A large-scale comparison of conventional and molecular methods for the evaluation of host–parasitoid associations in non-target risk-assessment studies

Authors


*Correspondence and present address: University of Hawaii at Manoa, Kauai Agricultural Research Center, 7370 Kuamoo Road, Kapaa, HI 96746, USA. E-mail: taradawn@hawaii.edu

Summary

  • 1Accurate identification of natural enemies is the cornerstone of biological control, and methods that can separate closely related species are essential in ecological studies of parasitoids. Conventionally, host rearing and dissection are used to define the ecological host range of candidate biological control agents and assess host-specificity of parasitoids. However, molecular methods may be more suitable for the evaluation of host–parasitoid associations.
  • 2To demonstrate the utility of molecular diagnostics in ecological host-range studies on parasitoids of Lygus plant bugs, host rearing, dissection and multiplex polymerase chain reaction (PCR) analysis were used to estimate parasitism levels and parasitoid species composition (genus Peristenus) in more than 26 000 field-collected target and non-target Miridae.
  • 3Parasitism levels estimated by conventional and molecular methods were similar but molecular analysis can detect parasitoids earlier than dissection and rearing. Parasitoid pupal mortality prevented the identification of more than 30% of individuals reared from non-target host material; however, paired samples analysed with the multiplex assay allowed the identity of these parasitoids to be inferred. Molecular methods can provide different, and generally more complete, parasitoid species composition information because the results are not confounded by the host and parasitoid mortality encountered in rearing. However, detection of a parasitoid in a host does not necessarily indicate survival to the adult stage. Further, molecular identification of parasitoid species may be restricted to those species for which PCR primers are available.
  • 4Synthesis and applications. For molecular diagnostic techniques to gain widespread adoption in ecological studies on natural enemy host range, they must provide information that is equivalent (or superior) to information obtained by conventional methods. Based on a large-scale case study, associations between Peristenus spp. and their mirid hosts were used to demonstrate the utility of molecular diagnostics in studies on parasitoid ecological host range; however, this approach can be extended to pre-release risk-assessment studies on other candidate biological control agents. Beyond agent identification, molecular diagnostics can facilitate and expedite pre- and post-release studies on the ecological host range of parasitoids, potential non-target effects, host–parasitoid associations and trophic interactions.

Introduction

In biological control, the level of parasitism in a host population is used as a measure of parasitoid efficacy and provides a quantitative evaluation of the impact of a parasitoid on a pest population (Van Driesche et al. 1991; Hawkins, Thomas & Hochberg 1993; Hawkins & Cornell 1994). Accurate estimation of parasitism levels is one of the main methodological problems encountered in the study of parasitoid efficiency (Höller & Braune 1988). Conventionally, the identification of parasitoids and quantification of parasitism levels in a host population rely on the rearing and dissection of field-collected insects. However, these two methods often provide different results, and failure to acknowledge that differences exist can result in data that are inaccurate (Day 1994). This can be particularly critical in non-target risk-assessment studies that address the host specificity of a candidate biological control agent. Ecological host-range studies are often used to evaluate the host specificity of a parasitoid, and rely on dissection and/or rearing of field-collected material to detect and identify the parasitoid community associated with target and non-target hosts (McEvoy 1996; Onstad & McManus 1996; Strand & Obrycki 1996; Greathead 1997). The failure to identify or detect candidate biological control agents accurately in non-target hosts may compromise the integrity of pre-release studies that seek to address the host specificity of candidate agents.

Molecular methods have been used to estimate parasitism levels and identify parasitoids in host populations (MacDonald & Loxdale 2004; Behura 2006; Greenstone 2006; Stouthamer 2006; Gariepy et al. 2007). Like dissection, molecular analysis provides a rapid estimate of parasitism level. However, using species-specific polymerase chain reaction (PCR) primers, molecular analysis has the potential to provide information that is not easily obtained by conventional rearing and/or dissection techniques and therefore may be useful in pre- and post-release studies on candidate biological control agents. These methods can identify parasitoids to species, which is often impossible by dissection alone (Tilmon et al. 2000; Ashfaq et al. 2004; Agustíet al. 2005; Gariepy et al. 2005; Jones et al. 2005). Molecular markers for parasitoids can detect minute quantities of DNA and allow the detection of parasitoid eggs within a host, which are often missed by dissection (Day 1994; Amornsak, Gordh & Graham 1998; Tilmon et al. 2000; Zhu et al. 2000, 2004; Zhu & Williams 2002; Ashfaq et al. 2004). Additionally, information is obtained more rapidly than by rearing, which is not only tedious but may be delayed by diapause prior to parasitoid emergence (Tilmon et al. 2000; Ratcliffe et al. 2002). Most importantly, host and/or parasitoid mortality (commonly encountered in rearing) does not affect the results obtained by molecular analysis (Ratcliffe et al. 2002).

Most studies that have compared parasitism levels obtained using molecular methods, dissection and/or rearing have shown no statistical difference (Ratcliffe et al. 2002; Agustíet al. 2005; Gariepy et al. 2005; Jones et al. 2005). None the less, some studies indicate that there are discrepancies in parasitism levels estimated by the different methods (Greenstone & Edwards 1998; Tilmon et al. 2000; Ashfaq et al. 2004). Many of these were based on small sample sizes that were disproportionate between methods and, as such, may have lacked sufficient statistical power to detect significant differences, particularly with low parasitism levels (Agustíet al. 2005).

Methods that provide the most complete species composition information are necessary to identify which parasitoids cause host mortality in a pest population. Pupal mortality in rearing often prevents the identification of individuals that successfully killed their host but failed to complete development to the (identifiable) adult stage. The inability to identify the parasitoid species that cause host mortality may lead to an underestimation of the impact a parasitoid has on a host population. This is of particular concern in non-target risk-assessment studies that evaluate the ecological host range of candidate biological control agents. The increased use of molecular techniques in parasitoid ecology may lead to increased adoption and success of biological control programmes, as these techniques provide taxonomic accuracy and give a thorough understanding of population genetics and gene flow in natural enemies (MacDonald & Loxdale 2004; Menalled, Alvarez & Landis 2004).

Species of Lygus Hahn (Hemiptera: Miridae) are serious pests of a wide variety of economically important crops in North America. Two European Peristenus Förster species (Hymenoptera: Braconidae), Peristenus digoneutis Loan and Peristenus relictus Ruthe, were introduced in the USA as part of a biological control programme for Lygus and are being considered for release in Canada. Other members of the genus Peristenus include holarctic Peristenus pallipes Curtis, a species complex associated with a variety of Miridae in Europe and North America (Brindley 1939; Goulet & Mason 2006). Molecular tools are available for the identification of Peristenus species and estimation of parasitism levels in Lygus (Tilmon et al. 2000; Erlandson et al. 2003; Zhu et al. 2004; Gariepy et al. 2005). One of these molecular tools is a single-step multiplex PCR assay for the detection of three Peristenus species simultaneously, which could provide an alternative to conventional methods used to identify Peristenus species and to evaluate the ecological host range of those species considered for release in Canada (Gariepy et al. 2005).

The objective of the present study was to conduct a large-scale comparative assessment of methods used to estimate the parasitism level and parasitoid species composition in target and non-target mirid populations. Four mirid species were used as case studies to compare rearing, dissection and multiplex PCR in the estimation of parasitism by three Peristenus species in Europe. The potential use and accuracy of these methods are discussed in the context of non-target risk assessment of exotic biological control agents.

Materials and methods

field collections

Four mirid species, Lygus rugulipennis Poppius (the target host for P. digoneutis and P. relictus), Closterotomus norwegicus Gmelin, Liocoris tripustulatus F. and Leptopterna dolobrata L. (potential non-target hosts for P. digoneutis and P. relictus) were collected from various locations in Germany (Schleswig-Holstein and Baden-Württemberg), Switzerland (Baselland, Bern, Jura and Solothurn) and France (Franche-Compté and Alsace). Lygus rugulipennis was collected on red clover Trifolium pratense L. (Fabaceae), alfalfa Medicago sativa L. (Fabaceae) and chamomile Matricaria recutita L. (Asteraceae). Liocoris tripustulatus was collected on stinging nettle Urtica doica L. (Urticaceae), Leptopterna dolobrata on orchard grass Dactylus glomerata L. (Poaceae), and Closterotomus norwegicus on chamomile and flowering plants in fallow habitats.

Lygus rugulipennis was collected from a total of 30 field sites during the last 2 weeks of August in 2003 and 2004. A total of 19 field collections was made for C. norwegicus throughout the month of June in 2003 and 2005. Liocoris tripustulatus was collected from a total of 20 field sites during the first 3 weeks of July in 2003 and 2004. Leptopterna dolobrata was collected from a total of 18 field sites during the last 3 weeks of June in 2003 and 2004. At each site, 300 nymphs of a given species were collected and subsamples of 100 nymphs were designated for rearing, dissection or molecular analysis. Mirids were collected in the third to the fifth nymphal instar, when maximum parasitism levels in the field are reached (Haye 2004).

parasitism level and parasitoid species composition

Rearing

Nymphs were reared in plastic cylinders (1·2 L) fitted with removable Petri-dish bottoms. Moist vermiculite (Opticulit, Optima-Werke, Oberwil, Switzerland) was added to the Petri dishes to serve as a pupation medium for parasitoid larvae. The vermiculite was separated from the rest of the cylinder by a mesh screen that allowed parasitoid larvae to pass through but excluded nymphs. Organic beans Phaseolus vulgaris L. (Fabaceae) were used as a food source for Lygus rugulipennis and C. norwegicus nymphs. Nymphs of Liocoris tripustulatus were provided with Urtica leaves and flowers, and nymphs of Leptopterna dolobrata were provided with D. glomerata blades and flowers. Freshly cut plant material was placed in plastic vials with moist florist foam to prevent wilting and was replaced every second day.

A maximum of 50 nymphs was reared in each cylinder and mirid adults were removed daily to prevent cannibalism. When there were no nymphs remaining, Petri dishes were removed and the parasitism level was recorded as the number of cocoons obtained from 100 nymphs reared from a given site. Parasitoid cocoons were overwintered in Petri dishes filled with vermiculite in a subterranean insectary where temperatures fluctuated with outside temperatures, ranging from 2 °C to 10 °C. In April of the following year, cocoons were brought into the laboratory, incubated at 20–22 °C, and emergence was monitored daily. Parasitoids that emerged from cocoons were preserved in 95% ethanol and sent to H. Goulet (Agriculture and Agri-Food Canada, Ottawa, Canada) for identification.

Parasitoid species composition and relative abundance were based on the identification of adults that emerged from cocoons obtained at each site. Parasitoid pupal mortality was calculated for each site based on the proportion of cocoons from which no parasitoid adults emerged.

Dissection

Nymphs were dissected in 70% ethanol under a compound microscope. The abdomen was opened using fine forceps, and the contents were examined visually for the presence of a parasitoid egg or larva. For each site, parasitism was recorded as the number of nymphs in which an immature parasitoid was found. Species composition information was not obtained from dissected host material, as immature Peristenus species are morphologically indistinguishable (Bilewicz-Pawinska & Pankanin 1974).

Molecular analysis

Nymphs were preserved in 95% ethanol. DNA was extracted and amplified using a multiplex PCR assay for P. digoneutis, P. relictus and P. pallipes, as described by Gariepy et al. (2005). The parasitism level was recorded as the number of nymphs that produced a positive PCR result, which indicated the presence of parasitoid DNA. Parasitoid species were distinguished based on the size of the PCR product generated in DNA amplification: 515 base pairs (bp) for P. digoneutis, 330 bp for P. relictus and 1060 bp for P. pallipes. Parasitoid species composition and relative abundance were estimated based on the proportion of species-specific PCR reactions that were positive for P. digoneutis, P. relictus, P. pallipes or combinations thereof (multiparasitism) from the total number of parasitoids detected.

statistical analysis

To test the null hypothesis that there was no significant difference in the mean parasitism level estimated by rearing, dissection or molecular analysis, a one-way anova (P = 0·05) was used. When a significant difference was detected, a Tukey HSD test (P = 0·10) was used to determine which means were significantly different. To test the null hypothesis that there was no significant difference between rearing and molecular analysis in the estimation of the proportion of each parasitoid species, a Wilcoxon paired samples test (P = 0·05) was used.

Results

In the current study, more than 26 000 mirid nymphs were collected, with 8700 nymphs analysed with each of the methods. In Lygus rugulipennis, parasitism levels among collection sites ranged from 4% to 66% by dissection, 8–63% by rearing and 7–70% by molecular analysis. By dissection, the minimum and maximum parasitism levels in C. norwegicus were 4% and 78%, respectively. Rearing provided parasitism levels that ranged from 4% to 43%, and parasitism levels estimated by molecular analysis ranged from 7% to 57%. In Liocoris tripustulatus, parasitism levels ranged from 0% to 56% in dissected samples, 6–53% in reared samples and 9–76% in samples analysed using PCR. Dissection, rearing and molecular analysis of field-collected Leptopterna dolobrata nymphs provided parasitism levels that ranged from 4% to 48%, 5–37% and 8–47%, respectively.

Mean parasitism levels for the four mirid species investigated are shown in Fig. 1. In three of the four species investigated, there was no significant difference between the mean parasitism levels determined using the three methods (Lygus rugulipennis F2,87 = 1·00, P= 0·37; C. norwegicus F2,54= 1·07, P= 0·35; Leptopterna dolobrata F2,51 = 1·29; P= 0·28). However, in Liocoris tripustulatus there was a significant difference (F2,57 = 3·58, P= 0·03); the mean parasitism level estimated by molecular analysis was significantly higher than by rearing and dissection.

Figure 1.

Mean parasitism levels estimated by dissection, rearing and molecular analysis of four mirid species collected in Europe. For a given host species, significant differences between methods are indicated by an asterisk (anova, P < 0·05; Tukey HSD, P < 0·10).

The mean proportion of parasitoids, hyperparasitoids, pupal mortality and multiparasitism obtained from rearing and molecular analysis of the target host, Lygus rugulipennis, is shown in Fig. 2. The mean proportion of P. digoneutis estimated by rearing and molecular analysis was not significantly different (Z = –1·59, P=  0·11). However, the mean proportion of P. relictus was significantly higher by molecular analysis compared with similar results obtained in the rearing of Lygus rugulipennis (Z = –3·25, P= 0·001). No significant difference in the mean proportion of P. pallipes estimated by rearing and molecular analysis was observed (Z = –1·34, P= 0·18). Hyperparasitoids of the genus Mesochorus are not detected by molecular analysis and information on multiparasitism is not available by rearing, thus comparisons for these categories were not possible.

Figure 2.

Mean proportion of parasitoids, hyperparasitoids, pupal mortality and multiparasitism from rearing and molecular analysis of the target host Lygus rugulipennis. Hyperparasitism by Mesochorus is not detected by molecular analysis and multiparasitism is not detected by rearing.

Species-composition information (based on the mean proportion of P. digoneutis, P. relictus, P. pallipes, hyperparasitoids, pupal mortality and multiparasitism) obtained by rearing and molecular analysis of the non-target hosts, C. norwegicus, Liocoris tripustulatus and Leptopterna dolobrata is shown in Fig. 3. In C. norwegicus, the proportion of P. digoneutis was not significantly different when estimated by rearing and molecular analysis (Z = 0·42, P= 0·68). However, the proportion of P. relictus detected using molecular analysis was significantly higher than in reared samples (Z = –2·55, P= 0·01). Similar results were obtained for P. pallipes (Z =–3·51, P < 0·001). There was no difference between rearing and molecular analysis of Liocoris tripustulatus nymphs in the proportion of P. digoneutis (Z = 0·45, P= 0·66) and P. relictus (Z = –1·60, P= 0·11). However, the proportion of P. pallipes was significantly higher when estimated by molecular analysis than by rearing (Z = –3·92, P < 0·001). In Leptopterna dolobrata, the mean proportion of P. digoneutis was not significantly different when estimated by rearing and molecular analysis (Z = –1·34, P=0·18) but there was a significant difference between the mean proportion of P. relictus reared and the mean proportion detected by molecular analysis (Z = –2·38, P= 0·02). Similarly, the mean proportion of P. pallipes was significantly higher by molecular analysis than by rearing (Z = –3·73, P < 0·001).

Figure 3.

Mean proportion of parasitoids, hyperparasitoids, pupal mortality and multiparasitism from rearing and molecular analysis of the non-target hosts Closterotomus norwegicus, Liocoris tripustulatus and Leptopterna dolobrata. Hyperparasitism by Mesochorus is not detected by molecular analysis and multiparasitism is not detected by rearing.

Discussion

The use of molecular methods in ecological studies on biological control agents is increasing (Symondson 2002; Sheppard & Harwood 2005; Behura 2006; Greenstone 2006; Stouthamer 2006; Gariepy et al. 2007). However, for molecular analysis to gain widespread adoption in ecological studies on natural enemy host range, it must provide information that is equivalent (or superior) to information obtained by conventional methods.

Two earlier studies on plant bug parasitoids suggested that molecular and conventional methods provide differing estimates of parasitism (Tilmon et al. 2000; Ashfaq et al. 2004). Both of these studies used rearing, dissection and molecular methods to estimate parasitism levels in Lygus plant bugs. Tilmon et al. (2000) indicated that molecular analysis and/or dissection provided higher estimates of parasitism than rearing; however, no statistical comparisons were made. Furthermore, collections were made at a single site, and the sample size was small and involved unequal sample sizes for the methods tested. Ashfaq et al. (2004) showed that PCR provided significantly higher estimates of parasitism level compared with dissection and rearing. However, sample sizes were again small and differed between methods.

The current study shows that, in most cases, all three methods give statistically equivalent estimates of parasitism level in a variety of host populations. However, in Liocoris tripustulatus rearing and dissection provided significantly lower mean parasitism values than molecular analysis. One possible explanation is that Liocoris tripustulatus nymphs were collected too early in the season, when parasitized nymphs probably contained the egg stage of the parasitoid. Although Peristenus eggs are easily detected by molecular analysis within hours of oviposition (Gariepy 2007), they are often overlooked in dissection because of their small size, which would result in an underestimation of the parasitism level. For rearing, it was probably the effect of host mortality during the extended rearing phase that led to a lower estimate of parasitism level. Therefore, the discrepancy between the data obtained using conventional and molecular methods could have resulted from the fact that molecular analysis was able to detect parasitism at an earlier stage than dissection and because molecular estimates were not confounded by mortality factors encountered in rearing. In general, parasitism levels estimated by conventional and molecular methods are equivalent in late-season field collections; however, parasitism levels in early season collections may be more accurately estimated by molecular analysis. Detection of a parasitoid in a host by molecular analysis does not necessarily indicate survival of the parasitoid to the adult stage, and caution should be exercised when the impact of the parasitoid on the host population is estimated solely on the presence of parasitoid DNA.

Parasitism levels (estimated by rearing) in these mirid species were comparable to those obtained in other studies in Europe (Bilewicz-Pawinska & Pankanin 1974; Bilewicz-Pawinska 1976; Haye et al. 2005; Haye et al. 2006). For each mirid species, the mean proportion of P. digoneutis estimated by rearing and molecular methods was not significantly different. This indicates that both methods provide equivalent estimates of the occurrence of this parasitoid species in the target and non-target hosts. However, in Lygus rugulipennis, C. norwegicus, and Leptopterna dolobrata, the mean proportion of P. relictus was significantly higher when estimated with molecular methods. In fact, the proportion of P. relictus in Lygus rugulipennis was approximately two times higher through molecular analysis than rearing. Although P. relictus is a minor contributor to the parasitoid species composition in the non-target hosts, the estimated proportion of this parasitoid species was approximately 10 times higher by molecular analysis than by rearing in C. norwegicus and Leptopterna dolobrata. This suggests that, when only rearing methods are used, the presence of P. relictus in mirid populations may be significantly underestimated.

The mean proportion of P. pallipes in C. norwegicus, Liocoris tripustulatus and Leptopterna dolobrata estimated by molecular analysis was significantly higher than the mean proportion estimated by rearing. This is clearly because of the levels of pupal mortality observed in reared samples, where 30–60% of the cocoons failed to produce a parasitoid adult.

Hyperparasitoids were obtained from reared samples but in fairly low proportions (usually less than 10%). Because the hyperparasitoid consumes the Peristenus larva within the cocoon, only the hyperparasitoid adult emerges in rearing and the identity of its Peristenus host is lost. This can have a large impact on abundance estimates of each species if one parasitoid species in a community of natural enemies is hyperparasitized more frequently than the others. In addition, it would render rearing data inaccurate, as the proportion of the preferred parasitoid species would be underestimated when only hyperparasitoids emerge from these individuals. Molecular analysis was unable to provide information regarding hyperparasitism because PCR primers for Mesochorus were not incorporated in the multiplex assay. This illustrates one of the limitations of molecular analysis; only those species for which primers are designed can be detected. However, molecular analysis of mirid nymphs should allow the detection and identification of Peristenus individuals even when hyperparasitized because DNA from the primary parasitoid would still be present. This prevents the loss of species composition information on primary parasitoids as a result of hyperparasitism but overlooks the impact of hyperparasitoids on primary parasitoids (Chen et al. 2006). The development of a system in which parasitoids and hyperparasitoids could be detected simultaneously would be ideal, as it would facilitate investigations on host–parasitoid–hyperparasitoid associations.

Detection of multiparasitism can provide information on competition between parasitoids that utilize the same host species. Interspecific competition for host resources can be detrimental to achieving biological control of a pest species (Levesque et al. 1993; Heinz & Nelson 1996; Urbaneja et al. 2003; De Moraes & Mescher 2005; Batchelor, Hardy & Barrera 2006; Muli et al. 2006). In the current study, multiparasitism was low and there appeared to be little interspecific competition between the Peristenus species investigated. However, there are very little data on the occurrence of multiparasitism in field-collected Miridae. Euphorine wasps are solitary endoparasitoids and, even when parasitized by multiple individuals, only one parasitoid (if any) will be reared from the host (Lachance, Broadbent & Sears 2001). Although host dissection may detect the presence of multiple parasitoid larvae, the inability to distinguish species prevents further information from being recorded. The use of the multiplex PCR assay allowed the detection of multiparasitism, as well as the identification of the species responsible. This may be useful in future studies on the occurrence of multiparasitism in Miridae, and would be of particular interest in areas where P. digoneutis and P. relictus have already been introduced, as they may compete with native parasitoids for host resources.

Direct non-target effects are often evaluated by rearing field-collected non-target hosts to determine whether the biological control agent of interest is present in the population. In this context, the loss of species-composition information as a result of host and/or parasitoid mortality may allow non-target parasitism to go undocumented. This is conceivable, as there could be fitness consequences for a parasitoid that develops in a non-target host. If this is the case, it may lead to an increased level of parasitoid mortality. These parasitoids may still result in the death of the host, and thereby in non-target effects, but the identity of the responsible parasitoid would not be obtained in the rearing process. It has been acknowledged that hosts that fail to produce a parasitoid adult may bias parasitism estimates obtained in rearing (Ratcliffe et al. 2002; Greenstone 2006). However, there appears to be no reference to the potential impact of parasitoid mortality on the accuracy of measurements for parasitoid species-composition and/or host-range data in non-target risk-assessment studies.

In the present study, a high proportion of cocoons reared from C. norwegicus, Liocoris tripustulatus and Leptopterna dolobrata failed to produce parasitoid adults. It is clear that the parasitoid pupal mortality observed in these reared samples is mainly attributable to P. pallipes, rather than P. digoneutis and P. relictus. In most cases, rearing alone would not allow the conclusive statement that unemerged parasitoids belong to a certain species because there is no way to identify such specimens. However, the identity of unemerged parasitoids can be inferred quite well using paired molecular data. Without this information, the possibility that some or all of the unemerged parasitoids were P. digoneutis or P. relictus could not be excluded. Correct estimation of the impact of candidate biological control agents on non-target populations is essential in risk-assessment studies. Molecular methods for parasitoid detection may facilitate and expedite identification of candidate biological control agents in non-target host populations and may be more suitable than conventional methods for ecological host-range studies on a given parasitoid species. Knowledge of parasitoid species and their specificity will enable more rigorous selection of candidate agents for the biological control of economically important insect pests.

According to Shaw (1994), methods that allow quantitative expression of the realized host range to be related to the performance of the parasitoid within a host population are essential. When sufficient rearing data are available on the parasitoid community associated with a given host species, molecular assays can be developed that encompass all members of this community. When this is the case, molecular methods for the estimation of parasitoid species composition have the potential to provide more complete information than rearing. However, rearing is essential to account for rare or unexpected species that may be overlooked by molecular analysis (Agustí et al. 2005). Although rearing provides important qualitative data on the parasitoid community associated with different host species, molecular methods are likely to provide better quantitative data in ecological studies on parasitoid host range and non-target risk.

Acknowledgements

The authors would like to thank the anonymous referees and associate editor for useful comments and suggestions to improve this manuscript. Many thanks to Henri Goulet for the identification of reared parasitoids specimens, and Tim Haye, Lars Andreassen, Leonore Lovis and Jake Miall for assistance in the field. This research was supported by the University of Saskatchewan and Natural Science and Engineering Research Council of Canada (scholarships for T. D. Gariepy) and Biocontrol Network of Canada.

Ancillary