Pollen feeding, resource allocation and the evolution of chemical defence in passion vine butterflies


  • M. Z. Cardoso,

    Corresponding author
    1. Section of Integrative Biology, University of Texas, Austin, TX, USA
    2. Botânica, Ecologia e Zoologia, Centro de Biociências, Universidade Federal do Rio Grande do Norte, Natal, Brazil
    • Correspondence: Márcio Z. Cardoso, Departamento de Botânica, Ecologia e Zoologia, Centro de Biociências, Universidade Federal do Rio Grande do Norte, Natal-RN, Brazil. Tel.: +55 84 3211 9205; fax: +55 84 3211 9205; e-mails: mzc@cb.ufrn.br; marciozikan@gmail.com

    Search for more papers by this author
  • L. E. Gilbert

    1. Section of Integrative Biology, University of Texas, Austin, TX, USA
    Search for more papers by this author


Evolution of pollen feeding in Heliconius has allowed exploitation of rich amino acid sources and dramatically reorganized life-history traits. In Heliconius, eggs are produced mainly from adult-acquired resources, leaving somatic development and maintenance to larva effort. This innovation may also have spurred evolution of chemical defence via amino acid-derived cyanogenic glycosides. In contrast, nonpollen-feeding heliconiines must rely almost exclusively on larval-acquired resources for both reproduction and defence. We tested whether adult amino acid intake has an immediate influence on cyanogenesis in Heliconius. Because Heliconius are more distasteful to bird predators than close relatives that do not utilize pollen, we also compared cyanogenesis due to larval input across Heliconius species and nonpollen-feeding relatives. Except for one species, we found that varying the amino acid diet of an adult Heliconius has negligible effect on its cyanide concentration. Adults denied amino acids showed no decrease in cyanide and no adults showed cyanide increase when fed amino acids. Yet, pollen-feeding butterflies were capable of producing more defence than nonpollen-feeding relatives and differences were detectable in freshly emerged adults, before input of adult resources. Our data points to a larger role of larval input in adult chemical defence. This coupled with the compartmentalization of adult nutrition to reproduction and longevity suggests that one evolutionary consequence of pollen feeding, shifting the burden of reproduction to adults, is to allow the evolution of greater allocation of host plant amino acids to defensive compounds by larvae.


Like other butterfly genera of Neotropical Heliconiini, and like their passifloraceous larval host plants, species of genus Heliconius are cyanogenic as adults (Nahrstedt & Davis, 1983; Zagrobelny et al., 2004). However, in contrast to other genera of passion vine butterflies, the genus Heliconius has diversified dramatically in terms of racial differentiation, intraspecific müllerian mimicry and species richness (Brown, 1981; Beltrán et al., 2007). The most distinct derived trait common to Heliconius but lacking in other heliconiine genera, or any other butterfly genus, is adult pollen feeding and correlated reorganization of how larvae and adults allocate nitrogenous resources to reproduction (Gilbert, 1972, 1975; Dunlap-Pianka et al., 1977; Boggs, 1981b). It was proposed that the innovation of pollen feeding provided a new adaptive zone for the radiation of Heliconius (Gilbert, 1991), but just how adult-acquired resources would fuel diversification is not obvious. One possibility is that increased amino acid intake via pollen feeding has allowed evolution of unpalatability since defensive compounds of Heliconius, cyanogenic glycosides, are derived from amino acids (Zagrobelny et al., 2008). Under this scenario, increased chemical defence of adults fueled expansion and diversification of the genus under intrageneric müllerian mimicry (Gilbert, 1991).

One clue is the finding of Chai (1990) and summarized by Gilbert (1991), that in a local community of 19 passion vine butterfly species in Costa Rica, pollen-feeding Heliconius were significantly more unpalatable to insectivorous rufous-tailed jacamar than other species from six other genera present in the system. This finding along with the fact that cyanogenic compounds are synthesized de novo from amino acid precursors in almost all heliconiines (Nahrstedt & Davis, 1981, 1983, 1985) raised the possibility (Gilbert, 1991) that adult palatability is increased in ecological time by success in pollen foraging. A similar adult-based system is seen in distasteful ithomiine butterflies that are known to acquire their unpalatability through adult acquisition of pyrrolyzidine alkaloids (Brown, 1984; Trigo, 2010). However, the distinction is that Heliconius, rather than collecting already synthesized toxins from plants, are acquiring precursors, amino acids that can be allocated to making babies or bullets. An alternative possibility is that by shifting allocation of reproductive effort to the adult stage in evolutionary time (Boggs, 1981a; Karlsson, 1994, 1995), Heliconius larvae are thus ‘free’ to allocate relatively more amino acids acquired from larval host plants to synthesis of relatively more cyanogens than related nonpollen-feeding species that must make babies as well as bullets during larval stages so to speak.

Our study is designed to distinguish between these alternatives. To do so, we tested the effect of adult diets with and without amino acids on cyanide production in Heliconius and contrasted cyanide production in freshly emerged Heliconius and two nonpollen-feeding species. Our results do not support the hypothesis that adult diet determines cyanide production and suggest that larval input is the main determinant factor in adult cyanogenesis in passion vine butterflies.

Materials and methods

Organisms and adult diets

We used Dryas julia and Agraulis vanillae (nonpollen feeders, palatable) and Heliconius charithonia, Hcydno, Hethilla and Hhecale (pollen feeders, unpalatable). Butterflies were reared on greenhouse cultures of Passiflora biflora in controlled conditions (25 °C, 80% R.H.). All Heliconius cultures had been established in the greenhouses for several years prior to the beginning of the project. Greenhouse populations of A. vanillae and D. julia were established solely for this study and were started from larvae collected outdoors in Austin, Texas (Agraulis) and adults shipped from Florida (Dryas). Adults were released in 2 × 2 × 2 m shaded experimental enclosures and hand-fed artificial diets. Afterwards, adults were frozen for chemical analysis. A subset was frozen after eclosion, without access to adult resource (freshly emerged experimental group). Adult diets consisted of an amino acid diet with 20% sugar (amino acid treatment) or a 20% sugar diet (control). Amino acid content mimicked pollen load composition collected by greenhouse butterflies (Table S1). Butterflies were given 60 μL of diet solution daily, an intake of 1.2 mg of amino acids for those in the experimental group, equivalent to a large pollen load, estimated from Gilbert (1972).

Amino acids vs. sugar diet

We kept adults of H. charithonia, H. ethilla and H. hecale in enclosures for 20 days and handfed them artificial diets. Trials for H. charithonia and H. ethilla were run at the same time. Due to logistical constraints during the trials, freshly emerged adults were available only for H. charithonia and trials with H. hecale were performed at a later date.

Comparison between freshly emerged butterflies

We raised numerous larvae of H. charithonia, H. ethilla and H. cydno to gather data on freshly emerged individuals and compare cyanide content between species. Because these were obtained much later, we analysed them separately from the diet experiments. We also compared cyanide concentration of freshly emerged adults of the pollen-feeding species, H. charithonia, and the nonpollen-feeding species, D. julia and A. vanillae. Because neither Dryas nor Agraulis feed on pollen we did not test these species with the diet treatments used in the other experiments.

Free-flying Heliconius

We collected adult Heliconius butterflies at Los Tuxtlas Tropical Biology Research Station in Veracruz, Mexico. Butterflies were killed by freezing and stored individually in vials containing pure methanol. Descriptions of the site and pollen load data are presented in Cardoso (2001).

Chemical analysis

Thawed butterflies were ground in microcentrifuge tubes, which were deposited in a stoppered scintillation vial with 0.5 mL of NaOH at the bottom and incubated for 24 h at 64 °C. Cyanide released from the butterfly was trapped as NaCN (sodium cyanide) and used in colorimetric analysis (Lambert et al., 1975; Brinker & Seigler, 1989, 1992; Schappert & Shore, 1995). Butterflies were dry weighed to the nearest 0.1 mg. For field caught butterflies, we modified the procedure to allow for methanol extraction; we dried a 0.5 mL aliquot of the extract inside a micro centrifuge tube and added ß-glycosidase (0.1 mL of linamarinase) to release cyanide. Subsequent steps were identical to those employed with the frozen butterflies.

Statistical analysis

Adult cyanide concentration (μmoles of cyanide per butterfly unity mass) and dry mass (mg) were examined using analysis of variance in Systat12 (Systat Software Inc., Chicago, IL, USA). Post hoc tests were performed when appropriate.


Amino acid vs. sugar diet

This analysis tested the hypothesis that a diet based on amino acids, compared to a sugar-only diet, would result in increased cyanide concentration in H. ethilla and Hhecale. We did not find support for the diet effect because there was no difference in cyanide concentration between sugar and amino acid-fed H. hecale (F1,10 = 0.096, R2 = 0.01, P = 0.76) or H. ethilla (F1,12 = 1.55, R2 = 0.114, P = 0.24) (Fig. 1a,b). Also, no significant differences between treatments were found in biomass (H. hecale, F1,10 = 1.01, P = 0.34; H. ethilla, F1,12 = 0.003, P = 0.96) or cyanide content (H. hecale, F1,10 = 1.36, P = 0.27; H. ethilla, F1,12 = 1.57, P = 0.23).

Figure 1.

Comparison of cyanide concentration in adult Heliconius butterflies fed experimental diets (‘sugar’ or ‘amino acid’) and in freshly emerged butterflies (‘fresh’), frozen soon after eclosion. (a) Heliconius hecale; (b) Heliconius ethilla; (c) Heliconius charithonia. Bars indicate standard error of the mean and numbers at the bottom are sample sizes. Letters indicate significant differences in post hoc test. NS = not significantly different.

We then compared cyanide concentration of the two diet types with the cyanide concentration of freshly emerged individuals. Similar to the previous experiment, our expectation was that amino acid intake in the adult stage would raise the cyanide concentration to levels higher than that provided by sugar-only diet or the larval intake. We found that cyanide concentration in amino acid-fed and freshly emerged individuals of H. charithonia were similar, whereas sugar-fed individuals had much lower cyanide concentration (F2,15 = 15.05, R2 = 0.67, P = 0.0002) (Fig. 1c). The post hoc test showed that the lower cyanide concentration in sugar-fed butterflies was significantly different from the other two treatments (Tukey's, P = 0.001 [sugar vs. amino acid fed]; P = 0.0001 [freshly emerged vs. sugar-only]; P = 0.845 [freshly emerged vs. amino acid fed]). Thus, similar cyanide concentrations in freshly emerged and amino acid fed H. charithonia refutes the hypothesis of increased concentration. Yet, amino acid loss in diet leads to decrease in cyanide content.

Comparisons between freshly emerged butterflies

Our goal in this analysis was two-fold. First, we wanted to characterize larval input to adult cyanide concentration and test for possible differences between Heliconius species. Second, we wanted to test if larval input differed between the pollen feeding and the nonpollen-feeding lineages. The between-Heliconius comparison showed that freshly emerged individuals of the three Heliconius species (H. charithonia, H. cydno and H. ethilla) did not differ in cyanide concentration (F2,57 = 0.58, R2 = 0.02, P = 0.56) (Fig. 2), total cyanide content (F2,57 = 0.296, R2 = 0.01, P = 0.74) or body mass (F2,57 = 0.46, R2 = 0.02, P = 0.63). The comparison between pollen feeding and nonpollen-feeding species showed that A. vanillae and D. julia eclosed with similar biomass and both were heavier than pollen-feeding H. charithonia (F2,34 = 4.7, R2 = 0.22, P = 0.016; Fig. 3a). Cyanide concentration, on the other hand, was significantly higher in H. charithonia than in the other two species (F2,34 = 17.7, R2 = 0.46, P = 2.5 × 10−5, Fig. 3b; Tukey's, P = 0.19 [Agraulis vs. Dryas]; P = 0.002 [Agraulis vs. Heliconius]; P = 0.0001 [Dryas vs. Heliconius]).

Figure 2.

Comparison of cyanide concentration between freshly emerged individuals of three Heliconius species (H. charithonia, H. cydno, H. ethilla). Larvae were raised under controlled conditions. All adults are freshly emerged butterflies and were frozen after eclosion. This procedure eliminates adult food intake and compares larval production. Bars indicate standard error of the mean and numbers at the bottom are sample sizes. NS = not significantly different.

Figure 3.

Adult dry mass (a) and cyanide concentration (b) in three Heliconiiti butterflies, representing nonpollen-feeding clades (Agraulis vanillae and Dryas julia) and the pollen-feeding clade (Heliconius charithonia). Larvae were raised under controlled conditions. All adults are freshly emerged butterflies and were frozen after eclosion. This procedure eliminates adult food intake and compares larval production. Bars indicate standard error of the mean and numbers at the bottom are sample sizes. Letters indicate significant differences in post hoc test. NS = not significantly different.

Free-flying butterflies

To better characterize cyanide dynamics in free-flying butterflies, we extracted and estimated cyanide concentration in three Heliconius species from Los Tuxtlas, Mexico. Cyanide concentration in wild H. charithonia and H. erato was higher than in H. ismenius (F2,85 = 66.7, R2 = 0.61, P = 1.0 × 10−8, Fig. 4).

Figure 4.

Cyanide concentration in free-flying Heliconius butterflies from the rainforest of Los Tuxtlas Biological Station, Veracruz, Mexico. Bars indicate standard error of the mean and numbers at the bottom are sample sizes. Letters indicate significant differences in post hoc test.


We did not find evidence to support the ecological hypothesis for increased cyanide concentration led by adult amino acid intake in Heliconius butterflies, as suggested in Gilbert (1991). In fact, lack of amino acids in the diet of adult butterflies did not lead to decrease in cyanide concentration, as was to be expected if cyanogenic glycosides were being produced via de novo synthesis from amino acid intake via pollen feeding. Moreover, the ecological hypothesis also predicted that cyanide concentration would rise when butterflies were offered amino acids. This was not observed. In fact, in almost all cases, cyanide concentration was similar among diet treatments. Previous work had shown that females tend to gather more pollen than males (Boggs et al., 1981), probably because of higher reproductive demands. Under the original hypothesis, we would expect females to be more unpalatable than males and that is not seen in palatability trials (Chai, 1990).

The comparison between freshly emerged butterflies is informative because all Heliconius species yielded similar cyanide concentration in young adults, indicating that cyanide levels are not a function of adult diet but are actually determined by larval input (see also Engler et al., 2000).

A strong phylogenetic component to cyanide dynamics in larvae and adults was suggested by comparison between pollen-feeding Heliconius and two nonpollen-feeding species, Dryas and Agraulis. Our analysis confirms previous reports that these basal heliconiines are cyanogenic and have lower cyanide concentration than Heliconius (Nahrstedt & Davis, 1981, 1983). This concentration difference suggested the possibility that the evolution of pollen feeding in Heliconius and the adult intake of amino acids contributed directly to increased cyanogenesis (Gilbert, 1991; Beltrán et al., 2007) as well as egg production. The data presented here eliminate the possibility that adult cyanogenesis is a proximate outcome of pollen feeding as we show that differences in cyanide concentration are already present in freshly emerged individuals, regardless of adult resource intake. Thus, the patterns of cyanide concentration, both within Heliconius species and between species of their tribe, place cyanogen defences along with growth and development as a function of larval feeding.

Larval input of defensive chemicals is common among aposematic butterflies because chemicals are sequestered from their toxic host plants (Nishida, 2002; Opitz & Müller, 2009). Indeed, the fact that most Passiflora species are highly cyanogenic (Spencer, 1988) would suggest a similar mechanism. However, most Heliconius species differ from this pattern in that they produce their cyanogenic defences from amino acids synthesis (Nahrstedt & Davis, 1983, 1985) and do not depend on the host plant toxins for their own chemical defence. Thus, in a new twist to the pollen feeding adaptive zone scenario, adult diet does not have a direct impact on butterfly palatability, but frees the larval stage to be creative with the amino acids extracted from its host.

A more direct impact of adult pollen feeding should not be completely dismissed, because in H. charithonia, we found that amino acid deprived individuals experienced a decrease in cyanide, while those fed amino acids maintained cyanide concentration similar to freshly emerged individuals. This suggests possible dynamics in nitrogen levels in which adults scrounge nitrogen reserves when devoid of pollen sources, perhaps from stored cyanogens. Whether this loss can be reversed by adding amino acids remains to be tested, but it suggests that recycling of cyanogenic glycosides can be taking place (O'Brien et al., 2002). This may explain why free-flying H. charithonia in Mexico are as cyanogenic as other Heliconius, because adults have plenty of access to pollen sources in the open habitats in Los Tuxtlas with little competition (Cardoso, 2001), whereas in Corcovado Park Costa Rica at Sirena, where they are excluded from high quality pollen resources by competition (Boggs et al., 1981), H. charithonia are relatively palatable (Gilbert, 1991) and presumably low in cyanogenesis. Perhaps, H. charithonia with less expectation of high quality pollen (Dunlap-Pianka, 1979) has retained the flexibility to reallocate cyanogens to the reproduction side of the ledger.

Our findings also bring into question the dynamics of the mimetic relationship between many Heliconius species and the highly unpalatable ithomiines (Brown, 1984; Chai, 1990), many of which are models to Heliconius mimicry (Brown & Benson, 1974; Gilbert, 1988, 1991; Joron et al., 1999). Because alkaloid defence in many ithomiines are only acquired after intake of pyrrolizidine alkaloid containing nectar or from withered leaves of borage plants by adults, freshly emerged individuals are completely palatable to predators (Brown, 1984; Trigo et al., 1996) as are individuals unable to find pyrrolizidine alkaloid sources due to decline in certain plants or butterfly numbers outstripping supply, as suggested by winter predation in ithomiines (Brown & Neto, 1976). Given that it is Heliconius that emerges fully loaded with their defensive chemicals, cyanogenic Heliconius may maintain the mimicry at least for some nontrivial interval.

Overall, our experiments indicate that most of the burden of chemical defence in Heliconiine butterflies is linked to larval resource acquisition with a less significant contribution of adult resources. Although an intensively studied subject, few studies have looked at consequences of life-history strategies on allocation of resources to defence in animals (Fordyce et al., 2006) as our study has. These results broaden our perspective on the evolutionary consequences of pollen feeding among the highly diverse Heliconius and should prove significant in opening new venues of investigation on the evolution of chemical defence among heliconiine butterflies. For example, attention should be given to the one clade of Heliconius (sara/sapho) known to sequester cyanogens from larval host plants (Engler-Chaouat & Gilbert, 2006). It appears that sequestration and de novo synthesis trade-off and that this trait is derived from an ancestral state of de novo synthesis. Does this innovation, which apparently increases adult cyanogenesis relative to those species studied here, allow more larval resources to be devoted to reproduction with less dependence on pollen feeding? There is much more to uncover in the biology of Heliconius.


We thank R. Ramakrishnan, C. Boggs, K. Kunte, C. Estrada, H. Engler, C. Penz and two anonymous reviewers for comments on the manuscript. M. Poenie, H. Engler, P. Schappert and K. Spencer cleared technical questions regarding cyanide chemistry. M. Beard, J. Bostick, J. Warr and A. Chang helped with the experiments. Logistical support and/or permits for collections and importation of butterflies were provided by: W.W. Benson (Brazil), MINENEM and the staff of Corcovado National Park (Costa Rica), Instituto Nacional de Ecología (Mexico; Oficio DOO700(2)-2778) and USDA-APHIS. Greenhouse facilities were developed through NSF grants DEB-79060332 and BRS-8315399 with the support of U.T. Austin. MZC acknowledges CAPES (Brazil) for a doctoral fellowship and the Tinker Foundation, Sigma-Xi and the Section of Integrative Biology (UT) for travel grants.