The Glanville fritillary, Melitaea cinxia, is a Eurasian butterfly that reaches its northern range limit in southern Finland. Adult females lay eggs in clusters of about 150 in June on the host plants Plantago lanceolata L. and Veronica spicata L. (Plantaginaceae), which grow in dry meadows. The butterfly larvae live gregariously through the summer, diapause gregariously through the winter, resume feeding in the spring, and pupate in the late spring [52]. The parasitoid Cotesia melitaearum is entirely specialized on M. cinxia, and occurs throughout the geographical range of the butterfly [53]. In the study area, the wasp has two or three generations in each year and per host generation, laying one to 40 eggs in an individual host larva, depending on the host larval instar [54,55].

A host larva is able to encapsulate a fraction of the parasitoid eggs laid in it. Encapsulation ability differs among host families [56] and with the environmental conditions. For instance, it has been shown, using nylon threads as an assay for encapsulation response, that encapsulation ability increases with host food plant quality [8] and with stress [57].

The insects in this study originate from five northern European localities (Figure 1, Table 1). Those from the Åland Islands (AL) in Finland have been under study for many years. In the Åland Islands, the host occurs as a metapopulation in a fragmented landscape [58]. In any one year, the parasitoid occupies only a small fraction (3 to 10%) of the local butterfly populations, persisting for the most part as small populations constrained by the population dynamics of the host, fragmentation of the landscape, and strongly density-dependent hyperparasitism [54,59]. Isolation of local populations leads to inbreeding in both the butterfly [60,61] and the parasitoid [62]. Uppland (UP), along the eastern coast of Sweden, has a fragmented habitat as in Åland, presumably with similar host and parasitoid population dynamics and degree of inbreeding. The Swedish island of Öland (OL) and the Estonian island of Saaremaa (SA) have large continuous areas of suitable habitat and large butterfly populations. Finally, the tiny island of Pikku Tytärsaari (PT) harbors a single population of the host and of the parasitoid within one coastal meadow of 10 ha. The PT population has been completely isolated for at least 75 years [51]. All the above butterfly populations belong to the same phylogeographic clade (the eastern clade; in Wahlberg and Saccheri [63], the OL population was erroneously assigned to the SE clade), but AL and UP are much more closely related to each other than the others [64]. All the wasppopulations are closely related to populations in western Europe [53]. The isolated population on PT originated from populations that are closely related to SA populations [51].

Melitaea cinxia larvae were collected from multiple sites within each locality in the autumn 2009. The insects used for the experiment were the progeny of field collected individuals. A fraction of the host larvae collected from all sites except OL were naturally parasitized by the wasp. The parasitoid occurs in OL [53], but by chance was not collected with the sample of host larvae. The host larvae were kept under winter-like conditions in the laboratory until late April 2010, when they were taken out of diapause and reared to adulthood or until parasitoids emerged. Adult butterflies were mated with individuals from the same population but different families. One egg cluster from each of five different mothers from each origin (AL, UP, OL, SA) was used for the experiment. No adult butterflies from the isolated PT island were available. Cotesia melitaearum emerged from the field-collected host larvae. In order to obtain the wasps for the experiment, five independent females from each origin (AL, UP, PT, SA) were individually mated with males the same origin but different families. The mated females then each oviposited into laboratory-reared host larvae originating from AL. Fourteen to 18 female progeny from each of these mothers were then mated to unrelated males from the same origin and used in the experiment.

Upon hatching, the host larvae were maintained in a growth chamber in family groups and fed fresh V. spicata leaves daily until their third instar, at which point 40 larvae from each of the five families per population were separated into groups of 10. The larvae were reared in groups because they are intrinsically gregarious. One group of 10 larvae from each of the five host families was parasitized by wasps from each of the four wasp populations. Each parasitism, defined as an insertion of the ovipositor, was observed. Three wasps were used to parasitize a set of 10 larvae, each of the three wasps parasitizing three or four host larvae. For every wasp population, this was repeated for each of the five independent host families, The three wasps used for each set of 10 larvae represented different combinations of the progeny of the five original female wasps from each locality. Thus, we replicated five times the interaction between a set of 10 host larvae from each host population with a set of three wasps drawn from the progeny of five female wasps from each wasp population. Notice that the same five host families interacted with all the wasp populations, which makes it possible to analyze the host family effect.

Once parasitized, the host larvae were placed back in the growth chamber and fed V. spicata leaves daily. After four days, when parasitoid eggs were hatching, three larvae from each group were dissected and the numbers of live and encapsulated parasitoid eggs and larvae were counted. The remaining host larvae were left to develop until parasitoids emerged, or the host larvae reached diapause. Those that reached diapause without parasitoids emerging from them were dissected, and the numbers of live and encapsulated first, second and third instar parasitoid larvae were counted. The numbers of the third (final) instar parasitoid larvae and pupae were considered together as the individuals emerging before diapause. Many host larvae in the diapause stage contained first or second instar parasitoid larvae, which were designated as destined to overwinter in the host to reach adulthood in the following spring. Host larvae that died without parasitoids coming out of them were dissected and the encapsulated and live parasitoid larvae were counted. Finally, after parasitoid larvae had left the host, the host was dissected and the remaining early instar parasitoid larvae, which would die along with the host, and encapsulated parasitoid larvae were counted.

In the above experiment, the host larvae were dissected at diapause before the pupation of most C. melitaearum, and hence the adult size and longevity of the parasitoid could not be measured. To compare the parasitoid populations in these characteristics, we used another set of parasitoids in the following spring to measure brood size at pupation, pupal weight, and adult longevity. To rear this set of wasps, the parent generation from the previous experiment was used to parasitize additional overwintering host larvae from Saaremaa (SA). Twenty-one to 31 host larvae were parasitized by five mothers from each origin. The parasitized hosts were allowed to diapause. Upon breaking diapause, the larvae were reared until wasps pupated. Five females from each family were mated to random males from the same origin and allowed to parasitize five post diapause (6th instar) host larvae from Åland (AL). The parasitized larvae were reared under laboratory conditions until the parasitoidsemerged and pupated. The brood size, individual pupal weight at six hours, number of days to adulthood, and sex were recorded. To measure adult longevity, each wasp was kept individually in a temperature-controlled chamber and fed honey water (3:1) daily until death.

The statistical analyses were conducted using JMP v9 (SAS, Cary, NC, USA). In the parasitoid by host origin experiment (Section 2.3), initial and live brood size, fraction of brood encapsulated at four days, rate of development at four days, live brood size at diapause and the fraction maturing in autumn were analyzed with respect the wasp origin, host origin, host family nested within origin, and the interaction between wasp and host origins using least squares ANOVA. Logistic regression was used to analyze the presence/absence of live parasitoids in a host, with respect to the same variables. In the experiment on parasitoid pupae and adult longevity (Section 2.4), least-squares ANOVA was used to evaluate brood size, pupal weight, pupal development time, and adult longevity. Brood size and the interaction of brood size with population origin were included in the model of pupal weight because the latter increases with decreasing brood size due to competition among the parasitoid larvae sharing the same host individual (linear regression R² = 0.08, p = 0.024). Pupal weight and the interaction between pupal weight and population origin were included in the model of adult longevity, since the latter tends to increase with wasp size (linear regression R² = 0.13, p ≤ 0.0001). Wasp mother was initially included as a factor nested in wasp origin, but the data were unbalanced and as no trend was observed upon inspection, the mother was left out of the final model. For each analysis, the distributions of the variables were tested for normality; no transformations were needed. Post hoc independent contrasts were used where needed to compare within multilevel factors.

Four days after parasitism most parasitoid eggs had hatched and some had already been encapsulated as eggs or larvae. The initial brood size, including both live and encapsulated individuals, was 2.5 on average, ranging from 0 to 7. There were significant differences among wasp origins (Table 2), with PT wasps producing the largest initial and live broods (Figure 2a,b). Host family had a highly significant effect, but host origin did not (Table 2). The initial brood size of PT wasps was large, primarily due to their performance in OL larvae (Figure 2a), but the overall interaction between wasp and host origins was not quite significant (Table 2). At this early stage, about 30% of parasitoid eggs and larvae were encapsulated with no systematic variation among wasp or host origins (Table 2).The fraction of living parasitoids that were still eggs at day 4 was used as an estimate of the rate of development of the wasps. This fraction ranged from 9 to 25% and varied significantly among the host and wasp origins (Table 2). Wasps from UP and wasps in hosts from OL developed slowly (Figure 3). The rate of development also depended on the specific wasp-host combination (Table 2: significant interaction term). Thirty-five out of the 240 larvae (15%) dissected at four days contained no wasp eggs or larvae. The absence of the wasp did not differ between wasp or host origins nor between host families.

The total parasitoid brood size at host diapause (fifth instar), including both live and encapsulated wasp larvae, was 2.17 on average, which is only slightly smaller than at four days (2.5). Similarly, the same fraction of hosts (16%) as at four days (15%) had no parasitoid progeny in them. However, in clear contrast to early development, at the host diapause stage, a large fraction of the parasitoid individuals in each brood was encapsulated, and the mean live brood size was only 0.91, against 1.75 at four days (Figure 2b,c). At diapause, PT wasps had large live broods. The size of the brood was unrelated to host origin (Table 3, Figure 2c), but it differed between host families (Table 3). The fraction of the brood encapsulated differed among host families and host origins, but was unrelated to wasp origin (Table 3). The largest fraction was encapsulated by UP host and the smallest by OL hosts (UP mean = 0.70(0.04), post hoc test F_1,382 = 4.83. p = 0.028; OL mean = 0.52(0.04), post hoc test F_1,382 = 9.08, p = 0.029). About 40% of the host larvae that had been parasitized successfully encapsulated the entire brood. Encapsulation of the entire brood differed among host families and host origins but not among wasp origins (Table 4). Host individuals from OL were the least successful in overcoming parasitism (27% encapsulation rate; χ² = 8.76, p = 0.003).

At the host diapause stage, some C. melitaearum larvae develop into their last (third) instar and leave the host, while others remain small (first and second instar) and stay in the host through diapause to become adults in the following spring [55,65]. In this experiment, too, some parasitoid larvae pupated at host diapause, whereas the remaining host larvae were dissected. Those parasitoids that left the host to pupate and those that were in the 3rd instar (though still inside the host) were interpreted as maturing in the autumn. About 35% of all parasitoid larvae matured in the autumn. The fraction of maturing larvae was highest in PT and lowest in UP wasps (Table 3, Figure 4a), and generally, a larger fraction of wasps from all origins matured in OL hosts than in hosts from the other origins (Table 3, Figure 4b).

Parasitoid brood sizes in the spring generation, the progeny of the overwintering individuals, ranged from 1 to 10 with the mean of 5.1. Wasps from AL had the largest broods (Table 5; post hoc test F_1,460 = 24.79, p < 0.0001). The wasp pupae took about 6 days to develop into adults, with no differences among populations. Female pupae were larger than males and pupal weights differed among the wasp populations (Table 5), with female AL pupae being the largest (1.77 ± 0.54 mg) (Table 5; post hoc test F_1,456 = 5.73, p = 0.017) and UP males being the smallest (1.19 ± 0.67 mg) (Table 5; post hoc test F_1,456 = 19.41, p < 0.0001). Female wasps lived on average twice as long as males (41 vs. 19 days), and AL females lived the longest (Table 5, Figure 5). Adult longevity increased significantly with pupal weight (Table 5).

Rate of encapsulation is clearly affected by attributes of the host. Four days after parasitism about 30% of the wasp eggs and small larvae had been encapsulated. This fraction was unrelated to host and wasp origin. By the time the host larvae had reached the diapause stage (5th instar), the encapsulation rate had increased to 60%. At this stage the fraction of the brood that was encapsulated depended on host origin but was unrelated to wasp origin. For the parasitoid this result contrasts what has been found for a related species, Cotesia sesamiae, which shows variation in encapsulation ability among populations [66]. However, in the latter case, there was partial reproductive isolation between populations using different hosts due to Wolbachia infection [12,67]. Population-level differences in the encapsulation ability of the host are not surprising as such differences have been found previously, even in Cotesia-Lepidoptera interactions [45], and encapsulation ability of M. cinxia is known to vary among families and with stress [8,56].

About 40% of the host larvae successfully encapsulated the entire brood of parasitoids that was laid in them. Thus, a large fraction of the parasitism attempts by the wasp failed, which is surprising because the species is entirely specialized to M. cinxia [68]. Host larvae from OL were the least successful at overcoming parasitism, with only 27% of them eliminating the parasitoid by the time they reached the diapause stage.

In the reciprocal parasitism experiment, the summer generation of wasps from PT had the largest initial brood size (number of eggs laid in a host), as well as the largest brood at host diapause. In parasitoid wasps, brood size is often related to female size and fecundity, as well as the ability of the parasitoid to overwhelm the host immune responses [24,69,70]. In parasitoids from PT, however, large broods were not associated with large female body size [71].

Parasitoid broods were larger in the spring generation, when the adult traits were measured, than in the summer generation, when the reciprocal parasitism experiment took place. This is to be expected because, after winter diapause the host larvae are larger than before diapause and, hence, female wasps typically oviposit more eggs into them [72]. In the summer generation PT wasps produced large broods (Figure 3c), whereas in the spring AL wasps produced the largest broods (Table 5). The spring AL wasps also had the heaviest pupae and the longest adult female lifespan. Because of their greater longevity, AL wasps are best adapted in late spring, when they must wait for the next generation of host larvae, which become available about four weeks after the wasps become adult. Great longevity is much less significant in the other two wasp generations, in which host larvae of the same generation are immediately available for parasitism by newly hatched adult parasitoids [55].

Rate of parasitoid development within a host individual is significant both for the parasitoid and the host. Parasitoids that develop slowly experience prolonged exposure to the host immune system, and they are vulnerable to host mortality and hyperparasitism [65]. Slow parasitoid development may be advantageous to the host because it spreads out the resource depletion caused by the parasitoid. On the other hand, for a koinobiont parasitoid such as Cotesia, developing slowly may be advantageous as it allows the host to grow, providing greater food resources during wasp larval development [69]. In this study, wasps developed from eggs to first instar slowly in hosts from OL, especially in the case of PT wasps. One explanation for variation in the development rate of gregarious parasitoids is food limitation, which can be related to brood size and host size [73]. In the present case, direct competition among brood mates for food is unlikely when the parasitoids are eggs or first instar larvae. Some other physiological mechanism(s) must play a role in affecting parasitoid development rate in the hosts. It is noteworthy that parasitoid development rate was lowest in OL hosts, which were least able to encapsulate the entire parasitoid brood, suggesting that slow parasitoid development does not necessarily lead to effective host defenses.

When the host larva reaches the diapause stage, the parasitoid larva can stay in the first or second instar and go into diapause, or develop to the third instar, exit the host, become an adult in the autumn, and parasitize nearby host larvae. Normally, a fraction of the progeny of summer wasps become adults in August to produce an additional autumn generation, while the majority stay within the host over the winter [55]. In this experiment about 35% of the wasps matured in the autumn. The fraction and number of wasps maturing in the autumn was highest in PT and lowest in UP hosts. The decision to stay within the host may be affected by resource availability, such that parasitoid larvae within small hosts choose to stay over the winter and finish development when the hosts grow larger in the spring. If this were the case, one might expect larger broods to stay within the host longer. Instead, we found that PT wasps, which had the largest broods, also matured most frequently in the autumn. For any parasitoid it would be advantageous to have three rather than two generations per host generation. This is especially relevant for C. melitaearum, which parasitizes only a small fraction of the hosts [55] that have a high rate of overwintering mortality, often exceeding 50% [60,74].

Considering the different host populations, the largest fraction of wasps from all origins matured in the autumn in OL hosts. The high rate of parasitoid population growth due to three rather than two generations per host generation makes the OL host population vulnerable to high rates of parasitism in the spring, Combined with the low rate of encapsulation of entire parasitoid broods by OL host larvae, this makes them appear to be particularly vulnerable to parasitism, potentially leading to population decline [54,55].

The conventional view of hosts and parasites is that, under a wide range of conditions, host populations evolve to become more resistant to local parasites, and the latter become more virulent, at least up to a point, against local hosts [3]. In the coevolutionary context, one species or the other may be ahead in the “arms race” and there may be temporal fluctuation in host resistance and parasite virulence [75]. Nonetheless, some systematic variation in resistance and virulence traits is expected when comparing local versus non-local interactions [2,76]. Local adaptation of parasites to hosts, and hosts to their parasites, has been detected in many organisms [76], but only a few studies have been conducted on local adaptation in parasitoid wasps and their hosts. Henter and Via [31,32] demonstrated the potential for coevolution by quantifying genetic variation of both resistance by the pea aphid Acyrthosiphon pisum and of virulence of its parasitoid Aphidius ervi within a single pea aphid population. Sandrock [77] found genetic variation in resistance within each of four populations of another aphid, Aphis fabae, and virulence of its parasitoid, Lysiphlebus fabarum, but no genetic differentiation or genotype specific interactions, suggesting lack of local adaptation in spite of genetic variation. Dupas and colleagues have shown, using Drosophila and the parasitoid L. boulardi, surprisingly simple gene-for-gene style coevolution of encapsulation between the fly and the parasitoid [50,78]. In contrast, Kraaijeveld and colleagues compared susceptibility of D. melanogaster and virulence of another parasitoid, A. tabida, across Europe and found no local adaptation. Instead, host resistance was highest in south-central Europe and lower elsewhere, while virulence of the parasitoid showed a cline from low in northern Europe to high in southern Europe. They attributed the lack of local adaption in virulence and resistance to the costs of defense and resistance [5] relative to selection on other traits associated with local environments, and the composition of the communities in which the species occurred [79].

In the present study, we found that neither the host nor the parasitoid performed better or worse when interacting with local versus non-local hosts/parasitoids. Rather than exhibiting a pattern of local adaptation, host populations differed in their susceptibility to parasitism by wasps from all origins as measured by their ability to encapsulate parasitoid eggs and larvae. Thus UP hosts were able to encapsulate the largest fraction of parasitoid broods and OL hosts were the least able in this respect. In the latter, the result was a low rate of complete recovery from parasitism (encapsulation of all parasitoid eggs and larvae). These results appear to be in line with what has been found in the Drosophila-A. tabida system and in the pea aphid system, suggesting complex patterns of natural selection on quantitative traits. They differ from what has been found for the parasitoid C. sesamiae and its stem-boring moth hosts in Africa. While reciprocal local adaptation has not been tested explicitly, the parasitoid, which has a wide host range, was more virulent to local than novel host populations [80] and host species [81]. Each population in our study was represented by the progeny of only five mothers, therefore we could have missed sampling adapted alleles within the populations by chance. If this were the case, then a larger scale study could reveal local adaptation such as Chinwada [78] found, of relatively simple traits such as encapsulation ability [28].

All of the populations in this study area have relatively recent common ancestry. In the absence of strong selection, populations of recently shared ancestry should have similar characteristics. Among the host populations in this experiment, host butterfly populations from AL and UP have the most recent common ancestry, as assessed by genome-wide genetic variation [64]. Thus, we might expect these two populations to perform similarly when interacting with parasitoid populations from different origins. Though parasitoids’ development rate was similar in these two host populations (Figure 3), they were not particularly similar in other traits. This result indicates that, while there was not local adaptation, there was local genetic differentiation on a short-term scale, on the order of hundreds of generations at most, in traits related to interaction with the parasitoid.

Another specific comparison among populations in the study can be made between parasitoids from the populations of PT and SA. Those from the small island of Pikku Tytärsaari (PT) have been isolated from other populations for at least 75 years [51]. The colonizers originated from populations that are closely related to populations in Saaremaa (SA) among our study populations, and have subsequently experienced strong inbreeding. Inbreeding depression typically lowers the ability of a population to withstand environmental stress [82], and is a well-established phenomenon in insects, including the host butterfly M. cinxia in this study, both in PT [51] and in local populations in AL [61]. We expected that inbreeding would have a negative effect on PT parasitoids, but we found no such evidence, as the PT wasps had the highest fecundity and developed faster than wasps from any other origin. In particular, they did not perform worse than the SA wasp population in any measure. Inbreeding depression in parasitoids is apparently rare, at least in part because of haplodiploidy [30]. In Hymenoptera, males are haploid and, hence, recessive deleterious mutations are expressed and selected against in all populations. Though other potential negative effects of inbreeding apart from the expression of recessive mutations exist [83], purging of deleterious alleles must have a strong effect. In spite of this, it is interesting to note that that there was no evidence of inbreeding depression of encapsulation ability, which is a trait expressed only by females, as it is regulated by factors such as polyDNA virus injected into the host upon parasitism.