Polyphyletic origin of toxic Pitohui birds suggests widespread occurrence of toxicity in corvoid birds

Knud A Jønsson, Rauri C.K Bowie, Janette A Norman, Les Christidis, Jon Fjeldså


Pitohui birds from New Guinea have been found to contain a toxin otherwise only found in neotropical poison arrow frogs. Pitohuis have been considered to be monophyletic and thus toxicity is thought to have evolved once in birds. Here, we show that Pitohuis, rather than being a tight-knit group, are polyphyletic and represent several lineages among the corvoid families of passerine birds. This finding demonstrates that the ability to be toxic is widespread among corvoid birds and suggests that additional members of this radiation, comprising more than 700 species, could prove to be toxic. It is postulated that toxic birds ingest the toxin through their insect diet and excrete it through the uropygial gland, from where it is applied to the skin and feathers. Thus, the ability to become toxic is most likely an ancestral condition but variation in diet determines the extent to which toxicity is expressed among corvoid birds. Variability in toxicity levels further suggests that the main function of the toxin is that of a deterrent against ectoparasites and bacterial infection rather than being a defence against predators as initially proposed.

1. Introduction

Fifteen years ago, the discovery that New Guinean Pitohui birds contained a toxin in their feathers and muscle tissue received much attention (Dumbacher et al. 1992). This toxin was identified as a homobatrachotoxin, a kind of neurotoxic steroidal alkaloid previously known only from the skins of five neotropical frog species in the genus Phyllobates (Dendrobatidae; Dumbacher et al. 2000). This finding was particularly remarkable because it was the first time that a potentially defensive toxin was identified in a bird. All six species in the genus Pitohui were scrutinized, revealing considerable individual variation in toxicity (Dumbacher et al. 2000). Two taxa, Pitohui dichrous and P. kirhocephalus, were particularly toxic; P. cristatus, P. nigrescens and P. ferrugineus were mildly toxic; and no toxins were detected in P. incertus. Traces of the toxin were found in a putative close relative of the Pitohuis, Colluricincla megarhyncha, but none was found in other putative relatives in the New Guinea region (Pachycephala schlegelii, Rhagologus leucostigma and Eulacestoma nigropectus; Dumbacher et al. 2000).

More recently, it was discovered that the birds may obtain the batrachotoxins from their diet, specifically from the poisonous melyrid beetle Choresine (Dumbacher et al. 2004). It was also found that the bird Ifrita kowaldi contained a similar spectrum of batrachotoxins as Pitohui. At the time, however, no reliable phylogenetic framework was available to determine the relationship between Ifrita and Pitohui, and thus whether the ability to tolerate this toxin in the body is restricted to a specific avian lineage (Dumbacher et al. 2004).

By placing Pitohui birds in a broader phylogenetic context, we demonstrate that toxic birds span the phylogeny of the corvoid assemblage of bird families in the Papuan region, and we propose a widespread mechanism to use toxins from the diet as defence, mainly against ectoparasites or dermal infections.

2. Material and methods

(a) Taxon sampling, amplification and sequencing

We obtained sequence data of 26 taxa sampled across the Crown Corvida radiation (table 1) representing all major families and including some aberrant Australo-Papuan taxa that have been suggested to be closely related to Pitohui. Three nuclear gene regions, myoglobin intron 2 (Myo2), ornithine decarboxylase (ODC) introns 6–7 and glyceraldehyde-3-phosphodehydrogenase (G3PDH) intron 11, were sequenced. These genes have previously been shown to be useful for resolving phylogenetic relationships in birds (e.g. Irestedt et al. 2006; Jønsson et al. 2007). For R. leucostigma, I. kowaldi and E. nigropectus, we only obtained sequence data for Myo2.

View this table:
Table 1

Taxa used in the study. (Acronyms are AM, Australian Museum, Sydney, Australia; ANWC, Australian National Wildlife Collection, Canberra, Australia; FMNH, Field Museum of Natural History, Chicago, USA; MCSNC, Museo Civico di Storia Naturale di Carmagnola, Italy; MV, Museum Victoria, Melbourne, Australia; NRM, Swedish Museum of Natural History, Stockholm, Sweden; ZMUC Zoological Museum of Copenhagen, Denmark. All samples are vouchered.)

The combined alignment consists of 1661 bp. For more details of indel length and positions, see the alignments of the individual gene regions deposited in GenBank. For primers, alignment, extractions, amplifications and sequencing procedures, see Irestedt et al. (2006) and Jønsson et al. (2007).

(b) Phylogenetic inference

Owing to the rather low number of insertions in the introns, the combined sequences could easily be aligned. All gaps have been treated as missing data in the analyses. Bayesian inference (BI) and maximum likelihood (ML) were used to determine phylogenetic relationships. Models of nucleotide substitution used in the analyses were selected for each gene individually by applying the Akaike Information Criterion implemented in MrModeltest v. 2.2 (Nylander 2005) in conjunction with PAUP*b10 (Swofford 2001).

Posterior probabilities of trees and parameters in the substitution models were approximated with MCMC and Metropolis coupling using the program MrBayes v. 3.1.1 (Huelsenbeck et al. 2001; Ronquist & Huelsenbeck 2003). Analyses were performed for both the individual gene partitions and the combined dataset, where each gene region was unlinked allowing for independent estimation of parameters. The chains for the individual gene partitions and for the combined dataset were all run for 10 million generations, with trees sampled every 100 generations. The trees sampled during the burn-in phase were discarded after checking for convergence and the final inference was made from the concatenated outputs.

The priori selection of nucleotide substitution models suggested that the GTR+Γ model had the best fit for all three gene regions, but as the nucleotide state frequencies and gamma distribution differed between the partitions, we still applied a partitioned analysis of the combined dataset. After discarding the burn-in phase, the inference for the individual genes and the combined dataset were based on a total of 95 000 samples each. The posterior distribution of topologies is presented as a 50% majority-rule consensus tree from the combined analysis in figure 1.

Figure 1

The 50% majority-rule consensus tree of 23 Crown Corvida species obtained from Bayesian analysis of the combined dataset (G3PDH, Myo2 and ODC). Posterior probability values greater than 0.95 are indicated above nodes (asterisk marks 1.00 posterior probabilities) and ML bootstrap support values greater than 70 are indicated below nodes. Family names are indicated to the right; ‘P’ marks species that have erroneously been referred to as Pachycephalidae in traditional classifications. ‘Toxic’ denotes species that are known to contain batrachotoxins or traces of batrachotoxins and ‘non-toxic’ denotes species known not to contain batrachotoxins or traces of batrachotoxins as demonstrated by Dumbacher et al. (2000).

The trees obtained from the Bayesian analyses of the individual genes (not shown) are topologically congruent overall, and all gene trees support the same relationships for Pitohuis. In fact, there are no topological conflicts that are supported by posterior probabilities above 0.95, and the combined tree is also in good topological agreement with other molecular studies of major relationships among Crown Corvida passerines which includes shrikes, bush-shrikes, butcher-birds, drongos, fantails, monarchs, crows and birds of paradise (Jønsson & Fjeldså 2006a).

Maximum-likelihood analyses were performed using Garli v. 0.95 (Zwickl 2006). Five independent analyses were performed using a GTR+I+Γ model and default settings. Nodal support was evaluated with 500 non-parametric bootstrap pseudoreplications. The score of the best likelihood tree (−ln L 7224.84258) was within 0.025 likelihood units of the best tree recovered in each of the other four runs, suggesting that the five runs had converged. The ML tree topology was almost completely congruent with the BI topology. No differences were found for well-supported nodes.

3. Results and discussion

Analyses of the three markers clearly demonstrate that the genus Pitohui is highly polyphyletic, with individual species widely dispersed among Crown Corvida lineages (figure 1), which has its centre of origin in the Australo-Papuan region (Jønsson & Fjeldså 2006b) and counts more than 700 species worldwide. This suggests that the propensity to assimilate batrachotoxins from dietary sources into feather structures and muscle tissues is widespread among the basal lineages of the Crown Corvida clade of songbirds (figure 1). The two most toxic taxa P. dichrous and P. kirhocephalus are sister species and closely related to Old World Oriolus. P. nigrescens, a mildly toxic species, is sister to all true Pachycephala whistlers, whereas the mildly toxic P. cristatus is closely related to Aleadryas rufinucha, a species distantly related to the Pachycephala radiation. P. ferrugineus, which has often been found to be devoid of batrachotoxins (Dumbacher et al. 2004), is closely related to the Australo-Papuan Colluricincla shrike-thrushes, which were also demonstrated to have traces of batrachotoxins (Dumbacher et al. 2004). I. kowaldi is not closely related to any Pitohui species.

We find little reason to believe that the toxin is produced de novo but rather that it is assimilated in birds, which, among a wide range of food items, feed on toxic melyrid beetles. We suspect that corvoid birds—notably the caterpillar-loving Campephagidae—eat poisonous insects and have developed a specific ability to tolerate various toxins and excrete them through the uropygial gland. This is probable considering the structural similarity between cholestanol (in the uropygial gland secretions) and one of the batrachotoxins (BTX-A-cis-O-cronate; Dan Stærk 2007, personal communication). Furthermore, the uropygial gland secretions of birds are known to contain a high diversity of biocidal/poisonous substances, which are potential agents for defence against ectoparasites, skin fungi, keratin-degrading bacteria, etc. (Jacob 1978; Poulsen 1993).

Given that batrachotoxins have been recorded in five different lineages of corvoid birds, it seems probable that the ability to ‘handle’ the toxin is widespread among members of the Crown Corvida, but that there is some variation in diet or the ‘risks’ associated with being exposed to melyrid beetles as prey. Where melyrid beetles (or other prey containing batrachotoxins) are abundant, it is probable that opportunistically foraging birds such as orioles and cuckoo-shrikes (Poulsen 1993) will feed on these prey items and become toxic. This could go undetected simply because these birds will feed on whatever items are easily accessible and thus toxicity levels are likely to vary in space and time. We predict that with additional field research many more species of corvoid birds will be found to harbour toxins.


We kindly thank the following institutions for providing access to fresh tissue and/or toe-pads; Australian Museum, Sydney, Australia; Australian National Wildlife Collection, Canberra, Australia; Field Museum of Natural History, Chicago, USA; Museum Victoria, Melbourne, Australia; Swedish museum of Natural History, Stockholm, Sweden; Zoological Museum of Copenhagen, Denmark. K.A.J. would also like to acknowledge the support from the Australian Museum Postgraduate Awards 2006/07.

  • Received September 11, 2007.
  • Accepted November 12, 2007.


View Abstract