Several ecological and genetic factors affect the diet specialization of insect herbivores. The evolution of specialization may be constrained by lack of genetic variation in herbivore performance on different food-plant species. By traditional view, trade-offs, that is, negative genetic correlations between the performance of the herbivores on different food-plant species favour the evolution of specialization. To investigate whether there is genetic variation or trade-offs in herbivore performance between different food plants that may influence specialization of the oligophagous seed-eating herbivore, Lygaeus equestris (Heteroptera), we conducted a feeding trial in laboratory using four food-plant species. Although L. equestris is specialized on Vincetoxicum hirundinaria (Apocynaceae) to some degree, it occasionally feeds on alternative food-plant species. We did not find significant negative genetic correlations between mortality, developmental time and adult biomass of L. equestris on the different food-plant species. We found genetic variation in mortality and developmental time of L. equestris on some of the food plants, but not in adult biomass. Our results suggest that trade-offs do not affect adaptation and specialization of L. equestris to current and novel food-plant species, but the lack of genetic variation may restrict food-plant utilization. As food-plant specialization of herbivores may have wide-ranging effects, for instance, on coevolving plant–herbivore interactions and speciation, it is essential to thoroughly understand the factors behind the specialization process. Our findings provide valuable information about the role of genetic factors in food-plant specialization of this oligophagous herbivore.
Food-plant specialization of insect herbivores can ultimately lead to speciation, and therefore, it is considered one of the most important factors responsible for the enormous diversity of insects (Berlocher & Feder, 2002; Matsubayashi et al., 2011). Most herbivores are specialized to feed on few plant species that belong to a single plant family, and less than 10% of herbivore species feed on plants from more than three different plant families (Schoonhoven et al., 2005). Variation in plant secondary chemistry may drive food-plant specialization process in two ways. Firstly, plant secondary chemistry may function as a direct selective factor that determines the preference, performance and specialization of the herbivores to different host species (e.g. Rank, 1992; Becerra, 1997; Rasmann & Agrawal, 2011). Secondly, if herbivores use plant chemicals as defence against their natural enemies, specialization to a toxic host plant that provides predator avoidance for the herbivore could evolve rapidly (Bernays & Graham, 1988; Singer, 2008). Insects may even prefer host plants that provide an ‘enemy-free space’ but are not nutritionally optimal (Singer et al., 2004). The degree of specialization may vary among herbivore populations, and even individuals within the same herbivore population might vary in their host plant preference and performance on different host plants (Fox & Morrow, 1981; Ueno et al., 2003; Schoonhoven et al., 2005; Singer, 2008). Indeed, polyphagy may be more common at the population level than at the individual level if individuals of particular ‘composite generalist’ species are feeding on different food-plant species (Singer, 2008). In general, specialization is a dynamic process that varies also in time (Thompson, 1994).
Food-plant specialization is affected and constrained by several ecological and genetic factors (Fox & Morrow, 1981; Bernays & Graham, 1988; Futuyma et al., 1995; Forister et al., 2007). Food availability varies spatiotemporally across herbivore populations and therefore may restrict specialization of an herbivore to a single food-plant species. By conventional view, being a specialist may be beneficial in predictable conditions where plant resources are abundant over space and time, enabling adaptation to plant nutritional quality and defensive compounds. Spatiotemporal variation in food availability may favour, in turn, generalist herbivores (e.g. Futuyma, 1976; Fox & Morrow, 1981; Fox & Caldwell, 1994). For example, seeds are extremely nutritious food, but have considerable spatiotemporal variation in their distribution. Such variation in seed production may force seed predators to feed on other than their optimal host plant species.
In this study, we focused on two genetic factors that may affect herbivore specialization. Firstly, the evolution of specialization is influenced by the level of genetic variation in herbivore performance on different food plants for selection to act on. Ultimately, lack of genetic variation may prevent adaptation and specialization to current and novel food-plant species (e.g. Futuyma et al., 1995; Forister et al., 2007). Secondly, one of the most popular approaches to study herbivore food-plant specialization has been to test for fundamental trade-offs, that is, negative genetic correlations between the performance of the herbivores on different food-plant species. A trade-off exists if a genotype well suited to one food plant has relatively poor performance on others due to antagonistic pleiotropy or linkage equilibrium of genes (Agosta & Klemens, 2009). By traditional view, the existence of trade-offs favours the evolution of specialization (Fry, 2003; Scheirs et al., 2005). There is evidence both for and against such trade-offs from several plant–herbivore systems (e.g. Gould, 1979; Fry, 1990; Karowe, 1990; Fox & Caldwell, 1994; Futuyma et al., 1995; Mackenzie, 1996; Thompson, 1996; Keese, 1998; Via & Hawthorne, 2002; Forister et al., 2007; Agosta & Klemens, 2009; García-Robledo & Horvitz, 2011). The reproductive mode of organism may influence on detecting and existence of trade-offs (Fry, 1990; Scheirs et al., 2005). Indeed, the majority of earlier studies that have found trade-offs between the performances of the herbivores on different food-plant species have studied asexually reproducing species, such as mites and aphids (e.g. Gould, 1979; Fry, 1990; Via, 1991; Mackenzie, 1996; Via & Hawthorne, 2002). For instance, Mackenzie (1996) found trade-offs between fecundity of the black bean aphid, Aphis fabae, on two host plants, whereas Agosta & Klemens (2009) found no trade-offs between growth and survival of the moth Rothschildia lebeau on different host plants. Moreover, trade-offs in host use of insect herbivores have been studied using limited variety of oligophagous or specialist herbivores, and the results are contradictory (Via, 1991; Thompson, 1996; Keese, 1998; Via & Hawthorne, 2002; Forister et al., 2007; García-Robledo & Horvitz, 2011).
To investigate whether genetic factors affect food-plant specialization of the seed-eating true bug, Lygaeus equestris, we conducted a feeding trial in laboratory using seeds of four plant species, Vincetoxicum hirundinaria, Crepis tectorum, Tanacetum vulgare and Verbascum thapsus. Although L. equestris most commonly occurs on V. hirundinaria, and is somewhat specialized on this species, it occasionally feeds on several alternative food-plant species (Solbreck & Kugelberg, 1972; Kugelberg, 1973a, 1974; L. Laukkanen, pers. obs.). However, feeding on alternative plant species may lead to higher predation risk, higher mortality, longer developmental time, reduced adult biomass and lower reproductive output of L. equestris (Kugelberg, 1973a,b; Tullberg et al., 2000). On the other hand, seed production of V. hirundinaria varies considerably among years and populations, mainly due to variation in sun exposure (Solbreck and Sillén-Tullberg, 1986a,b; Ågren et al., 2008) and seed predation by the specialist tephritid fly, Euphranta connexa (Leimu & Syrjänen, 2002; Leimu & Lehtilä, 2006; Solbreck & Ives, 2007; Muola et al., 2010). These factors may cause remarkable spatiotemporal variation in the population density of L. equestris and in intraspecific competition of L. equestris for V. hirundinaria seeds. Therefore, it is likely that selection caused by ecological factors affects specialization of L. equestris to V. hirundinaria. However, genetic factors, that is, level of genetic variation and trade-offs, may fundamentally influence specialization of L. equestris on V. hirundinaria.
We addressed the following specific questions: (1) Does the performance of L. equestris vary among the food-plant species? (2) Is there genetic variation in the performance of L. equestris on different food-plant species? and (3) Are there trade-offs between the performance of L. equestris on the different food-plant species?
Materials and methods
Lygaeus equestris L. (Heteroptera: Lygaeidae) is a granivorous (seed-eating) true bug with aposematic colouration. Lygaues equestris is specialized to feed on Vincetoxicum hirundinaria Med. [= Cynanchum vincetoxicum (L.) Pers.] (Apocynaceae, former: Asclepiadaceae) seeds, and in Finland, it is found merely in V. hirundinaria populations (Rintala & Rinne, 2010). Lygaeus equestris is locally common in the distribution area of V. hirundinaria, but its population sizes vary considerably among years and populations (Solbreck & Sillén-Tullberg, 1990a,b; L. Laukkanen, pers. obs.). Lygaues equestris is both a pre- and a post-dispersal seed predator of V. hirundinaria seeds. Both larvae and adults feed on the green ovulae, developing seeds and mature seeds, whereas the younger instars tend to feed more often on already dispersed seeds (Solbreck & Kugelberg, 1972; Kugelberg, 1977). Lygaeus equestris is usually monovoltine and overwinters as adult. The female L. equestris oviposits on the ground-layer vegetation in June and July. Adults of the new generation commonly appear from late-July onwards (Solbreck & Kugelberg, 1972).
Vincetoxicum hirundinaria is a highly poisonous long-lived perennial plant that contains diverse types of secondary compounds, such as antofine and phenolic compounds (Muola et al., 2010), which may delimit the number of herbivores feeding on it. The secondary compounds of V. hirundinaria may be essential in the nutrition of specialized herbivores, and they may also function in the chemical defence of the herbivores against their natural enemies.
Both the observations from natural populations and laboratory experiments confirm that L. equestris can occasionally also feed on several other plant species than V. hirundinaria (Solbreck & Kugelberg, 1972; Kugelberg, 1973a,b, 1974; L. Laukkanen, pers. obs.). In the present experiment, we used Crepis tectorum L. (Asteraceae), Tanacetum vulgare L. (Asteraceae) and Verbascum thapsus L. (Scrophulariaceae) as alternative food-plant species. These plant species occur in our study area, and our field observations confirm that L. equestris feeds on these species (L. Laukkanen, R. Leimu & A. Muola, pers. obs.). The three alternative plant species used in the experiment contain several specific chemical compounds that may confer resistance to herbivores. Water extract, essential oil and particular components (camphor, thymol and umbellulone) of the oil of T. vulgare have been found to repel several herbivore species (Brewer & Ball, 1981; Schearer, 1984). Crepis tectorum contains sesquiterpene clycosides (Kisiel & Kohlmünzer, 1989), and V. thapsus contains iridoid glycosides, sesquiterpenes and flavonoids (Hussain et al., 2009). However, we do not have any information about the possible chemical defence on the seeds of these species.
To investigate how feeding on the alternative food-plant species affects life history characteristics of L. equestris and the potential costs of alternative host use, we conducted feeding trials in laboratory using seeds of the four plant species. We collected seeds of V. hirundinaria, C. tectorum, T. vulgare and V. thapsus from several natural populations located in the south-western archipelago of Finland from June–September 2007 and dried them at 25 °C. The seed origins of each species were mixed before used in the experiment.
Lygaeus equestris individuals were collected from one natural population in the south-western archipelago of Finland in late August 2007 (parental generation). The individuals were stored in plastic vials with a supply of husked Helianthus annuus L. (Asteraceae) seeds. Lygaeus equestris individuals were overwintered at +5 °C for 3 weeks. Twenty L. equestris females and males were then randomly paired, and each pair was allowed to mate in a glass jar. The mating pairs were fed with H. annuus seeds and kept at 30 °C and 22L-2D photoperiod. The rearing methods were adopted from Solbreck & Sillén-Tullberg (1981). The egg batches produced were removed from the jars and transferred to Petri dishes.
After hatching, the larvae were assigned to feed on seeds of the four plant species (split-brood design with 20 families, 10–19 individuals per family per plant species, 40–56 per family and 895 individuals in total). The larvae were reared individually on Petri dishes (diameter 9 cm) at 24 °C and 22L-2D photoperiod. Distilled water and surplus of seeds of the food-plant species were supplied during the whole larval development.
The Petri dishes were examined daily to record mortality and moults. After the last moult, the sex of L. equestris adults was determined. As insect females and males may have different nutritional requirements (Lee, 2010), we studied whether food-plant species affects the sexes differently. We measured the fresh weight of adults on the day of the last moult. The individuals were allowed to feed and drink after the last moult until measured.
We used developmental time, adult biomass and mortality as estimates of herbivore performance (Eyles, 1964; Kugelberg, 1973b). The distribution of mortality is binomial, as larvae either died during the larval period or stayed alive. To examine the effects of food-plant species on and variation among L. equestris families in the probability of mortality, a generalized linear mixed model for binomial regression was conducted. Food plant was included as a fixed factor and L. equestris family and the interaction between family and food plant as random factors in the analysis. The analysis was conducted using the GLIMMIX procedure of the sas Statistical Package (version 9.2; SAS Institute Inc., Cary, NC, USA).
To investigate whether variation in the developmental time and adult biomass of L. equestris was associated with their diet, we conducted two separate mixed model anovas with developmental time and adult biomass as dependent variables. All except one L. equestris individual feeding on T. vulgare died before reaching the adulthood, and thus, they were all exluded from these analyses. The total number of individuals in these analyses was therefore 678. Food plant and sex were included as fixed factors and L. equestris family as a random factor in the analyses. We also included the interactions between family, food plant and sex in the model. The error terms were determined following Zar (1984). Because the assumptions of parametric anova were not fulfilled for developmental time, we conducted an anova using ranked values. These analyses were conducted using spss (SPSS Inc., Chicago, IL, USA).
We conducted Spearman correlation analyses to investigate genetic correlations, that is, trade-offs, between the different food-plant species in life history traits (mortality, developmental time and adult biomass). We only tested correlations between traits with observed among-family variation, as a proper genetic correlation cannot be defined in the absence of genetic variation (Ueno et al., 2003). We used family means in these analyses. The correlation analyses were conducted with sas 9.2 (SAS Institute Inc., 2002–2007). Bonferroni correction was used to counteract the problem of multiple comparisons.
The mortality, developmental time and adult biomass of L. equestris varied significantly among the four food-plant species, indicating variation in their quality as food for L. equestris. All except one larva fed on T. vulgare died during the larval period (total mortality 99.5%), whereas the larvae grown on V. hirundinaria, C. tectorum and V. thapsus were more likely to reach the adult stage. Mortality of L. equestris on V. hirundinaria (21.7 ± 6.8%; mean ± SE) and C. tectorum (20.0 ± 5.2%) was rather low compared to mortality on V. thapsus (91.9 ± 1.9%; Fig. 1). The differences in probability of mortality among these four food-plant species were statistically significant (F3,54.81 = 63.05, P <0.001). Variation in probability of mortality among the L. equestris families was marginally significant (main effect of family: χ2 = 2.25, d.f. = 1, P =0.067). The effect of food plant on probability of mortality also varied among the L. equestris families (family × food plant interaction: χ2 = 4.81, d.f. = 1, P =0.014; Fig. 1). Given the significant two-way interaction, we tested for variation in probability of mortality among the L. equestris families separately for each of the four food-plant species. There were significant differences in probability of mortality among the families when L. equestris was fed on V. hirundinaria (χ2 = 26.10, d.f. = 1, P <0.001) and marginally significant differences when fed on C. tectorum (χ2 = 2.16, d.f. = 1, P =0.071), suggesting genetic variation in probability of mortality on these food-plant species. We found no statistically significant genetic variation in the probability of mortality of L. equestris on T. vulgare or V. thapsus: there were no significant differences among the L. equestris families on these food plants (χ2 < 0.01, d.f. = 1, P =0.497 and χ2 = 0.86, d.f. = 1, P =0.176, respectively).
Food-plant species had also a significant effect on the developmental time of L. equestris (Table 1; Fig. 2). Developmental time was shortest on V. hirundinaria (39.9 ± 0.15 days; mean ± SE) and longest on V. thapsus (56.1 ± 1.11 days). On C. tectorum, the average developmental time was 43.8 ± 0.27 days. Developmental time varied significantly among L. equestris families, which indicates genetic variation in this trait (Table 1; Fig. 2). The two-way interaction between food plant and family was marginally significant, suggesting that the effect of food-plant species on developmental time possibly varied among the L. equestris families (Table 1). Given the marginally significant two-way interaction, we tested for variation in developmental time among the L. equestris families separately for each of the three food-plant species on which L. equestris reached the adult stage. We included sex as a fixed factor and family and the interaction between family and sex as random factors in these 2-way anovas. When L. equestris was fed on V. hirundinaria, there was significant variation among the families in developmental time (χ2 = 5.00, d.f. = 1, P =0.013). On the contrary, the genetic, that is, among-family variation in developmental time was not significant when the larvae were fed on C. tectorum or V. thapsus (χ2 = 0.20, d.f. = 1, P =0.327 and χ2 = 0.60, d.f. = 1, P =0.219, respectively). On V. thapsus, the level of among-family variation differed between females and males (family and sex: χ2 = 6.8, d.f. = 1, P =0.005). The level of among-family variation in developmental time did not differ between sexes when L. equestris was fed on V. hirundinaria or C. tectorum (family and sex: χ2 = 0.50, d.f. = 1, P =0.240 and χ2 = 0.70, d.f. = 1, P =0.201, respectively). When fed on V. hirundinaria, males had slightly longer developmental time than females (40.3 ± 0.19 and 39.4 ± 0.21 days, respectively; F1,17.9 = 8.91, P =0.008). The differences in developmental time between sexes were nonsignificant when L. equestris fed on C. tectorum or V. thapsus (F1,19 = 3.24, P =0.088 and F1,12.5 = 0.05, P =0.819, respectively).
Table 1. Results of mixed model anova testing the differences in developmental time of Lygaeus equestris among 20 families, three food-plant species and between sexes in a feeding trial. Lygaeus equestris was able to reach the adult stage on these three food-plant species
Source of variation
Family × Food plant
Family × Sex
Food plant × Sex
Family × Food plant × Sex
Food-plant species had a significant effect on adult biomass, but there were no statistically significant differences among the L. equestris families (Table 2; Fig. 3). The adult biomass was 22.5 ± 0.70 mg when fed on V. thapsus, 34.5 ± 0.31 mg on C. tectorum and 40.8 ± 0.33 mg on V. hirundinaria. The adult biomass did not differ between female and male L. equestris (Table 2).
Table 2. Results of mixed model anova testing the differences in adult biomass of Lygaeus equestris among 20 families, three food-plant species and between sexes in a feeding trial. Lygaeus equestris was able to reach the adult stage on these three food-plant species
Source of variation
Family × Food plant
Family × Sex
Food plant × Sex
Family × Food plant × Sex
As genetic variation of L. equestris was observed only in mortality on V. hirundinaria, mortality on C. tectorum and developmental time on V. hirundinaria, we were able to investigate only two genetic correlations between the different food-plant species in life history traits. Mortality of L. equestris on V. hirundinaria correlated strongly positively and statistically significantly with mortality on C. tectorum (r = 0.58, n = 20, P =0.015), whereas the mortality on C. tectorum correlated positively, but not significantly with developmental time on V. hirundinaria (r = 0.13, n = 20, P >0.999).
We did not find any significant negative genetic correlations between mortality, developmental time and adult biomass of L. equestris on the studied food-plant species, suggesting that fitness trade-offs do not constrain alternative host use in L. equestris. Our results are thus in agreement with most studies that have found no evidence for trade-offs in insect performance on different host plants (Futuyma, 2008). A positive correlation may indicate a generalist strategy in food-plant utilization, that is, ‘a master of all trades’ genotype that has high fitness on several host species (Ueno et al., 2003; Forister et al., 2007; García-Robledo & Horvitz, 2011). Positive correlations between performance on different food-plant species may further suggest that generalization has evolved by inclusion of sets of plant species to the diet rather than by independent adaptation to single plant species (Ueno et al., 2003; Forister et al., 2007). We were able to test for two correlations in the performance of L. equestris on the different food plants. Both tested correlations were positive, although only the correlation between the mortality of L. equestris on C. tectorum and that on V. hirundinaria was statistically significant. Therefore, our results are in agreement with several previous studies suggesting that positive genetic correlations are more commonly found than negative ones (Futuyma, 2008).
By traditional view, trade-offs in host use promote herbivore food-plant specialization (Rausher, 1988; Scheirs et al., 2005). However, the significance of trade-offs in specialization and coevolution is much disputed, as in practice, most studies have not found significant negative correlations (Gould, 1979; Fry, 1990; Karowe, 1990; Fox & Caldwell, 1994; Futuyma et al., 1995; Mackenzie, 1996; Thompson, 1996; Keese, 1998; Via & Hawthorne, 2002; Forister et al., 2007; Agosta & Klemens, 2009; García-Robledo & Horvitz, 2011). The lack of significant negative genetic correlations in host use of phytophagous insects has been explained by numerous methodological problems and confounding genetic factors. Firstly, the performance measures used may not be adequate to detect the negative correlations (Rausher, 1988; Scheirs et al., 2005). According to a review by Scheirs et al. (2005), negative genetic correlations are more likely to be found in studies based on both adult and offspring performance. In here, we studied offspring performance using three traits (mortality, developmental time and adult biomass). Lygaeus equestris females oviposit in the ground-layer vegetation and the larvae move among the vegetation. Therefore, the host plant choice of the ovipositioning female may not be as important for the success of the developing larvae in this species as in other herbivorous insects with more sedentary larvae. Thus, although we did not examine host choice or reproductive success of the adults, we believe that our performance measures were sufficient for the detection of ecologically relevant trade-offs. Secondly, systematic environmental effects contribute to broad-sense genetic correlations (Falconer & Mackay, 1996) and may obscure detection of the trade-offs. In here, we examined broad-sense genetic correlations using a split-brood design, and therefore, the correlations may be obscured by environmental effects, such as nongenetic effects of maternal or paternal environment (Fox et al., 1995). Furthermore, other genetic factors may hinder the expression of trade-offs, and direct selection may modify the variance–covariance structure that produced the trade-offs and purge the trade-offs (e.g. Rausher, 1988; Fry, 1993; Thompson, 1996; Whitlock, 1996). Our results suggesting lack of trade-offs should be interpreted with caution due to the somewhat low sample size. Of course, food-plant specialization may also arise when trade-offs are not detected or do not exist (Rausher, 1988; Fry, 2003). For instance, the specialization of herbivores may evolve without trade-offs when the level of predation by the natural enemies of the herbivore differs between food-plant species (Rausher, 1988). We are currently not able to distinguish between the alternative explanations for why we did not detect trade-offs in this study. However, we are also studying the trade-offs in host use of L. equestris using a selection experiment that might be able to reveal trade-offs not revealed by split-brood designs (Fry, 2003).
Although most studies have found no evidence for trade-offs, heritable variation has often been found in the use of alternative food-plant species (Karowe, 1990; Futuyma et al., 1995; Thompson, 1996; Ueno et al., 2003; García-Robledo & Horvitz, 2011). Fox & Caldwell (1994) studied food-plant specialization of another Lygaeus species, the small milkweed bug, L. kalmii, feeding on two species of Asclepias and Taraxacum officinale. In contrast to our result, they found genetic variation of L. kalmii in all studied performance traits on the different food-plant species. Fox & Caldwell (1994) also found several positive correlations between the performance traits on the different host plants, which suggest that L. kalmii may be more generalist in its food-plant use than L. equestris. Our study species L. equestris may lack genetic variation in central performance traits when utilizing different food-plant species, as we did not detect genetic variation in mortality on T. vulgare and V. thapsus, developmental time on C. tectorum and V. thapsus and in adult biomass. It is worth notice that when L. equestris was fed on its primary food plant V. hirundinaria, there was significant genetic variation in both mortality and developmental time. This existing genetic variation on V. hirundinaria may further promote adaptation and enable further specialization to this food-plant species and suggests that specialization may be evolutionarily dynamic rather than a dead end (Thompson, 1996). Furthermore, the lack of significant genetic variation in mortality and developmental time on the other food-plant species may restrict adaptation of L. equestris to these nonoptimal present food plants. Besides this study, other studies have also not detected genetic variation in performance on different host species, suggesting that the lack of genetic variation may potentially affect the evolution of diet breadth of insect herbivores (Futuyma et al., 1995; Keese, 1998).
The food-plant use of an herbivore presents ‘a compromise’ between being a specialist and being a generalist (Fox & Morrow, 1981). In addition to the genetic constraints, to understand the host breadth of an herbivore, we need to take into account multiple ecological factors, such as spatiotemporal variability in host availability, variation in nutritional quality or secondary chemistry of the different food-plant species, and predator refuge. Seed production of V. hirundinaria is spatiotemporally highly variable due to abiotic factors (Solbreck and Sillén-Tullberg, 1986a,b; Ågren et al., 2008). Furthermore, seed predation by the specialist fly Euphranta connexa can be close to 100% in some populations and certain years (Leimu & Lehtilä, 2006; Solbreck & Ives, 2007), and thus, it also contributes to the spatiotemporal variation in seed availability. Given this variation in food availability, being strictly monophagous might not be a prospective strategy for L. equestris. Our recent, still unpublished results suggest that the number of vascular plant species in the habitat seems to have an effect on genetic variation of L. equestris populations (L. Laukkanen, P. Mutikainen, A. Muola and R. Leimu, unpubl. data). This result further emphasizes the importance of the availability of alternative food plants to the population viability of L. equestris.
Seasonal and annual variation in the relative abundance of food plants commonly hinders food-plant specialization of herbivorous insects (Berlocher & Feder, 2002). However, the fact that L. equestris performed best on V. hirundinaria compared to the three alternative food plants, together with the previous findings on the feeding preference of L. equestris for V. hirundinaria (Kugelberg, 1974) and the occurrence of L. equestris only in distribution areas of V. hirundinaria in Scandinavia (Solbreck & Sillén-Tullberg, 1990a; Hämet-Ahti et al., 1998; Rintala & Rinne, 2010), suggest that L. equestris is rather specialized on V. hirundinaria in this region. However, the level of food-plant specialization may vary spatially (Fox & Morrow, 1981; Schoonhoven et al., 2005). For instance, in Sicily, L. equestris is not dependent on V. hirundinaria, as the plant species does not occur there (Solbreck et al., 1989). In addition, the temporal variability in the quality of food may change the specialization pattern (Kugelberg, 1973b; Bar-Yam & Morse, 2011). Taken together, ours and previous results suggest that the degree of food-plant specialization of L. equestris varies across geographical landscapes and in time (Thompson, 1994).
Food-plant specialization is strongly affected by food-plant quality. For example, food-plant specialization of the red milkweed beetle, Tetraopes tetraophthalmus, is driven by the variable concentrations of toxic cardenolides in the roots of its main host, the common milkweed (Asclepias syriana) and other Asclepias species (Rasmann & Agrawal, 2011). Vincetoxicum hirundinaria has much in common with the species in the genus Asclepias: they belong to the same subfamily (Asclepiadoideae) and are highly toxic perennials. There are many similarities between their herbivores as well, as L. equestris and T. tetraophthalmus are both aposematic and are to some degree specialized in their food-plant use. It seems therefore likely that the toxic secondary compounds may drive the food-plant specialization of L. equestris on its primary food plant, V. hirundinaria. Because of the high mortality, T. vulgare was the least suitable food plant for L. equestris. This is probably due to high levels of secondary compounds that are characteristic to this plant species and known to repel herbivorous insects (Brewer & Ball, 1981; Schearer, 1984; Dembitskii et al., 1985). Our results suggest that L. equestris is able to tolerate or detoxify especially the chemical compounds of V. hirundinaria, but is unable to process plant chemicals of other species, such as T. vulgare.
Bernays & Graham (1988) have suggested that the refuge provided by the host plants against the natural enemies of the herbivore determines the host breadth of the herbivore. The chemical compounds of food plants may affect the colouration of herbivorous insects and consequently their defence against predators (Ojala et al., 2007; Lindstedt et al., 2010). The colour of L. equestris varies depending on the food; individuals feeding on alternative food plants have a less intensive shade of red with a faint suggestion of grey or orange compared to those feeding on V. hirundinaria (Tullberg et al., 2000; L. Laukkanen, unpubl. data). Furthermore, birds prefer L. equestris fed on the seeds of sunflower, Helianthus annuus, to those fed on V. hirundinaria (Tullberg et al., 2000). The higher attack rate by bird predators and higher mortality of L. equestris after an attack are probably related to changes in aposematic colouration, decreased adult biomass and modifications in the chemical composition of L. equestris (Scudder & Duffey, 1972; Gamberale & Tullberg, 1996; Tullberg et al., 2000). These results might therefore suggest that feeding on alternative food plants may involve a cost in terms of lower antipredatory defence in L. equestris and that refuge from predators may partly determine the host breadth of this seed predator. However, there are no observations of predation or parasitism in natural populations of L. equestris in Scandinavia (Solbreck & Sillén-Tullberg, 1990a; L. Laukkanen, R. Leimu & A. Muola, pers. obs.).
Food-plant specialization accelerates coevolutionary arms race between a plant and its herbivore, as specialist species evolve faster than generalist species (Whitlock, 1996). Coevolution between L. equestris and V. hirundinaria is highly probable regardless of the fact that L. equestris is not strictly specialized to this plant species. Lygaeus equestris reduces the fitness of V. hirundinaria (R. Leimu, unpubl. data), and V. hirundinaria individuals vary in their quality as food for L. equestris (Muola et al., 2010; L. Laukkanen, unpubl. data). Knowledge on the relative importance of V. hirundinaria and the alternative host plants for L. equestris and the factors that affect specialization or alternative food-plant utilization are central for understanding the potential coevolution between L. equestris and V. hirundinaria. Furthermore, such knowledge is necessary to understand the potential coevolution between V. hirundinaria and the other herbivore species because selection and evolutionary changes due to one of the herbivores are likely to affect the other herbivores using the same food resource.
We thank Tuija Koivisto, Janika Ulenius, Anu Veijalainen and Eero J. Vesterinen for assistance in establishing and conducting the experiment and Aino Kalske for constructive comments on the manuscript. This study was financially supported by the Academy of Finland (grants 8109859 and 138308 to R.L.), Ella and Georg Ehrnrooth Foundation and Oskar Öflund Foundation (grants to L.L.).