Na⁺/K⁺-ATPase resistance and cardenolide sequestration: basal adaptations to host plant toxins in the milkweed bugs (Hemiptera: Lygaeidae: Lygaeinae)

Abstract

Despite sequestration of toxins being a common coevolutionary response to plant defence in phytophagous insects, the macroevolution of the traits involved is largely unaddressed. Using a phylogenetic approach comprising species from four continents, we analysed the ability to sequester toxic cardenolides in the hemipteran subfamily Lygaeinae, which is widely associated with cardenolide-producing Apocynaceae. In addition, we analysed cardenolide resistance of their Na⁺/K⁺-ATPases, the molecular target of cardenolides. Our data indicate that cardenolide sequestration and cardenolide-resistant Na⁺/K⁺-ATPase are basal adaptations in the Lygaeinae. In two species that shifted to non-apocynaceous hosts, the ability to sequester was secondarily reduced, yet Na⁺/K⁺-ATPase resistance was maintained. We suggest that both traits evolved together and represent major coevolutionary adaptations responsible for the evolutionary success of lygaeine bugs. Moreover, specialization on cardenolides was not an evolutionary dead end, but enabled this insect lineage to host shift to cardenolide-producing plants from distantly related families.

1. Introduction

Coevolution in plant–herbivore interactions involves the production of plant defences and insect counter strategies. From the insects' perspective, diet specialization, resistance to plant toxins, sequestration of these toxins and aposematic coloration may often be coupled as an adaptive strategy [1,2]. Yet, quite remarkably, we have little understanding of the macroevolution of such strategies and how interacting traits facilitate tolerance and sequestration of host plant toxins. In particular, the genetic, physiological and ecological basis of host use in closely related species needs to be assessed in a phylogenetic context in order to understand the evolution of insect resistance traits, potential costs and patterns of host shifts.

Plants in the Apocynaceae produce toxic cardenolides (aka cardiac glycosides) [3], which are specific inhibitors of the ubiquitous animal enzyme Na⁺/K⁺-ATPase, a cation carrier essential for a variety of physiological functions [4,5]. Due to the ubiquity of their target, cardenolides are considered universal toxins potentially affecting any animal species. On a broad scale, several studies have shown that a community of distantly related insects evolved the same strategies to overcome two major defence traits of Apocynaceae, latex and cardenolides [6–8]. By contrast, within insect lineages, our knowledge on the evolution of these insect traits is very limited.

The milkweed bugs (Hemiptera: Lygaeidae; Lygaeinae, ca 550 species) are seed predators which are well known for their aposematic black and red coloration and seem to be commonly associated with apocynaceous host plants on five continents [9–13]. The thick walls, wide spacing and rapid seed dispersal of Asclepias seed pods (and probably also fruits of other genera of Apocynaceae like Gomphocarpus and Calotropis) may be adaptations which have been selected for by seed predators like milkweed bugs [14], supporting the ecological significance of the interaction between lygaeine bugs and Apocynaceae. The close relationship between these insects and the Apocynaceae is furthermore supported by the assumption that North American Asclepias and the associated Lygaeus species share the same evolutionary origin as their distribution matches [15].

The large milkweed bug (Oncopeltus fasciatus (Dallas, 1852)) and the small milkweed bug (Lygaeus kalmii Stål, 1874) are well known to sequester cardenolides from Asclepias seeds [16] which protect them against predator attacks [17]. Oncopeltus fasciatus has been shown to possess a specialized, double layered epidermis forming the so-called dorsolateral space where cardenolides absorbed into the body are concentrated and stored. Upon squeezing (e.g. caused by a predator attack), special weak areas of the cuticle rupture and droplets of cardenolide-rich fluid are released [18]. Besides O. fasciatus and L. kalmii, several other Lygaeinae have also been shown to sequester [19] or contain (as dried museum specimens) cardenolides [9]. In addition, O. fasciatus was shown to possess a Na⁺/K⁺-ATPase which is highly resistant to cardenolides (target site insensitivity) [20]. Both, O. fasciatus and L. kalmii, share a modified form of Na⁺/K⁺-ATPase carrying specific amino acid substitutions which lead to reduced sensitivity to cardenolides [7,8,21]. Both adaptations, cardenolide sequestration and target site insensitivity may be involved in interactions across three trophic levels, allowing for use of Asclepias seeds as a food resource and facilitating storage of these toxins in the body cavity as a defence against predators. Accordingly, the milkweed bugs provide an ideal model to study the macroevolution of these two traits involved in adaptation and probably coevolution.

To trace the macroevolution of both traits, we constructed a molecular phylogeny of 20 lygaeine species plus four outgroups (two Lygaeidae, one Oxycarenidae and one Pyrrhocoridae). Using in vitro assays of hemipteran Na⁺/K⁺-ATPase, we tested seven species of Lygaeinae for target site insensitivity. The ability to sequester cardenolides was assessed by feeding assays with radioactively labelled cardenolides in nine species. Based on these data, we reconstruct the origin of cardenolide resistance and sequestration in the Lygaeinae and investigate whether target site insensitivity and cardenolide sequestration are maintained in species no longer encountering the toxins in their host plants. By combining our data, we conclude that the secondary use of unrelated but cardenolide-containing host plants in other families, such as Adonis vernalis (Ranunculaceae), Digitalis purpurea (Plantaginaceae) or Urginea maritima (Asparagaceae) can be explained by pre-existing adaptations to cardenolides.

2. Material and methods

(a). Construction of molecular phylogeny

Our sampling included 20 species of Lygaeinae from four continents (electronic supplementary material, table S1) which based on our taxonomic knowledge represent the phylogenetic relationships within the subfamily. Species were selected so as to represent well-established host associations and assumed taxonomic breadth but was limited by species availability. Our sampling comprises species using Apocynaceae as hosts as well as species which exclusively use host-plants from other families. Some species were represented by more than one individual (indicated by Roman numerals in figure 1). In addition, DNA was sequenced from Kleidocerys resedae (Panzer, 1797) and Belonochilus numenius (Say 1832), species belonging to closely related subfamilies in the Lygaeidae (Ischnorrhynchinae and Orsillinae), and Pyrrhocoris apterus (Linnaeus, 1758) and Oxycarenus lavaterae (Fabricius, 1787) (Pyrrhocoridae and Oxycarenidae) as representatives of more distant relatives [22]. The target sequences were 1714 bp from the 3′ half of the mitochondrial cytochrome oxidase subunit I and II genes (COI/II) including the tRNA leucine gene (tRNA^Leu) between them and 507 bp of the large nuclear ribosomal subunit (28S rDNA). DNA was extracted from fresh (preserved at −20°C), ethanol preserved (98%) or dried specimens either by the DNeasy Tissue Kit (Qiagen) or a DNA extraction system for dried museum material described by Gilbert et al. [23]. DNA vouchers were deposited at the Zoological Research Museum Alexander Koenig, Bonn, Germany.

Figure 1.

Maximum-likelihood (ML) tree of the subfamily Lygaeinae based on the combined dataset of COI, COII, tRNA^Leu and 28S genes. Values above branches indicate ML bootstrap support values (1000 replicates) and Bayesian posterior probabilities (500 000 generations): only values more than 50% are shown. The black branches represent three lygaeid species from subfamilies other than the Lygaeinae (Ischnorrhynchinae, Orsillinae) and P. apterus (Pyrrhocoridae) as well as O. lavaterae (Oxycarenidae) which were used as outgroups. Capital letter A marks the evolutionary origin of target site insensitivity and cardenolide sequestration, capital letter B the loss of the ability to sequester cardenolides. Evidence for sequestration indicated by (S) was taken from the literature [9], experimental evidence for sequestration from this study is given by S+, S− indicates lack of sequestration ability. (Online version in colour.)

The target sequences were amplified by standard polymerase chain reaction protocols. To generate homologous sequences for the 28S rDNA fragment, we used the primers described by Muraji & Tachikawa [24]. Amplification of the target gene region COI/II was achieved by amplifying two or three smaller overlapping fragments using primers previously reported by Maus et al. [25] and Weller et al. [26].

Complementary strands of single individuals were edited and aligned using Sequencher v. 4.6 (Gene Codes Corporation, Ann Arbor, MI). The final 2221 bp alignment consisted of 1714 bp of the mitochondrial COI/II and tRNA^Leu genes and of 507 bp of the nuclear 28S gene obtained for 35 individuals.

Phylogenetic reconstructions were carried out using maximum-likelihood (ML) and Bayesian inference. Prior to likelihood analyses, best-fit models of nucleotide substitution were selected with likelihood ratio tests as implemented in Modeltest v. 3.7 [27]. Models of sequence evolution and parameters were estimated for each gene partition separately. ML analyses were performed with Treefinder [28] using a GTR + G model of sequence evolution for all partitions. Tree searches were started from five trees derived by a random walk of 10 nearest neighbour interchange steps around a centre tree (neighbour joining tree) generated by PAUP v. 4.0b10 [29]. The robustness of the ML tree was evaluated by bootstrap analyses with 1000 replicates using the same programme.

Bayesian analysis was conducted with MrBayes v. 3.0b4 [30], fitting a GTR + I + G model to each of the four data positions. Substitution parameters were estimated separately for each gene partition. Two independent runs were carried out with four parallel Markov chain Monte Carlo chains of 1 million generations and trees sampled every 200 generations.

(b). Lygaeinae specimens for sequestration assays and in vitro analysis of Na⁺/K⁺-ATPase

Adult specimens of Lygaeinae were obtained from the field or if the number of specimens was not sufficient maintained and propagated for a few generations in laboratory cultures. As in the phylogenetic analysis, we used species of the families Phyrrhocoridae, Berytidae and also Lygaeidae from a non-lygaeine subfamily as outgroup comparisons. All outgroup species used here and in the phylogeny are not reported to use Apocynaceae as hosts.

For sequestration and Na⁺/K⁺-ATPase assays with Cosmopleurus fulvipes (Dallas, 1852), Horvathiolus superbus (Pollich, 1781), Lygaeus equestris (Linnaeus, 1758), L. kalmii, Lygaeus simulans Deckert, 1985, O. fasciatus and Spilostethus pandurus (Scopoli, 1763), we used individuals reared in the laboratory on husked sunflower seeds and water only. Colonies were maintained at a 16 L : 8 D photoperiod at 26°C (C. fulvipes, H. superbus, O. fasciatus and S. pandurus) or 30°C (L. equestris, L. kalmii and L. simulans).

For Arocatus longiceps (Stål, 1872), Arocatus melanocephalus (Fabricius, 1798), K. resedae, Metatropis rufescens (Herrich-Schaeffer, 1835) (Berytidae) and P. apterus field collected individuals were used. Arocatus longiceps and A. melanocephalus were maintained on sunflower seeds and seeds of Platanus (2 days) under ambient conditions (A. l.) or at 26°C (A. m., 16 L : 8 D cycle). Kleidocerys resedae, M. rufescens and P. apterus were maintained on sunflower seeds for 2–3 days as well (26°C, 16 L : 8 D cycle).

Tropidothorax leucopterus (Goeze, 1778) was reared on sunflower seeds and cut branches of Vincetoxicum hirundinaria (26°C, 16 L : 8 D cycle). See the electronic supplementary material, table S1, for the origin of hemipteran specimens used in this study.

(c). In vitro assay of Na⁺/K⁺-ATPase

To test for the occurrence of target site insensitivity to cardenolides, we assayed Na⁺/K⁺-ATPase of seven lygaeine species and one lygaeid outgroup species in vitro. Na⁺/K⁺-ATPase assays were performed as described in Petschenka et al. [31]. Briefly, brains and thoracic ganglia of hemipteran specimens (killed and stored at −80°C) were dissected under deionized water, pooled (see table 1 for numbers of individuals used) and homogenized in deionized water (500 µl) using an all glass grinder (Wheaton). Extracts were frozen at −80°C, lyophilized and stored frozen until use. Prior to assay, lyophilisates were reconstituted by adding 200 µl water, vortex stirring and incubation for 10 min in a chilled ultrasonic bath. Undissolved residues were removed by centrifugation at 5000g (3 min). Protein content of Na⁺/K⁺-ATPase preparations was determined using the method of Bradford [32] and subsequently adjusted to provide a total amount of 6 µg protein per Na⁺/K⁺-ATPase reaction. Cardenolide sensitivity of Na⁺/K⁺-ATPase was determined by photometric quantification of inorganic phosphate released from ATP by Na⁺/K⁺-ATPase at different concentrations of ouabain over a period of 20 min at 37°C. To test for linearity, we measured a time course of P_i release over the period of incubation (reactions stopped after 0, 5, 10, 15 and 20 min) using a Na⁺/K⁺-ATPase preparation of O. fasciatus under all reaction conditions (10⁻³ to 10⁻⁸ M plus controls) and found that P_i release was always linear (see [31] and the electronic supplementary material, figure S2). Linearity under all conditions ensures that ouabain inhibition curves are not biased by nonlinear P_i release over time under different incubation regimes.

Table 1.

IC₅₀ values of ouabain for hemipteran Na⁺/K⁺-ATPase tested in vitro. (IC₅₀ values are calculated based on the fitted curves as lygaeine Na⁺/K⁺-ATPase does not reach 50% inhibition within the range of inhibitor concentrations used.)

species	IC80 value for ouabain [M]	IC50 value for ouabain (calculated)	number of individuals used per replicate of Na+/K+-ATPase in vitro assay
Arocatus longiceps	1.89 × 10−4	1.17 × 10−3	19–23
Horvathiolus superbus	1.82 × 10−4	1.87 × 10−3	16
Lygaeus equestris	8.68 × 10−5	1.36 × 10−3	8
Lygaeus kalmii	3.71 × 10−4	2.50 × 10−3	10–15
Oncopeltus fasciatus	2.19 × 10−4	2.22 × 10−3	6
Spilostethus pandurus	5.58 × 10−4	3.13 × 10−3	5
Tropidothorax leucopterus	1.54 × 10−4	1.46 × 10−3	7–9
Kleidocerys resedae	2.06 × 10−7	1.14 × 10−6	ca 20

(d). Sequestration assay

To assess the ability to sequester cardenolides, we fed nine lygaeine species in six genera and three outgroup species (K. resedae, M. rufescens and P. apterus) with radioactively labelled cardenolides. As plants typically produce several cardenolides with a wide polarity range, we used the polar [³H]-ouabain and the relatively non-polar cardenolide [³H]-digoxin (both Perkin Elmer LAS GmbH, Rodgau, Germany) to cover a part of this range. Both cardenolides probably do not occur naturally in the host plants of Lygaeinae but were used due to their commercial availability. Individuals were immobilized with a lasso made of dental floss (figure 2, inset) and their proboscis was manually introduced into a 2 µl droplet of 5% sucrose solution containing 5 µM ³H-cardenolide dissolved in ethanol on parafilm (final concentration of ethanol 17.7%). After feeding, specimens were kept for 10 days at 26°C and supplied with water and sunflower seeds ad libitum to allow for gut clearance of cardenolides. The incubation period of 10 days was chosen as we observed that in individuals kept for only 72 h before analysis differences were resolved less clearly and data showed larger variation (see the electronic supplementary material, figure S1) which might be due to cardenolides still present in the gut. After 10 days, specimens were frozen in liquid nitrogen and homogenized with a pestle (glass or stainless steel). To evaluate stored cardenolides, samples were extracted with 1 ml methanol by vortex stirring. After centrifugation, an aliquot (200 µl) of the supernatant was added to 3 ml scintillation cocktail (Ultima Gold, Perkin Elmer) to quantify radioactivity (amount of [³H]-cardenolide absorbed into the body cavity) with a liquid scintillation counter (Wallac 1409). In addition, the residual radioactivity on the parafilm used as a feeding support and the radioactivity of the drinking solution (2 µl aliquots) were determined to calculate the amount taken up and the percentage of radioactivity actually stored by the hemipteran specimens. Ninety-five per cent confidence intervals of means were calculated in JMP^® Pro v. 10.0.2 (SAS institute Inc.). All feeding experiments (each species was tested for ouabain and digoxin, separately) were repeated with three to 13 specimens (see figure 2 for sample sizes).

Figure 2.

³H-cardenolide-sequestration in nine Lygaeinae and three outgroup species 10 days after oral administration. Bars indicate the proportion of stored cardenolides (digoxin, blue (dark), ouabain, green (light); means ± 95% CI, error bars projecting below zero were cut for clarity). The total amount ingested was set to 100%. The picture of a feeding L. kalmii illustrates the method used to force the hemipteran specimens to drink the test solutions. (Online version in colour.)

3. Results

(a). Phylogeny of Lygaeinae

In the Bayesian and the ML analyses, the subfamily Lygaeinae was recovered as a monophyletic group. The ML tree generated with a GTR + G model fitted to each gene partition (figure 1) and the tree derived from the Bayesian analysis supported without any conflicts the same topology. Of the 11 genera, all individuals of the same species from different populations clustered together and were supported by high bootstrap values. In all genera represented by more than one species, species clustered together. However, the genera Spilosthetus and Lygaeus are paraphyletic with respect to other species. In Spilostethus, Haemobaphus concinnus (Dallas, 1852) is included while in Lygaeus, L. equestris + L. simulans are strongly supported as monophyletic group but the new world species L. kalmii branches off before the remaining Lygaeus and Spilostethus species and does not seem to belong to the same genus. In all analyses, the subfamily Lygaeinae is split into two well-supported sister groups. The smaller one consists of a monophyletic genus Arocatus (A. aenescens, A. rusticus, A. longiceps and A. melanocephalus) with Caenocoris nerii (Germar, 1847) at the base of this clade. In the second major group, C. fulvipes appears with strong support as the most basal lineage and sister group to the remaining Lygaeinae. The relationship between the genera Horvathiolus, Melanocoryphus, Graptostethus, Tropidothorax and O. fasciatus, on the other hand, is less well supported. No previous phylogenetic analyses with a similar coverage of the Lygaeinae exist that could provide additional support for the resolution of this part of the tree [22,33].

(b). Ouabain resistance of Na⁺/K⁺-ATPase in vitro

Na⁺/K⁺-ATPases of all seven Lygaeinae tested here showed a highly similar characteristic of in vitro inhibition by ouabain (figure 3). Their Na⁺/K⁺-ATPase was highly resistant to cardenolides and nearly not affected over three orders of magnitude of ouabain concentration (10⁻⁸ to 10⁻⁵ M). At 10⁻³ M ouabain, where non-adapted Na⁺/K⁺-ATPases are typically completely inhibited [34], the enzyme preparation of all Lygaeinae tested still showed about 70% remaining activity. IC₅₀ values for all species are presented in table 1.

Figure 3.

In vitro inhibition of Na⁺/K⁺-ATPase by ouabain. Each data point represents the mean of three biological replicates ±s.d. (a) Na⁺/K⁺-ATPase of seven species of Lygaeinae (top set of curves) versus Na⁺/K⁺-ATPase of the outgroup species K. resedae (dashed line). A.l., A. longiceps; S.p., S. pandurus; L.k., L. kalmii; O.f., O. fasciatus; L.e., L. equestris; T.l., T. leucopterus; H.s., H. superbus. (b) Comparison of the non-adapted Na⁺/K⁺-ATPase of Drosophila melanogaster (dashed line, G. Petschenka 2012, unpublished data) with the cardenolide-resistant Na⁺/K⁺-ATPase of D. plexippus (dotted line, data from [31]) and O. fasciatus (solid line).

Lygaeine Na⁺/K⁺-ATPase is much more resistant to ouabain (cardenolides) than Na⁺/K⁺-ATPase of the monarch butterfly (figure 3). In comparison, the lygaeid (non-lygaeine) outgroup species used in this study, K. resedae, is highly sensitive to ouabain, starting to show inhibition at ouabain concentrations as low as 10⁻⁷ M. Our data thus support that pronounced target site insensitivity to cardenolides probably is a common feature of all Lygaeinae and evolved at the base of the subfamily ((A) in figure 1).

(c). Cardenolide sequestration in the Lygaeinae

The percentage of cardenolides recovered from individual bugs fed with 2 µl 5 µM [³H]-ouabain (1.46 ng) or [³H]-digoxin (1.56 ng) was determined 10 days after feeding (figure 2).

With the exception of A. longiceps and A. melanocephalus, which stored less than 7% of the initial amount taken up orally for both cardenolides, all other Lygaeinae stored 83–98% of the applied digoxin. The amount of stored ouabain in these species ranged from 55% to about 100%, with the exception of T. leucopterus which stored 10% only. Two of the three outgroup species only held marginal amounts of both compounds (less than 2%), whereas P. apterus, the most distant outgroup species, still had 39% of the initially imbibed ouabain and 6% digoxin. Our results suggest that throughout the sequestering Lygaeinae (all species except for Arocatus), digoxin as a non-polar cardenolide is preferentially stored (five out of seven cases) compared with ouabain. Mapping the evolution of sequestration on our phylogeny unambiguously supports an evolution at the base of the Lygaeinae correlated with the origin of target site insensitivity ((A) in figure 1) and a secondary loss of sequestration in A. longiceps and A. melanocephalus ((B) in figure 1).

(d). Host plant records

The majority (80%) of the species of Lygaeinae represented in our phylogeny (figure 1) are reported to use hosts in the Apocynaceae ([9,35,36], J. Deckert 2014, personal observation). By contrast, A. longiceps and A. melanocephalus do not use Apocynaceae throughout their life cycle but are associated with Platanus (Platanaceae) or Ulmus (Ulmaceae) [36]. Although species of Lygaeinae can be found on a variety of plants (up to 29 plant families in South African species), species of the Apocynaceae are probably used as hosts more frequently than any other plant family. Moreover, in contrast to adults, nymphs of many Hemiptera are rather restrictive in the range of suitable host plants and it seems likely that many species of Lygaeinae even breed on a variety of genera and species of Apocynaceae [10]. Thus, the association with Apocynaceae seems to be a general pattern in this group. In addition to Apocynaceae, several species are known to be closely associated with cardenolide-producing plants from non-related families. Lygaeus equestris for example depends either on V. hirundinaria (Apocyncaceae) as host or on A. vernalis [36], a Ranunculaceae-producing cardenolides [37]. Furthermore, H. superbus is commonly found on D. purpurea ([12,36], G. Petschenka 2012, personal observation), a Plantaginaceae which is rich in cardenolides [38] and S. pandurus uses the bufadienolide (a structurally related class of compounds with the same mode of action as cardenolides) producing Asparagaceae U. maritima as host [39,40].

4. Discussion

We reconstructed the macroevolution of two adaptations to host plant toxins, cardenolide resistance and sequestration. We found that both traits are probably basal features of the Lygaeinae. Our findings demonstrate that the coevolutionary response of herbivores to plant toxins and their use as an acquired defence involves a combination of traits not just a single adaptation. As several insect species with cardenolide-resistant Na⁺/K⁺-ATPase were also shown to sequester the toxins, the two traits might often be linked and are probably co-adaptive (e.g. Chrysochus auratus: Chrysomelidae; Poekilocerus bufonius: Pyrgomorphidae; Danaus plexippus: Nymphalidae; [41–45]). However, the leaf beetle Labidomera clivicollis which feeds on Asclepias species without sequestering cardenolides but still has a Na⁺/K⁺-ATPase bearing resistance conferring substitutions [7] and cardenolide-sequestering arctiid moths (e.g. Empyreuma pugione) with sensitive Na⁺/K⁺-ATPases [34] indicate that both traits are not obligatorily linked.

We found that seven of the nine Lygaeinae tested here store the orally ingested cardenolides ouabain and digoxin in their body. This supports earlier findings of Scudder & Duffey [9] who detected cardenolides in dried museum specimens of more than 20 genera of Lygaeinae. The lack of sequestration in A. longiceps and A. melanocephalus found here is in line with the life history of these two species which live on plants not known to produce cardenolides (Platanus: Platanaceae, or Ulmus: Ulmaceae, respectively). Feeding on non-apocynaceous hosts is most likely a derived state in this genus, as Arocatus species in Australia are well known to use apocynaceous plants (A. aenescens, A. chiasmus, A. continctus, A. montanus and A. rusticus feed on Araujia, Asclepias, Gomphocarpus, Nerium and Parsonsia species, respectively [35]). As A. longiceps and A. melanocephalus are recovered as sister species in a monophyletic group with A. aenescens and the cardenolide-sequestering A. rusticus [9] and this group is moreover in a sister group relationship with the cardenolide-sequestering C. nerii [19], we assume that the ability to sequester cardenolides was lost in the branch leading to the two European Arocatus species. This finding may indicate that physiological adaptations which are necessary to sequester cardenolides might be costly and are reduced if not needed. Comparatively small amounts of ouabain (but only marginal amounts of digoxin) sequestered by the most distant outgroup species P. apterus may possibly derive from adaptations to other plant compounds and do not go along with reduced sensitivity of its Na⁺/K⁺-ATPase (see also sequence data in [7]).

For the second cardenolide-related trait investigated here, target site insensitivity of Na⁺/K⁺-ATPase, an opposite pattern appeared. Both species, A. longiceps and A. melanocephalus, which are not exposed to dietary cardenolides and do not store the toxins, still possessed a cardenolide insensitive Na⁺/K⁺-ATPase. As target site insensitivity was maintained in A. longiceps and A. melanocephalus, this adaptation may not cause significant costs. Our findings suggest that cardenolide-resistant Na⁺/K⁺-ATPases and sequestration of dietary cardenolides are likely to be synapomorphic features of the Lygaeinae which originated at the basis of this group. Given that Lygaeinae are the most species-rich lineage within the Lygaeidae [46,47] with more than 500 species compared with Orsillinae (250 species) and Ischnorhynchinae (75 species), the adaptations to dietary cardenolides may represent a key innovation of this group.

Milkweed bugs on five continents use Apocynaceae as hosts (e.g. North America: Asclepias, South America: Asclepias, Eurasia: Nerium, Vincetoxicum, Africa: Gomphocarpus, Australia: Parsonsia [9,12,16,35]) suggesting a very old relationship between these two taxa. The common occurrence of cardenolides in the Apocynaceae was thus probably the evolutionary driver for the observed adaptations in Lygaeinae. This ancestral tolerance for cardenolides then obviously enabled several species of Lygaeinae to switch to distantly related families which also produce cardenolides. Examples include L. equestris feeding on A. vernalis, a cardenolide-producing Ranunculaceae [37] or H. superbus which often is associated with D. purpurea (Plantaginaceae) ([12,36], G. Petschenka 2012, personal observation). Furthermore, S. pandurus uses U. maritima as a host [40], an Asparagaceae known to contain bufadienolides. Most likely, these lygaeine species also sequester cardenolides from these sources as has been shown for L. equestris (G. Petschenka 2009, unpublished data) collected from A. vernalis. Such a use of plants from non-apocynaceaous families in the Lygaeinae certainly is an example for host shifts which were driven by pre-existing adaptations to specific secondary compounds.

Specialization on a specific host was often assumed to trade off with the ability to use other hosts thus leading to an evolutionary dead end [48–50]. The adaptive radiation of Lygaeinae on Apocynaceae and their frequent host shifts to plants bearing similar secondary compounds, rather represent an example for host shifts along phytochemical bridges that were enabled by pre-existing adaptations as has also been observed in pyrrolizidine alkaloid-sequestering species [51–54]. The abundance of cardenolides in the Apocynaceae might even be the basis for the high diversity of the Lygaeinae indicating that highly specific adaptations like cardenolide-resistant Na⁺/K⁺-ATPase also can lead to evolutionary success if toxins occur commonly in a highly diverse plant family. Successful development of several species on cardenolide-free hosts (e.g. A. longiceps and A. melanocephalus on Platanus and Ulmus, or Graptostethus servus on Ipomoea [10]) also prove that adaptations which are highly specific for a certain class of plant toxins do not necessarily lead to an evolutionary dead end.

Our in vitro investigation of lygaeine Na⁺/K⁺-ATPase extracted from hemipteran nervous tissue showed a very uniform pattern. Na⁺/K⁺-ATPase preparations from all seven species investigated here showed the same ouabain resistance. The characteristic of Na⁺/K⁺-ATPase inhibition by ouabain observed here resembles the one described by Moore & Scudder [20]. Remarkably, we have not found a stepwise pattern of evolution of Na⁺/K⁺-ATPase resistance as detected in milkweed butterflies [31] but rather strong resistance even in the species placed on the earliest branches in our phylogeny. Interestingly, cardenolide resistance of milkweed bug Na⁺/K⁺-ATPase is about 10-fold higher than cardenolide resistance of D. plexippus Na⁺/K⁺-ATPase (based on the IC₅₀ of ouabain). This remarkable difference could be an adaptation to seed feeding of milkweed bugs as Asclepias seeds in some species have been reported to have much higher cardenolide concentrations than leaves [55]. Moreover, the higher level of resistance might provide the physiological basis for the much higher concentrations of sequestered cardenolides in milkweed bugs as compared with monarch butterflies [3].

Previous analyses of Na⁺/K⁺-ATPase α gene sequences (the subunit of Na⁺/K⁺-ATPase where cardenolides bind to) of L. kalmii and O. fasciatus have shown that both species possess amino acid substitutions at positions known from structural analyses and mutagenesis of the mammalian gene to confer resistance to cardenolides [7,8]. Furthermore, transcriptome analyses of both species evidenced the presence of three copies of the relevant Na⁺/K⁺-ATPase α1 gene. Based on divergence estimates, it was assumed that many lygaeine species possess the same set of Na⁺/K⁺-ATPase copies [8]. This is in good agreement with the homogeneity of our in vitro ouabain inhibition assays in seven species. All of the three ATPase α1 gene copies share a replacement of the conserved asparagine by histidine at position 122 [8], a substitution which has been shown to significantly lower the sensitivity of the Na⁺/K⁺-ATPase to ouabain [7,21,56]. Our preliminary analyses of gene transcripts in the Lygaeinae investigated here (see the electronic supplementary material, table S1) corroborate the presence of this histidine residue in all species, however, only a combination of four substitutions, glutamine¹¹¹threonine + asparagine¹²²histidine + phenylalanine⁷⁸⁶asparagine + threonine⁷⁹⁷alanine, as observed in the ATPase α1C copy [8] results in an enzyme showing similar resistance to ouabain as observed here (S. Dalla & S. Dobler 2014, unpublished data). Which of the three Na⁺/K⁺-ATPase α copies is predominantly expressed in the nervous tissue is still an open question, yet it is plausible that all species express the same set of genes with identical amino acid substitutions in their Na⁺/K⁺-ATPases. We can assume that the series of gene duplications and amino acid substitutions must have arisen by stepwise evolution just as in milkweed butterflies, but only a broader sampling of other basal lygaeines and more closely related outgroups can potentially reveal traces of this process.

Our finding of insensitive Na⁺/K⁺-ATPase(s) expressed in the nervous system of the Lygaeinae suggests that Na⁺/K⁺-ATPase in the hemipteran nervous system is not protected by the perineurium as was suggested for cardenolide-sequestering Lepidoptera that nevertheless possess sensitive Na⁺/K⁺-ATPases [34,57]. As Na⁺/K⁺-ATPase is essential for the generation of neural action potentials and is heavily expressed in insect nervous tissues [5,34], selection pressure by cardenolide-containing host plants must probably result in either of these two protective mechanisms.

We have shown that two traits associated with the beneficial use of a host plant toxin most likely form a co-adaptive strategy. Our findings indicate that sequestration of toxins need to be accompanied by other adaptations to compensate for the cost which otherwise would arise when plant toxins are incorporated into an insect's body. Given the evolutionary success of milkweed bugs, the ecological benefit achieved by sequestration not only seems to outweigh potential costs of these additional adaptations but also could, on the other hand, lead to a dual increase of fitness. For the milkweed bugs, it was suggested that sequestration of host plant toxins may have superseded the role of the metathoracic scent gland which is generally involved in defence in the Hemiptera. In this case, the insects not only save the production of endogenous defensive compounds but the gland even secondarily adopted a sexual function [58]. Evolutionary trade-offs like this could well exist in other groups of Hemiptera and have also been shown to occur in other groups of herbivorous insects [1].

Data accessibility

Data from Na⁺/K⁺-ATPase and sequestration assays were deposited on Dryad (doi:10.5061/dryad.1707a). All sequences generated for this study have been deposited under GenBank accession no. LN623642–LN623676 and LN614550–LN614584.

Funding statement

This work was supported by the German Research Foundation (DO 527/5-3 and PE 2059/1-1) and the Templeton Foundation.