Major processes involved in the transition from the undefended to defended phenotype: 1) toxin intake, here visualized with two discrete points representing low and high rates; 2) toxin elimination rate (Elim. Rate), e.g., via toxin metabolism; 3) toxin sequestration rate (Seq. Rate), i.e., the active transport of toxins for storage in a specific location such as the skin; and 4) toxin accumulation rate (Acc. Rate), or the rate at which toxins are accumulated in the animal. Defense phenotypes are ultimately a result of how these processes interact over time, here arbitrarily from 0 h, immediately after toxin ingestion, to 24 h following ingestion. Although toxin intake influences the total possible amount of toxin accumulation, it cannot fully explain the defensive phenotype. We hypothesize that the “no accumulation” phenotype is characterized by the absence of any ability to sequester toxins in combination with a high rate of elimination, resulting in 0 toxin accumulation (solid grey lines); this phenotype is a likely ancestral state for many animals. In contrast, we hypothesize that “passive accumulation” is characterized by lower elimination rates than the no accumulation phenotype, leading to a low amount of toxin accumulation (solid yellow lines); however, some mechanisms of toxin transport could also exist, in which case a low sequestration rate could result in a passive accumulation phenotype when elimination rate is high (dashed yellow lines). We hypothesize that the “active sequestration” phenotype evolves from an intermediate passive accumulation phenotype through the addition of novel sequestration mechanisms that result in high sequestration rates (solid purple lines). However, elimination rates could still modulate the amount of toxins ultimately accumulated, with lower elimination rates resulting in a higher proportion of toxin accumulation overall (dashed purple lines).

Evolutionary models of toxin sequestration in Dendrobatidae have changed over time.

A) When several species of aposematic dendrobatids (purple lines) were found to have narrower dietary niches than undefended dendrobatids and other frogs [10,15,16], researchers hypothesized that diet specialization may have driven the radiation of aposematic dendrobatids [17]. B) Chemists hypothesized that aposematic dendrobatids sequester dietary alkaloids via an alkaloid uptake system [11]. Daly [18] postulated that an alkaloid uptake system was present in the ancestor of Dendrobatidae (here denoted as passive accumulation) and that it is “overexpressed” in defended dendrobatids (here denoted as active sequestration). C) A phylogenetic analysis of Dendrobatidae revealed that chemical defense and diet specialization evolved independently several times [9]. The new information helped generate the diet-toxicity hypothesis, which posits that shifts from a generalist to a specialist diet drove the multiple origins of alkaloid uptake through enhanced resistance and/or more efficient sequestration systems [4,8]. D) Here we propose a combination of these hypotheses, i.e., that passive accumulation, alkaloid consumption, and some level of alkaloid resistance was present in an early dendrobatid lineage; enhanced resistance and active sequestration mechanisms then arose later, resulting in the chemical defense phenotype. This hypothesis places less emphasis on dietary changes and more strongly emphasizes novel molecular mechanisms (e.g., binding proteins and target-site insensitivity [19–21]). Phylogenies in each subpanel highlight how increasing resolution impacted our understanding of phenotypic diversification in Dendrobatidae. All images of frogs were taken by RDT.

Range and median of alkaloid quantity (estimated by the sum of integrated areas) and alkaloid diversity (number of different compounds) by species from the GCMS assessment. The presumed chemical defense phenotype for poison frogs is given according to Santos and Cannatella [4]. Purple rows highlight defe species. *Data are from UHPLC-HESI-MS/MS, which does not provide quantitative data.

From left to right: an ultrametric tree showing phylogenetic relationships inferred previously [55] among sampled species with the three defended poison frog clades highlighted in purple, the undefended clades in dark gray, and non-dendrobatids in light gray (Bufonidae: Amazophrynella siona and Atelopus aff. spurrelli; Leptodactylidae: Lithodytes lineatus). Tile color indicates the log of the total quantity of alkaloids in each class as measured by the sum of integrated areas of alkaloids of that class from GCMS data per individual. The number in each tile indicates the number of alkaloids (including isomers) detected in each individual for each class. On the right are prey items recovered from the stomach of each individual, colored by arthropod group and scaled to 1 (total number of prey identified are shown under N). Note the large proportion of ants (Formicidae, dark purple) and mites (Acari, light purple) in many of the individuals compared to other prey types. See Table S3 for alkaloid-level data and Table S4 for raw diet data. Poison frog genera names are abbreviated as follows: All., Allobates; Ame., Ameerega; And., Andinobates; D., Dendrobates; E., Epipedobates; H., Hyloxalus; Le., Leucostethus; O., Oophaga; P., Phyllobates; R., Rheobates; S., Silverstoneia; Alkaloid class abbreviations are based on [50,56] and are as follows: HTX, histrionicotoxins; PTX, pumiliotoxins; PTXB, Pumiliotoxin B; aPTX, allopumiliotoxins; DeoxyPTX, deoxypumiliotoxins; hPTX, homopumiliotoxins; deoxy-hPTX, deoxy-homopumiliotoxins; DHQ, decahydroquinolines; NMeDHQ, N-Methyldecahydroquinolines; HO-DHQ, hydroxy-decahydroquinolines; 3,5-P, 3,5-disubstituted pyrrolizidines; HO-3,5-P, hydroxy-3,5-disubstituted pyrrolizidines; 5-I, 5-substituted indolizidines; 3,5-I, 3,5-disubstituted indolizidines; 5,6-I, 5,6-disubstituted indolizidines; 5,8-I, 5,8-disubstituted indolizidines; Dehydro-5,8-I, Dehydro-5,8-Indolizidines; 5,6,8-I, 5,6,8-trisubstituted indolizidines; HO-5,6,8-I, Hydroxy-5,6,8-trisubstituted indolizidines; 1,4-Q, 1,4-disubstituted quinolizidines; 4,6-Q, 4,6-disubstituted quinolizidines; 3,5-Q, 3,5-disubstituted quinolizidines; 1,3,4-Q, 1,3,4-trisubstituted quinolizidines; Lehm, lehmizidines; Epiquinamide, epiquinamide; 2-Pyr, 2-substituted pyrrolidine; 3-Pyr, 3-substituted pyrrolidine; 2,5-Pyr, 2,5-disubstituted pyrrolidines; Pyr, pyrrolizidine of indeterminate substitution; 2,6-Pip, 2,6-disubstituted piperidines; Pip, other piperidines; Pyri, pyridines (including epibatidine); GTX, gephyrotoxins; Tricyclic, coccinelline-like tricyclics; SpiroP, spiropyrrolizidines; Necine, unspecified necine base; Unclass, unclassified alkaloids without known structures.