Quick viewing(Text Mode)

Moore2009chap65.Pdf

Moore2009chap65.Pdf

, , barbets, and allies ()

William S. Moorea,* and Kathleen J. Migliab and divergence times of the Piciformes and its aDepartment of Biological Sciences, Wayne State University, Detroit, constituent to the level of families. MI 48202, USA; bDuke University Medical Center, Molecular As many as eight nominal families have been included Genetics & Microbiology, 210 Jones Building, Box 3020 Durham, in the Order Piciformes under various classiA cations: NC 27710, USA Picidae (, , and woodpeckers; ~28 gen- *To whom correspondence should be addressed (wmoore@biology. biosci.wayne.edu) era, 216 ), Indicatoridae (; ~4 genera, 17 species), Megalaimidae (Asian barbets; ~3 genera, 26 species), Lybiidae (African barbets; ~7 genera, 42 spe- Abstract cies), Capitonidae ( barbets; ~2 genera, 13 species), Ramphastidae (toucans; ~7 genera, 36 species, The avian Order Piciformes comprises two major lineages, the and Galbulae, which diverged as early as 70 million years ago (Ma). The and (Galbulae) also diverged relatively early, ~53 Ma. Later diversifi cation of the Pici gave rise to six additional families, beginning ~44–38 Ma. Molecular clock estimates for the origins of piciform clades, some estimated here, are consistent with the chron- ology of the sparse record. With the exception of the woodpeckers, species of several piciform families were abundant on the northern continents during the Paleogene (66–23 Ma) but are now restricted to the tropics.

7 e Order Piciformes is a diverse assemblage of species that vary greatly in size, appearance, distri- bution, ecology, and life history. To the extent one can make descriptive generalizations, they tend to be stocky, brightly colored with disproportionately large bills—taken to an extreme in the toucans—and arbor- eal habits (Fig. 1). 7 eir distributions are restricted to the Asian, African, and New World tropics, with the excep- tion of the woodpeckers, which collectively have a more expansive distribution that includes the Old and New World temperate regions. Most species are insectivorous but many eat fruit at least occasionally and the barbets are primarily frugivorous. Cavity nesting is pervasive in the order, as is particularly well known for the woodpeckers, which have adaptations of the bill, , and associated musculature and enervation that renders them extra- ordinarily eB ective at excavating nest cavities in wood. 7 e honeyguides are nest parasites, but parasitize only Fig. 1 A Red-headed ( erythrocephalus), cavity-nesting species. Here we review the relationships Picidae, from North America. Photo credit: R. Moul.

W. S. Moore and K. J. Miglia. Woodpeckers, toucans, barbets, and allies (Piciformes). Pp. 445–450 in e Timetree of Life, S. B. Hedges and S. Kumar, Eds. (Oxford University Press, 2009).

HHedges.indbedges.indb 445445 11/28/2009/28/2009 11:29:32:29:32 PPMM 446 THE TIMETREE OF LIFE

Ramphastidae 7 6 Capitonidae 4 Lybiidae

Megalaimidae Pici 3 Indicatoridae 5 1 Picidae Bucconidae 2 Galbulidae Galbulae

Paleogene Neogene CENOZOIC 50 25 0 Million years ago Fig. 2 The timetree of woodpeckers, toucans, barbets, and allies (Piciformes). Divergence times are shown in Table 1.

including two species of ), Galbulidae (jaca- were subsequently corroborated in a study by Lanyon mars; ~5 genera, 18 species), and Bucconidae (pu‡ irds; and Zink (7) based on protein electromorph characters, ~12 genera, 35 species). Collectively, the species assigned although the latter study did not include a to these families group into two clades that diverged rela- (Indicatoridae). 7 e tree hypothesized in these studies is tively early in the diversiA cation of (1, 2). One ((((Indicatoridae, Picidae), (Ramphastidae, Capitonidae)), comprises the Galbulidae and Bucconidae and the (Bucconidae, Galbulidae)), Outgroup). Although these other comprises the Picidae, Indicatoridae, and barbets studies produced congruent results, some intraordinal and toucans, regardless of how the barbets and toucans details remained to be determined, and monophyly of are assigned to nominal families. 7 ese two ancient lin- the order as well as identiA cation of its closest relative eages are considered distinct orders, Galbuliformes and remained contentious. Piciformes, in some classiA cations, but suborders, Pici Olson (8), Burton (9), and Sibley and Ahlquist (4) and Galbulae, of Piciformes in others. Regardless of the argued that Piciformes is polyphyletic with the Galbulae nomenclature, the evidence is strong that the Pici and (Bucconidae and Galbulidae) related to the coraciiforms. Galbulae are both monophyletic. Until very recently, Lanyon and Zink (7) were not able to test the hypoth- however, it was less certain whether the Pici and Galbulae esis of monophyly of Piciformes directly, but they noted were closest relatives, but three recent studies based on that their distance data supported a closer relation- DNA sequence data from nuclear-encoded genes strongly ship of Galbulae to a coraciiform ( Momotus) than to support this hypothesis (1, 2, 10). 7 us, it is reasonable to other piciforms, which suggests . However, recognize the Order Piciformes comprising two major recent DNA sequence-based studies consistently sup- lineages, the Pici and Galbulae, totaling ~403 species. port monophyly of Piciformes at statistically signiA cant Using Peters (3) classiA cation as a reference list for levels (1, 2, 10). Johansson and Ericson’s (1) study, based species usually included in an order called Piciformes, on exon segments from the nuclear RAG1 and c-myc the history of systematic groupings of those species genes and intron II from the myoglobin gene totaling into families, superfamilies, and suborders is a kaleido- 3400 nucleotides, was speciA cally designed to test the scope through the nineteenth and much of the twen- close relationship of Pici and Galbulae and thus mono- tieth centuries. 7 e early systematic history has been phyly of Piciformes. Combining all sequences from the thoroughly reviewed (1,4–6). Considerable stability of three genes, monophyly was supported by maximum inferred relationships was established with the cladistic parsimony (90% bootstrap) and Bayesian (100% poster- studies of Swierczewski and Raikow ( 5), and Simpson ior probability) analyses. 7 e maximum likelihood ana- and CracraJ (6) based on myological and osteological lysis based on the “Early Bird” data set also inferred a characters. 7 ese studies reached identical conclusions monophyletic Piciformes with 100% BS support (2). 7 e regarding relationships among piciform families, which “Early Bird” data set is more comprehensive and includes

HHedges.indbedges.indb 446446 11/28/2009/28/2009 11:29:35:29:35 PPMM Eukaryota; Metazoa; Vertebrata; Aves; Piciformes 447

Table 1. Divergence times (Ma) and their confi dence/credibility intervals (CI) among woodpeckers, toucans, barbets, and allies (Piciformes).

Timetree Estimates Node Time Ref. (10) Ref. (17) Ref. (22) Time Time CI Time CI

161.85370.589–5793.6107–80 255.05356.974–4473.692–53 340.943.638.151–2873.289–59 431.532.530.542–21– – 529.930.928.840–20– – 6 24.6 24.6 24.6 – 50.9 65–36 7 13.4 13.4 13.4 – 48 63–34

Note: Node times in the timetree represent the mean of time estimates from two studies (10, 17). PATHd8 analysis of DNA sequences from fi ve nuclear genes was conducted in ref. (10). Ref. (17) presents a reanalysis of the same data using the Bayesian program Multidivtime. Some divergence times were estimated here, which is detailed in the text.

~32,000 nucleotides from 19 nuclear gene regions and likelihood and Bayesian) and a dense taxon sample ena- 169 species. 7 us the inferred descent of species, listed bled him to resolve nearly complete relationships among by Peters, from a common ancestor is probably true; that species comprising the traditional Capitonidae and is, Piciformes is monophyletic and, thus, a valid taxon. Ramphastidae, including enigmatic lineages and/or With regard to relationships within and among fam- convergent lineages such as , Calorhamphus, ilies, the greatest uncertainty has involved the barbets Semnornis, and . and toucans. Traditionally the barbets, which occur in Moyle’s analyses strongly support the existence of three the African, Asian, and New World tropics, were con- geographically deA ned clades: Asian barbets, African sidered a single family, the Capitonidae, and the tou- barbets, and New World barbets, the last one including cans, which are restricted to the New World, were put the toucans. 7 is is consistent with the earlier A nding of in a separate family, the Ramphastidae (3). A series of Sibley and Ahlquist (4), and their recognition of three recent studies, however, have been consistent in show- families is warranted: Lybiidae (African), Megalaimidae ing that the traditional Capitonidae is paraphyletic with (Asian), and Capitonidae (New World). Recognition of a close relationship between the New World barbets and the Family Ramphastidae also seems warranted, but toucans and that the Asian and African barbets, each a the placement of the -barbet, Semnornis ram- distinct clade, are more distantly related. 7 ese recent phastinus, then becomes an issue. 7 is species, which studies are based on anatomical (9, 11) as well as molecu- appears very much like what one might imagine as the lar characters (4, 12–14). common ancestor of toucans and New World barbets, is 7 e most deA nitive study is Moyle’s (14) based on a data not strongly supported as the closest relative of either. set that combines 1045 nucleotides from the mitochon- In Moyle’s model-based analyses, Semnornis was clos- drial encoded cyt b gene with 938 nucleotides from intron est to the toucans but the ML bootstrap and estimated 7 of the nuclear-encoded β-A brinogen gene (β-A bint7). Bayesian posterior probabilities supporting this node Combining the rapidly evolving cyt b sequences with were only 54% and 67%, respectively. 7 e morphological the more slowly evolving β-A bint7 sequences produced intermediacy of the toucan-barbet, of course, is not a data set with phylogenetically informative characters the cause of uncertainty in the molecular analyses, but at low levels of homoplasy across the full time spectrum rather the internode that represents the common ances- of barbet . Moyle also had a relatively com- tor of Semnornis, with either the toucans or New World plete representation of taxa. 7 e combination of genes barbets, is short and deep in the tree. Regardless of the evolving at appropriate rates, analytical methods based exact placement of Semnornis, it, along with the toucans on detailed nucleotide substitution models (maximum and New World barbets, comprise a clade supported by

HHedges.indbedges.indb 447447 11/28/2009/28/2009 11:29:36:29:36 PPMM 448 THE TIMETREE OF LIFE

100% ML bootstrap and posterior probabilities (14). For hummingbird stem was used as a A xed calibration point simplicity of presentation, we will consider the toucans and the remaining 21 established minimum ages (including Semnornis) a family, Ramphastidae, and simi- for the ancestral lineages they were thought to represent. larly we consider the New World barbets (Capitonidae), 7 e PL dates averaged older than the corresponding Asian barbets (Megalaimidae), and African barbets PATHd8 dates. Moreover, the authors described a “ghost (Lybiidae) as individual families. range” in the PL analysis where the origins of lineages Monophyly of the other families and their inferred represented by fossils averaged 21 million years older relationships have been more certain. Bucconidae (puB - than the fossils themselves. Ericson et al. (10) thought birds) and Galbulidae (jacamars) are clearly monophy- the PATHd8 dates to be more reliable, but pointed out letic and related (4–7, 15). 7 e DNA–DNA hybridization that the systematic disparity between the ages of nodes data support the close relationship of Galbulidae and estimated by PATHd8 and PL leJ the answer to the ques- Bucconidae but indicate that the divergence between tion of whether diversiA cation of Neoaves came before these two lineages is ancient (4). 7 e Picidae includes the or aJ er the Mesozoic/Cenozoic boundary (66 Ma) wrynecks, piculets, and woodpeckers, each comprising a ambiguous, but they presented only the estimates for subfamily, and is closest to the Indicatoridae (5, 6, 16). the PATHd8 analysis (10). Brown et al. (17) reanalyzed In summary, the Order Piciformes as deA ned here is Ericson et al.’s (10) data using their revised fossil calibra- monophyletic, comprising eight families (Fig. 2). 7 e tions, a somewhat diB erent sequence alignment, and a closest relative of the Piciformes is probably a clade com- Bayesian methodology (Multidivtime, 21). 7 e conten- prising a subset of species traditionally assigned to the tion and disparate results between Brown et al.’s com- Order (2). 7 us, Coraciiformes is para- ment (17) and Ericson et al.’s (10) initial study and reply phyletic but Piciformes is not. 7 e deepest split within (18) stem from some misunderstanding of details in the Piciformes separates the Suborder Pici from Galbulae. original paper (18) but also from diB erent fossil calibra- 7 e Galbulae soon bifurcated to give rise to two fam- tions and analytical methods. 7 e last two illustrate how ilies, the Bucconidae and Galbulidae. In the Pici, the A rst important these factors are in estimating divergence bifurcation gave rise to two clades, one comprising the times based on molecular clocks. 7 ese estimates are Indicatoridae and Picidae and the other the geographical summarized in Table 1. clades of barbets plus toucans. Within the latter clade, In addition, we estimated times for the divergence the Asian barbet lineage (Megalaimidae) is basal and the of Ramphastidae, Capitonidae, and Lybiidae as follows next split gave rise to the common ancestor of African (Fig. 2). We downloaded 63 β-A bint7 sequences from barbets (Lybiidae) and New World barbets plus toucans. 57 piciform species representing all eight families. 7 e Finally, the toucan lineage (Ramphastidae) diverged sequences were aligned and average genetic distances from the New World barbets (Capitonidae). To date, no computed using the MEGA soJ ware (Tajima-Nei dis- study including a molecular clock analysis has focused tance with gamma parameter = 1.0) for all clades shown speciA cally on Piciformes. However, the broad-based in Table 1. Two regression lines were then determined for phylogenetic study of 75 families representing essen- the β-A bint7 distances as functions of the estimated ages tially all of Neoaves by Ericson et al. (10), and the ensu- inferred using (a) PATHd8 ( 10) and (b) Multidivtime ing comment (17) and reply (18) include molecular clock (17) for all other nodes in Fig. 2. 7 e slopes of the regres- estimates for A ve of the seven piciform nodes (Table 1); sion lines were estimated such that each line was forced the remaining two nodes can be estimated from relevant through the origin (i.e., genetic distance equals zero at sequences archived in the National Center for Biological the time the lineages diverged). 7 e slopes of the least- Information (NCBI) nucleotide data base. squares regression lines are the substitution rates as esti- Ericson et al. (10) determined sequences totaling 5007 mated by each of the two methods. 7 ese rates were then nucleotides for A ve nuclear gene regions: c-myc (exon interpolated to estimate the ages of the divergence of the 3), RAG-1, myoglobin (intron 2), β-A brinogen (intron 7, two nodes in the ((Ramphastidae, Capitonidae), Lybiida) β-A bint7), and ornithine decarboxylase (introns 6 and tree using the genetic distances for β-A bint7. 7, and exon 7). 7 ey used two computer programs with 7 e regression line for the Multidivtime data was a distinct rate smoothing algorithms, PATHd8 (19) and remarkably good A t with no apparent deviation from r8s, a penalized likelihood (PL) method (20), to esti- linearity or outliers (slope = 0.00456 nucleotide substitu- mate divergence dates. Calibration was based on 22 fos- tions/million years). By this method, the estimated age sils mapped onto the Bayesian tree estimated by Ericson for the split between Asian barbets (Megalaimidae) and et al. (10). A 47.5 million-year-old fossil representing the the clade of African and New World barbets (common

HHedges.indbedges.indb 448448 11/28/2009/28/2009 11:29:36:29:36 PPMM Eukaryota; Metazoa; Vertebrata; Aves; Piciformes 449

ancestor of Ramphastidae, Capitonidae, and Lybiida) is nuclear gene estimates. 7 us, diB erent estimation pro- 24 and 13.4 Ma for the split between Capitonidae and cedures might contribute a small amount to the discrep- Ramphastidae. 7 e regression analysis based on the ancy, but it is unlikely to be the major factor. 7 is leaves PATHd8 method is slightly more complex, because an the diB erences between the data sets, mitochondrial vs. aspect of the PATHd8 program is that it collapses inter- nuclear gene sequences, as a likely cause. Based on simu- nodes that are short or where there is uncertain phylo- lation studies, Moore et al. (23) showed that the mito- genetic resolution (10). 7 us, the two oldest splits in chondrial encoded cyt b gene would not perform well as Fig. 2 were collapsed to the same level and both dated a molecular clock for birds beyond ~10 Ma, whereas the at 53.0 Ma (Table 1) in Ericson et al.’s (10) PATHd8 ana- nuclear-encoded β-A brinogen intron 7 would perform lysis. It was apparent from the A tted regression line that well even to estimate time nodes older than 60 Ma. 7 e the common ancestor of all piciforms is a salient outlier rapid evolution of mtDNA and saturation by multiple and its age is underestimated by the PATHd8 analysis. substitutions leads to underestimation of the substitu- Discarding this point and recalculating the slope of the tion rate and over estimation of the age of nodes. For this regression line gives, remarkably, a value identical to reason, we did not include the time estimates based on that A tted to the Multidivtime slope. 7 us, the ages esti- the mtDNA in calculating the average age of piciform mated for the Ramphastidae, Capitonidae, and Lybiida nodes, although we did include the estimates in Table 1. divergences by the two methods are identical. A better It is clear that much work remains to resolve sources of estimate for the common ancestor of Piciformes, based uncertainty and that more attention needs to be paid to directly on PATHd8 calculated from the regression equa- diB erences in the “clocking” accuracy of diB erent genes tion, is 70.8 Ma. at diB erent time depths. Coincidental with the A nal revision of this paper, With the exception of the woodpeckers, the distribu- Brown et al. (22) published an additional temporal ana- tions of modern piciform species are restricted to tropical lysis of Aves based on 4594 bp of mtDNA sequence from regions of both the New and Old Worlds. 7 is might sug- 135 avian species (ND1, ND2, 12S rRNA, and nine tRNA gest that their distributions resulted from the breakup of genes). Although this study focused on diversiA cation Gondwanaland and the tectonic riJ ing of the southern of major lineages in relation to the Mesozoic/Cenozoic continents. 7 e African and South American barbets are boundary, estimated dates and conA dence intervals closest relatives (14), for example, and their divergence for A ve of the seven piciform nodes can be extracted may be attributed to the separation of from their timetree. 7 ese have been added to Table 1. from ~100 Ma (24). However, the timetree dates Disparities between the ages of nodes based on the are at odds with this hypothesis because the inferred mtDNA analysis and the earlier nuclear gene analyses, dates for all nodes are much too recent; speciA cally, the including those reported in the earlier Brown et al.’s (17) split between African and Asian barbets occurred only paper, are striking with the A ve mtDNA-based estimates 24.6 Ma. 7 e piciforms as a whole are relatively weak ranging 1.3–3.6 times older. A thorough exploration of P iers and dispersers. 7 e Asian barbets, for example, the cause of this apparent bias is beyond the scope of have not crossed Wallace’s Line, and only three species this paper; however, some tentative inferences are evi- of woodpeckers have crossed the line—and then barely. dent: 7 e fossil calibrations used in the mtDNA study Trans-Atlantic raJ ing of terrestrial vertebrates appears are nearly identical to those used in the nuclear gene to have occurred in rare instances ( 25) but would seem study of Brown et al. (17), which was a modiA cation of highly improbable in the case of barbets because they the set used by Ericson et al. (10). 7 is suggests that dif- have high metabolic rates and it is doubtful that even a ferences in calibration are not the cause of the dispar- large oceanic raJ could support their ecological needs for ity. DiB erences in estimation methods suggest a second a sustained journey. Nonetheless, the hypothesis of dir- potential cause. Both sets of nodal times were estimated ect dispersal of an ancestral barbet from Africa to South by Bayesian methods, but the nuclear gene estimates America via raJ ing cannot be absolutely rejected. A more were computed by the program package Multidivtime; plausible hypothesis is that the ancestral piciform species whereas, the mitochondrial-gene estimates were com- dispersed across Beringia and that they were once dis- puted by BEAST. Brown et al. (22) tested several com- tributed broadly across the temperate regions of Eurasia putational methods on the mtDNA data set: in general and North America, but these ancestral forms were sub- the BEAST estimates are higher than the Multidivtime sequently extinguished from the northern continents. estimates, but not as consistently, or of nearly the mag- 7 is is exactly the pattern observed in the fossil nitude, as the diB erences between the mitochondrial and record, albeit a sparse fossil record. Moreover, dates

HHedges.indbedges.indb 449449 11/28/2009/28/2009 11:29:36:29:36 PPMM 450 THE TIMETREE OF LIFE

associated with various fossil piciforms accord well 3. J. L. Peters, Check- of the World, Vol VI with the molecular clock estimates for the chronology (Harvard University Press, Cambridge, 1948). of nodes. 7 e fossil record shows that early piciforms 4. C. G. Sibley, J. E. Ahlquist, Phylogeny and ClassiA cation occurred in both the New and temper- of Birds: A Study in Molecular Evolution (Yale University ate zones in the Lower (~54 Ma) and perhaps Press, New Haven, London, 1990). 5. E. V. Swierczewski, R. J. Raikow, Auk 98, 466 (1981). the Paleocene (>54 Ma). Fossils assigned to the extinct 6. S. F. Simpson, J. CracraJ , Auk 98, 481 (1981). piciform Family Gracilitarsidae have been described 7. S. M. Lanyon, R. M. Zink, Auk 104, 724 (1987). for the Lower Eocene of , North America, and 8. S. L. Olson, Auk 100, 126 (1983). the Paleocene of (26). Middle Eocene to Lower 9. P. J. K. Burton, Bull. Brit. Mus. Nat. Hist. Zool. 47, 331 European fossils assigned to the extinct (1984). Family Sylphornithidae exhibit combinations of char- 10. P. G. P. Ericson et al., Biol. Lett. 2, 543 (2006). acters representative of the two major lineages, the Pici 11. R. O. Prum, Zool. J. Linn. Soc. 92, 3131 (1988). and Galbulae, of the Piciformes, and a cladistic analysis 12. S. M. Lanyon, J. G. Hall, Auk 111, 389 (1994). based on osteological characters established monophyly 13. F. K. Barker, S. M. Lanyon, Mol. Phylogenet. Evol. 15, 215 of a group comprising Gracilitarsidae, Sylphornithidae, (2000). and the crown Piciformes (26). 7 e oldest fossil repre- 14. R. G. Moyle, Mol. Phylogenet. Evol. 30, 187 (2004). sentative of the Pici dates to the early Oligocene, ~34–30 15. E. HoP ing, H. M. F. Alvarenga, Zool. Anzeiger 240, 196 (2001). Ma (27). 7 ese fossil dates are reasonably good matches 16. D. M. Webb, W. S. Moore, Mol. Phylogenet. Evol. 36, 233 to those inferred in the timetree. 7 e timetree indicates (2005). that the ancestral piciform lineage bifurcated ~70.5 Ma 17. J. W. Brown, R. B. Payne, D. P. Mindell, Biol. Lett. 3, 257 to give rise to the Pici and Galbulae lineages. One would (2007). expect to A nd the earliest fossils for the lineage leading 18. P. G. P. Ericson, C. L. Anderson, G. Mayr, Biol. Lett. 3, to living Piciformes and its two daughter lineages in the 260 (2007). Paleocene, Eocene, and Oligocene, which is born out by 19. T. Britton, C. Anderson, D. Jacquet, S. Lundqvist, the fossil record. 7 e inferred bifurcation establishing K. Bremer, PATHd8: A Program for Phylogenetic Dating of the Pici and Galbulae predates the Paleocene slightly, but Large Trees without a Molecular Clock, www.math.su.se/ the conA dence interval (89–57 Ma) is large and includes PATHd8/ (Stockholm University, Stockholm, Sweden, much of the Paleocene. 2007). 20. M. J. Sanderson, Mol. Biol. Evol. 19, 101 (2002). 21. J. L. 7 orne, H. Kishino, Syst. Biol. 51, 689 (2002). Acknowledgments 22. J. W. Brown, J. S. Rest, J. Garcia-Moreno, M. Sorenson, D. P. Mindell, BMC Biol. 6, 6 (2008). We thank P. Ericson and J. Brown for information. 7 is 23. W. S. Moore, S. M. Smith, T. Prychitko, in Proceedings of work was supported in part by the U.S. National Science 22nd International Ornithological Congress, N. Adams, Foundation. R. Slotow, Eds. (University of Natal, Durbin, 1999), pp. 745–753. 24. B. C. Storey, Nature 377, 301 (1995). References 25. N. Vidal, A. Azvolinsky, C. Cruaud, S. B. Hedges. Biol. 1. U. S. Johansson, P. G. P. Ericson, J. Avian Biol. 34, 185 Lett. 4, 115 (2007). (2003). 26. G. Mayr, Biol. Rev. 80, 515 (2005). 2. S. J. Hackett et al., Science, 320, 1763 (2008). 27. G. Mayr, Auk 122, 1055 (2005).

HHedges.indbedges.indb 450450 11/28/2009/28/2009 11:29:36:29:36 PPMM

Top View