Research ArticleMolecular Evolution

Interacting amino acid replacements allow poison frogs to evolve epibatidine resistance

See allHide authors and affiliations

Science  22 Sep 2017:
Vol. 357, Issue 6357, pp. 1261-1266
DOI: 10.1126/science.aan5061

Poison frogs resist their own chemical defense

Poison frogs produce a neurotoxin that protects them from predation. The frogs, however, run the risk of intoxicating themselves. Studying the frog neurotoxin epibatidine, which binds to acetylcholine receptors, Tarvin et al. found a single amino acid substitution. The substitution changes the configuration of the acetylcholine receptor, so that it decreases its sensitivity to the toxin. But acetylcholine signaling is essential for normal life. Expressing frog receptors in human cells revealed that different amino acid substitutions have occurred in different lineages that allow the frog to resist its own toxins while still letting target neurotransmitters function effectively.

Science, this issue p. 1261


Animals that wield toxins face self-intoxication. Poison frogs have a diverse arsenal of defensive alkaloids that target the nervous system. Among them is epibatidine, a nicotinic acetylcholine receptor (nAChR) agonist that is lethal at microgram doses. Epibatidine shares a highly conserved binding site with acetylcholine, making it difficult to evolve resistance yet maintain nAChR function. Electrophysiological assays of human and frog nAChR revealed that one amino acid replacement, which evolved three times in poison frogs, decreased epibatidine sensitivity but at a cost of acetylcholine sensitivity. However, receptor functionality was rescued by additional amino acid replacements that differed among poison frog lineages. Our results demonstrate how resistance to agonist toxins can evolve and that such genetic changes propel organisms toward an adaptive peak of chemical defense.

Acquiring chemicals from the environment and recycling them for antipredator defense is a survival strategy that has evolved in nearly every major branch of life (1). Exposure to toxic chemicals may have high physiological costs, but it can also be an opportunity for organisms to capitalize on these substances as new resources. Organisms that accumulate these chemicals risk self-intoxication unless they can resist their own defenses through compartmentalization, metabolic detoxification, or target-site insensitivity—i.e., changes in the molecular target of the toxin that affect its ability to bind (2). Many toxins target evolutionarily conserved proteins such as ion channels, which govern key nervous system functions. Thus, revealing the mechanistic basis of toxin resistance deepens our understanding of protein function and provides insights into nervous system evolution (3, 4). Moreover, the physiology of toxin resistance is a crucial aspect of chemical defense, and characterizing the evolution of resistance might elucidate how and why organisms acquire toxic defenses (5).

Neotropical poison frogs (Dendrobatidae) have independently evolved chemical defenses at least four times (6). The origins of chemical defense are usually accompanied by shifts toward bright coloration, resulting in a complex phenotype or syndrome known as aposematism (6). Theoretically, aposematic and nonaposematic poison frogs represent alternative peaks on an adaptive landscape that arose as a result of disruptive selection that favored more extreme phenotypes over intermediate ones (e.g., conspicuous but not well defended, or defended but not aposematic) (7). The multiple origins of aposematism within dendrobatids suggest that the switch from nonaposematic to aposematic phenotypes is easily attained within this group. Characterizing the evolution of toxin resistance, a key step in this phenotypic transition, may reveal pathways between these adaptive peaks in which toxin resistance facilitates origins of toxin sequestration.

Chemically defended dendrobatids take up from their diet over 800 types of lipophilic alkaloids (8), many of which modulate nervous system function (9). Their effects vary from benign to lethal (10), but most are bitter-tasting and thus generally aversive to predators (11). Epibatidine, one of the best known of these alkaloids, was first isolated from the phantasmal poison frog Epipedobates anthonyi in 1974 (12). Epibatidine has an analgesic effect 200 times that of morphine, yet it targets a specific subset of nicotinic acetylcholine receptors (nAChRs) rather than opioid receptors (13). Because of these qualities, epibatidine has inspired pharmacological innovations, although its toxicity has prohibited its successful development as a pharmaceutical (14).

Toxic animals, including poison frogs, often evolve resistance to their toxins via amino acid replacements in toxin-binding sites (target-site insensitivity) (15, 16). The location of these replacements is constrained by protein function, leading to predictable and convergent mechanisms of resistance (17). For example, resistance to tetrodotoxin (TTX), a NaV1 voltage-gated sodium channel blocker, evolved many times in toxic pufferfish, newts, and snakes that feed on newts via various amino acid replacements at residues in NaV1 proteins that interact with TTX (1719). Similarly, resistance to cardiac glycosides, which inhibit the sodium-potassium pump, has evolved at least 14 times in toxic insects and amphibians, as well as their predators, via amino acid replacements in the cardiac-glycoside binding site (4, 20).

Evolving epibatidine resistance involves different strategies at the molecular level, as epibatidine is an agonist that shares a binding site with ACh, the endogenous ligand of nAChRs, whereas TTX and cardiac glycosides act on receptors that are not ligand-gated (21, 22). Resistance to epibatidine thus requires decreased sensitivity to epibatidine while preserving sensitivity to the endogenous agonist ACh that interacts with many of the same amino acids all without disrupting the normal receptor function.

Phylogenetic identification of amino acid replacements in the poison frog nAChR

Based on what is known about the toxin and ligand, we hypothesized that epibatidine-bearing frogs would have nAChRs that resist epibatidine yet display normal ACh sensitivity and that the basis of resistance would involve genetic changes in the ligand-binding site. To test this hypothesis, we sequenced genes in poison frogs encoding the primary molecular target of epibatidine in the brain, the α4β2 nAChR (chrna4 and chrnb2) (23). Epibatidine has been detected in two distinct lineages of dendrobatids, Epipedobates and Ameerega (12), so we predicted two origins of resistance. Consequently, we sequenced these genes from 9 species of these genera, as well as 19 other species of poison frogs, including 8 species of Dendrobatinae (Dendrobates and Phyllobates), a clade of chemically defended poison frogs lacking epibatidine, and 11 nondefended species (table S1) (24).

Four sites in the β2 subunit (F106, S108, A110, and I118, numeration of the mature human protein) have unique amino acid replacements in the alkaloid-sequestering dendrobatids Epipedobates, Ameerega, and Dendrobates [subgenus Oophaga sensu (25)] (Fig. 1A), the last of which is not known to have epibatidine defenses. These replacements are near the epibatidine-binding site in the α+–β interface: between loops A and E and in loop E (Fig. 1, B to E) (21, 22). Each of these replacements involves a single nucleotide change in the first or second codon position (table S2) (24), suggesting non-neutral evolution. Five additional sites in α4 were found to have amino acid replacements unique to these poison frogs, but only one of these (D176N) was near the epibatidine-binding site (table S3 and fig. S1) (24).

Fig. 1 Amino acid replacements in β2 associated with alkaloid-defended poison frogs.

(A) Alignment of dendrobatid (black), nondendrobatid (gray), and outgroup (gray) β2 sequences (table S4). Yellow branches in the phylogeny [adapted from (25)] indicate alkaloid-defended lineages; asterisks indicate clades in which epibatidine has been detected; the unit of the scale bar is the number of expected substitutions per site. Focal species names are in bold and colored by their amino acid replacement pattern. The amino acid replaced only in Epipedobates poison frogs is in red (F106L); amino acids replaced only in Ameerega are in purple (A110I and I118V); the convergently evolved replacement is in cyan (S108C). Genotypes of clades with replacements are indicated to the left of the alignment (see Table 1). (B and C) Structure of the human (α4)2(β2)3 nAChR (22) (B) from the side and (C) from extracellular space. α4 subunits are in light gray, β2 subunits are in gold, and the ligand-binding sites are indicated by gray spheres. (D and E) Closer view of the binding site from (D) extracellular space and (E) viewpoint indicated by labeled arrow in (C). Amino acid residues identified with gray and gold arrows are known to be involved in ACh and/or epibatidine binding (21, 28, 29). Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

Electrophysiology of amino acid replacements in the poison frog nAChR

The α4β2 nAChR is a pentameric protein that exhibits two different stoichiometries: a high-ACh sensitivity conformation (HS), (α4)2(β2)3, and a low-ACh sensitivity conformation (LS), (α4)3(β2)2 (26, 27). To determine experimentally whether the identified amino acid replacements provide resistance to epibatidine, we used site-directed mutagenesis to introduce poison frog amino acid replacements into human nAChRs. We then coexpressed the wild-type and mutated β2 subunits with human α4 nAChR subunits in Xenopus laevis oocytes and measured acetylcholine and epibatidine concentration-response curves (CRCs) through two-electrode voltage clamp (parameters, results, and statistical analyses from all CRCs are shown in tables S5 to S16) (24). For each subunit combination, we injected different ratios of α4 and β2 transcripts to favor the formation of either HS or LS conformations (24). For brevity, we describe only HS nAChRs in the main text, because we found the same general pattern of channel sensitivity to ACh and epibatidine in both stoichiometries. For LS nAChR results, see fig. S2 and tables S5 to S7 (24). We also performed electrophysiology experiments to investigate whether the one replacement in the α4 subunit near the ligand-binding site (D176N) affected LS nAChR function, but we found no evidence for an effect (figs. S1 and S4 and table S8) (24).

For clarity, we denote all nAChR genotypes with four letters indicating the amino acid residue at each of the four sites of interest (106, 108, 110, and 118) (see Table 1). Bold letters in each genotype indicate amino acid replacements introduced into a transcript via site-directed mutagenesis.

Human-to-frog mutants

The Epipedobates and Ameerega replacement patterns (LCAI and FCVV genotypes) produced by mutagenesis showed ACh CRCs identical to that of the wild-type human FSAI genotype, whereas the subgenus Oophaga replacement pattern (FCAI) showed a decrease in sensitivity to ACh (Fig. 2A). All three nAChRs with poison frog amino acid replacement patterns (LCAI, FCVV, and FCAI) were less sensitive to epibatidine than the wild-type receptor (Fig. 2B), indicating that these replacement patterns are sufficient to produce epibatidine-resistant phenotypes (Table 1 and tables S5 to S7) (24). Interestingly, the ACh CRC is biphasic for the Oophaga replacement pattern, suggesting that in the human genetic background the S108C replacement may induce assembly of LS nAChRs. Because the LS stoichiometry possesses two kinds of binding sites, application of increasing concentration of ACh results in a biphasic curve that reflects activation of the two HS binding sites at low ACh concentrations and of the single LS binding site at high ACh concentrations (24). Thus, resistance to epibatidine conferred by the S108C replacement incurs a cost of ACh sensitivity in the human β2 subunit.

Fig. 2 ACh and epibatidine concentration-response curves in high-sensitivity α4β2 nAChRs.

Left panels show responses to ACh, and right panels show responses to epibatidine. (A and B) Human α4β2 nAChRs: wild-type genotype (FSAI) and receptors containing the amino acid patterns identified in Epipedobates, Ameerega, and Dendrobates (Oophaga) poison frogs (LCAI, FCVV, and FCAI genotypes, respectively). (C and D) Human α4β2 nAChRs: wild-type (FSAI) and Ameerega genotypes (FCVV, FCAI, FSVI, and FSAV). (E and F) Human α4β2 nAChRs: wild-type (FSAI) and Epipedobates genotypes (LCAI, LSAI, and FCAI). (G and H) Epipedobates α4β2 nAChRs: wild-type (LCAI) and human genotypes (FSAI, FCAI, and LSAI). Dotted lines correspond to human FSAI and LCAI curves from (C) and (D). Error bars smaller than the symbols are not visible. Data were fitted to either monophasic (solid line) or biphasic (dashed line) curves. (Inset) Schematic of HS α4β2 nAChR stoichiometry; ligand-binding sites are indicated by arrows.

We then characterized the physiological effect of each individual replacement in poison frogs by generating human α4β2 nAChR transcripts with single amino acid replacements (LSAI, FSVI, and FSAV). As with the S108C replacement, human transcripts with the I118V replacement (FSAV, derived in Ameerega) provided moderate resistance to epibatidine at a cost of ACh sensitivity, possibly because this amino acid replacement also induced assembly of LS nAChRs (Fig. 2, C and D, and Table 1). In contrast, human receptors with either F106L (Epipedobates) or A110V (Ameerega) displayed no change in ACh and epibatidine sensitivity, indicating that these replacements probably do not contribute to epibatidine resistance (Fig. 2, E to H, and Table 1). Instead, these replacements appear to compensate for the decrease in ACh sensitivity incurred by the replacements that provided resistance (Table 1), because human receptors with the LCAI genotype (Epipedobates) or the FCVV genotype (Ameerega) both showed normal ACh response (Fig. 2, A and B, and Table 1) (24).

Epipedobates-to-human mutants

We synthesized and expressed the wild-type Epipedobates α4β2 nAChR (LCAI genotype) and a double mutant replicating the plesiomorphic human genotype (FSAI) in Xenopus laevis oocytes and performed electrophysiology assays. The Epipedobates-to-human mutant (FSAI) showed greatly increased sensitivity to epibatidine but no change in sensitivity to ACh (Table 1, table S7, and Fig. 2, G and H), indicating that the replacements in Epipedobates were necessary for resistance.

To understand the contributions of each replacement when it occurs in the poison frog genetic background, we expressed the single-mutant genotypes FCAI and LSAI in the Epipedobates β2 subunit. Whereas S108C incurred a drastic cost in ACh sensitivity in the human genetic background (Fig. 2A; compare FSAI and FCAI), the Epipedobates-to-human FCAI mutant demonstrated only a minor (but significant) decrease in sensitivity to ACh (compared to FSAI), suggesting that some other aspects of the poison frog genetic background ameliorate the large cost of this replacement in the human FCAI genotype (Table 1). This difference may be explained by the observation that the S108C replacement in human receptors appeared to induce formation of LS nAChRs (Fig. 2A), which are less sensitive to ACh than HS nAChRs. However, the Epipedobates nAChR never appeared to form the LS stoichiometry, even when the injected complementary RNA subunit ratio favored its formation (compare Fig. 2, G and H, to fig. S2, G and H) (24). Little is known about the poison frog α4β2 nAChR, but the apparent absence of the LS stoichiometry in Epipedobates (evidenced by the lack of a biphasic, right-shifted curve) lessens the cost of the S108C replacement, and might be related to epibatidine exposure and resistance.

As predicted, the Epipedobates FCAI receptor displayed a decrease in sensitivity to epibatidine compared with FSAI (Table 1 and table S7) (24), confirming the role of S108C in epibatidine resistance (Fig. 2, G and H). As with the human-to-frog LSAI receptor, the Epipedobates-to-human LSAI receptor affected neither ACh nor epibatidine sensitivities compared with FSAI (Table 1). The LCAI genotype (wild-type in Epipedobates) displayed normal responses to ACh and decreased sensitivity to epibatidine (Table 1). Thus, as in the human receptor, C108 provides epibatidine resistance and L106 appears to compensate by normalizing α4β2 receptor function in Epipedobates poison frogs.

Amino acid replacements in poison frog nAChR are proximal to the epibatidine binding site

We found that amino acid replacements in the poison frog β2 subunit (Fig. 1) alter α4β2 nAChR sensitivity to epibatidine (Fig. 2B). We propose that this is in part due to the proximity of the amino acid replacements to the epibatidine binding site. Namely, the β2C108 residue directly contacts the side chain of α4W156, one of the main determinants in stabilizing epibatidine binding (Fig. 1, D and E) (21, 28). The sulfur-containing side chain of C108 is bulkier than that of serine, and it could modify the epibatidine-W156 interaction. The I118V replacement in Ameerega, which also contributes to epibatidine resistance (Fig. 1, D and E), is next to F119, a residue that interacts with the epibatidine chloropyridine ring and stabilizes the epibatidine chlorine atom through its backbone carboxyl group. Moreover, the A110V replacement is next to V111, another amino acid residue that interacts with epibatidine via van der Waals forces (21, 28, 29). These replacements are located in β sheets that are involved in epibatidine binding but are less involved in ACh binding (21, 28, 30). The β2 side of the binding pocket is further from ACh than is the α4+ side and thus forms looser interactions with ACh, such that amino acid replacements in the β region that allow changes in epibatidine binding may be less likely to affect ACh sensitivity. This structure-function problem was apparently solved via an identical genetic change three times within poison frogs and refined via different genetic changes at least twice in these lineages.

Evolutionary pathways toward epibatidine resistance

Toxin resistance often evolves in response to recurrent exposure to toxins (2, 4, 31, 32); thus, patterns of resistance should reflect the evolutionary history of toxin exposure. The evolutionary patterns of amino acid replacements in the poison frog β2 nAChR subunit suggest that in each of the Epipedobates, Ameerega, and Dendrobates (Oophaga) clades (Fig. 1A), an ancestral species was likely exposed to epibatidine, resulting in selection for and evolution of epibatidine resistance about 5, 10, and 8 million years ago, respectively (25). Although no clade of poison frogs that has epibatidine defense lacks amino acid replacements in the β2 nAChR, epibatidine has only been detected in 2 of 3 sampled species of Epipedobates, in 2 of 12 sampled species of Ameerega, and in none of 9 sampled species of Dendrobates (Oophaga) (9, 12). It is possible that some populations with epibatidine defense are extinct or have not been detected or that the dietary source of epibatidine, presumed to be an arthropod, is not as available as it was long ago (12). Although epibatidine resistance may have arisen as a side effect of some other change to the protein, mutations in the ligand-binding domain are uncommon (Fig. 1A) and presumably evolve under strong selective pressures. Regardless of the apparent rarity of epibatidine in poison frogs, the epibatidine-resistant phenotype (determined by electrophysiology) does not appear to have been lost in any resistant lineages (Oophaga, Epipedobates, or Ameerega), suggesting that lack of resistance has a high cost, that reversion to a nonresistant phenotype is physiologically difficult, or that maintenance of epibatidine resistance is not costly.

The evolutionary patterns underlying origins of epibatidine resistance in poison frogs reflect an adaptive landscape with two peaks that maximize fitness of alternative phenotypes: toxin-resistant and defended or toxin-sensitive and undefended. Given that S108C provides epibatidine resistance and that it is found in all three resistant clades, we argue that it provides a substantial selective advantage. We suggest two possible evolutionary pathways for acquisition of toxin resistance. In the first, initial replacements may provide a small selective benefit of resistance yet carry some physiological cost in receptor function. For example, the S108C replacement arose independently in all three lineages and is sufficient to produce an epibatidine-resistance phenotype. However, it also incurs decreased sensitivity to ACh in both the human and the Epipedobates backgrounds (Table 1), and the fitness cost of this replacement in living organisms is not clear. We speculate that yet unidentified mutants in the poison frog nAChR sustained receptor functionality—i.e., by inducing nAChR expression changes—until other replacements such as F106L evolved to rescue receptor sensitivity to ACh. Disruptive selection on populations with both genotypes may have propelled the populations with S108C toward a new adaptive peak.

In the second possible trajectory, certain mutants already present in the gene pool provide a genetic background in which resistance arises without cost. For example, the artificial genotype LSAI (F106L) shows no reduction in either ACh or epibatidine sensitivity (Table 1). Thus, a frog species with F106L has evolved a novel genotype (LSAI), intermediate between FSAI (plesiomorphic) and LCAI in Epipedobates, without incurring a cost, which subsequently allows the C108 replacement to also evolve without cost. However, the LSAI genotype does not exist in any taxa we sampled. It is not present in Silverstoneia, the sister group of Epipedobates (two of eight species sampled), nor in the closely related taxa Ameerega and Colostethus (Fig. 1A). Thus, this second pathway, in which a novel genotype evolves without apparent cost, is not found in poison frogs. However, this pathway is known in the brown plant-hopper (Nilaparvata lugens) (33), in which two amino acid replacements confer resistance to fipronil, a noncompetitive antagonist of γ-aminobutyric acid receptors. This occurs in an apparently sequential process in which the second amino acid change provides high resistance yet has a high fitness cost and never occurs without the first (33). It is unclear how common such preexisting compensatory mutations are, although it appears that mutations providing incremental increases in resistance are quite common. In Danainae butterflies, newts, garter snakes, and poison frogs, toxin resistance tends to increase over evolutionary time via additional amino acid replacements that occur in parallel with increased concentrations of chemical defenses (15, 16, 34, 35). It is possible that preadaptive mutations that allow resistance to evolve with little cost are present in these organisms and have not been identified. The presence of such preadaptive mechanisms would imply a shallow, “neutral” valley on the adaptive landscape that facilitates the movement from one adaptive peak to another.

The Epipedobates, Ameerega, and Dendrobates (Oophaga) clades, which are evolutionarily young (6, 36), are an example of rapid and ongoing diversification possibly driven by the evolution of resistance to antipredator toxins (15). We demonstrate that resistance to epibatidine involves finely tuning a highly conserved binding site without disrupting receptor function, providing insights into evolutionary pathways culminating in chemical defenses. Thus, evolution, with millions of years and subjects, can solve complex problems in systems biology that may otherwise seem impossible.

Table 1 Effects of amino acid replacements on ligand responses in human and Epipedobates high-ACh sensitivity (HS) nAChRs.

Relative fold change in sensitivity induced by amino acid replacements was calculated as (mutant EC50)/(FSAI EC50) for each genetic background [i.e., (mutant)/(reference); see tables S5 to S7] (24). Values greater than 1 indicate that relatively more ligand is required to elicit the same response; thus, higher values indicate lower sensitivity. ACh assessments for biphasic curves (>1) are qualitative, but both cases result in lowered sensitivity (see tables S5 and S6) (24). *P < 0.01; **P < 0.001 (two-way analysis of variance, corrected for multiple comparisons using Tukey’s test; see tables S9, S11, S13, and S14) (24).

View this table:

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Tables S1 to S16

Data S1 and S2

References (3751)

References and Notes

  1. See the supplementary materials.
Acknowledgments: Data are available in the supplementary materials; sequences are archived in GenBank under accession numbers MF598757 to MF598762, MF580080 to MF580125, and MF619959 and MF619960. R.D.T. is indebted to the many herpetological societies that funded this project: American Society of Ichthyologists and Herpetologists, Herpetologists’ League, Texas Herpetological Society, Minnesota Herpetological Society, Society for the Study of Amphibians and Reptiles, Chicago Herpetological Society, and North Carolina Herpetological Society, in addition to the National Geographic Society Young Explorer’s Grant 9468-14, the Society for the Study of Evolution, the Society of Systematic Biologists, and NSF DEB-1404409. R.D.T. was supported by a Graduate School Continuing Fellowship at University of Texas at Austin and an NSF Graduate Research Fellowship. J.C.S. was supported by start-up funds at St. John’s University and by the National Science and Engineering Research Council Collaborative Research and Training Experience Training Program in Biodiversity Research at University of British Columbia. J.C.S. also thanks J. W. Sites Jr., for his support as a postdoctoral advisor at Brigham Young University with grants NSF EF-1241885 and EF-1241848. W.S. thanks D. R. Dietrich of University of Konstanz for his support as a thesis advisor. L.A.O. was supported by a Bauer Fellowship at Harvard University, a William F. Milton grant from Harvard Medical School, and a L’Oreal For Women in Science Fellowship. D.C.C. was supported by NSF DEB-1556967. H.H.Z. was supported by NSF IOS-1557857 and PLR-1443637. C.M.B. and R.A.H. were supported by the University of Texas at Austin Waggoner Center for Alcohol and Addiction Research and NIH R01-AA006399. We thank the two anonymous reviewers and P. Andolfatto for comments that greatly improved the manuscript. We also thank L. Coloma, A. Amézquita, M. Betancourth, S. Ron, and T. LaDuc for facilitating specimen collection and access to museum tissues. All specimens were collected under approved protocols (Harvard University IACUC 12-10 and UT Austin IACUC 2012-00032 and 2015-00205) and valid permits [001-13 IC-FAU-DNB/MA and 001-11 IC-FAU-DNB/MA (Ecuador), IBD0359 Res 1177-2014 (Colombia), and SPR-1097-912 (USA)].

Stay Connected to Science

Navigate This Article