ABCB transporters in a leaf beetle respond to sequestered plant toxins

Abstract

Phytophagous insects can tolerate and detoxify toxic compounds present in their host plants and have evolved intricate adaptations to this end. Some insects even sequester the toxins for their defence. This necessitates specific mechanisms, especially carrier proteins that regulate uptake and transport to specific storage sites or protect sensitive tissues from noxious compounds. We identified three ATP-binding cassette subfamily B (ABCB) transporters from the transcriptome of the cardenolide-sequestering leaf beetle Chrysochus auratus and analysed their functional role in the sequestration process. These were heterologously expressed and tested for their ability to interact with various potential substrates: verapamil (standard ABCB substrate), the cardenolides digoxin (commonly used), cymarin (present in the species's host plant) and calotropin (present in the ancestral host plants). Verapamil stimulated all three ABCBs and each was activated by at least one cardenolide, however, they differed as to which they were activated by. While the expression of the most versatile transporter fits with a protective role in the blood–brain barrier, the one specific for cymarin shows an extreme abundance in the elytra, coinciding with the location of the defensive glands. Our data thus suggest a key role of ABCBs in the transport network needed for cardenolide sequestration.

1. Introduction

The coevolutionary arms race between herbivores and their host plants gave rise to a plethora of defensive strategies and counter-adaptations [1]. An impressive number of plants evolved an arsenal of bioactive secondary metabolites to repel their natural enemies. In turn, many insect species have adapted to tolerate these phytotoxins and exploit them for their anti-predator defence, often by transferring them into specialized defence glands [2,3]. This strategy, however, necessitates adaptations to tolerate the toxins and to traffic them within the body in a coordinated and controlled way [4,5].

Cardiac glycosides (consisting of cardenolides and bufadienolides) represent a class of phytotoxins that evolved convergently in multiple plant lineages and target Na,K-ATPase, an essential transmembrane ion carrier that all animal cells depend on but plants do not possess. Despite their general toxicity, cardiac glycosides provided the first example of a plant toxin recycled as an anti-predator defence by monarch butterflies [6,7]. This strategy is shared by many other insects [8] and they all share the challenge of avoiding autointoxication when accumulating cardiac glycosides in their body. One strategy for avoiding the toxic effect of cardiac glycosides involves the target site insensitivity of the Na,K-ATPase. Several studies demonstrated that specific amino acid substitutions in this enzyme evolved convergently as an adaptation to cardenolide-containing host plants [9–11]. Yet not all sequestering insects rely on target site insensitivity, and even in those that do, additional adaptations are usually necessary to protect the sensitive nervous tissue from the toxic effects of cardiac glycosides [12,13]. These additional adaptations remain largely unexplored; our goal was therefore to identify and characterize the carrier proteins responsible for the transport and compartmentalization of cardiac glycosides.

The dogbane beetle Chrysochus auratus (Coleoptera, Chrysomelidae) offers an ideal model to investigate the molecular mechanisms and physiological adaptations necessary for the evolution of host plant specialization and a sophisticated defence system [14]. The adults of this species sequester polar and apolar cardenolides from their host plant Apocynum cannabium into their defensive secretions. As in many other leaf beetle species, these secretions ooze from glands on the pronotum and elytra when the beetles are attacked [15]. This selective accumulation of cardenolides in specialized glands requires directed transport, possibly detoxification and protection of the nervous tissue, where a sensitive form of Na,K-ATPase is expressed (a cardenolide resistant form is expressed in other tissues [10]).

Although recycling of host plant-derived cardenolides by specialized insects is widespread, the decisive adaptations necessary for selective uptake, transport and accumulation of cardenolides remain largely unresolved. The blood–brain barrier surrounding the nervous tissue can form a diffusion barrier for polar compounds but exclusion of apolar cardenolides requires additional mechanisms in the form of transport proteins [12,16]. Promising candidate transporters involved in cardiac glycoside trafficking are ABCB (ATP-binding cassette, subfamily B) proteins, which in mammals as well as arthropods shield the sensitive nervous tissue against intruding cardenolides [12,17–19]. An ABC transporter was previously shown to be involved in the sequestration of plant compounds into defensive secretions in a leaf beetle [20]. We therefore focused on these carriers to disentangle the cardenolide transport network in C. auratus. Searching transcriptomic data, we identified three different C. auratus ABCB transporters and expressed them heterologously in cell culture. ATPase activity assays were employed to investigate the capacity of the harvested proteins to mediate the transport of different cardenolides. Additional tissue-specific qRT-PCR analyses located the expression sites of the three transporters. Our results show that the ATPase activity of all three C. auratus ABCB transporters is stimulated by at least one of the tested cardenolides, indicating their involvement in active cardenolide transport, and give strong hints to their tissue-specific functional roles.

2. Material and methods

(a). Transcriptome assembly and annotation

Total RNA of one specimen of C. auratus collected in Ithaca, New York, USA, was isolated with the RNeasy Plus Mini kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. An mRNA library was constructed and the transcriptome sequenced on an Illumina HiSeq 2000 (Illumina, San Diego, USA) by GENterprise (Mainz, Germany). Paired-end Illumina reads were processed and assembled de novo using T-IDBA [21]. The resulting database of 91 487 sequences was searched using the BLAST+ command-line tool [22]. Several known or predicted insect ABCB sequences (see electronic supplementary material, table S1) were used as input for tblastn commands. Sequence hits with e < 10⁻⁵⁰ were selected, aligned with the AlignX module of Vector NTI Advance (11.5.1, Invitrogen, LifeTechnologies; Carlsbad, USA) and clustered into three isoforms, each containing an open reading frame (ORF) of approximately 4 kb length, potentially encoding C. auratus ABCB transporters. These isoforms were checked and verified by PCR.

(b). Phylogenetic analysis

Sequences of 43 ABC full transporters (40 ABCBs and 3 ABCCs) including the three C. auratus sequences were aligned using ClustalX in VectorNTI. All positions containing gaps and missing data were eliminated, leaving a total of 419 positions in the final dataset. A maximum-likelihood tree generated in MEGA7 [23] and based on the LG model [24] with gamma-distributed rates (α = 1.0179) and invariable sites (I = 4.95%) was rooted on ABCC transporters as outgroup. Initial trees for the heuristic search were obtained automatically by Neighbor-Joining and BioNJ algorithms based on a JTT model.

(c). Homology protein structure modelling

We modelled the tertiary protein structure of the open reading frames of the three identified C. auratus ABCB transporters (Ca_ABCB) based on the available crystalline structures of mouse and C. elegans ABCB1 using the Phyre2 protein-modelling server (www.sbg.bio.ic.ac.uk/phyre2), [25]. The predicted protein tertiary structures were then visualized using PyMOL (www.schrodinger.com/pymol) and graphs further improved with Adobe Illustrator CC 2018 (Adobe Inc.).

(d). Amplification, cloning, and generation of recombinant baculovirus

We used RNA of the same specimen used for transcriptome sequencing for first-strand cDNA synthesis (SuperScript III reverse transcriptase, Invitrogen) and RT-PCR. The initial amplification of potential transcripts of Ca_ABCB transporter was carried out with primers located in the untranscribed regions of the previously identified transcripts. PCR products of expected length were reamplified with primers including start and stop codon of the ORFs, suitable restriction sites for cloning, and a short terminal loading ramp (electronic supplementary material, table S2). PCR products were purified (QIAquick Gel Extraction Kit, Qiagen), digested with the selected enzymes and cloned into pFastBacDual (Invitrogen). Clones showing matching insert lengths were checked by sequencing (GATC Biotech; Konstanz, Germany).

To produce high levels of heterologously expressed proteins, we used the Bac-to-Bac expression system (Invitrogen) following previously described methods [26]. For each construct, separate rounds of infections with Ca_ABCB-baculovirus were conducted to produce three to five independent membrane vesicle preparations following the protocol of Sarkadi et al. [27], with slight modifications. Cell pellets were resuspended in 8 ml TEMP buffer, pH 7.0 (50 mM Tris, 50 mM mannitol, 2 mM EGTA-Tris pH 7.0, 0.05 µg µl⁻¹ PMSF (AppliChem; Chicago, USA) + 1 tablet/50 ml buffer cOmplete Protease Inhibitor Cocktail (Roche; Basel, Switzerland) + 2 mM dithiothreitol (DTT) (Fluka; St Louis, USA), transferred into a glass homogenizer and pounded 100 times. Homogenates were centrifuged (500g, 4°C) and the supernatants were decanted into new centrifugation tubes [27]. The supernatants of each sample were pooled and centrifuged (100 000g, 4°C) to get rid of cytosolic Sf9 cell compartments. Pelleted crude membrane fractions were resuspended in TEMP (without DTT) and pressed 25 times through a syringe needle (27 × 3/4, B. Braun; Melsungen, Germany) to generate Ca_ABCB-membrane vesicles.

(e). Immunodetection and western blot of heterologously expressed Ca_ABCB proteins

The expression patterns of Ca_ABCB proteins in Sf9 cells 72 h post-infection were monitored by immunocytochemistry using the monoclonal ABCB antibody C219 (Thermo Scientific) and revealed with a Cy3-conjugated goat anti-mouse secondary antibody following standard procedures [26].

For detection on western blot equal volumes of each sample (containing 100 µg total protein) were denaturated (5 min, 55°C) in Laemmli buffer, resolved on a 7.5% SDS-PAGE (1.5 h, 100 V) and then transferred onto nitrocellulose membranes (Whatman, PROTEAN). Western blotting was carried out on a Mini-PROTEAN Tetra Handcast System (Biorad; Hercules, USA). Ponceau S solution (0.1% Ponceau S + 5% acetic acid) was used to check the amount of protein and to fix proteins to the membrane and removed by washing with ddH₂O. Membranes were subsequently incubated in TBS-T blocking buffer (tris-buffered saline + 0.05% Tween20) plus 5% milk powder for 1 h at room temperature, washed and then incubated overnight at 4°C with 2 µg µl⁻¹ C219 antibody in TBS-T (+1% milk powder). The primary antibody was removed and membranes were incubated in TBS-T (+1% milk powder) with 0.4 µg ml⁻¹ horseradish peroxidase-coupled antibody (Dianova) for 2 h at room temperature. After washing (3 × 10 min), bound antibodies were visualized by DAB staining (0.08% 3, 3′-diaminobenzidine tetrachloride, (Carl Roth) +0.04% cobalt chloride, 0.09% H₂O₂ in 50 mM Tris–HCL, pH7.5).

(f). Ca_ABCB ATPase activity assays

Crude Sf9 cell membranes overexpressing Ca_ABCBs were isolated to determine the substrate-dependent ATPase activity by measuring the release of inorganic phosphate following the method described by Sarkadi et al. [27] with slight modifications. We examined the interaction of the Ca_ABCB transporters with verapamil (Merck; Darmstadt, Germany), a well-known stimulator of Pgp ATPase activity, and three different cardenolides, cymarin (Phytolab; Vestenbergsgreuth, Germany), digoxin (Merck) and calotropin (isolated by preparative high performance liquid chromatography from monarch caterpillars raised on Asclepias curassavica, following the protocol in Petschenka et al. [28]). All cardenolides were first dissolved in dimethyl sulfoxide that was further diluted to a final concentration of 2%. Assays were run in 96-well plates in duplicates. Finally, we produced 12–32 technical replicates for each Ca_ABCB transporter-cardenolide combination. Ca_ABCB membrane vesicle preparations (5 µg protein per well) were incubated (20 min, 27°C) in 30 µl assay buffer (final concentration 50 mM 3-(N-morpholino)propanesulfonic acid(MOPS)-Tris, pH 7.0, 50 mM KCL, 5 mM Na-acid, 100 mM egtazic acid(EGTA)-Tris, pH 7.0) with different cardenolide concentrations in the presence and the absence of 500 µM sodium orthovanadate (Merck), a non-competitive P-type ATPase inhibitor. The reaction was started by adding 20 µl Mg-ATP (Merck; final concentration 3 mM) and stopped (after 20 min, 27°C) by adding 20 µl 5% sodium dodecyl sulfate. To reveal the amount of phosphate released the following reagents were added: 50 µl P-reagent (2.5 M H₂SO₄ + 1% ammonium molybdate (Carl Roth) + 0.014% antimony potassium tartrate (Merck), 20 µl 20% acetic acid, 25 µl 1% ascorbic acid). The colorimetric reaction took place at room temperature (15 min) and was measured in a microplate reader at 655 nM (Model 680, Bio-Rad). Based on the estimated raw data and phosphate standard curves of K₂HPO₄, we calculated the release of inorganic phosphate and the ATPase activity (nmol Pi mg⁻¹ protein min⁻¹) in each well. Dose-dependent effects of cardenolide treatment (0–800 µM) on Ca_ABCB ATPase activities were calculated as the difference between ATPase activity (nmol Pi mg⁻¹ protein min⁻¹) in the presence and absence of sodium orthovanadate. Vanadate-sensitive ATPase activity of membrane vesicles of uninfected Sf9 cells in the absence of cardenolides was defined as baseline activity.

(g). Kinetics/statistics

Experimental data were fitted to the expanded Michaelis–Menten equation including Hill coefficients (OriginPro 2016G; OriginLab Corporation):

$$V=V_{min}\frac{V_{max}-V_{min}}{1+{10}^{(\log \mathrm{E}{\mathrm{C}}_{50}+[c])n}}$$

where V = velocity (nmol pi mg⁻¹ protein min⁻¹), V_min = minimal velocity, V_max = maximal velocity, EC₅₀ = ligand concentration producing 50% of maximal response, c = the actual test drug concentration and n = Hill slope, the parameter characterizing the degree of substrate-binding site cooperativity.

Ca_ABCB ATPase transporter-specific activities at 800 µM cardenolide treatment were tested for statistically significant differences by two-way ANOVA (OriginPro 2016G) followed by Tukey post hoc tests. Since no dose–response curves could be fitted in cases where the transporters did not react to a cardenolide, data were compared by additionally fitting curves with nonlinear mixed effects models in R (procedure nime) and evaluating statistical differences between the slopes.

(h). qRT-PCR of Chrysochus auratus tissues

To investigate the tissue distribution of Ca_ABCB transporters, eight different tissues of nine C. auratus beetles (collected in Ithaca, New York, USA) were dissected: crop to midgut, midgut to rectum, Malpighian tubules, reproductive organs, muscles, tracheae, nervous tissue and elytra. Tissue samples of three beetles were pooled (resulting in one biological replicate) before RNA extraction (RNeasy Mini Kit; Qiagen). RNA quality was checked spectrophotometrically (Nanodrop 2000, Thermo Fisher Scientific) and concentrations equilibrated before cDNA synthesis (ProtoScript II Synthesis Kit, New England Biolabs). Exon-spanning gene-specific primers (electronic supplementary material, table S2) were designed based on the Ca_ABCB sequences and the Ensembl Metazoa database.

Quantification of the transcript levels was based on a standard curve of linearized pFastBacDual plasmids containing the entire coding region of Ca_ABCB1,2 or 3. Threshold cycle (Ct) values were estimated on a StepOne Real-Time PCR System and pre-analysed with the StepOne Software v. 2.3 (Thermo Fisher Scientific). cDNA products were amplified in triplicates using the EvaGreen qPCR Mix II (ROX) (Bio-Budget Technologies, Germany) in a volume of 20 µl with a final cDNA amount of 30 µg total RNA by following the manufacturer's protocol. Amplification conditions were 90°C for 15 min followed by 40 cycles consisting of 15 s at 95°C, 20 s at 60°C, and 20 s at 70°C. Control samples without cDNA were loaded on each 48-well plate to check for unspecific amplification products via melting curve analysis. Specificity and amplification efficiencies of the qPCR-primers were first evaluated by duplicate standard curve reactions with 10-fold serial dilutions (from 10⁸ to 10²) of recombinant plasmids [29].

Means of three biological replicates were estimated and statistically analysed using one-way ANOVA (OriginPro 2016G, OriginLab Corporation).

3. Results

(a). Phylogenetic reconstruction and homology-based three-dimensional structure models of Chrysochus auratus ABCB proteins

The three identified C. auratus ABCB transporter sequences clustered with all other beetle ABCB sequences in a monophyletic clade divided into two well-supported branches, comprising sequences designated as ABCB1 or ABCB2 in most other beetle species. A recent gene duplication gave rise to Ca_ABCB2 and Ca_ABCB3, similar to the situation observed in Dendroctonus ponderosae. The only non-beetle sequence in this monophyletic clade belongs to Drosophila melanogaster MDR50 (figure 1). Protein structure modelling of Ca_ABCBs shows that the mouse ABCB1 and the human ABCB1 share the common architecture consisting of two transmembrane domains and cytosolic nucleotide binding domains (electronic supplementary material, figure S1a), supported with 100% prediction confidence.

Figure 1.

Maximum-likelihood tree (LG + Γ + I) of ABC transporter amino acid sequences (cf. electronic supplementary material, table S1); Chrysochus sequences marked with stars (colours according to figure 2); numbers indicate bootstrap proportions in 1000 replicates. (Online version in colour.)

To identify key residues decisive for the interaction with cardenolides, we focused on nine amino acids (Ile306, Phe336, Ile340, Phe343, Phe728, Phe942, Thr945, Leu975, and Val982) that were previously identified to be promiscuous residues in the ligand-binding pocket of mouse ABCB1 [30,31]. Eight of these (underlined) were strongly linked to interactions with different cardenolides [32] (figure 2). These amino acids are conserved at the homologous positions between the mouse and human ABCB1 transporter (electronic supplementary material, figure S1a). To better visualize the residues situated on the inward-facing binding pocket of the transporters, we placed a cut plane through all Ca_ABCB transporters and highlighted these nine residues and the corresponding Ca_ABCB residues (figure 2a; adapted from [32]). Four of these residues were identical (orange residues) between human and Ca_ABCB3 (figure 2d), three in Ca_ABCB2 (figure 2c) and one in Ca_ABCB1 (figure 2b). Five of the nine residues do not match in any Ca_ABCB transporters (electronic supplementary material, figure S1b). However, the conserved human Phe336 at the M-site is present in all Ca_ABCBs at the corresponding positions (figure 2). Likewise, phenylalanine is present in Ca_ABCB2 and Ca_ABCB3 at the positions corresponding to the human Phe728 and human Phe942.

Figure 2.

Three-dimensional structural superposition of Ca_ABCB proteins (a). Horizontal sections through the Ca_ABCB transporters' periplasmic area affording a view of the potentially important residues discussed in the text for Ca_ABCB1 (b), Ca_ABCB2 (c), Ca_ABCB3 (d). Amino acids identical to the human_ABCB1 are shown in orange, yellow indicates dissimilar residues. Val982 highlighted in red is essential for the transport of convallatoxin [31]. (Online version in colour.)

(b). Expression of Chrysochus auratus ABCB transporters in Sf9 cells

All three Ca_ABCB transporter transcripts were successfully amplified, cloned and expressed in Sf9 cells. Immunocytochemistry verified that the ABCB proteins were strongly expressed in the cell membrane. A weaker fluorescence signal was detected in nuclear membranes and the Golgi apparatus (electronic supplementary material, figure S2a). This signal was almost imperceptible in the empty baculovirus control (electronic supplementary material, figure S2b). The western blot analysis confirmed the overexpression of all three Ca_ABCB transporters in membrane vesicle preparations and a molecular size of approximately 130 kDa (electronic supplementary material, figure S2c). Bioinformatic analysis of the amino acid sequences of the Ca_ABCB proteins indicated that all Ca_ABCB transporters have two potential binding sites for the C219 antibody, yet Ca_ABCB1 and Ca_ABCB2 harbour one amino acid replacement in the first C219 epitope (VQEALD), closer to the N-terminus of the proteins (electronic supplementary material, figure S2d).

(c). Effects of cardenolides on Ca_ABCB transporter ATPase activity

The ATPase activity of all three Ca_ABCB transporters was stimulated in a dose-dependent way by at least one of the three tested cardenolides. To validate the functionality of the three proteins we first confirmed that they were stimulated by verapamil, which was strongest at a concentration of 60 µM (electronic supplementary material, figure S3a). By contrast, the ATPase activity of vesicle preparations from uninfected Sf9 cells was not affected by verapamil or any of the tested cardenolides (data not shown). At the lowest tested cardenolide concentration of 5 µM, the mean of Ca_ABCB ATPase activities (25 nmol Pi min⁻¹ mg⁻¹ protein, s.d. ± 1.54) matched the baseline ATPase activity (mean of 0 µM cardenolides = 26.6 nmol Pi min⁻¹ mg⁻¹ protein, s.d. ± 1.94). The strongest increase in transporter-specific ATP hydrolysis was in all cases reached at the maximum cardenolide concentration of 800 µM (electronic supplementary material, figure S3b). At this concentration, the specific activity of all three tested transporters was raised by calotropin. The ATPase activity of Ca_ABCB1 and 2 was also significantly stimulated by digoxin, whereas Ca_ABCB3 did not react to this substrate. Only Ca_ABCB2 showed significantly increased ATPase activity in the presence of 800 µM cymarin, while Ca_ABCB1 and 3 were not stimulated. The estimated Hill slopes from the dose–response curve fitting (figure 3) show that velocities of ATP hydrolysis were similar for all three tested cardenolides (electronic supplementary material, table S3). An analysis of exponential curve slopes reveals that digoxin markedly raised the ATPase activity of Ca_ABCB1 in a dose-dependent manner and significantly more so than the one of Ca_ABCB2 (figure 3d). By contrast, Ca_ABCB3 did not respond to digoxin (figure 3a). While the ATPase activity of Ca_ABCB3 was only significantly stimulated by calotropin (figure 3b), the Ca_ABCB2 transporter-specific ATPase activity was significantly stimulated by all three cardenolides. Calotropin and cymarin stimulated Ca_ABCB2 significantly more than the other two transporters. Most intriguingly, an analysis of exponential curve slopes confirms that the ATPase activity of the latter transporters was not (Ca_ABCB3) or only marginally (Ca_ABCB1) stimulated by cymarin (figure 3c,d).

Figure 3.

Ca_ABCB mediated ATP hydrolysis (nmol pi mg⁻¹ protein min⁻¹) in the presence of (a) digoxin, (b) calotropin or (c) cymarin (host plant of C. auratus). Mean ± s.d. and a dose–response fit of the data are shown as solid lines, broken lines indicate data where no dose–response curves could be fitted. (d) For statistical comparisons, exponential curves were fitted to all data by nonlinear mixed effect models and the slopes of the resulting curves compared. Different letters indicate statistically significant differences (p < 0.05) between the slopes. (Online version in colour.)

(d). Tissue-specific expression of Ca_ABCB transporters

RT-qPCR data revealed that the three Ca_ABCB genes have a strongly differentiated tissue distribution. The most remarkable finding is the high expression level of Ca_ABCB2 in the elytra (figure 4) where the defensive glands are situated and total RNA levels were extremely low. Also, this transporter was strongly expressed in the nervous tissue, which otherwise showed by far the highest expression of Ca_ABCB1 of all tissues. The expression of both transporters in the other tested tissues was much lower (some Ca_ABCB2 in muscles and reproductive organs) to negligible. The expression of Ca_ABCB3 was an order of magnitude lower in all tissues but highest in the two gut tissues. Comparing the transcript levels of the three Ca_ABCBs across the eight tested tissues we found that Ca_ABCB1 had significantly higher expression in the nervous tissue (F_7,16 = 4.1, p < 0.05) while Ca_ABCB2 transcripts were strongly expressed in the nervous tissue (4.67 × 10⁵; s.d. ± 3.80 × 10⁵ copies µg⁻¹ RNA) and was significantly more abundant in the elytra (F_7,16 = 16.72, p < 0.05).

Figure 4.

Quantification of Ca_ABCB transporters expression levels in eight adult tissues of C. auratus. Bars represent the copies of Ca_ABCB transcripts per µg RNA. All data are means of three biological replicates ±s.d. Different letters indicate statistically significant differences (p < 0.05) between expression levels of each transporter across tissues (ANOVA, with post hoc Tukey tests). (Online version in colour.)

4. Discussion

Our functional tests with the three ABCB transporters of C. auratus support their pivotal role for the beetles' handling of cardenolides from their host plants: all three Ca-ABCB transporters interacted with at least one of the tested cardenolides. Combined with their tissue-specific expression, our data support a conserved protective function in the blood–brain barrier, on one hand, and a derived function involved in sequestration of the host compound cymarin into the defensive glands, on the other hand.

ABCB proteins have been extensively studied in vertebrate systems in the context of drug resistance, mainly relating to cancer therapy [33]. Yet they also play a vital role in insect adaptations to insecticides [16,34,35] and have previously been implied in herbivores' adaptation to host plant toxins [12,36]. Our present data broaden our understanding of the role of ABCB transporters in this context. All three Ca_ABCBs strongly reacted to the presence of verapamil, a standard substrate of ABCB transporters, but differed in substrate specificities for the employed cardenolides digoxin, calotropin and cymarin, with only partly overlapping substrate specificities. Most strikingly, only Ca_ABCB2, resulting from recent gene duplication, was stimulated by the presence of cymarin, a cardenolide present in the beetles' host plant. By contrast, calotropin, a cardenolide found in ancestral hosts of C. auratus relatives [37], stimulated the ATPase activity of all three Ca_ABCBs. Digoxin, which was shown to be a substrate for ABCB1 (P-gp) transporters in both vertebrates and insects [12,32,38], interacted with both Ca_ABCB1 and Ca_ABCB2 while the ATPase activity of Ca_ABCB3 was not stimulated.

Ligand docking predictions and site-directed mutagenesis studies suggest that polyspecificity of ABCBs to physico-chemically diverse substrates is determined mostly by characteristics of the transmembrane binding domain-associated R-site (rhodamine binding site) and H-site (Hoechst binding site), as well as the M-site (modulatory site) located at the inward-facing top of the transporter [30,39–41]. Studies on a P-gp overexpressing MDCK-cell line revealed that a single mutation in the substrate-binding pocket of the transporter (Gly185Val; H-site) inhibited the transport of digoxin [42]. However, digoxin and other ligands can alternatively act as substrate, competitive inhibitor or modulator of P-gp [43]. Previous studies assessed several residues among ABCB transporters to be evolutionarily conserved in distantly related organisms and to be involved in ligand-binding and substrate specificity [41,44–48]. In silico predictions of P-gp-ligand interactions by Klepsch et al. suggest that human Phe336 belongs to the inhibitor binding residues [49]. Gozalpour et al. [32] reported that the affinity of cymarin to human Ile340 plays a crucial role in P-gp inhibition. We found that this amino acid is replaced by methionine in Ca_ABCB1, by threonine in Ca_ABCB2 and by asparagine in Ca_ABCB3 suggesting that inhibition of Ca_ABCBs by cymarin might be disabled by these substitutions. Mutagenesis studies provided evidence that Val982 (M-site) is essential for the transport of convallatoxin, but did not affect the affinity for cymarin and digoxin [31]. The human Val982 (figure 2b–d, red residue) is replaced by isoleucine in Ca_ABCB1 and Ca_ABCB3 and leucine in Ca_ABCB2. Loo et al. [50] reported that Val982 constitutes an important residue for verapamil binding yet its replacement in Ca_ABCBs does not corrupt stimulation by verapamil. We thus found that at least the discussed residues do not hint at a distinct pattern reflecting the observed substrate profiles of the Ca_ABCBs. We conclude that, although the selected amino acids may be involved in cardenolide interactions, they are not decisive for determining how cardenolide specificity of Ca_ABCB transporters evolved.

The tissue-specific distribution of ABCB transporters is directly related to their specific physiological functions in mammals and arthropods [51,52]. By quantifying the absolute amount of mRNA we elucidated the tissue-specific expression patterns of Ca_ABCBs that in combination with their substrate profiles give strong hints to their functional roles. Ca_ABCB2, the only transporter showing interaction with the host cardenolide cymarin in the ATPase assays, was highly expressed in the elytra. This strongly suggests that Ca_ABCB2 plays a key role in the selective transport and sequestration of host cardenolides into the defence glands of C. auratus. In the leaf beetle Chrysomela populi, a member of the ABCC transporter family, CpMRP, was likewise shown to be involved in the transport of plant glucosides from the haemolymph into the beetle's defensive secretions [20]. While transport into defence fluids represents a highly derived mechanism, the protective role of ABCB transporters in the blood–brain barrier against cardenolides like digoxin is well established in vertebrates and insects alike [12,17,38]. In C. auratus this neuroprotective role is apparently realized by Ca_ABCB1 together with Ca_ABCB2, thus covering the full spectrum of cardenolides tested here. In addition, exclusion of toxins by ABCB transporters is an important process at the gut membrane and in the Malpighian tubules [13,38,53]. Nevertheless, only Ca_ABCB3 shows an appreciable expression in these tissues leaving the possibility that excretion of cardenolides is not a prime task of Ca_ABCBs and that members of other ABC gene families may play a more important role in preventing the uptake of xenobiotics at the gut membrane or enabling their excretion via Malpighian tubules [53].

Taken together our data strongly support that ABCB carrier proteins play a decisive role in determining the trafficking of sequestered cardenolides while safeguarding the sensitive nervous tissue. We assume that the evolution of diversified Ca_ABCB substrate and tissue specificities that coincide with a recent gene duplication paved the way in the adaptation of C. auratus to its host plant and its specific secondary compounds.

Data accessibility

Sequences have been deposited to the European Nucleotide Archive ENA, study no. PRJEB39596 and accession numbers LR862432–LR862436. Raw data of enzyme assays and qPCR analyses are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.r4xgxd29h [54].

Author contributions

P.K. designed and carried out the experiments, analysed the data and wrote the paper; M.B. carried out the bioinformatic study, analysed the sequence data, supervised the cloning of the ABCB genes and co-wrote the paper; A.D. assembled the transcriptome; M.K. cloned the ABCB genes; S.D. conceived and supervised the study and wrote the paper.

Competing interests

We declare we have no competing interests.

Funding

This work was funded by grants from the Deutsche Forschungsgemeinschaft (grant no. Do527/5-3) and the John Templeton Foundation (ID no. 41855) to S.D.