, evidence of direct genetic trade-offs (i.e., negative cross-host fitness correlations) using quantitative genetics techniques have been found in only a few studies (Gould 1979; Karban 1989; Fry 1990; Karowe 1990; Via 1991; MacKenzie 1996; Agrawal 2000) and more frequently have not been found (e.g., Rausher 1984; Hare and Kennedy 1986; Bernays and Graham 1988; Fox 1993; Thompson 1996). However, evidence of negative genetic cross-host fitness correlations does not preclude the existence of trade-offs. As Fry (1996) illustrates, specialization is promoted whenever fitness norms cross, indicating that genotypes have different fitness rankings on different host species. This pattern is observed in many phytophagous insects and is a plausible yet controversial explanation for why there is so much diversification and specialization in host use by herbivorous insects.

In a system in which genotypes have different fitness rankings on different hosts, the advancement of specialization is greatest when gene flow is reduced between populations (Fry 1996). The population level outcome of processes driving diversification, such as trade-offs or selection, is therefore dependent on the system-specific mechanisms that impact gene flow. Host fidelity, is one such mechanism that can greatly reduce gene flow between populations because many insects feed, mate, and oviposit on or near their hosts (Hawthorne and Via 2001; Funk et al. 2002). Locally adapted gene pools are therefore degraded when gene flow is high between populations and conserved when gene flow is limited. The conservation of gene pools thus preserves the differences generated by trade-offs or selection and is an important factor in the maintenance of genetic differentiation in insect populations.

Although there has been great progress in understanding the ecology and evolution of host use in phytophagous insects, little is known about the processes that mediate host-range evolution in another diverse and highly specialized group, the insect parasitoids. Parasitoid lifestyles are predominantly found in the orders Hymenoptera and Diptera, but are also found in many other holometabolous orders of Insecta. Parasitoids have a particularly intimate relationship with hosts because a single host harbors the parasitoid's offspring until maturity (Godfray 1994). Within families, parasitoids have undergone extensive adaptive radiation as evidenced by the vast number of species in the major parasitoid clades (Godfray 1994; Godfray and Shimada 1999; Irwin et al. 2003). Although little is known about the mechanisms that mediate population and species divergence in parasitoids there is evidence that suggests host affiliation may be responsible for driving differentiation of lineages (Powell and Wright 1988; Fellowes and Kraaijeveld 1998; Pike et al. 1999; Morehead et al. 2001; Aldrich and Zhang 2002; Dupas et al. 2003; Stireman et al. 2006). Many parasitoids have specialized traits designed to cope with host defenses and are highly sensitized to chemical cues that they use to locate particular host species (Vet and Dicke 1992: De Moraes et al. 1998). The intimate nature of this relationship promotes coevolutionary dynamics, while generating disruptive selective pressures associated with specialization on different host species (Thompson 1994). These factors suggest that local host species utilization may play an important role in the differentiation and diversification of parasitoid communities.

In the following experiments we investigate the potential for adaptation to a novel host by an insect parasitoid, as well as the costs associated with the host shift, using a replicated quasi-natural selection experiment under controlled laboratory conditions. The aphid parasitoid Aphidius ervi Haliday (Hymenoptera: Aphidiidae) is considered a generalist species although some host-range specialization has been reported. Aphidius ervi from Europe uses a variety of host species and is considered predominantly a generalist, whereas A. ervi from Japan primarily specializes on pea aphids and exhibits low levels of fitness on other host species commonly used by the European variety (Hajimu and Tada 2000). Our experiments were designed to address the following questions: (1) Are host species potential agents of directional selection in Aphidius parasitoids? Furthermore, can a population of parasitoids adapt to a novel, initially low-quality host as demonstrated through an increase in population fitness? (2) What costs are associated with adaptation to a novel host and are fitness trade-offs evident in the ability to use an ancestral host species? (3) What are the behavioral and physiological mechanisms that contribute to, or confine, host-range expansion? Further, are these traits selected over successive generations or are they plastically induced within a generation?

Materials and Methods


The colony of A. ervi was originally collected from Acyrthosiphon pisum (Harris), (Hemiptera: Aphididae) (i.e., the “ancestral” host) in alfalfa fields at London, Ontario and maintained at the Southern Crop Protection and Food Research Centre, Agriculture Canada for several generations, on broad beans Vicia faba L. (cv. ‘Broad Windsor’), before being transferred to Simon Fraser University (Burnaby, British Columbia). The source parasitoids were arbitrarily assigned to nine colonies (>300 individuals each), and were maintained on A. pisum, feeding on broad beans at Simon Fraser University for approximately three to four parasitoid generations to allow the parasitoids to acclimate to laboratory conditions prior to the start of the experiment. All insect colonies were maintained at 19 ± 2.0 °C daytime, 17 ± 2.0 °C nighttime temperature, 50–60% RH, and a L16:D8 photoperiod. These environmental conditions maintained the aphids in a continual parthenogenetic cycle throughout the experiment. Replicate selection lines were initiated by haphazardly removing approximately 1000 parasitized aphids as mummies from the nine stock colonies and placing them in a common emergence cage. Each mummy harbored a single Aphidius parasitoid in pupal form. Upon emergence, adults were allowed to mate. The mated females were then subdivided into replicate populations maintained on either of two different host species; pea aphids, A. pisum (the “ancestral” host) or foxglove aphids, Aulacorthum solani (Kaltenbach) (Hemiptera: Aphididae), (the “novel” host). These are hereafter called P-line and F-line for the pea and foxglove selection lines, respectively. The original population of A. solani was established via collections from commercial pepper greenhouses in Abbotsford, British Columbia. These were transferred to broad bean plants and maintained for 2 months prior to the start of the experiment to allow the aphids to adapt to the new host plant and laboratory conditions. Acyrthosiphon pisum was collected from alfalfa fields in Southern British Columbia and maintained in 12 separate colonies at Simon Fraser University. Acyrthosiphon pisum are a large, high-quality host, and are a primary host of A. ervi in many regions (Hajimu and Tada 2000) whereas A. solani are a small, low-quality host; in that A. ervi generally produces substantially fewer mummies when initially exposed to this host species (Henry et al. 2005). Both aphid species in experimental and stock colonies were maintained on broad beans to remove plant effects on parasitoid behavior or reproductive performance. Replicate selection populations were initiated sequentially over 8 weeks as parasitoid females became available from the stock colonies. Eight populations were maintained on A. pisum and 14 populations were maintained on A. solani. Replicate populations were used for each host species to reduce the stochastic effects of genetic drift on the overall mean values. Each replicate population was initiated using approximately 50 mated female parasitoids, which were exposed to 24 broad bean plants infested with several thousand host aphids. Dilute honey (10% solution), water (nutrition for parasitoids), and fresh bean plants (food plants for aphids) were added each week. Aphids were added to each replicate population on a per cage basis if aphid populations dropped below a specified density determined by subsampling several leaves in each population per week (inline image= 50 ± 10 aphids per apical bean leaf). Furthermore, no aphids were ever removed from the cages, which prevented mixing of aphids and parasitoids between replicates, and allowed aphids to coevolve to some degree with the parasitoid populations. Parasitoid populations were maintained in isolation on each host species for 2 years (∼50 parasitoid generations).


Assay parasitoids, reared on A. pisum or A. solani, were always 2- to 3-day-old females that had been given continuous access to dilute honey (10% solution), water and males from their replicate cage for fertilization. Parasitoid females were naïve at testing meaning that they had no contact with any hosts except the mummy casings from which they have emerged. Each parasitoid female was used only once then discarded.


Parasitoid fitness and host-affiliated trade-offs

F-line and P-line parasitoids were assayed for fitness on their natal host (i.e., selection line host species) and non-natal host every generation for the first four parasitoid generations to determine the initial response in trait values to the hosts, then again after 40 and 50 generations. Therefore, the four treatment groups sampled at each parasitoid generation were: F-line parasitoids assayed on pea aphids; F-line parasitoids assayed on foxglove aphids; P-line parasitoids assayed on pea aphids, and P-line parasitoids assayed on foxglove aphids. For each of the parasitoid generations assayed, a subsample of 60–80 mummies was removed from each population and isolated in glass emergence containers as a group. Upon emergence, parasitoids were allowed to mate with individuals from their own replicate population and fed on honey and water for 24 h, after which 20 females were randomly selected from each cohort and individually assayed on natal and non-natal hosts (10 females on foxglove and 10 on pea). Parasitoids that were not assayed were returned to their appropriate replicate populations.

Individual parasitoid females were allowed to forage on 40 second instar (46- to 48-h old), A. pisum or A. solani on a bean leaf for 4 h. Aphid instar was determined by age, which correlates strongly with size for each species (mean volume at 46–48 h is 0.031 ± 0.006 mm3 for A. solani and 0.109 ± 0.011 mm3 for A. pisum). Second instar aphids were chosen for this assay because they represent a relatively high-quality host that does not display behavioral defenses that would affect parasitoid handling time. After 4 h, parasitoids were removed and the aphids were carefully transferred to an excised bean leaf, which was kept fresh by inserting the leaf petiole into a water-filled glass vial capped with Parafilm (Pechiney Plastic Packaging, Chicago, IL). Excised leaves with aphids were then sealed in petri dishes in which they remained until mummies formed.

Proxies chosen to represent parasitoid fitness for each selection line were mean number of hosts parasitized (mummies formed per 40 aphids), and mean proportion of parasitoid adults eclosing from mummies (Roitberg et al. 2001). The mean number of hosts parasitized was analyzed using all of the above-noted generations. Proportion eclosing from mummies was compared at parasitoids generations 1 and 50 only. Traits were compared for each assay host species separately (i.e., F-line vs. P-line fitness on foxglove and F-line vs. P-line fitness on pea) to reduce the complexity of the statistical model. The P-line was always used as the control population, with deviations in trait values indicating a response to selection in the F-line parasitoids.

A generalized linear model (GLM) was used to analyze the results from Experiment 1 with an overdispersion parameter applied to account for nonindependence of sampling units (McCullagh and Nelder 1989). GLM model effects included: parasitoid host selection line, parasitoid generation, and parasitoid generation × selection line cross. All model effects were treated as fixed. For mean hosts parasitized, a Poisson distribution was used with a log link function. Proportion eclosion was analyzed using a logistic regression with a binomial distribution. Data were analyzed using JMP 6.0 statistical software (SAS Institute, Cary, NC).


Host fidelity and parasitoid virulence

Factors influencing host-range evolution in insect parasitoids generally involve both behavioral and physiological components that work in conjunction to mediate acceptance and subsequent reproductive success on a given host. The following experiments were used to explore the influence of host fidelity of the female parasitoid (i.e., willingness to accept a host species) and the physiological virulence of the parasitoid larvae, when adapting to a novel host species. In the following experiments equal numbers of individual females were sampled from each replicate population. Individuals used for each experiment were pooled for the analyses.

Host fidelity

Host species fidelity was investigated using parasitoids from each host selection regime (F-line and P-line) assayed on either a natal or non-natal host species at generations 1 and 50. Individual parasitoids were allowed access to a single second instar host in a gel capsule. Parasitoids were allowed to sting each host only once, and the behaviors leading to the initial sting were recorded for each individual. Host fidelity was assessed by measuring oviposition latency, and host rejection. Oviposition latency was defined as the time from the parasitoid's first inspection of the host (i.e., antennal contact with host) until the host was accepted, indicated by a successful oviposition. Parasitoids that did not oviposit in the host after 5 min were considered to have rejected the host. Host fidelity was investigated using parasitoids that had remained on a given host species for 50 generations. To determine whether the parasitoid's host fidelity had evolved under the host selection regime or was a plastic response to the current natal host, latency was assessed at generation 1 and compared to generation 50 using F-line parasitoids only. Oviposition latency was analyzed using an ANOVA for all selection and assay host combinations. Differences in the number of hosts rejected and accepted between assay hosts were compared separately for each selection line using a chi-square analysis.

Parasitoid virulence

The physiological component of adaptation to a novel host was explored by assessing the performance of the developing larvae for each combination of host selection regime and assay host. Individual parasitoids were given access to a single host from either species in a gel capsule as described above. After a single sting, the aphid was removed and transferred to an excised bean leaf. Aphids were monitored daily to determine parasitoid larval development. Parasitoid virulence and host resistance were used to determine the overall success rate of larval development until pupation. Parasitoid virulence was defined as the loss of fitness of a parasitized host (i.e., aphid mortality due to infection with the parasitoid), because the death of the host is required for successful parasitism. Virulence was therefore ascertained by comparing the number of aphids that died due to parasitism, as indicated by the formation of a parasitoid mummy compared with the aphids that resisted parasitism. This measure was tested for each parasitoid selection line on each assay host species using a chi-square analysis (i.e., F-line assayed on pea vs. foxglove and P-line assayed on pea vs. foxglove). A parasitoid's fitness was therefore determined by its success in transmitting offspring to a particular host species.

To establish that a sting was equivalent to an egg being laid, subsamples of hosts (∼40 per selection/assay host cross) from this experiment were dissected and inspected for parasitoid eggs. Stung aphids from both species were injected with methyl blue, which made the parasitoid eggs easier to identify. Egg presence was compared for each assay host and host selection line combination using a chi-square analysis.



Parasitoid fitness and host-affiliated trade-offs

There was significant variation in the number of hosts successfully parasitized from P-line and F-line parasitoids assayed on foxglove aphids (GLM, χ2(11)= 175.42, P < 0.0001, overdispersion = 4.71). Selection line influenced the number of hosts successfully parasitized (χ2(1)= 23.50, df = 1, P < 0.001). F-line females parasitized an overall greater number of hosts when assayed on foxglove aphids (inline image= 8.43 ± 0.46) than P-line females (inline image= 5.03 ± 0.36). There was a significant effect of generation (χ2(5)= 94.84, P < 0.0001). There was no interaction between generation and treatment (χ2(5)= 3.94, P > 0.05). This latter result is most likely due to the large variance created by sampling the mean hosts parasitized at six different parasitoid generations. To determine if the F-line parasitoids increased their mummy production over the 2 years on foxglove aphids, a contrast analysis was applied to compare the difference in the number of hosts parasitized at generation 1 versus generation 50. This analysis demonstrated that the selection lines differed in the number of hosts parasitized when assayed on foxglove aphids at generations 1 and 50 (Contrast analysis, χ2(1)= 15.52, P < 0.0001). F-line females produced significantly more mummies when assayed on foxglove at the end of the 50-generation period than they did at generation 1 (Fig. 1).

Figure 1.

Mean hosts parasitized per parasitoid generation for parasitoids assayed on 40 second instar foxglove aphids, A. solani. Black line with diamond markers represents foxglove selection line (F-line) and gray line with square markers represents pea aphids, A. pisum, selection line (P-line) parasitoids. Black bars signify standard error for each parasitoid generation.

The number of hosts parasitized differed significantly when comparing P-line and F-line females assayed on pea aphids (GLM, χ2(9)= 68.09, P < 0.0001, overdispersion 3.67). Individual effects indicated that selection line affected the number of hosts parasitized (χ2(1)= 24.32, P < 0.0001). The P-line produced a greater number of mummies over all generations combined (hosts parasitized: F-line inline image= 13.5 ± 0.6; P-line inline image= 17.7 ± 0.6). There was an effect of generation (χ2(4)= 25.56, P < 0.0001) and a selection line by generation interaction, with the F-line parasitizing fewer hosts over the generational period than P-line females (Fig. 2) (χ2(4)= 17.12, P < 0.005). Generation 3 was not included in the analysis and figure due to missing data.

Figure 2.

Mean hosts parasitized per parasitoid generation for parasitoids assayed on 40 second instar pea aphids, A. pisum. Black line with diamond markers represents foxglove aphid, A. solani, selection line (F-line) and gray line with square markers represents pea aphid selection line (P-line) parasitoids. Black bars signify standard error for each parasitoid generation.

When comparing the proportion eclosion from mummies for the F- and P-line parasitoids assayed on pea aphids a whole model effect was detected (GLM, χ2(3)= 12.89, P < 0.005, overdispersion 2.26). Individual model effects indicated that there was no effect of selection line (χ2(1)= 1.64, P > 0.05), or of selection line within generations (χ2(1)= 1.93, P > 0.05). However, an effect was detected between generations 1 and 50 in that generation 50 had fewer parasitoids successfully eclosing from pea aphid mummies (χ2(1)= 10.07, P < 0.005) (Fig. 3A). The proportion eclosion from mummies was not different between the two selection lines assayed on foxglove aphids (GLM, χ2(3)= 2.90, P > 0.05, overdispersion 1.51). Furthermore, no individual model effects were detected for the proportion eclosion from foxglove mummies (selection line, χ2(5)= 0.01, P > 0.05; generation, χ2(1)= 0.93, P > 0.05; generation × selection line, χ2(1)= 0.73, P > 0.05) (Fig. 3B). A general lack of variation in eclosion between selection lines over the generational period indicates that this trait is most likely not under selection within this system.

Figure 3.

Mean proportion eclosion from parasitized pea (A) and foxglove (B) aphids at generations 1 and 50 for F-line (black diamonds) and P-line (gray squares) parasitoids. Black bars signify standard error.


Host fidelity

A similar pattern in host fidelity arose for both selection lines when parasitoids were assayed on natal versus non-natal hosts. There was a significant difference between natal and non-natal assay groups in the latency period before a host was accepted (analysis of variance [ANOVA, F3, 285= 12.80, P < 0.0001]). No effects of selection host (P > 0.05) or assay host (P > 0.05) were detected; however there was an interaction between selection line and assay host (F1= 41.99, P < 0.0001). A Tukey HSD analysis revealed that parasitoids accepted a natal host (i.e., selection regime and assayed on the same host species) more quickly than when offered a non-natal host (Table 1). A similar pattern arose for the difference in the number of hosts accepted or rejected, with a greater number of non-natal hosts being rejected. There was a significant difference in the number of accepts and rejects for F-line parasitoids (χ2(1,162)= 10.60, P < 0.005) with females rejecting more pea (21.7%) than foxglove aphids (4.9%). P-line females showed a similar pattern with more foxglove aphids being rejected (14.6%) than natal pea aphids (5.5%), and the analysis only marginally failed to demonstrate significance (χ2(1,159)= 3.7, P= 0.055).

Table 1.  Mean latency period (sec) prior to oviposition for parasitoid females from each selection line assayed on natal and nonnatal host species.
Selection hostAssay host

The increase in the number of hosts parasitized observed in the F-line parasitoids assayed on foxglove aphids over the selection period indicated that a trait influencing parasitism, such as host fidelity or virulence, was under selection. To determine if a behavior that mediated host fidelity was under selection, oviposition latency was compared at generations 1 and 50 on the two host species using F-line parasitoids. The analysis revealed that there was no change in oviposition latency for the F-line parasitoids over the 2-year selection period (generation × assay; F3,171= 9.55, P > 0.05). The F-line parasitoids took significantly longer to oviposit in pea aphids than in foxglove aphids, and this response did not change over the selection period indicating that the behavior was plastically induced after developing in a host for a single generation.

Parasitoid virulence

The differential survival of single-stung aphids was assessed using parasitoids from both host selection lines to determine if different host species act as selective agents on parasitoid larvae. To eliminate the potential confound of parasitoids not laying eggs when they sting a subsample of stung aphids from the selection line/assay crosses were dissected, stained, and inspected for a parasitoid egg. There were no differences in the number of eggs found in any of the hosts, (χ2(3,113)= 1.16, P > 0.05) demonstrating that a sting was equivalent to an egg being laid regardless of selection line or assay host species.

Larval virulence was compared between selection lines separately from each assay host species. Early aphid mortality was recorded but not used in the analysis because mortality occurred, in almost all cases, within the first 24 h after oviposition due to aphids walking off the leaves and refusing to feed, prior to parasitoid egg hatching. F-line parasitoids had a greater virulence when assayed on foxglove aphids than P-line parasitoids in that they successfully parasitized a natal host species more often than a non-natal host (χ2(1,123)= 5,74, P < 0.05). A similar pattern emerged from the P-line parasitoids, which also had a greater virulence on their natal host than the F-line parasitoids (χ2(1,102)= 8.06, P < 0.005) (Fig. 4).

Figure 4.

Proportion of pea (A) and foxglove (B) aphids parasitized (gray), resisted (black), or deceased (white) at 24 h after receiving a single oviposition from either F-line or P-line parasitoids.


Selection experiments are a powerful tool in the study of evolutionary biology because they allow the investigator to vary the environmental context in a controlled manner. This approach is particularly useful in the study of local adaptation when exploring trade-offs associated with niche breadth, providing an alternative method to classic quantitative genetics techniques (Fry 2003). In the present study, a trade-off in the ability to maintain a high level of fitness on more than one host suggests a genetic basis for host utilization that may limit host-range expansion in Aphidius parasitoids. Trade-offs associated with utilization of different hosts are thought to be an important mechanism generating genetic diversity among populations of insects and encourages local adaptation when combined with limited gene flow (Kawecki and Ebert 2004).

In this study, adaptive phenotypic evolution was investigated using a single population of parasitoids exposed to pea aphids (A. pisum), a high-quality host, and to foxglove aphids (A. solani), a low-quality novel host. Reciprocal trials were performed on both hosts in each of several generations to determine if adaptation to the novel host resulted in a trade-off in fitness on the ancestral host. When initially exposed to the novel host, parasitoid fitness was low as shown through the modest mummy production in experiment 1 at generation 0. However, mummy production increased substantially after a single generation on the novel host species. This change in fitness is most likely due to a plastic shift in host fidelity that conditions parasitoids toward the host species in which they developed. Evidence for this behavioral plasticity is demonstrated in the greater affinity A. ervi has for a host species after developing in it for a single generation, which decreased oviposition latency (Table 1) and host rejections, thereby facilitating elevated mummy production through increased host exploitation (see also Henry et al. 2006). If under selection, host fidelity could be responsible for changes in mummy production over a generational period as parasitoids become more willing to accept natal than non-natal host species. If host fidelity was under selection, oviposition latency should have changed over the selection period. The induced response in host fidelity did not change over the 2-year selection period, indicating that the mechanism is plastically induced and not under direct selection. A similar plastic response has been documented for host plant preference in Aphidius parasitoids in that parasitoids have a greater attraction for the host plant complex in which they developed previously, regardless of their population of origin (Poppy et al. 1997; De Moraes et al. 1998; Daza-Bustamante et al. 2002). The following three generations did not show a significant change in fitness on either host, however the initial fitness increase was maintained on foxglove aphids in the F-line parasitoids. Behaviors that encourage continual use of the same host species, such as the increase in host fidelity observed in Aphidius parasitoids, promote further adaptation through natural selection and the conservation of locally adapted gene pools. Where behavioral modifiers cause avoidance of feeding or oviposition on a host, mean population fitness has the potential to increase by adapting to one host free of countervailing selection from the other (Fry 1996). Therefore trade-offs in the presence of a trait that limits gene flow have the potential to stabilize specialization and destabilize generalization under idealized environmental conditions.

After remaining on foxglove aphids for 2 years, parasitoid fitness on foxglove aphids increased dramatically compared to P-line parasitoids (Fig. 1). A concurrent decrease in fitness was observed in the performance of the F-line parasitoids on the ancestral host, the pea aphid (Fig. 2). This pattern suggests that simultaneous maximization of fitness on pea and foxglove aphids was constrained within the initial population of parasitoids used in this experiment, possibly due to the antagonistic pleiotropic action of one or more genes. A local maximum appears to have been reached in the F-line parasitoids at generation 40 as subsequent sampling revealed a plateau in mummy production on both host species. When controlling for parasitoid behavior, parasitoids similarly retained greater virulence (i.e., more parasitoids successfully completed development) developing in a natal as opposed to a non-natal host (Fig. 4). These results provide strong evidence that there is a physiological mechanism involved in overcoming host defenses and that this process is under directional selection, indicated by the increase in virulence beyond the initial plastic shift, over the selection period. After an initial plastic response to a novel environment, there is often a period of latency before a subsequent fitness increase is a pattern consistent with genetic change through the rearrangement of allele frequencies, with noticeable phenotypic change occurring only as the population moves toward homozygosity in the genes under selection. This pattern is consistent with our data in that noticeable fitness trade-offs did not occur until parasitoids spent four generations on foxglove aphids. Although our design did not take into account the possibility of maternal effects such as viruses, virus-like particles, endosymbionts and other substances known to be maternally transmitted that increase virulence in ichneumonoid parasitoids or resistance in hosts (reviewed in Vinson 1990; Oliver et al. 2005), the existence of fitness trade-offs on the ancestral host supports the notion of a genotypic response to a selective agent.

Genetic variation in virulence within natural populations of parasitoids has been measured in several systems. Leptopilina boulardi, a parasitoid of Drosophila simulans provided evidence that a single population contained significant heritable variation in the ability to parasitize their hosts (Carton et al. 1989). Similarly, a selection study using laboratory populations of parasitoids showed an increase in their ability to overcome Drosophila encapsulation (Kraaijeveld et al. 2001). Considerable genetic variation in parasitoid virulence and host resistance has been demonstrated in several well-studied endoparasitoid systems (Henter and Via 1995; Hufbauer 2002; Kraaijeveld et al. 2002; Dupas et al. 2003). Within aphid parasitoid systems Henter (1995) found significant additive heritable variation in the ability to parasitize pea aphids within a single population of A. ervi. In a related study, Henter and Via (1995) found significant genetic variability within a population of pea aphids in their susceptibility to parasitization by A. ervi. The presence of additive genetic variation in aphid resistance and parasitoid virulence indicates that these populations have the potential to coevolve in response to selection (Henter 1995). Variation in the ability to parasitize natal and non-natal host species has been recorded for several aphid parasitoids suggesting a genetic basis for virulence and resistance across a variety of systems (Powell and Wright 1988; Pennacchio et al. 1994; Pike et al. 1999; Antolin et al. 2006). Research by Oliver and associates (2005, 2006) on the A. ervi—pea aphid system has revealed that aphid endosymbionts confer varying degrees of resistance to parasitism. This discovery indicates that we may need to broaden our outlook on the coevolutionary dynamics of host-parasitoid systems to include the function and frequency of endosymbionts, or other organisms influencing parasitism rates, within populations of both hosts and parasitoids.

Host-range evolution in insect parasitoids is typically governed by both behavioral and physiological components. We investigated the traits of adults and offspring, as both can potentially influence overall fitness levels and the detection of trade-offs associated with using different hosts (Scheirs et al. 2005). Our work demonstrates that adult behavior remains plastic within this system but larval virulence is genetically determined, as indicated by the accrued fitness of F-line parasitoids and inability of parasitoids to simultaneously maximize fitness on more than one host, a pattern suggested to occur in other aphid–parasitoid systems (Powell and Wright 1988; Antolin et al. 2006). Theory predicts that in a system in which genotypes have different fitness on different hosts, selection should promote specialization as long as parasites are capable of host choice and the cost of being “choosey” is low (Fry 1996). The logic behind this theory is that specialization is favored because parasites that remain on a single host can evolve faster in response to the evolution of a single host's defenses (Kawecki 1998). Our research, as well as research in several other Aphidius systems has demonstrated that parasitoids are “choosey” in that they have evolved a tendency to prefer the same host type in which they developed (Pennacchio et al. 1994; Poppy et al. 1997; Daza-Bustamante et al. 2002). This process is thought to inherently destabilize generalization and promote specialization, even if trade-offs are absent (Balkau and Feldman 1973). When combined with a trait under hard selection, such as parasitoid virulence, fitness trade-offs associated with the use of different host species have the potential to rapidly progress populations toward specialization on different hosts as alternate stable states (Fry 1996). However, it should be noted that intraspecific genetic variation for performance on different hosts is just one mechanism that can potentially lead to differentiation and diversification of insect populations. The evolution of a specialist or generalist strategy is mediated not only by mechanisms generating and maintaining genetic variation, but also how these traits interplay with the many different factors that are specific to each environment, such as competition for hosts with other organisms and the abundance and distribution of host species that parasitoids are exposed to at both a spatial and temporal scale. Aphid populations follow classic meta-population dynamics and are prone to successions of colonization and extinction of local populations. This process is thought to inherently destabilize specialization through the obliteration of locally adapted gene pools (Kawecki and Ebert 2004). However, Aphidius parasitoids may counteract this process by the use of a highly sensitized host location mechanism or through the “hitchhiking” of parasitoid larvae within colonizing aphids. Furthermore, local adaptation could be diluted if constant migration from other hosts occurs, although migration between host species could slow evolution but not prevent a response to selection or diffuse coevolution between species (Henter 1995).

Host-associated genetic divergence has been reported in several parasitoid species (Morehead et al. 2001; Stireman et al. 2005; Hayward and Stone 2006), although few studies have addressed host-based genetic divergence in aphid parasitoids. Vaughn and Antolin (1998) found that Diaeretiella rapae from two adjacent fields containing Russian wheat aphids and cabbage aphids had extensive genetic differentiation between six host-associated populations tested (using heritable RAPD markers). Two subsequent studies on D. rapae populations concluded that although genetic differentiation and local adaptation does occur by host species, analyses of mitochondrial DNA suggested that sufficient gene flow prevents populations from becoming completely isolated and that local selection rather than isolation creates genetic subdivisions between populations on different hosts (Baer et al. 2004; Antolin et al. 2006). Our work supports the notion that local host-affiliated selective pressures could potentially generate genetically distinct subpopulations that correspond to different host species with fitness trade-offs limiting the number of hosts a single populations could optimally use at a given time.

In conclusion, models of genetic differentiation associated with ecological specialization emphasize divergence as a consequence of different selective pressures between differing environments (Schluter 2001). Empirical evidence is growing in support of the many proposed mechanisms thought to contribute to the widespread patterns of host affiliation in herbivore communities (for reviews see: Bush 1975; Price 1980; Mopper 1996; Poulin and Morand 2000; Berlocher and Feder 2002). However, far less research has been directed at the insect parasitoids, which are also an immensely diverse, potentially rich taxa for studying host-affiliated ecological divergence. Our results suggest that a number of behavioral and physiological mechanisms could promote specialization in Aphidius parasitoids as well as limit the number of hosts a single population can effectively use at any one time. These mechanisms are characteristic of traits thought to contribute to host-associated differentiation in parasite populations. However, local adaptation has not been detected in a number of natural populations of other, widely studied parasitoid systems exhibiting similar traits (Hufbauer 2001; Kraaijeveld et al. 2002; Dupas et al. 2003) but has been detected in others (Vaughn and Antolin 1998; Althoff and Thompson 2001; Morehead et al. 2001; Stireman et al. 2005; Antolin et al. 2006; Hayward and Stone 2006). It should be noted that the insects and hosts studied to date represent only a fraction of the tremendous diversity of parasitic relationships that exist. Only now are we acquiring evidence that suggests that host-associated differentiation and cryptic speciation may be more common than previously thought (Stireman et al. 2005), especially for the insect parasitoids (Stireman et al. 2006; Smith et al. 2007). Although traits facilitating directional selection and specialization exist in many parasite systems as a means to constantly improve fitness or counteract host defenses, the evolution and maintenance of differentiation between populations requires specific environmental conditions, such as stable host populations and limited gene flow between host-affiliated populations. Host–parasite systems that are prone to disturbances or those that experience extensive gene flow may have selection slowed or disrupted. Our work has demonstrated the potential for a single population of Aphidius parasitoids to differentiate based strictly on host species utilization. However, to gain a better understanding of how these traits function in nature, detailed studies are required that link mechanisms that drive differentiation to the genetic structure of natural parasitoid populations.

Associate Editor: P. Abbot


The authors would like to thank E. Meuser and J. Page for their hard work and dedication to this project. We would also like to acknowledge Dr. C. Schwarz for statistical advice, and NSERC and the BC Greenhouse Growers Association for funding.