To understand the origin of Pantepui montane biotas, we studied the biogeography of toucanets in the genus Aulacorhynchus. These birds are ideal for analyzing historical relationships among Neotropical montane regions, given their geographic distribution from Mexico south to Bolivia, including northern Venezuela (Cordillera de la Costa), and the Pantepui. Analyses were based on molecular phylogenies using mitochondrial and nuclear DNA sequences. Topology tests were applied to compare alternative hypotheses that may explain the current distribution of Aulacorhynchus toucanets, in the context of previous hypotheses of the origin of Pantepui montane biotas. Biogeographic reconstructions in RASP and Lagrange were used to estimate the ancestral area of the genus, and an analysis in BEAST was used to estimate a time framework for its diversification. A sister relationship between the Pantepui and Andes+Cordillera de la Costa was significantly more likely than topologies indicating other hypothesis for the origin of Pantepui populations. The Andes was inferred as the ancestral area for Aulacorhynchus, and the group has diversified since the late Miocene. The biogeographic patterns found herein, in which the Andes are the source for biotas of other regions, are consistent with those found for flowerpiercers and tanagers, and do not support the hypothesis of the geologically old Pantepui as a source of Neotropical montain diversity. Based on the high potential for cryptic speciation and isolation of Pantepui populations, we consider that phylogenetic studies of additional taxa are important from a conservation perspective.
Citation: Bonaccorso E, Guayasamin JM (2013) On the Origin of Pantepui montane biotas: A Perspective Based on the Phylogeny of Aulacorhynchus toucanets. PLoS ONE 8(6): e67321. doi:10.1371/journal.pone.0067321
Editor: Carles Lalueza-Fox, Institut de Biologia Evolutiva - Universitat Pompeu Fabra, Spain
Received: January 30, 2013; Accepted: May 16, 2013; Published: June 26, 2013
Copyright: © 2013 Bonaccorso, Guayasamin. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by NSF (Dissertation Improvement Grant DEB-0508910) and Universidad Tecnológica Indoamérica. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Given their old geological origin, geographic isolation, and endemic biota, the highlands of the Guianan Shield have been a source of inspiration for explorers and naturalists alike. This biogeographic region, known as ‘Pantepui’ , , is formed by mountains derived from the Precambrian sandstone rocks of the Roraima Group in southern Venezuela, western Guyana, and northern Brazil . Since the Jurassic–Cretaceous period, progressive erosion of the Roraima sandstone has shaped the landscape, leaving behind spectacular tabletop mountains (tepuis) surrounded by savannas and tropical forests .
The ancient age of the Pantepui has long been associated with the notion of undisturbed and continuous processes of biological diversification and isolation , . This idea, known as the Plateau Theory, proposes that the ‘Pantepui fauna is the remnant of a fauna formerly widespread on a plateau now dissected by erosion into separate tepuis’ . Initially, this hypothesis was supported by remarkable levels of plant and animal endemism , . However, thorough analysis of plant distributions revealed that endemism was lower than previously estimated . Also, topographic studies showed that most tepuis are effectively connected to each other and to lowland ecosystems in the present  and were likely connected in the past . In fact, palaeoecological studies based on Quaternary sediments indicate vertical displacement of vegetation in the middle and lower elevations of these table mountains . These results do not deny the plausibility of the Plateau Theory as an explanation for the origin of some Pantepui endemics, but show that evolution in isolated tepuis might not be the generality for all such organisms.
Interestingly, many Pantepui animals and plants are neither closely related to those of the surrounding lowlands, nor endemic to particular tepuis, but have taxonomic affinities to populations found in other Neotropical montane regions , , , , . Chapman  concluded that, although a quarter of the distinctive birdlife of mounts Roraima and Duida derived from tropical elements of the surrounding Amazonian habitats, about half of the species might have origins among the Andean tropical and subtropical avifaunas. Historical relationships between these avifaunas were established by the sharing of species within genera, subspecies within species, or even the same subspecies.
Regarding the origin of biogeographic affinities between Pantepui and Andean species, Chapman  argued that this pattern ‘may be explained by the disappearance of their common ancestor, or connecting forms, in the intervening area’, and that ‘the cause of their disappearance is attributed chiefly to the influence of climatic changes.’ Tate  was more explicit about the mechanism of such a scenario. He proposed that during a time of lower temperatures (such as the Last Glacial Maximum), lower mountain ranges may have served as bridges of suitable habitat that allowed biotic exchange from the Andes to the Pantepui and vice versa. Following Chapman’s argument, common ancestors or connecting forms that lived in these bridges would have disappeared with increasing temperatures.
An alternate hypothesis, proposed by Mayr and Phelps , states that the subtropical avifauna of Pantepui is derived from that of the Andes and other subtropical regions by ‘island hopping,’ not only during the Quaternary period but in multiple occasions in the past. Here, the flexibility of the time premise is based on the idea that animals have different dispersal abilities that may be expressed depending on external or internal stimuli. Whereas some species may disperse with changing conditions in the environment (e.g., lowering of life zones during glaciations), others may disperse in response to the evolution of morphological, physiological, and behavioral traits that facilitate dispersal . Haffer  supported the idea of ‘island hopping,’ but restricted the time-window for dispersal events to the Pleistocene.
Croizat  opposed the idea of dispersal between the Andes and the Pantepui, but suggested that, if any chance of dispersal existed, the Pantepui should be the source of migrants, given its ancient age. A similar proposition based on the old age of the tepuis was applied to propose a Pantepuian origin for Andean members of the Diglossa carbonaria and D. lafresnayi species groups . More recently, Givnish et al. ,  identified the Guianan Shield as the center of origin for bromeliads, which dispersed subsequently to the Brazilian Shield, the Andes, the Amazon Basin, and Central America, and re-colonized the Guianan Shield. Bromeliads arose ∼100 million years ago (Mya) , when the Guianan Shield was still experiencing erosion (up to70 Mya; ) and before significant uplift of the Andes occurred in the Miocene–Pliocene , .
In summary, current hypotheses for the origin of Pantepui biotas with close affinities to other montane regions state that: (1) they derived from taxa with broad distributions in the Neotropical montane regions that differentiated (via vicariance) into the Pantepui regional biota, sometime in the past ,  including the time after the Pleistocene ; (2) they have their origin in multiple dispersal events from the Andes and other montane regions at different points in time , including the Pleistocene ; and (3) they have derived from an ‘ancient tepui’ biota that has contributed elements to younger montane regions, including the Andes , , . Neither parsimony analysis of endemicity based on distributions of >400 Neotropical montane species , nor comparisons of area cladograms of 43 co-distributed species complexes , have been able distinguish among these hypotheses. Although a plethora of comparative studies will be needed to answer these questions, only very limited phylogenetic information is currently available , , , , .
To understand biological endemism in the Pantepui highlands, we studied the phylogeny of toucanets in the genus Aulacorhynchus. These toucanets are ideal for understanding historical relationships among Neotropical montane regions because they inhabit subtropical and temperate forests from Mexico south to Bolivia in a fragmented distribution (Figure 1). Former taxonomy  recognized six species: A. prasinus, a species complex , , that is widely distributed across the Mesoamerican highlands and the Andes; A. coeruleicinctis and A. huallagae in the Central Andes; A. haematopygus in the Northern Andes; A. sulcatus in the mountains of northern Venezuela (Cordillera de la Costa), west to the Venezuelan Andes and northeastern Colombia; and, most relevant, A. derbianus in the Andes from Bolivia to northern Ecuador, and discontinuously in the Pantepui. More recently, a phylogenetic study by Bonaccorso et al.  revealed that A. derbianus is paraphyletic: Andean populations of this species (A. d. derbianus) are more closely related to A. sulcatus, than to the Pantepui populations (A. d. duidae, A. d. whitelianus, and A. d. osgoodi). Based on this work, a new taxonomy consistent with the evolutionary history of the group was proposed, where Andean populations of A. derbianus retain the species name and the Pantepui populations were designed as A. whitelianus , .
Subspecies in the Aulacorhynchus prasinus complex are shown as independent evolutionary units following Puebla-Olivares et al.  and Bonaccorso et al. ; taxonomy of South American species, other than A. prasinus, follows Remsen et al. .
Herein, we assess historical hypotheses for the origin of Pantepui Aulacorhynchus using molecular evidence generated by Bonaccorso et al. . Specifically, we aim to (1) test the best estimate of phylogeny against topologies congruent with other hypotheses for the origin of the Pantepui species, (2) reconstruct the biogeographic history of the group and its pattern of colonization across the Neotropics, and (3) place the diversification of Aulacorhynchus in a temporal framework. We present our results in the context of previous hypotheses regarding the origin of Pantepui biotas to understand patterns of distribution across Neotropical montane regions.
Materials and Methods
Taxon and Gene Sampling
Genetic sequence data are available from GenBank (accession numbers AY959855, AY959828 , and JF424372–JF424596 ). Hypothesis testing and biogeographic reconstructions were performed on phylogenetic trees estimated using 34 samples that cover all species and subspecies of South American Aulacorhynchus, and includes Mesoamerican and South American representatives of the A. prasinus species group (Table S1). To polarize character states, we included samples from all other toucan genera (Andigena, Selenidera, Pteroglossus, and Ramphastos) as outgroups.
Molecular character sampling consisted of the mitochondrial genes NADH Dehydrogenase Subunit 2 (ND2) and cytochrome b (cytb), and the nuclear genes β-Fibrinogen intron 7 (βfb7) and Transforming Growth Factor beta 2 intron 5 (TGFβ2.5). Because DNA from Aulacorhynchus huallagae was obtained from dry toe pads that resulted in low yields, only ND2 sequences were available for analyses. Also, nuclear sequences of Andigena hypoglauca were concatenated with mitochondrial sequences of Andigena cucullata published by Weckstein . These combined sequences are technically chimeric, however, their appropriateness to root the trees is justified because sequences from both taxa are, in all probability, more closely related to one another than to those of other species in the combined dataset (i.e., we assumed that Andigena is monophyletic). This later affirmation was supported by a phylogenetic analysis of cytb sequences for species of all Ramphastidae available on GenBank (results not shown). With the aforementioned exceptions, all individuals were represented by the four loci.
A hypothetical temporal frame for the diversification of Aulacorhynchus was generated using a broader sampling of 63 individuals sequenced for cytb (Table S1). Laboratory procedures for obtaining all sequences are detailed elsewhere , .
Best-fit models of evolution were estimated for each gene in jModeltest  under the Akaike Information Criterion. These preliminary analyses indicated the following best-fit models: TIM+I+Г, for ND2; TrN+I+Г, for cytb; HKY+Г, for βfb7; and GTR+I, for TGFβ2.5. Maximum likelihood trees were obtained in GARLI (ver. 2.0; ), to allow phylogeny estimation using data partitions. We implemented a partition by gene, assigning each gene its best-fit model ‘family.’ Individual solutions were selected after 5000 generations with no significant improvement in likelihood (significant topological improvement set at 0.01). Final solutions were selected when the total improvement in likelihood score was <0.05, using default values for the remaining GARLI settings . Ten independent runs were completed to assure consistency of likelihood scores.
Bayesian trees were obtained in MrBayes 3.2 , also partitioning data by gene. Model parameters were unlinked between partitions, except topology and branch lengths. Analyses consisted of two independent runs of 10×106 generations and four Markov chains (temperature = 0.20), with trees sampled every 1000 generations. Of the 10,000 trees resulting, the first 2500 were discarded as burn-in. The remaining trees were combined to calculate the posterior probabilities in a 50% majority-rule consensus tree.
The best estimate of phylogeny, in which Aulacorhynchus whitelianus (Pantepui) is sister to A. sulcatus+A. derbianus (Cordillera de la Costa+Andes), was tested against two other topologies depicting alternate relationships between the Pantepui lineage and other Aulacorhynchus lineages. The first alternate hypothesis was based on the idea of ancient origins of Pantepui endemics (‘ancient tepui’ hypothesis), with a sister relationship between Pantepui taxa and a clade uniting taxa from all other Neotropical montane regions (Figure 2A). This possibility was represented by a topology in which A. whitelianus was sister to all other Aulacorhynchus. If that relationship were to hold, additional analyses should be used to test among Patepui origin and posterior dispersal, Andean origin and posterior dispersal, or vicariance. The second hypothesis was based on the findings of a previous phylogenetic study , in which the Pantepui lineage of Myioborus redstarts was sister to a lineage from the Cordillera de la Costa in northern Venezuela. This sister relationship between Pantepui and Cordillera de la Costa was represented by a tree in which A. whitelianus was placed as sister to A. sulcatus, the only Aulacorhynchus in the Cordillera de la Costa (Figure 2B).
A: Topology congruent with the ‘ancient tepui’ hypothesis, showing a sister relationship between Pantepui taxa and a clade uniting taxa from all other Neotropical montane regions. B: Topology congruent with a hypothesis where species from Pantepui and Cordillera de la Costa are more closely related to each other than to species from other regions.
We applied a Bayesian approach to hypothesis testing. This procedure consisted of taking the post-burn-in trees from the posterior probability distribution and filtering to detect trees compatible with the alternate hypothesis in PAUP. The number of the trees retained estimates the posterior probability that the hypothesis is correct .
Reconstruction of Ancestral Areas
To infer the biogeographic history of Aulacorhynchus, we followed two approaches: Dispersal Vicariance Analysis (DIVA) in a Bayesian framework, implemented in RASP (ver 1.107; ), and a Dispersal-Extinction-Cladogenesis model (DEC) developed for a likelihood framework, in Lagrange (ver. 20120508; ). Both methods allow estimating the probability of optimizing ancestral areas at each node of a phylogenetic tree. In RASP, the Bayesian binary MCMC analysis takes into account the inherent uncertainty of the phylogenetic inference by optimizing ancestral areas over multiple trees . In Lagrange, the DEC model specifies a instantaneous transition rate matrix between ranges along branches of a phylogenetic tree, and applies it to estimating likelihoods of ancestral states .
Geographic regions analyzed, included: Northern Andes (from the Huancabamba Depression north to western Venezuela), Central Andes (from the Huancabamba Depression south to Bolivia), Cordillera de la Costa, Pantepui, and Mesoamerica. The segregation among geographic regions takes into account the existence of major barriers that may limit dispersal and gene flow  and separate the geographic ranges of some populations or species in Aulacorhynchus. Also, treatment of the Northern and Central Andes as different biogeographic units takes into consideration phylogenetic studies suggesting that patterns of geographic diversification in the region may proceed in a south-to-north direction, matching the sequential uplift of the Andes , .
Areas were assigned to species and populations, as follows: (1) Central Andes–Aulacorhynchus prasinus from South America, A. derbianus, A. coeruleicinctis, A. huallagae; (2) Northern Andes–Aulacorhynchus prasinus from South America, A. derbianus, A. haematopygus, A. sulcatus from the Andes of Venezuela; (3) Cordillera de la Costa–A. sulcatus from Cordillera de la Costa; (4) Pantepui–A. whitelianus; and (5) Mesoamerica–Aulacorhynchus prasinus from Mesoamerica. Minor ridges such as Sierra de Perijá (between northern Colombia and Venezuela) and Sierra Nevada de Santa Marta (northeastern Colombia) were considered as part of the Northern Andes, to limit the number of areas included in analyses. Guided by current species distributions, we restricted the number of maximum areas to two, in both (RASP and Lagrange) analyses.
Bayesian binary MCMC (RASP) and DEC (Lagrange) analyses were run based on a set of Bayesian trees obtained using the Bayesian Markov chain Monte Carlo (MCMC) procedure implemented in BEAST (v.1.6.2; ). In this analysis, outgroups were excluded and species in Aulacorhynchus were constrained to be monophyletic. The justification behind this procedure was that species monophyly was supported by Bayesian posterior probabilities >0.95 in the previous phylogenetic analyses run over the same dataset (see Phylogeny, above). BEAST was run with a relaxed-clock model in which each gene was assigned its best-fit model of evolution, uncorrelated rates at each branch (estimated from a log-normal distribution), and a Yule speciation process tree prior. The program was run for 10×106 generations, sampling every 1000 generations. Convergence of chains to stationary was confirmed by inspection of posterior density of parameters in Tracer ; topological convergence was inspected in AWTY . After discarding 2000 trees as burn-in, a maximum clade probability tree was obtained using the remaining 8000 trees.
Bayesian binary MCMC analysis in RASP was performed over the 8000 post-burnin BEAST trees using the BEAST maximum clade probability tree as a ‘condense tree’. The analysis was conducted by setting the evolutionary model to F81+ G, running it for 50,000 cycles using 10 chains, sampling every 100 cycles and discarding 100 trees, and setting root distribution to null.
Likelihood DEC analyses in Lagrange were performed over the maximum clade credibility tree obtained in BEAST. In range constraints', adjacency of areas was allowed only between areas that were geographically contiguous (i.e., between the Northern Andes and all other areas, and between Cordillera de la Costa and Pantepui). Maximum range size was set to two areas, and ranges allowed in the analysis included all possible combinations within those imposed by adjacency and maximum range size. Dispersal was only allowed between adjacent areas. Baseline rates of dispersal and local extinction were estimated by the program.
Hypothetical Time of Diversification in Aulacorhynchus
To estimate an approximate time frame for the speciation of Aulacorhynchus toucanets, we applied a second BEAST analysis over the cytb mitochondrial gene dataset. We used a mean substitution rate guided by previous estimates for the avian genome (0.008–0.0095 Ma–1, for cytb  and ∼0.0105 for an average rate among birds ). The prior value for the Euclidean mean was set at 0.0105 Ma–1 and the prior standard variation of this parameter was set at 0.0034; both values were allowed to vary between zero and infinity. Because applying such general calibration disregards rate heterogeneity of the mitochondrial molecular clock across avian lineages , , this analysis was not intended to produce precise estimations of diversification times. Instead, we present the results of the analysis as a first approximation that should be tested against independent sources of temporal evidence when they become available.
Analyses were run using the GTR+Г+I model under a relaxed clock. Uncorrelated rates at each branch were estimated from a log-normal distribution. Species in Aulacorhynchus were constrained to be monophyletic. We used a Yule speciation process tree prior, running the program for 50×106 generations, and sampling every 5000 generations. Convergence of chains to stationary was confirmed by inspection of posterior density of parameters in Tracer  and topological convergence in AWTY . After discarding 2000 trees as burn-in, a maximum clade probability tree was obtained using the remaining 8000 trees.
Phylogenetic trees derived from ML and Bayesian analyses, based on the concatenated matrix of mitochondrial and nuclear data, were highly congruent, and showed significant nodal support for major clades. Given these results, we proceeded to analyze biogeographic information in the context of phylogenetic inference.
Hypothesis Testing and Reconstruction of Ancestral Areas
None of the topologies consistent with alternate biogeographic hypotheses was found among the set of Bayesian post-burn-in trees. This result indicates that the Bayesian posterior probability of these topologies is close to zero (contingent on the model, data, prior probabilities, and convergence of the MCMC ).
The Bayesian binary MCMC analysis in RASP and the likelihood DEC analysis in Lagrange produced different results regarding the ancestral area of Aulacorhynchus (Figure 3). RASP, inferred the Central Andes as the most probable ancestral area for the genus (Node I, P = 0.64); other probable areas included the Northern Andes (P = 0.23) and a combination of the Central and Northern Andes (P = 0.1). Also, the Central Andes was the most probable ancestral area reconstructed at Nodes II (P = 0.43) and III (P = 0.7), followed by the Northern Andes, and a combination of both areas.
Colors indicate geographic areas and combinations of up to two areas, and roman numerals indicate nodes of interest. A: Bayesian Binary MCMC Analysis in RASP; pie charts indicate the marginal probability of each area at nodes of interest; nodal support expressed as Bayesian posterior probabilities are showed above pie charts, with asterisks indicating Bayesian posterior probabilities = 1.00. B: Dispersal-Extinction-Cladogenesis analysis in Lagrange; numbers above splits indicate their relative probability; for simplicity, only splits summing ≥70 relative probability are shown.
Lagrange, on the other hand, showed a first split in which the Northern Andes is the most probable ancestral state for Aulacorhynchus (Node I, P = 0.3), followed by two equally probable splits (P = 0.2): (1) Central Andes+Northern Andes, and Northern Andes; (2) Northern Andes, and Northern Andes+Mesoamerica. Also, at Nodes II and III an ancestral origin in the Northern Andes is always more probable than an origin in the Central Andes.
RASP and Lagrange showed the Northern Andes as the most probable ancestral state for Node IV. Also, both analyses showed that species distributed in Cordillera de la Costa and the Pantepui, originated from dispersal of an ancestor from the Northern Andes. Presence of A. sulcatus in the Northern Andes is most likely explained by secondary dispersal from Cordillera de la Costa (RASP) or persistence in an area composed by the Northern Andes and Cordillera de la Costa (Lagrange).
Hypothetic Time of Diversification in Aulacorhynchus
According to the BEAST analysis (Figure 4), and considering the highest posterior density (HPD) interval at each node, the origin of taxa in the genus Aulacorhynchus occurred between the late Miocene and the early Pleistocene (95% HPD = 9.2–1.6 Mya). The analysis indicates that all currently recognized species likely originated prior to the Pleistocene, with one exception: A. sulcatus and A. derbianus split at some time between the mid Pliocene and the early Pleistocene (95% HPD = 3.6–1.6 Mya). Interestingly, the divergence between the Mesoamerican and South American lineages of A. prasinus (95% HPD = 6.9–3.8 Mya) is estimated as being older than the origin of all currently recognized species except A. coreuleicinctis. In other Aulacorhynchus species, subspecific taxa originated mainly within the last 2 million years. The separation of the Pantepui lineage, A. whitelianus, from its sister clade, A. derbianus+A. sulcatus (Andes+Cordillera de la Costa), occurred within the Pliocene (95% HPD = 4.9–2.6 Mya).
Biogeography of Aulacorhynchus
A hypothesis for the evolutionary origin of species in the genus Aulacorhynchus was first presented by Haffer  in his landmark work Avian Speciation in Tropical America. Therein, he proposed a sequential dispersal of an ancestral species from the Andes (A. derbianus) into the Pantepui (A. whitelianus), the eastern Cordillera de la Costa (A. sulcatus erythrognathus), the central Cordillera de la Costa (A. s. sulcatus), and west to the Venezuelan Andes and the mountains of northeast Colombia (A. s. calorhynchus). This sequence would have occurred during cool, glacial periods of the Pleistocene that lowered life zones across the humid montane forests. Haffer’s ideas were settled upon the prevailing taxonomy that suggested a common origin of the Andean and Pantepui populations of A. derbianus, and the observation of progressive changes in bill morphology (shape and coloration) in populations from contiguous but isolated montane areas. Molecular data showing that populations of A. derbianus from the Andes are more closely related to A. sulcatus than to A. whitelianus  indicate that Haffer’s proposal needs revision.
Topology tests showed that the phylogenetic arrangement found by Bonaccorso et al.  in which A. derbianus (Andes) and A. sulcatus (Cordillera de la Costa) are sister species, is significantly more likely than those implying a sister relationship between A. whitelianus and the remaining species in the genus, or a sister relationship between A. whitelianus and A. sulcatus. These results reject a biogeographic origin of Aulacorhynchus in the ‘ancient tepui’, as well as a close relationship between lineages from the Pantepui and Cordillera de la Costa.
Furthermore, ancestral area reconstructions showed an Andean origin for Aulacorhynchus (Figure 3). Whether the ancestral area of the genus is the Northern or the Central Andes, or both, is debatable. First, the RASP analysis (were the Central Andes had the highest probability of being the ancestral area) allowed dispersal among all biogegoraphic areas. In the Lagrange analysis (were the Northern Andes had the highest probability of being the ancestral area), we were able to restrict dispersal among non-adjacent areas. Thus, since RASP imposes no restrictions on dispersals, it allows inferring the Central Andes as an ancestral area in Node II. In RASP, this configuration in Node II has an important influence on the optimization of the Central Andes as the most likely ancestral area in Node I, which represents the ancestor of all Aulacorhynchus. This reconstruction is much less likely in Lagrange, because dispersal between Mesoamerica and the Central Andes was not allowed.
Second, the DEC optimization of ancestral states allows incorporating information from branch lengths to inform biogeographic inferences. This procedure avoids underestimating evolutionary change in terms of range expansions or contractions (dispersal vs. local extinction; ). Thus, in Lagrange, a proportionally long branch such as that conducting to A. coeruleicinctis (from the Central Andes), may have an important influence in optimizing other ancestral states over the Central Andes at a Node III, reducing the probability of optimizing the Central Andes at Node I.
Nevertheless, clear biogeographic patterns emerge within Aulacorhynchus. The Northern Andes has a higher probability of being the source of lineages that now inhabit the Cordillera de la Costa and the Pantepui (both analysis), and even the Central Andes (Lagrange). Most likely, it is also the source of the Mesoamerican lineage, since dispersal between the Central Andes and Mesoamerica is highly improbable. Our results are consistent with recent multi-species analyses in tanagers (Thraupini) indicating that the Northern Andes has been a source for lineages in other regions, with more dispersal events happening out of, than into this region .
On the other hand, examination of geographic patterns of Aulacorhynchus toucanets leaves an unresolved question. Considering that A. derbianus is distributed from Bolivia to northern Ecuador, and its sister species, A. sulcatus, is distributed from northwestern Colombia (Sierra de Perijá and Sierra Nevada de Santa Marta) to eastern Venezuela (Figure 1), what is the process explaining their absence from the Colombian Andes? Similar discontinuous distributions bridging the main Andes of Colombia are seen in other Andean birds such as Grallaria haplonota, Turdus olivater , and Buthraupis montana , among others. Whereas poor sampling across the region is still a possibility, in the case of the conspicuous Aulacorhynchus toucanets, we are inclined to support a scenario of local extinction. Whether the presence of A. haematopygus along the Andes of Colombia might have played a role in the potential competitive exclusion of a population of A. derbianus or A. sulcatus (or their ancestor) in this region, remains uncertain and difficult to test.
In the temporal realm, BEAST analyses indicated that diversification of Aulacorhynchus as a whole has been a process that started at ∼9.2–5.9 Mya, with all major lineages and most currently recognized species originating before the Pleistocene (Figure 4). Separation between A. whitelianus (Pantepui) and A. sulcatus+A. derbianus (Cordillera de la Costa+Andes) occurred sometime between 5.9 and 2.6 Mya, which negates the possibility of separation between Andean and Pantepui populations after the Pleistocene. Still, patterns of genetic differentiation seen in virtually all populations of Aulacorhynchus (except those in the A. prasinus complex) have a recent origin (∼2 Mya; Figure 4), and may have resulted from the sequential expansion and isolation of forest regions during the glacial cycles of the Quaternary. However, adequate testing of the processes behind this genetic pattern requires more in-depth population-level sampling and analyses.
Separation of South American and Mesoamerican ligeages of Aulacorhyncus prasinus occurred at ∼6.9–3.8 Mya. This estimate suggests that dispersal from South America northward into Mesoamerica may have taken place prior to the accepted time range estimate for the completion of the Isthmus of Panama ∼3.5–2.5 Mya . However, the closure date of the isthmus is based on evolutionary divergence of marine organisms and therefore must be considered as a minimum age . Other organisms might have dispersed across the Isthmus before the accepted 3.5–2.5 Mya estimate. In fact, exchange of vertebrate lineages between continents prior to the completion of the Isthmus has been documented for mammals , based on fossil data, and Diglossa flower piercers , Campylorhynchus wrens , some tanagers , Pristimantis frogs , and viperid snakes , among others, based on molecular data.
Biological exchange between North America and South America (or vice-versa) prior to the formation of the isthmus, may be explained by the existence of a string archipelago that connected both continents ,  and allowed dispersal by island hopping. However, a recent study based on stratigraphic data suggests that southern Central America had coalesced into a peninsula connected to North America by 19 Mya . Although this study does not provide a probable date for the closure of the isthmus, it seems that dispersal over a singular see channel would be more plausible than dispersal along a string archipelago. For Aulacorhynchus toucanets, as well as for other forest birds, movement across continents must have been constrained by the potential existence of suitable habitat along the peninsula and by their ability to disperse across the sea channel.
Regardless of interesting aspects that still need careful consideration and study, biogeographic analyses conducted on the phylogeny of Aulacorhynchus toucanets fit the predictions of a particular biogeographic hypothesis for the origin of Pantepui biotas. They support an origin in the Andes and posterior dispersal into the Pantepui, as first proposed by Mayr and Phelps , in a time window before the Pleistocene (contra Haffer ). However, we concede that our time estimates are only as good as the current methods for dating evolutionary events. Placing biogeographic processes in an evolutionary time framework across avian lineages may well have limitations in the calibration of evolutionary rates , , unexpected rate variation among lineages, and genetic polymorphism in ancestral populations .
The Origin of Pantepui Diversity
Our results support previous analyses based on shared distributions of birds , that pointed to Andean origins of many subtropical Pantepui endemics. To our knowledge, only two phylogenetic analyses have included sufficiently dense species sampling to uncover the evolutionary origins of Pantepui montane lineages. The first, by Pérez-Emán , was a pioneering analysis of the Neotropical redstarts in the genus Myioborus. Therein, the Pantepui clade (M. albifacies, M. cardonai, and M. castaneocapillus) was sister to the endemic species from the Paria Peninsula of the Cordillera de la Costa (M. pariae). This result contrasts with ours in indicating close relationships between Pantepui and coastal elements. Unfortunately, further biogeographic interpretation regarding routes of colonization of Myioborus between these areas was precluded by low phylogenetic resolution at interior nodes .
The second study, by Mauck and Burns , focused on Diglossa flowerpiercers. In this study, species from the Pantepui (D. major and D. duidae) form a clade that is sister to species from the Andes and Cordillera de la Costa. They identified the Andes as the ancestral area for the genus, with Pantepui and Central American species founded via single colonization events. Although they assigned the species from Cordillera de la Costa (D. venezuelensis) to the Andean region, it seems clear from their tree that it would represent yet another independent dispersal event. Their results are similar to ours regarding areas of origin and direction of dispersal. Also, in a broader biogeographic reconstruction for tanagers, Sedano and Burns  identified the Northern Andes as one of the sources of Pantepui taxa, with a low probability for dispersal events from the Pantepui to the Andes.
The emerging picture of the origin of Pantepui biota is complex, with some taxa dispersing from the Pantepui (bromeliads , ), others dispersing from the Andes (flowerpiercers ; tanagers ; Aulacorhynchus toucanets, this paper), or having close affinities with the Cordillera de la Costa (redstarts ). Phylogenetic studies in frogs  and Rapateaceae plants  have identified the surrounding lowlands as yet another source of diversity. The question of why only some species reach and colonize the Pantepui may require a deeper and more precise understanding of species’ dispersal capacities, competitive abilities, niche breath, and adaptability to new conditions, among others.
Conservation of Pantepui Populations
It is clear that morphological similarity must not preclude careful examination of isolated populations of the Pantepui region. Although diagnosable from one another, Aulacorhynchus derbianus and A. whitelianus are morphologically very similar, and yet, do not form a monophyletic lineage. The tepui redstart, Myioborus castaneocapilla, was formerly considered a northern population of the brown-capped redstart M. brunniceps, from western Argentina , . Thus, likely, Pantepui populations of many other organisms are evolutionary significant units  and may represent endemic species level taxa as well. This consideration is important given recent phylogenetic studies suggesting high rates of homoplasy in Neotropical birds, caused by natural  and sexual selection , processes that may conceal good biological and evolutionary species.
Excepting the highlands in the Brazilian Shield, the Pantepui is the most isolated montane region in the Neotropics. Separated from the Andes and the Cordillera de la Costa by the extensive lowlands of the Amazon and Orinoco basins , the region possess incredible potential for speciation and genetic diversification. Moreover, biotic  and phylogenetic differences  between eastern and western tepui groups, imply additional within-Pantepui divergence. Considering the high potential for cryptic speciation and genetic isolation of Pantepui populations, phylogenetic studies of these taxa are important from a conservation perspective. This information is particularly relevant when new taxonomic assessments may indicate a reduction in geographic range (or Extent of Occurrence), which is one of the criteria determining species’ conservation status .
List of tissue samples and GenBank accession numbers for sequences of species included in the present study.
Previous versions of this paper benefitted from insightful comments by A. Townsend Peterson, Charles Barnes, and three anonymous reviewers. Diego Páez kindly assisted in the installation of the Python plataform necessary for running Lagrange.
Conceived and designed the experiments: EB JMG. Performed the experiments: EB. Analyzed the data: EB. Contributed reagents/materials/analysis tools: EB JMG. Wrote the paper: EB JMG.
- 1. Mayr E, Phelps WH (1967) The origin of the bird fauna of the south Venezuelan highlands. Bulletin of the American Museum of Natural History 136: 269–328.
- 2. Huber O (1987) Consideraciones sobre el concepto de Pantepui. Pantepui 2: 2–10.
- 3. Huber O (1995) Geographical and physical features. In: Steyermark JA, Berry PE, Holst BK, editors. Flora of the Venezuelan Guyana. Saint Louis, Missouri: Missouri Botanical Garden Press. 1–61.
- 4. Briceño HO, Schubert C (1990) Geomorphology of the Gran Sabana, Guyana Shield, southeastern Venezuela. Geomorphology 3: 125–141.
- 5. Huber O (1988) Guayana highlands versus Guayana lowlands, a reappraisal. Taxon 37: 595–614.
- 6. Rull V (2004) Is the Lost World really lost? Palaeoecological insights into the origin of the peculiar flora of the Guayana Highlands. Naturwissenschaften 91: 139–142.
- 7. Maguire B (1970) On the flora of the Guayana highland. Biotropica 2: 85–100.
- 8. Hoogmoed MS (1979) The herpetofauna of the Guianan region. In: Duellman WE, editor. The South American herpetofauna: Its origin, evolution and dispersal. Lawrence, Kansas: The University of Kansas. 241–280.
- 9. Berry PE, Huber O, Holst BK (1995) Floristic analysis and phytogeography. In: Steyermark JA, Berry PE, Holst BK, editors. Flora of the Venezuelan Guayana. Portland: Timber Press. 161–191.
- 10. Steyermark J (1979) Flora of the Guayana Highland: Endemicity of the generic flora of the summits of the Venezuela tepuis. Taxon 280: 45–54.
- 11. Steyermark J (1979) Plant refuge and dispersal centers in Venezuela. In: Larsen K, Holm-Nielsen LB, editors. Tropical Botany. London, UK: Academic Press.
- 12. Sánchez-González LA, Morrone JJ, Navarro-Sigüenza AG (2008) Distributional patterns of the Neotropical humid montane forest avifaunas. Biological Journal of the Linnean Society 94: 175–171.
- 13. Weir JT (2009) Implications of genetic differentiation in Neotropical montane forest birds. Annals of the Missouri Botanical Garden 96: 410–433.
- 14. Chapman FM (1931) The upper zonal bird-life of Mts. Roraima and Duida. Bulletin of the American Museum of Natural History 63: 1–135.
- 15. Tate GHH (1938) Notes on the Phelps Venezuelan expedition. Geographical Review 28: 452–474.
- 16. O’Riain M, Jarvis JUM, Faulkes CG (1996) A dispersive morph in the naked mole-rat. Nature 380: 619–621.
- 17. Haffer J (1974) Avian Speciation in Tropical South America. Cambridge, Masachusetts: Nuttall Ornithological Club.
- 18. Croizat L (1976) Biogeografía analítica y sintética (“Pangeografía”) de las Américas. Biblioteca de la Academia de Ciencias Físicas, Matemáticas y Naturales 15: 1–890.
- 19. Graves GR (1982) Speciation in the Carbonated Flower-piercer (Diglossa carbonaria) complex of the Andes. Condor 84: 1–14.
- 20. Givnish TJ, Millam KC, Berry PE, Sytsma KJ (2007) Phylogeny, adaptive radiation, and historical biogeography of Bromeliaceae inferred from ndhF sequence data. In: Columbus JT, Friar EA, Porter JM, Prince LM, Simpson MG, editors. Monocots Comparative biology and evolution–Poales. Claremont, CA.: Rancho Santa Ana Botanic Garden. 3–26.
- 21. Givnish TJ, Barfuss MHJ, Ee BV, Schulte K, Horres R, et al. (2011) Phylogeny, adaptive radiation, and historical biogeography in Bromeliaceae: insights from an eight-locus plastid phylogeny. American Journal of Botany 98: 872–895.
- 22. Briceño HO, Schubert C, Paolini J (1990) Table-mountain geology and surficial geochemistry: Chimanta Massif, Venezuelan Guayana Shield. Journal of South American Earth Sciences 3: 179–194.
- 23. Gregory-Wodzicki KM (2000) Uplift history of the Central and Northern Andes: A review. Geological Society of America Bulletin 112: 1091–1105.
- 24. Hoorn C, Wesselingh FP, Steege Ht, Bermudez MA, Mora A, et al. (2010) Amazonia through time: Andean uplift, climate change, landscape evolution, and biodiversity. Science 330: 927–931.
- 25. Croizat L, Nelson G, Rosen DE (1974) Centers of Origin and Related Concepts. Systematic Zoology 23: 265–287.
- 26. Pérez-Emán J (2005) Molecular phylogenetics and biogeography of the Neotropical redstarts (Myioborus: Aves, Parulinae). Molecular Phylogenetics and Evolution 37: 511–528.
- 27. Mauck WM, Burns KJ (2009) Phylogeny, biogeography, and recurrent evolution of divergent bill types in the nectar-stealing flowerpiercers (Thraupini: Diglossa and Diglossopis). Biological Journal of the Linnean Society 98: 14–28.
- 28. Sedano RE, Burns KJ (2010) Are the Northern Andes a species pump for Neotropical birds? Phylogenetics and biogeography of a clade of Neotropical tanagers (Aves: Thraupini). Journal of Biogeography 37: 325–343.
- 29. Remsen JV, Cadena CD, Jaramillo A, Nores M, Pacheco JM, et al. (2012) A classification of the bird species of South America. American Ornithologists’ Union.
- 30. Navarro AG, Peterson AT, López-Medrano E, Benítez-Díaz H (2001) Species limits in Mesoamerican Aulacorhynchus toucanets. Wilson Bulletin 113: 363–372.
- 31. Puebla-Olivares F, Bonaccorso E, Monteros AEdl, Omland KE, Llorente-Bousquets JE, et al. (2008) Speciation in the Emerald Toucanet (Aulacorhynchus prasinus) complex. Auk 135: 39–50.
- 32. Bonaccorso E, Guayasamin JM, Peterson AT, Navarro-Siguenza AG (2011) Molecular phylogeny and systematics of Neotropical toucanets in the genus Aulacorhynchus (Aves, Ramphastidae). Zoologica Scripta 40: 336–349.
- 33. Weckstein JD (2005) Molecular phylogenetics of the Ramphastos toucans: Implications for the evolution of morphology, vocalizations, and coloration. Auk 122: 1191–1209.
- 34. Posada D (2008) jModelTest: Phylogenetic model averaging. Molecular Biology and Evolution 25: 1253–1256.
- 35. Zwickl DJ (2006) Genetic algorithm approaches for the phylogenetic analysis of large biological sequence datasets under the maximum likelihood criterion. Austin, Texas: The University of Texas at Austin.
- 36. Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19: 1572–1574.
- 37. Huelsenbeck JP, Rannala B (2004) Frequentist properties of Bayesian posterior probabilities of phylogenetic trees under simple and complex substitution models. Systematic Biology 53: 904–913.
- 38. Yu Y, Harris AJ, He X-J (2011) RASP (Reconstruct Ancestral State in Phylogenies) 2.0 beta.
- 39. Ree RH, Smith SA (2008) Maximum likelihood inference of geographic range evolution by dispersal, local extinction, and cladogenesis. Systematic Biology 57: 4–14.
- 40. Yu Y, Harris AJ, He X-J (2010) S-DIVA (statistical dispersal-vicariance analysis): A tool for inferring biogeographic histories. Molecular Phylogenetics and Evolution 56: 848–850.
- 41. Doan TM (2003) A south-to-north biogeographic hypothesis for Andean speciation: Evidence from the lizard genus Proctoporus (Reptilia, Gymnophthalmidae). Journal of Biogeography 30: 361–374.
- 42. Chaves JA, Weir J, Smith TB (2011) Diversification in Adelomyia hummingbirds follows Andean uplift. Molecular Ecology 20: 4564–4576.
- 43. Drummond AJ, Rambaut A (2007) BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evolutionary Biology 7: 214.
- 44. Rambaut A, Drummond AJ (2004) Tracer. Oxford, UK: University of Oxford.
- 45. Wilgenbusch J, Warren D, Swofford D (2004) AWTY: A system for graphical exploration MCMC convergence in Bayesian phylogenetic inference. Available: http://king2scsfsuedu/CEBProjects/awty/awty_startphp.
- 46. Fleischer RC, McIntosh CE, Tarr CL (1998) Evolution on a volcanic conveyor belt: using phylogeographic reconstructions and KAr-based ages of the Hawaiian Islands to estimate molecular evolutionary rates. Molecular Ecology: 533–545.
- 47. Weir JT, Schluter D (2008) Calibrating the avian molecular clock. Molecular Ecology 17: 2321–2328.
- 48. Lovette I (2004) Mitochondrial dating and mixed support for the ‘2% rule’ in birds. Auk 121: 1–6.
- 49. Peterson AT (2006) Application of molecular clocks in ornithology revisited. Journal of Avian Biology 37: 541–544.
- 50. NatureServe (2012). Arlington, VA: NatureServe Web Service.
- 51. Leal CA, Meneses HS, Gereda O, Cuervo AM, Bonaccorso E (2011) Ampliación de la distribución conocida y descripción del plumaje juvenil del Azulejo de Wetmore (Buthraupis wetmorei, Thraupidae). Ornitología Colombiana 11: 91–97.
- 52. Coates AG, Obando JA (1996) The geologic evolution of the Central American Isthmus. In: Jackson J, Budd AF, Coates AG, editors. Evolution and Environment in Tropical America. Chicago, IL: University of Chicago Press. 21–56.
- 53. Farris DW, Jaramillo C, Bayona G, Restrepo-Moreno SA, Montes C, et al. (2006) Fracturing of the Panamanian Isthmus during initial collision with South America. Geology: 1007–1010.
- 54. Marshall LG, Butler RF, Drake RE, Curtis GH, Tedford RH (1979) A radioisotope chronology for the Late Tertiary interchange of terrestrial faunas between the Americas. Science 204: 272–279.
- 55. Barker FK (2007) Avifaunal interchange accross the Panamanian isthmus: Insights from Camphylorhynchus wrens. Biological Journal of the Linnean Society 90: 687–702.
- 56. Weir JT, Bermingham E, Schluter D (2009) The Great American Biotic Interchange in birds. Proceedings of the National Academy of Science 106: 21737–21742.
- 57. Pinto-Sánchez NR, Ibáñez R, Madriñán S, Sanjur OI, Bermingham E, et al. (2012) The Great American Biotic Interchange in frogs: Multiple and early colonization of Central America by the South American genus Pristimantis (Anura: Craugastoridae). Molecular Phylogenetics and Evolution 62: 954–972.
- 58. Zamudio K, Greene H (1997) Phylogeography of the bushmaster (Lachesis muta: Viperidae): Implications for neotropical biogeography, systematics, and conservation. Biological Journal of the Linnean Society 62: 421–442.
- 59. Coates AG, Jackson JBC, Collins LS, Cronin TM, Dowsett HJ, et al. (1992) Closure of the Isthmus of Panama: The near-shore marine record of Costa Rica and western Panama. Geological Society of America Bulletin 104: 814–828.
- 60. Kirby MX, Jones DS, MacFadden BJ (2008) Lower Miocene stratigraphy along the Panama Canal and its bearing on the Central American Peninsula. PLos ONE 3: e2791.
- 61. Arbogast BS, Edwards SV, Wakeley J, Beerli P, Slowinski JB (2002) Estimating divergence times from molecular data on phylogenetic and population genetic timescales. Annual Review of Ecology and Systematics 33: 707–740.
- 62. Salerno PE, Ron SR, Señaris JC, Rojas-Runjaic FJM, Noonan BP, et al. (2012) Ancient tepui summits harbor young rather than old linages of endemic frogs. Evolution 66: 3000–3013.
- 63. Givnish TJ, Evans TM, Zjhra ML, Paterson TB, Berry PE, et al. (2000) Molecular evolution, adaptative radiation, and geographic diversification in the Amphiatlantic family Rapateaceae: Evidence from ndhF sequences and morphology Evolution. 54: 1915–1937.
- 64. Moritz C (1994) Defining ‘Evolutionary Significant Units’ for conservation. Trends in Ecology and Evolution 9: 373–375.
- 65. Cadena CD, Cheviron ZA, Funk WC (2010) Testing the molecular and evolutionary causes of a ‘leapfrog’ pattern of geographical variation in coloration. Journal of Evolutionary Biology 24: 402–414.
- 66. Omland KE, Lanyon SM (2000) Reconstructing plumage evolution in orioles (Icterus): repeated convergence and reversal in patterns. Evolution 54: 2119–2133.
- 67. IUCN (2001) IUCN Red List Categories and Criteria. Gland, Switzerland Cambridge, UK: IUCN Species Survival Commission.